Speakers, Voice Recognition, and Star Audio Fest 2009 Will Take Place More

SPEAKER CROSSOVER STUDY

June 2009 US $7.00/Canada $10.00

Tube, Solid State, Loudspeaker Technology

HEADPHONE AMP UPGRADE How to Achieve A Better Preamp Uncovering a Class D Amp Secret Designing a Studio-Quality Mike

BETTER Building a Single-Ended Transformer www.audioXpress.com

Cover-609.indd 1 4/27/2009 3:53:19 PM AudioNews

New from Stanton, the T.92 USB or go to http://embassysuites.hilton. features two two- turntable features a high-torque drive com/en/es/groups/personalized/ pole Butterworth motor, an S-shaped tonearm, and built- DALTXES-LSA-20090616/index. high-pass lters in DSP with key lock. Another newly jhtml to register online. (-12dB/octave). upgraded DJ deck, the T.55, includes For more, please a straight e AMP-15, from TERK Technologies go to www.tdl-tech.com. tonearm. Both (www.terk.com), is a high-performance are built on 15dB ampli er designed to improve the e Audio Engineering Society a rugged and signal strength needed for broadband (AES) 36th International Conference, sturdy base, cable and antenna reception, and “Automotive Audio—Sound in Motion” and come guarantees quality signal levels for will take place from June 2-4, at the equipped with a slipmat, dustcover, and longer cable runs. Supporting two-way Hyatt Regency in Dearborn, Mich. e Stanton’s 500.V3 cartridge. For more communication required for broadband program will cover the latest technology information, go to www.stantondj.com. cable devices, the AMP-15 includes a and techniques used in the industry power adapter and RG-59 connection such as active noise cancellation, For its third annual meeting, the Lone cable. loudspeakers, voice recognition, and Star Audio Fest 2009 will take place more. is presents an opportunity for in Dallas, Tex. at the Embassy Suites TDL Technology’s new Model 603A the automotive audio community to Dallas Park Central June 19-21. ere Tuneable Stereo Rumble Filter removes share experiences and de ne in-car audio is no registration cost for exhibitors, low-frequency noise from the turntable for the next generation. ose interested and no admission price, with the only and records without adding any audible in sponsorship or exhibiting should participant prices being to the hotel. di erence in the music. Operating contact [email protected]. To make reservations by phone call from a 24V DC “wall wart,” the 603A 972-234-3300 and tell them you are is tunable in 5Hz steps from 15 to Phiaton (www.phiaton.com) has coming to the Lone Star Audio Fest, 65Hz, has input resistance of 47kΩ, and unveiled the PS 300 NC noise-canceling

6 audioXpress 6/09 www.audioXpress.com

audionews609.indd 6 4/27/2009 3:41:20 PM headphones, which play even if the noise e Butter y, from Vestalife, is an e DP-200USB, from Denon canceling battery dies and features a iPod speaker dock which includes three Electronics, is a fully automatic turntable mute button. Wearers of the PS 300 feel interchangeable that allows users to easily convert LPs to the ear pads contract, while pressure is colored faceplates. It modern MP3 digital audio les. Featuring exerted on the front features 3.5mm Aux a see-through ip-up cover and built from of the speakers. input, LED volume high-quality aluminum, the DP-200USB Also featured with indicators, and dock includes an internal MP3 encoder and a these headphones connectors for any front panel USB port. e PC software is an adaptive iPod sizes. To learn contains Audio Waveform Recognition acoustic impedance more, please visit www.vesta-life.com. as well as Auto Track Divide. To nd out control, 40mm electrodynamic drivers, more, visit www.dm-holdings.com. 98dB sensitivity, and a maximum input Paradigm Electronics Inc. introduces power of 1000mW. the SUB 12, a 12″ driver which e Sonus is a pocket-sized sound level includes mineral- lled co-polymer meter range which includes Designed to o er a quick, easy way polypropylene cones with Resonance a combined sound- and for users to power up their circuits, Control Ribs, dual voice-coil designs, dose-meter sporting dual- SchmartModules are six single voltage and proprietary Air ow Ventilation frequency measurement, regulator devices from SchmartBoard. System cooling. e SUB 15 is the data-logging to a sizeable e company’s prototyping system 15″ model with all these features; both memory, and measurements is used to hand solder components also include switched mode power to support noise at work and onto SchmartBoards, and these new supplies, high-power transformers, environmental applications. power modules help ease the process noise suppression networks, glass/ ese new meters feature three Class 1 and even further. ese Regulated Power epoxy circuit boards, and advanced three Class 2 instruments to the IEC61672 Modules come in 1.5, 1.8, 2.5, 3.5, 5, short-circuit protection. Low frequency Standard, and two dosimeters. Measurement and 9V, and suggested retail price is extension in the SUB 15 is 12Hz, while range is 30-140dB, and parameters include $15. For more information, go to www. the SUB 12 has 16Hz. For more, go to Leq, Lmax, LCpeak, and Lepd. For more, schmartboard.com. www.paradigm.com. visit www.scantekinc.com. aX

audioXpress June 2009 7

audionews609.indd 7 4/27/2009 3:41:29 PM solid state By Luciano Peccerini Beyond G4OEP: Improving a Flea-Powered Headphone Amp This headphone amp upgrade adds four relay-selectable inputs, a numeric display, a phono-preamp stage, and an auxiliary output.

irst of all, I wish to thank Aren van Waarde (and Dr. Smith, a.k.a. G4OEP) for his article on F the G40EP (March ’08, p. 36). Yes, I accepted Mr. van Waarde’s challenge and built the headphone amp. After powering up the circuit, I was amazed what I heard! This amp is so simple and clean—no magic—with a straightforward circuit PHOTO 1: The G40EP. (a little bit different from TDA2003 manufacturer suggested topology) that works very well. I must confess that the circuit did not hold my attention at first look. After two weeks, I decided to search for the parts in my junk boxes and put everything together. The components are easy to find, with neither one “white-fly” nor esoteric “audio grade” component. In a few hours of heat and some smoke, the circuit was working very well, to my surprise. In fact, this simple circuit sur- passed my expectations. So, I decided to add the following improvements: • Multiple selectable inputs • Phono (MC) preamplifier with RIAA passive network • Volume control • Visual indication to selected input (a LED seven-segment subminiature display) • Auxiliary output (to feed my single-ended “A-class” 47 triode mode amplifier) • A reasonably fine-looking cabinet CIRCUIT FEATURES You can analyze these improvements separately in five parts: 1. The G4OEP, a TDA2003-based low- FIGURE 1: G40EP. gain amplifier (see the aX article for more details) 8 audioXpress 6/09 www.audioXpress.com

peccerini3040.indd 8 4/27/2009 3:43:17 PM 2. Power supplies: two separate single +12V DC to each MCap® RXF ± G4OEP and one dual 12V DC for the phono stage Radial Xtra Flat Capacitor 3. The phono stage itself 4. The MUX and digital section 5. The cabinet G4OEP This is the same circuit as in the aX article. But, due to the difficulty of finding R1/R101 (130Ω), I mounted a 270Ω in parallel (135Ω), which worked fine. Basically it’s a TDA car audio power amplifier operating at low gain so the quiescent point leads the amp to work at “A-class” MCap® RXF Oil in a very linear region of its transference curve. See Fig. 1 Table 1 Featuring the ultimate winding geometry (edgewise) for and . • extremely short, low-loss signal transmission, • extremely reduced residual-resistance (ESR), POWER SUPPLIES • remarkable low residual-inductivity (ESL). As I said before, I used three power supplies. Two of them Polypropylene capacitor-foil, alu metallized. Table 2 Fig. 2 Grouted winding against microphonic effects. are identical ( and ), and a symmetrical (positive • Fit-In-Adaptors now available. and negative) one for the phono stage (Table 3 and Fig. 3). These small transformers rated to 500mA are so simple, like MCap® Supreme all the circuitry discussed here. For the positive ones I used transformers rated to 110V AC/15V AC, single secondary coil. For the phono stage power supply, I used an 110V AC/15+15V AC center- tapped. In the symmetrical power supply for the phono stage, all transformers are in parallel with the mains, S0 and F1 MCap® Supreme Oil and Supreme Silver/Gold/Oil circuit. TubeCap® - Optimized High Voltage MKP MCap® Supreme PHONO STAGE MCap® Supreme Silver/Oil I have found many interesting circuits at the ESP website. MCap® Supreme Silver/Gold While searching for an RIAA passive network preamp, I encountered Mr. Elliot’s web pages, which included an MResist™ interesting topology that I changed by substituting the Supreme original Burr-Brown chips to the common TL074 (Texas Audiophile Resitors Instruments). My objective was to simplify the original cir- TABLE 1: Circuit Parts List Component Attributes Designation 10µF × 25V Polyester C1, C101 470µF × 16V Electrolytic C2, C102, C6, C106 Varied Foil Coils & Air Core Coils 1000µF × 16V Electrolytic C3, C103 220nF × 100V Polyester C4, C104 Silver/Gold 2.2nF × 100V Polyester C5, C105 Internal Wirings Exclusive Argentinean Distributor 100nF × 100V Polyester C7, C107 SK NATURAL SOUND www.naturalsound.com.ar TDA2003 Power car audio chip U1, U101 130Ω Metal film ¼W R1, R101 2k2 Metal film ¼W R2, R102, R3, R103 1Ω Metal film 1W R4, R104 See more audio innovations on www.mundorf.com TABLE 2: POSITIVE POWER SUPPLY and subscribe for our newsletter inf o @ mundorf.com Components Attributes Designation OEM and dealer inquiries invited 100nF × 250V Polyester C1-4, C6, C101-4, C106 MSolder™ Silver/Gold 4700µF × 25V Electrolytic C5, C7-9, C105, C107-9 MSolder™ Supreme 1N4007 Si rectifier diode D1-8 7812 1A positive regulator U1, U2 1A Glass 5 × 20mm F1 5 × 20mm Fuse-holder panel mount F1 110/220V-0/15V × 7.5VA Single secondary T1, T2 IEC male plug 3-pin Panel mounting V AC High End Components Made In Germany Since 1985

audioXpress June 2009 9

peccerini3040.indd 9 4/27/2009 3:43:17 PM TABLE 3: SYMMETRICAL POWER SUPPLY cuitry that uses four single audio grade Component Attributes Designation op amps from TL family (quad in a 14 100nF × 250V Polyester C1-4, C101-4 DIP package). 4700µF × 25V Electrolytic C5-7, C105-7 The result was very good: No PCB 1N4007 Si rectifier diode D1-4 layout, everything was mounted in a 78L12 400mW pos regulator U1 perfboard (pad pattern, instead of track 79L12 400mW neg regulator U2 pattern) with ground plane provided by 110/220V 15/0/15V × 7.5VA Center-tapped secondary T1 aluminum foil (used for cooking and glued with glue-stick inside the cabinet TABLE 4: PHONO STAGE PARTS LIST and correct grounded in a single point). Components Attributes Designation The schematics I used are in Fig. 4. See 22µF × 16V Electrolytic C1, C3, C101, C103 also Table 4. 22nF × 100V Polyester C2, C102 At Mr. Elliot’s website (http://sound. 82nF × 100V Polyester C4, C104 westhost.com/project06.htm), you can 1µF × 25V Bipolarized C5, C105 find more information about RIAA ac- 47k 1% metal film ¼W R1, R101 curacy and other significant parameters. 2k2 1% metal film ¼W R2, R102 If you want to improve the phono pre- 4k7 1% metal film ¼W R3, R103 180k 1% metal film ¼W R4, R104 amplifier according to your needs, drop 10k 1% metal film ¼W R5, R105 a note to aX and let us know what you 2k7 1% metal film ¼W R6, R106 did. The phono circuit has deep clean 100k 1% metal film ¼W R7, R9, R107, R109 bass and affordable mids (voice is very 820Ω 1% metal film ¼W R8, R108 clean), the highs are very honest, too. TL074 Quad-operational amp U1 You will soon note the huge differences of recordings among your vinyl records.

FIGURE 2: Positive power supply.

10 audioXpress 6/09 www.audioXpress.com

peccerini3040.indd 10 4/27/2009 3:43:18 PM FIGURE 3: Symmetrical power supply for phono stage.

FIGURE 4: Phono stage.

audioXpress June 2009 11

peccerini3040.indd 11 4/27/2009 3:43:20 PM FIGURE 5: Mux and digital section.

FIGURE 6: Connections diagram.

12 audioXpress 6/09 www.audioXpress.com

peccerini3040.indd 12 4/27/2009 3:43:21 PM MUX AND DIGITAL SECTION I then applied glue to a ground-plane from the backside when it is powered-up. I needed to use my headphones with made from aluminum cooking foil and The small white square in Photo 2 is the following different sources: a Ma- covered the inner box surface with glue the display, showing the selected input rantz CD-63SE CD player, a Pioneer from a glue-stick. Some sanding work and from the back-panel. Finally, a grounded VSX-4800 tuner, a Marantz SD-535 carnauba-wax application guarantees a 2mm aluminum plate covers the bottom dual tape deck, and a turntable (Garrard natural opaque finish. side of the MDF box. Small rubber feet B2-FG and/or a Thorens TD-166). So I chose the chicken-head style knobs glued on this plate sustain everything. I decided to build a relay circuitry to do to give a vintage look to the G4OEP Table 6 lists the final parts. this and more. Small numbers around headphone amplifier. I also built a pair of Figure 6 shows the system connec- the selector switch were not an option, acrylic round washers (30mm diameter, tions (one channel shown). I hope this because I needed to see the selections 2mm thickness, and 5mm in length) to little challenge encourages other aX read- from a distance. Figure 5 and Table 5 put around the knobs, just for aesthetics; ers to build their own units and share show the MUX and digital section. my intention was to illuminate with LEDs the results. aX I used three small sealed reversible re- TABLE 5: MUX AND DIGITAL SECTION PARTS LIST lays with two sections NO-NC each, to do the job, commanded by the S1 rotary Components Attributes Designation switch and the “OR logic” provided by 10µF × 25V Electrolytic C1 some diodes in a simple and old-fash- 1N4007 Si rectifier diode D1-5 ioned way. I used the same approach to 1N4148 Si signal diode D6-D18 7-seg display cath comm Subminiature red type D19 command the D19 (seven-segment CC 78L12 400mW positive regulator U1 display). Depending on the S1 position, 1k8 metal film ¼W R1-7 D19 displays “P” for phono input or “1,” 12V DC 2 reversible sealed JX2RC2 Metaltex RL1-3 “2,” or “3” for auxiliary direct inputs. Rotary switch EV4X1 S1 Note: If the switch is in the P (phono) position, all relays are displayed “off ” in the schematics. This is to facilitate construction, regarding normally open (NO) and normally closed (NC) contacts, depending on the kind of relays you are using. D19 is an LED seven-segment common-cathode (CC) device. I used a subminiature one (10mm × 7mm). As you can see, I used a LM78L12 TO-92 small regulator, because I supplied the power for this perfboard from unregulated +12V DC (in fact, 18V DC unregulated—taken before U1 or U2) from the G4OEP power supply. You can choose any one unregulated source to feed the MUX; left or right will work fine. I ex- perienced no audible interference. CONSTRUCTION PHOTO 2: Completed headphone amp. The cabinet is a simple MDF box with TABLE 6: MISCELLANEOUS PARTS five faces, no bottom end, covered with Components Attributes Designation Curupixa (pronounced Kooroopy- 100k LOG dual potentiometer P1 shah, a Brazilian wood) darkened foil. On-off switch SPDT switch 220V AC at 1A S0 I machined all holes in the MDF box Aluminum plate 150 × 260 × 2mm from the inside before mounting. After Aluminum foil Half square meter Ground plane machining, I covered the box and the 4 rubber feet 20 × 10mm wooden foil with contact glue (beware of RCA connector plate 4 dual female 4 selectable inputs RCA connector plate 1 dual female Auxiliary output poisonous vapors when using), let them 2 knobs Chicken-head style/beige Volume and Input selector dry for ten minutes, and carefully ap- P-10 stereo jack output Connector to headphones plied the foil over the box with a metal- Power cord for IEC connector lic, smooth, rounded mass to eliminate MDF 15mm thick box 190 × 95 × 270mm (DHW) 5 sides (without bottom side) air bubbles between the MDF and the 2 Acrylic washers 30mm diameter 2mm thick covering. audioXpress June 2009 13

peccerini3040.indd 13 4/27/2009 3:43:22 PM solid state By Ed Simon Removing the Six “Straight Wire” Kinks: Prologue to a Better Preamp Here are some—perhaps forgotten—tips on preamp design.

finally needed to replace my “pre- pedance. The phase shift may run from as added noise or grit to the sound. amp,” which was nothing more just under 1° to more than 7° at the Capacitive coupling of noise can be than a stepped attenuator and four upper frequencies. treated by field theory or approximated Ijacks in a small aluminum frame. In the worst case this is something a by circuit theory. For the circuit theory It just didn’t look right, and was a bit critical listener could hear. When you approach, assume there is a small capaci- limiting when changing sources. Plug- consider that the signal probably goes tor hidden in the circuit that couples the ging and unplugging cables have just through two or more cables (they may “straight wire” to the noise source. You too many disadvantages. be longer than 3′), the signal source can also model a capacitor to ground It is often said that the best preampli- probably does not drive capacitive loads from the conductor and even allow for fier is a “straight wire with gain.” But I as nicely as pure resistive loads, and the some attenuation from a shield. do not think we should set our sights capacitance may even be nonlinear, then Magnetic coupling depends on the that low. you begin to see where one “straight area of the conductor and strength of Electric circuit theory is a simplifi- wire” problem lies. the field. Moving the power transformer cation of how circuits behave. Even in The issue of inductance is a bit differ- away from sensitive areas is always a the limited scope of this theory, wire ent. Resistive losses do not vary with fre- good idea. Making circuit paths shorter has three characteristics of concern: re- quency. Capacitive losses increase with also provides benefits. If the “straight sistance, capacitance, and inductance. frequency, as do series inductive losses. wire” is a real cable grounded at both Many audio effects are not well covered The difference is that capacitive losses ends, it is almost certain that you will by this approach—for example, what load the source, while inductive losses have a magnetically induced voltage. happens when a semi-conducting crystal raise the load impedance. Inductive loss- Common source impedance is why forms in a conductor—but circuit theory es are also greater at high frequencies. lights flicker when a motor starts. When is a good starting place. In the 3′ cable you see a loss of about the desired signal shares a signal path 0.00785dB at high frequencies com- with something else, the other signal FACTORS pared to 0.00000000785dB at the lowest causes a voltage drop in that path. If It may be simplistic to assume that frequencies. Is it possible some could the other signal varies—say, from noise you want the lowest resistance possi- hear this? Probably not! or even signal—that variation and its ble without considering the actual use. companion voltage drop shows up in the The source impedance of most con- MORE CONCERNS signal path. sumer audio gear is about 100Ω at low The next problem of the “straight wire” Vibration is always present. Insulated frequencies rising at higher frequen- are the six sources of noise it is sus- wires will have triboelectric noise. Ca- cies, designed to be loaded by at least a ceptible to: radio frequency interference pacitors will react by modulating the 10,000Ω load. When you have resistance (RFI), capacitive coupling, magnetic spacing between electrodes and hence in the “straight wire” conductor, there coupling (or mutual inductance), com- their capacitance, resulting in an AC is a loss of signal level. Three feet of mon source impedance, vibration, and voltage from any bias placed across 24-gauge wire loaded by 10,000Ω has a moisture. them. Inductors may have movement loss of about 0.0000779dB. Increasing I treat RFI as a separate source be- in their windings, causing value shifts. the wire to 10 gauge drops the loss to cause it is usually from a far-off source. Even some resistors may generate noise 0.00000258dB. In audio there are gifted In reality it is a combination of capaci- from movement. people who do hear more than others, tive coupling and magnetic coupling. Moisture, of course, can cause cor- but not in this case. The cure is to shield the wire and gear, rosion. Whenever two different metals The second concern is capacitance even though that adds capacitance and touch and moisture is present, you pro- loading the signal source. The 3′ cable signal loss. Shields also have the prob- duce a small noisy battery. Many folks loads the source causing a loss of about lem known as skin effect—the lower the find always leaving their gear on makes 0.011dB for low source impedance to frequency of the RFI, the farther it can it sound better. Some of this is due to 1.16dB with a rising to high source im- penetrate the shield. RFI often shows up heat reducing the moisture in compo- 14 audioXpress 6/09 www.audioXpress.com

Simon3066.indd 14 4/27/2009 3:48:30 PM nents such as resistors and capacitors. as the ground reference, this will help to Both are affected to a small degree by reduce magnetic coupling. You get some moisture. Of course, the biggest problem of the advantages of a balanced line in can occur in switches, where the mois- the single-ended cable. ture may cause oxidation of the contacts. Of course, to keep the RFI out of the Sulfur or other compounds in the air ground, you should add a third shield. To Solen is bringing you the first audiophile grade two-way make this even worse. minimize the undesired energy, the sec- monitor amplifiers for the DIY The final issue in the “straight wire ond and third shields are grounded only market. Using all Polystyrene or with gain” is gain. If you are using a CD at the driven end. This prevents mutual Polypropylene capacitors in the player or tuner as your music source, impedance of the shielding system from signal path, gold plated RCA both of these are capable of delivering entering the signal path. socket, removable IEC power 1V RMS. Some high-powered profes- You can add a buffer amplifier at your chord and a high output power sional amplifiers will produce full power source to power the driven shield. Of transformer. with less than that. Even a low gain course, if you do not want to modify (26dB) power amplifier will deliver 50W your existing gear, an easier way is to AP-016 with that input. So for many cases you put all the buffers in a new preamp. The do not need gain—you need loss! Of buffers need to be decent, but not great course, if you have inefficient loudspeak- amplifiers. After all, they will be attenu- ers or perhaps want tone control, then ated by the amount that used to be ad- you may want some gain. equate just for noise. This is a good place for amplifiers such as the TL072 or any SHIELD SOLUTIONS FET input amplifier, because they are It is misleading to think of gain as only less susceptible to RFI. Of course, you voltage amplification. You may wish to probably can even get away with a gold- increase the current available to the load. en oldie such as an NE1558. I mixed This helps to reduce the effect of capaci- them up to see whether I could hear a tance or other loads inside the power difference. amplifiers or connecting cables. In my design I used several small It turns out there is a nice simple so- cards to become input or output circuits lution that will almost completely elimi- (Fig. 1). The sending source can be a CD $94.50 nate capacitive loading of interconnect- player, DVD, tuner, and so on. ing cables, increase the effective attenu- I designed two versions of the buf- The AP-016 is available in three ation of the shield, and add almost no fer card. One is a send card designed to versions. The AP-016A crossover noise of its own. bridge the output and provide a modest point is at 2Khz, the AP-016B is at That solution is the driven shield. In- current signal to the first active shield. It 2.7KHz and the AP-016C is at th stead of surrounding the signal carry- also tries to make sure all common mode 3.5KHz. All of them are 4 order ing wire with a grounded shield, use currents are accounted for properly. It Linkwitz-Riley crossover. one that carries the same signal! Now does this by grounding the source of the the capacitance is not to ground but to active shield’s current at the source. The Specifications: the same voltage. The insulation can be return signal wire, which was the ground HF Power Output: 30Wrms nonlinear, because there is no voltage and shield in a normal cable, is now just LF Power Output: 80Wrms across it, no nonlinearity. The inductance a wire going to the common of both THD: 0.03% S/N ratio @ rated W: 90db of the wire is very slightly reduced by so- connectors. The final outer third shield Input sensitivity: 1V lenoid effect and some reverse capacitive is there to try and scare RFI away. Input impedance: 22Kohms coupling, although this is not really of The send card can take either a direct 4th order X-over: 2KHz AP-016A practical value. wired input or a signal from an optional 2.7KHz AP-016B For RFI the interference now must RCA jack. The output is on a stereo ″ 3.5KHz AP-016C travel through an extra isolated shield, so ¼ jack with a case grounded binding Weight: 2.6Kgs (5.7lbs) this one change multiplies the loss of a post for the outer shield. The connecting Dimensions W x H x D: ″ single shield. Same deal for any capaci- cable has a matching ¼ stereo phone W: 137mm (5.4") tively coupled noise. plug and a wire lead on one end and a H: 218mm (8.6") To get even trickier, you can now standard RCA on the other. D: 81mm (3.2") cover this shield with a second insulat- The receive version of the card works Cut-Out W x H: ed shield, which will prevent the signal by making it easier to literally suck more W: 108mm (4.25") from inside leaking out and contaminat- signal from the source. It requires a spe- H: 190mm (7.5") ing the ground or return reference sig- cial cord that has an RCA plug for the AC Voltage: 115V / 230V nal. If you twist a single conductor wire input and a phone plug for the output. around this second shield and use this The received center conductor signal MAKE YOUR SPEAKERS ACTIVE! audioXpress June 2009 15

Simon3066.indd 15 4/27/2009 3:48:30 PM is buffered and sent backwards on the active shield. The next shield is now grounded at the receive end, where the induced current originated. The outer shield is connected to the send side because any noise from an outside source is best dealt with from the stronger signal side. However, the receive end of this shield is connected to the case through a small capacitor. The capacitor to the case keeps any RFI shielded, or at least all of the components at the same potential inside a Faraday cage. There is also a capacitor (C4) from the buffer common to the case. PARTS SELECTION One problem quickly showed up. It seems that some designers have tweaked their designs to work into a capacitive load! Some outputs are of such high impedance that the shield buffer acts as FIGURE 1: Input and output shield drivers. positive feedback. So placing a small capacitor 47 to 330pF silver mica (C1 and C2) between the signal lead and buffer/ preamp common when needed will restore the source signal to flat. A second problem is that a rail-to-rail op amp does not want to work into a capacitive load, so don’t use one. I used gold-plated Neutrik NRJ6HF- AU phone jacks for the ¼″ phone jacks. These come without mounting nuts, so be sure to order some. They fit nice- ⁷⁄₁₆ ly into a ″ hole. Neutrik also makes gold-plated NP3Xb phone plugs. I did not use anything fancy for the third FIGURE 2: Tone control circuit. shield binding posts. For the RCA connectors I tried three

FIGURE 3: Power supply regulator.

16 audioXpress 6/09 www.audioXpress.com

Simon3066.indd 16 4/27/2009 3:48:33 PM different types: all gold-plated audio- heat shrink and use it to pull the cable from the audio circuits by using two phile types. You can pick your favorite through. On long runs you can use a 10Ω resistors for each card. here; just be sure it can handle a large brass wire and solder it to the silver wire. You can use the cards and cables as diameter cable. After you clean the connectors with De- standalone devices. I used one as a guitar To construct some of the cables I used oxit, the cables are complete. cable buffer, which resulted in noticeably silver or silver-plated stranded Teflon Of course, if you are using audio cable better high end. I also chose to build insulated wire. For others I cheated and for the core, you can skip a step. these cards into my new preamp. started with a piece of shielded audio I positioned the send and receive In my preamp the signal lead from cable cut from a decent premade RCA- modules so that two receive and one each input card goes to a selector switch RCA cord. The quality of the signal send fit on an Expresspcb mini-card. via a piece of un-insulated dead soft conductors does affect the final product. There is nothing special about the cards, silver wire, which is attached to nylon I have found that thinner better metals which do not handle high-quality sig- standoffs to minimize vibration-induced do have an advantage. nals. I did isolate their power supplies signals. I used a 12-pole double throw The shield braid used to be a common item. I could only find it in the larger size at McMaster Carr (part #5537K26 and 5537K14). I used a third-hand jig to hold the connector and cable while soldering. I did not want any motion while the solder cooled because theory suggests that this promotes crystal formation. After soldering the RCA end center silver 22-gauge conductor, I slipped ¹⁄₈ over ″ heat shrink, cooked it, and then placed the rear cover. Next to go on was a thin shield braid that did not quite extend into the cover. I added a small FIGURE 4: Power supply top input, middle after regulator, bottom after shunt regulator. jumper connected to this at the non RCA end. Tape kept it in place so it would not short to the center conductor. A layer of ½″ heat shrink that did go NOW AVAILABLE! into the cover was next. More thin braid went over this, but again not into the The NEW cover. The jumper went to the RCA end for the send cables, the far end for the others. I wrapped a silver or silver-plated OLD COLONY insulated wire around this and soldered SOUND L AB it to the shell terminal. If this is a send cable, you need to connect the jumper and a short bit of uninsulated wire. The CATALOG next piece of ½″ heat shrink covered this. I connected the phone plug to the other end. I needed to cut the strain You’ll ﬁnd: Over 150 books and CDs relief short and ream it a bit to fit the t cable. The larger shield braid goes over on audio technology FIND THE ENTIRE the completed connectors, and then New test CDs t Software for design and trimmed and soldered to the appropri- t PRODUCT SELECTION ate jumpers. A bit of tape held this in measurement ON-LINE AT www. Sound Strobe and more place and Techflex tubing in red or black t audioXpress.com covered the outer shield. This was held test equipment in place at each end by more ¾″ heat shrink tubing that also covered the rear Old Colony Sound Laboratory, PO Box 876, Peterborough, NH 03458-0876 USA shells of the connectors. Toll-free: 888-924-9465 Phone: 603-924-9464 Fax: 603-924-9467 It was easy to push the braid and E-mail: [email protected] www.audioXpress.com Techflex over the cable, but the heat FIND THE ENTIRE PRODUCT SELECTION ON-LINE AT shrink sometimes was difficult. It was www.audioXpress.com easier to push a stiff wire through the audioXpress June 2009 17

Simon3066.indd 17 4/27/2009 3:48:35 PM FIGURE 5: Not so good DIY cable top, bottom preamp and cables. FIGURE 6: Preliminary tone control curves. Note HF rolloff and noise level hides THD inherent in test equipment.

(12PDT) rotary switch for the selec- time I will buy an attenuator already as- the trickiest part. In the late 70s, when tor. It is a military surplus silver-plated sembled, which is easier, works the same mixing boards were becoming common, large diameter ceramic rotary switch. I or a bit better, and actually costs less! the tone control issue re-emerged. Ini- treated the contacts with Deoxit. The The output of the attenuator original- tially it was more a question of how to signal return or “ground” leads from each ly went to the output card. An outboard make an effective low-cost tone control, input go directly to the output send buf- ±36V DC power supply fed an inter- but once the knockout punch was pub- fer card, where I establish my common nal shunt regulator to power the buffer lished by P. J. Baxandall in 1952, there ground reference point. A better point cards. A direct run went to each card was only quibbling about variations. would be at the return end of the at- establishing a second common ground at By the 70s, with cheap parts, ICs, and tenuator, but I did not figure that out the power supply, reducing any mutual biquad or other new filter topologies, until I finished. impedance signals. A single connection the issue became how should the knob The output of the selector switch goes coupled the output card to the power change the sound? to a stepped attenuator. I had a leftover supply. The problem now was not so much 24P4T Alps switch made for this pur- I used a shunt regulator modified to designing the circuit, but getting the pose. I made PC cards to hold 1206 size have more AC response; in this use with “feel” right. When you give the bass 1/3W surface mount resistors, a size that a known load and good shielding, RFI knob a quarter turn, you expect certain can be soldered without magnification. should not be a problem. Measurements changes. More or less change is awk- I used 94 of the IRC PFC-W1206LF- showed the shunt had more than 20dB ward. As loudspeakers improved, the ap- 03-XXXX-B tantalum nitride 0.1% 25 additional noise attenuation than with parent effect of the control was greater. ppm/C non-magnetic resistors to make just the regulator. The power transform- What was once a fine treble control the attenuator. I selected these because er, bridge rectifier, and first filter caps are now is too harsh on a modern tweeter. A as high-end audio parts go, these are in a small steel box 6′ away from the rest bass control that was fine on a tight sus- reasonably priced. I like the 1206 size of the gear. pension woofer becomes tubby on many because they are easier to solder. My The downside of this shunt regulator around today. attenuator has 10K input impedance is that it draws a higher startup current The standard Baxandall circuit for and varies on the output. I learned from than a series regulator. So it’s best to use a two-knob tone control is three ca- my previous “preamp” that most of my a time delay fuse. pacitors, two variable resistors, and a listening is done about 20dB down, so I few fixed resistors. To get smooth ac- jiggled my step sizes to have more reso- TONE CONTROL CIRCUIT tion the controls needed to be linear. A lution in that range. You need a tone control to complete the later takeoff used log pots and different After doing all this, I realized next preamp. The tone control was actually capacitor values to get a better feel. The

FIGURE 7: Bottom PC layout. FIGURE 8: Top copper layer.

18 audioXpress 6/09 www.audioXpress.com

Simon3066.indd 18 4/27/2009 3:48:39 PM PHOTO 1: Third-hand holding parts to be soldered.

FIGURE 9: PC cards silkscreen layer.

disadvantage of the Baxandall was that change the switches later. The PC cards RESULTS it required a center-tapped pot and a for the tone control are stacked on top of Measurements of the preamp showed pentode to provide the negative feed- each other to maintain a similar level of that it was actually better than one web back. wire lead path length and vibration level. DIY cable design. It was as good as a Today negative feedback is not a novel To keep out vibration I made some jumper cable across the test equipment application. Stereo, however, has placed a feet out of two aluminum discs and two for most of the audio band. There was few more demands on a tone control. A different foam rubbers. Both are quite a higher level of 60Hz hum than from stereo potentiometer with good track- soft—40 00 durometer for one and a a short jumper. This was interesting ing, high resistance, and a center tap is bit firmer for the other. The impedance because the fundamental of the power really hard to find. Using switches and mismatch helps to reduce vibration. supply was 120Hz, which was 105dB resistors gives me a great advantage. I I mounted the feet to a sub-plate ½″ down at the output of the regulator. So can select the resistor values to give me deep and added a bit of foam to seal I suspect this is due to the longer cables logarithmic control and still use match- this chamber to the chassis. I used 6-32 and greater area in the possible magnetic ¹⁄₈ ing capacitors. screws with pieces of auto store ″ ID field. I used regular carbon film resistors rubber hose as flexible spacers to attach When I first turned on the new pre- for the control switches. Once my crew the plate. amp, I heard a noticeable buzz in the and I have used the control a bit, it will I punched a 1″ matrix of holes in the sound. Then I realized the preamp was be practical to home in on the useful bottom of the chassis to make mounting set to tuner for a sports-talk station. range and pick exact values. Then I will components easy. The sub-plate now The buzz was just normal for them! On change to better switches and resistors. sealed the chassis. Inside was silica gel to everything else the system seems dead The circuit I used is deliberately very keep moisture down and as a bonus the quiet. close to the original design and is shown loose fill absorbs vibrations. The final So it is possible to do better than a in Fig. 2. environmental control was a strip of sil- “straight wire with gain.” You can have I used an LM4562 op amp, silver mica ver tarnish paper to remove sulfur and less noise, lower losses, higher output capacitors, and precision resistors for the keep the silver shiny new. current, selection ability, gain, and tone tone control cards because I need to only controls. aX

PHOTO 3: Bottom pan with gasket and silica gel, foot parts, and rubber PHOTO 2: Chassis wiring. tubing spacers.

audioXpress June 2009 19

Simon3066.indd 19 4/27/2009 3:48:42 PM tubes By Ari Polisois Three Novel Transformers for SE Amps Introducing the next generation of output transformers.

or many reasons, I am convinced place the gapped1 system that remained secondary2 (Fig. 1). that in the sound reproduction the best alternative to the push-pull. To save the AC of the primary, you amplifiers field the single-ended Being deeply interested in using the SE need to suppress the AC voltage that F topology, combined with tri- topology, but with more power, I re- fights against it, in the opposed winding odes, produces a better sound than any sumed the search to find a solution to (that I will name the “tertiary”). A suit- other. Many readers will agree with this unacceptable lack of efficiency. able capacitor shunting the tertiary can this statement, others will not. I must do the job, for it hits only the AC, not specify that most of those who support SC-OPT (Self-Compensated the DC, but you soon realize that the the push-pull layout are motivated by Output Transformer) primary’s inductance is severely affected, the power output they can get from this Who hasn’t thought about a system as you will see later. Consequently, if, on kind of amp, which can easily reach 50, based on an extra winding generating a one hand, it is quite easy to achieve the 70, 100 or more watts with just a pair magnetic field of the same strength as DC offset, and short-circuit the opposed of strong output valves per channel. On the primary’s, but in the opposite direc- AC potential, another problem arises, the other hand, the single-ended (SE) tion, in order to accomplish a complete and you must restore the primary’s in- units, in order to deliver such power, and automatic offset of the DC fields? ductance that has collapsed, due to this require very large and expensive output It is certainly the first idea that comes short circuit. transformers. to mind, if you overlook the fact that You know, however, that the reduction This explains why SE amps in excess if this works for the direct current, the of inductance is greater if the coefficient of 20W are not the rule. The cause is same offset also applies to an AC signal of coupling between the opposed wind- well known: the idle DC (direct current) applied to the primary. Consequently, ings is high3. This coefficient, in prac- at which the valves operate in their still no voltage would be transferred to the tice, ranges from very close to 1 (if the condition in a SE layout. coupling is tight) and falls below 0.9 When designing a standard SE (if the coupling is loose). It becomes transformer, you must consider that obvious that if you want to limit the part of its core magnetic headroom damages of the tertiary’s AC short- (that depends on the size: cross-sec- circuiting solution, you must reduce tional area and length of the mag- the coupling between, on one side, netic path in which the flux travels, the tertiary and, on the other side, as well as the quality of the ferrous the primary with its secondary(ies). material) will be “occupied” by the Irrespective of the coefficient of magnetization due to this direct cur- coupling between the main opposed rent. Enough availability must be re- windings, tertiary and primary, the served in the core for the alternating latter must have an extremely high current signals to prevent an excess coupling coefficient with its second- of flux density to cause its satura- ary, which means a good winding tion, in which case the distortion of geometry and interleaving. the sound increases. I started bench experiments on Many well-known scientists have this basis and included the physical spent some time trying to find a so- separation of the conflicting wind- lution to the SE transformer’s draw- ings, using suitable C or UI cores. backs, starting with the non-produc- But the real solution came when, tive DC field, but their suggestions FIGURE 1: The self-compensated output transformer. encircling the tertiary’s bobbin with were not considered justified to re- a few turns of magnetic band on 20 audioXpress 6/09 www.audioXpress.com

polisois2957.indd 20 4/27/2009 3:45:42 PM PHOTO 2: SC-SCC- SET showing common iron core.

PHOTO 1: SC-OPT with flux escape.

the outer side, I noticed a significant when equipped with an SC-OPT, de- Despite its use, for best results, of low reduction in the loss of inductance of livers much more power than with a internal resistance valves, this special the primary. Photo 1 shows the “mag- gapped transformer of the same size, is a SC-OPT features several advantages netic flux escape.” This device can be real, substantial improvement. with respect to sound quality5. considered an auxiliary magnetic core A pleasant surprise was the discovery An amp I listen to very often at home that absorbs part of the magnetic lines, that the sound it produces is extremely is fitted with two paralleled 2A3s per thus reducing the effective number of clean and airy, as many listeners noted. channel (equal to 400Ω internal resis- lines that the tertiary and primary have However, no scientific explanation was tance). It uses a self-compensated output in common (and, consequently, the cou- found for this sonic feature. This unit transformer and even if its bass range pling factor). was presented at the AES 118th conven- extends to only 46Hz, at –3dB, it is my As an example, I consider the case tion (Spain, May 2005, paper 6346). preference, because of the unequalled of an output transformer built for a 6C33C-B valve, the powerful tube with about 100Ω internal resistance. Initial primary inductance (measured, without load, with an inductance meter) = 4.8H (1200 turns – core area 7.6cm2 – magnetic length 0.43m). This inductance, with the tertiary (same number of turns) shorted, falls to 0.48H (1/10 of the initial). I calculated the coupling coefficient to be 0.95. The –3dB point at low frequencies results (theoretically) in 35Hz4. By adding six turns of a 0.35mm thick magnetic band around the tertiary bobbin, the primary inductance was restored (due to the magnetic lines leaking from the tertiary field in the main core to this additional external magnetic circuit) to over 1H. In spite of an improved bass response, the advantages I was seeking (reduced bulkiness and weight, compared to a standard SE output transformer) had not been fully reached, and, from an efficiency point of view, you could not consider the unit able to compete successfully against a push-pull transformer. Nevertheless, the fact that a SE amp using low internal resistance valves, audioXpress June 2009 21

polisois2957.indd 21 4/27/2009 3:45:44 PM midrange and high frequency sound ity, lower content of odd harmonics, ence, still being positively impressed. No quality. simplicity of the circuits, and so on). doubt that an SC-SCC OPT is worth a try, not only for the special sound, but SC-SCC-SET This transformer was also designated as also for its advantages (lower price, less In May 2006, engineer Giovanni Mari- “double-ended,” as opposed to the clas- than half of the two SE transformers ani, technical director of Graaf, Modena, sic SE transformer. it replaces, improved simplicity, and ef- Italy, specializing in output transform- Its disadvantage, however, is that some ficiency)6-7. er-less (OTL) amps, and I presented a [crosstalk]6 is present, but limited to the The curve in Fig. 2 pertains to a pro- paper describing another novel trans- low frequencies (thus not affecting the totype of an amp that has been duplicat- former, the SC-SCC-SET (split core, stereophonic effect). It is practically un- ed eight times by eight young students stereo common circuit, single-ended noticeable, if the loudspeakers are three (13- to 15-years old) in Italy (Photo transformer) at the 120th AES Conven- or more meters apart. 3), under the supervision of my friend tion held in Paris, France (paper 6831). As stated, the two primaries (left and Antonio Tucci8. This result proves that, Considering the fact that a loose cou- right channels) in this setup produce with this transformer, the SE topology pling in the same magnetic core allowed opposed fields (as in a push-pull trans- compares favorably to the push-pull, still some freedom for the primary winding former). The DC magnetic fields pro- keeping its appreciable character. to work, producing an ideal separation duced cancel each other, provided the with the help of the opposed DC fields, idle currents are of the same intensity 4 × SC-SCC-SET device both the right and left channels (two (but, as mentioned, this is not automati- After my initial enthusiasm, a doubt primaries with their closely coupled, in- cally ensured), whereas the AC ones, crept into my mind. Was it possible to terleaved secondaries) were fitted on the thanks to the opposed phases, do not achieve further improvements? same common iron core (Photo 2). This interfere with each other. Consider Fig. 3 and Photo 4. As ex- solution required a phase inverter, as How does the sound of this OPT plained in the AES 122nd Convention with the push-pull transformers. compare with that of an SC-OPT? Paper #7151 (Vienna, Austria, May The DC offset was also ensured, as What strikes me immediately is its rich- 2007), the number of primary/second- in the p-p OPT, by feeding the left and ness, due to the healthy presence of the ary windings’ bobbins is increased (two right primary coils, having opposed bass range. At ease with a jazz quartet as for each channel in the basic setup, with winding senses, with the same amount well as with a philharmonic orchestra, it the possibility of more), connected in of anode DC current. Obviously, there does not miss any detail and it handles such a way that the currents that cross was no self-offsetting feature, and the powerful transients effortlessly. them lead to an automatic compensa- balancing of the DC fields must be The sonic scene is filled and uniform tion of the magnetism, reducing the accomplished with a suitable, manual and, depending on the recording ap- same very close to zero, independently device. proach and orchestras, can give the im- of the idle DC intensities flowing to The advantages were: pression that the instruments are there the valves. • extended bass range (as low as 5Hz at in front of you, playing next to each The advantage of this layout is that, –3dB) with a very good high-frequen- other, but distinctly. The sound does with proper connections, it does not re- cy limit (Fig. 2) not “jump” from one place to the other. quire an external direct current balanc- • saving space, weight, and cost, The instruments can be heard as in a ing device—everything is done auto- with regard to the former SE natural, wide angle space. People who matically inside the transformer itself. It transformers6-7 are unaware of the output transformer’s also eliminates the need for matching • getting as much power as the push- secret notice a difference with the char- parallel valves. pull configurations, but with the acteristics of the sounds they are used Figure 4 shows that, in every branch, properties of the SE layout (sensibil- to, but they cannot describe the differ- the direction of the DC flow is the same,

FIGURE 2: Frequency range of the SC-SCC-SET. PHOTO 3: Amp with SC-SCC-SET.

22 audioXpress 6/09 www.audioXpress.com

polisois2957.indd 22 4/27/2009 3:45:45 PM PHOTO 4: SC-SCC-SET configuration with two bobbins per FIGURE 3: Physical layout of the 4 × 4 output transformer. channel.

but the winding sense of each primary is apart, the primary windings concerned expectations are fulfilled also as regards opposed to the other and that the pri- operating in series aiding condition are the sound quality. mary coil’s terminals are connected in not antagonist and they work freely. An a crossed way, from one branch to the important point is that if one transformer One OPT Used for other. and a pair of valves (or more) are used for Both Channels The power supply current flows in each channel, no [crosstalk] takes place. The In this case, the [crosstalk] is present, the primaries 1 and 3, connected in se- 4x-SC-SCC is currently undergoing ex- but as with the SC-SCC, it concerns ries, but on different branches of the tensive tests in order to check that the only the low frequencies with a negli- magnetic circuit, until it reaches the anode of valve 2. Primaries 2 and 4 are in the same situation as the other valve. Four fields are produced by the DC, that, under certain conditions, fight each other. As a result, the overall magnetization is close to zero. The simplest case is when all windings have the same number of turns (N1-N2-N3-N4) as in cases 1 to 5 in Fig. 4. No matter how strong or weak the anode currents are—their relative levels—the cancellation of the magnetization takes place. In cases 6 to 9 of Fig. 4, as a result of the four fields strength (aiding and opposed) produced by the windings operating in the magnetic path, whose al- gebraic sum is zero9, the magnetization of the core is also cancelled. Some other alternatives—with the number of turns of the primaries equiva- lent in pairs—also produce the same result (cases 6 to 9). However, some com- binations do not work, such as cases 10 and 11. Thus, the DC offset can be ensured in the favorable cases. With regard to the alternating current, due to the fact that valve 1 and valve 2 are modulated with signals 180° audioXpress June 2009 23

polisois2957.indd 23 4/27/2009 3:45:46 PM FIGURE 5: The split core—stereo compensated circuit SE transformer FIGURE 4: 4× SC-SCC offsetting conditions. (one magnetic core for both channels).

gible impact on the stereophonic effect. The Newest Products and However, it would be possible to control the [crosstalk] by shunting the upper Technologies are Only a Click blocks 1 and 2 with suitable capacitors, Away! for instance, as is done in the SC-OPT mouser.com to the degree (depending upon the value of the capacitor) that it does not cause a collapse of the other primaries’ induc- tances. With this solution the [crosstalk] is moved to lower frequencies. Readers with comments or questions about the behavior of the 4X SC-SCC are invited to e-mail Ari.polisois@wanadoo. fr, [email protected], lpolisois@ yahoo.it, or, preferably, A2Belectronic@ wanadoo.fr. Upco ming Developments Tests are being conducted on bench units to utilize fully and advantageously

• Over A Million Products Online the principles discovered with the for- • More Than 390 Manufacturers mer versions (SC-OPT, SC-SCC-SET, • Easy Online Ordering and 4× SC-SCC) in order to: • No Minimum Order a) Reduce the [crosstalk] (when the 4× • Fast Delivery, Same-day Shipping is used for both channels on a com- (800) 346-6873 The Newest Products for Your Newest Designs mon core). As mentioned, in a layout × Mouser and Mouser Electronics are registered trademarks of Mouser Electronics, Inc. Other products, logos, and company names mentioned using one 4 transformer per channel, herein, may be trademarks of their respective owners. with parallel valves, obviously, there is 24 audioXpress 6/09 www.audioXpress.com Mouser_AudioXpress_5-10-09.indd 1 3/11/09 11:40:56 AM

polisois2957.indd 24 4/27/2009 3:45:49 PM no [crosstalk]. This corresponds to a coil is exactly the same as those passing in the ACKNOWLEDGMENTS push-pull having SE sound character, other, which is practically impossible) to less The 4× SC-SCC transformer is the latest but it does not require an auxiliary than 1. step in a program that started about ten offsetting circuit. 4. Because of the load and other actual work- years ago to improve the sound of valve b) Extend the high-frequency range ing conditions, the effective inductance is differ- audio amplifiers. beyond the current limits, the goals ent. With the invaluable support of engi- being 5-10Hz and 70-80kHz at 5. There are many valves having a low internal neers Menno van der Veen and Pierre –3dB. resistance (6C33 – 6336 – 6080 – 6550 – KT, and Touzelet as well as, most recently, Gio- c) Obtain the offset and balancing con- so on, triode connected) that can eventually be vanni Mariani of Graaf, Italy, sensible im- ditions automatically, without the use reduced by putting several tubes in parallel. Until provements have been achieved that led of any external device. now used for other applications, such as voltage to the application of several patents, five d) Improve the simplicity of the sin- regulators, this family of valves comes, luckily, of which have been granted by the French gle- or double-ended amps by inte- back to active life. Most of them do not require INPI (Institut National de la Propriété grating the loudspeakers’ crossover extremely high anode voltages. Industrielle). units into the output transformer, 6. This novel device was presented at the AES I am also grateful to Edward T. Dell, Jr., for instance. 120th convention, held in Paris in May 2006 Editor of Glass Audio and audioXpress mag- (paper 6831). More details on the transposing of azines, for publishing my first article on the This feature, which took years of re- the low frequencies are given in the above paper. subject (“A Universal Phase Splitter,” p. 32, search, is practically concluded and will 7. To better illustrate this result, the OPT GA 1/00) and subsequent ones, and also to be the subject of a presentation at a (bi-channel) in Photo 2 replaces two OPTs 50% the staff of the Italian magazine Costruire coming AES convention. bigger in size. HI FI, for printing several other articles, Today, the SE topology has a new 8. Thanks to Antonio, these kids have discov- and, recently, also to Elektuur and Elektor life and the future is promising. I have ered the sound of a valve amp and some of them Deutschland of the international periodical been striving to give DIY audiophiles will surely become audioXpress readers. of electronics, Elektor. Finally, I would like a wider choice of transformers (output 9. The electric fields strength of each winding, to mention engineer Menno van der Veen or interstage) they can fit to their amps, (producing aiding or opposing magnetism) result (worldwide known for his works in the according to their objectives. On be- from the product:- N (number of turns) × Ia (DC audio reproduction field), whose support half of engineers Mariani, Touzelet, and anode current). has been of the utmost importance. myself, many thanks for your interest in this field. Please do not hesitate to contact us with any questions or comments. aX REFERENCES 1. Why a gap in a SE OPT (or choke)? Just because you must avoid the saturation of the core due to an excess of permeability, which is an improvement on the core magnetization capacity obtained by inserting a ferrous or similar material inside the windings. However, permeability has a particular characteristic that you might consider a defect: It changes with the field strength, and when the latter is relatively too strong, it starts lowering, due to the saturation of the core. When this happens in an audio transformer, the sound becomes greatly distorted. 2. In practice, this is absolutely true as far as the low-frequency end is concerned, but you still get a weak one on the high-frequency side. 3. In the case of opposing coils in series, such as the primary and tertiary of the SC-OPT, the formula that applies is: Lt (total inductance) = Lp + Lt – 2M, where Lp is the primary inductance, Lt the tertiary’s, and M the resulting mutual inductance according to their coefficient of coupling, calculated as follows: M = k*√(Lp*Lt), where k is the coupling factor (equal to 1, if the number of lines of force passing through one audioXpress June 2009 25

polisois2957.indd 25 4/27/2009 3:45:51 PM solid state By Wim de Haan The Pragmatic Class D Amplifier With the "best-kept secret" in audiodom, you can construct your own Class D amp.

lass D amplifiers are gain- amplifier using B&O’s ICEPower tech- SMD components throughout. Because ing in popularity these days. nology based on the 200ASC board as the modules use an EMI/RFI PCB en- Hypex and Tripath offer great used by Cary Audio and Jeff Rowland. closure, you cannot determine compo- C products for DIY enthusiasts. Searching eBay is often part of my nent locations or any device numbers. Class D has some advantages over class daily routine, and one day I came across So the amplifier technique and circuit A, AB, B, or tubes—one of which is a tiny amplifier that was completely un- design remain a mystery. its high efficiency. At normal output known to me. Even the company’s name Pragmatic Communications Systems power there is no need for heatsinks, Pragmatic didn’t ring a bell, but it piqued has been selling these class D products simply because heat dissipation is al- my interest nonetheless. The ad listed since 1996. Prasanna Shah, who de- most nonexistent. At higher output the following features: class D, low dis- signed this class D module, says the fol- only a rather small heatsink is required. tortion of typical 0.01% to 0.05%, no lowing: “One reason for our anonymity Another advantage is the amp’s physical ripple in the passband, and audiophile is that we have not been able to spend size and weight. sound. money on magazine advertisements. We One of the very early class D designs Searching for “pragmatic amplifier” have invested money in technology and was the Sinclair X-10 amplifier, which in the Google search engine, I was dis- performance rather than advertisements. was sold in the early 60s as both a kit appointed by the number of related We have sold our amplifiers to lovers of and as a complete product. This power hits. Either this amp was a forgotten high fidelity all around the globe (Swe- amplifier yielded 10W of output power treasure or just not a good item, but I den, Norway, Denmark, Switzerland, but, in reality, produced only a couple of decided to take a risk anyway. I pur- UK, Germany, the Netherlands, Bel- watts. Those who successfully built the chased three of the ten pairs of ampli- gium, Italy, France, Hong Kong, Sin- kit reported that the amplifier radiated fiers the eBay auction offered. The final gapore, Australia, USA), and most of like a radio transmitter. price was $115 US—a good deal—and the sales have been word-of-mouth In the 80s I repaired high-end audio I later regretted that I did not buy all referrals.” for a specialized audio shop. Many rare ten pairs. Shah continues: “The Pragmatic am- amps passed through my hands, includ- plifiers have been tested against very ing the heavily-built Sony TA-N88 THE PRAGMATIC expensive Mark Levinson amplifi- audio power amplifier. This model was CLASS D MODULE ers at Madrigal Audio Labs’ listening one of the first commercial amplifiers The modules (Photo 1), measuring only station. . . at listening sound levels they to use class D pulse width modulation 8.7 × 5.1 × 2.0cm (L × W × H), use could not tell the difference between the (PWM) technology, along with Pragmatic amplifier and Mark the use of a switching power Levinson amplifiers, except that supply. the Mark Levinson amplifiers At that time I did not really were an order of magnitude pay attention to this. My inter- more expensive, weighed more ests in class D started in 2000 than 100 times more, and dis- after reading an article on Tri- sipated tremendous amount of path’s class D technology. I even heat energy into the room.” ordered some modules. How- ever, for some reason, I never THE PRACTICAL got around to using them in a REALIZATION project. Last year I listened to These tiny amps are very easy Tripath and Hypex UcD based PHOTO 1: Pragmatic mono class D module. to use. Each amp comes as a class D amps and built a class D complete working module, so 26 audioXpress 6/09 www.audioXpress.com

dehaan3067.indd 26 4/27/2009 3:42:42 PM there is no need for component sol- I first set up a laboratory power sup- off, with the speakers not suffering from dering or stuffing a circuit board using ply and two Pragmatic modules, with any dangerous pops or clicks. I built this SMD devices and no need for adjust- no enclosure—just some wires laying amp for a total cost of less than 70 euros ments. around on the floor. All worked without (about $65 US). Each amp comes with three pairs of hum or hiss. Because the first amplifier per- wires: power supply (red and black), line I built three amplifiers. Amp One formed so well, I built another, bet- input (orange and black), and a speaker (Photos 2-4) uses an ABS enclosure ter-looking, amp. Amp Two (Photos pair (positive output marked on PCB). and a Dell SX280 miniPC 12V DC 5-7) uses a switching power supply by All you need is to connect the 12V DC switching power supply, both of which XP Power (Farnell 1337158), which PSU (at least 5A per channel) and a fine I bought for not too much at a radio was a leftover from a former project. line input signal (plus, of course, speak- flea market. The standby function of A PSU by Mean Well, such as the ers) to have a full power amp. the PSU is used for switching on and SP-150-12 or SP-100-12, would be a better choice, pricewise, but the XP PSU fits the used enclosure perfectly, and the specs are great. Amp Two uses a San- sui VU meter, which I also bought at a radio flea market. The VU meters use a dedicated DIY circuit board by JLM Audio Australia, and these are directly mounted PHOTO 5: Front Amp Two. on the VU meters. This circuitry is powered by the PHOTO 2: Prototype Amp One. 12V DC from the switching power supply. The 12V scale lighting uses a flexible white LED strip. Because I still had one amplifier pair left, I built a third amplifier (Photos 8-10). This amp uses a con- ventional power supply with a toroidal mains transformer and a voltage regulator; I used one single 120VA transformer and a LM338 voltage regulator for each channel. The LM338 comes in TO-3 housing and is good PHOTO 3: Overview inside Amp One. PHOTO 6: Amp Two overview. for 5A. I also purchased all parts for this amp at a radio flea market. The LM338 goes for 1.50 euros and the 120VA mains transformer for 20 euros. I bought the ABS enclosure for just 15 euros. Use of a switching power supply is interesting, not only because of its price and compact size, but also because this kind of power supply works on a similar principle PHOTO 4: Amp One. PHOTO 7: Amp Two inside overview. as that of a class D (or pulse width modulation) ampli- audioXpress June 2009 27

dehaan3067.indd 27 4/27/2009 3:42:44 PM fier except that the reference voltage of COMPANY SPECS istics a PWM amplifier is the varying audio Power: 12-15V DC • Flat frequency response from 10Hz to signal instead of a fixed voltage as in a Continuous power: 24W per channel 50kHz switching power supply. Peak Power: 60W • No ripple in the passband Signal-to-noise ratio: >100dB • Very high signal-to-noise ratio TECHNICAL ASPECTS Dynamic range: >100dB • Very high dynamic range Its monoblock design provides extreme- Frequency response: 20Hz to 20kHz • 24W (peak 60W) ly clean state-of-the-art power. The • Only 12V DC power required SIWA-248 unit accepts any volume- Applications: Features and Benefits • Module works with 4 and 8Ω loud- controlled line level input and uses a • Compact monoblock amplifier speakers 12V-15V DC power supply. It has con- • Residential or commercial audio am- • Very low and uniform distortion on tinuous power of 24W and a peak power plification compact design all frequencies and very high signal- of 60W. Advanced protection circuitry • High efficiency to-noise ratio and dynamic range. guards against system damage from • High clock frequency overload, overheating, or short circuits. • Very linear gain and phase character- SOUND IMPRESSIONS TABLE 1 To evaluate the sound performance, How to connect the Pragmatic module I used a large Tannoy D700 and a power supply...... +12V DC = red...... ground = black small-size Dynaudio. The Dynaudio line input...... signal in = orange...... ground = black is based on a T330D Esotar tweeter speakers...... - output not marked on PCB...... + output is marked on PCB and 17WXL04 low mid unit, both by Please note: the amplifier circuitry is designed in such a way that you can not connect the negative loudspeaker Dynaudio. At first you notice the ease output to ground. with which music is reproduced. Sound

PHOTO 8: Front Amp Three.

PHOTO 10: Amp Three inside.

PHOTO 9: Amp Three overview. PHOTO 11: The Pragmatic SAA-224 stereo standalone amplifier.

28 audioXpress 6/09 www.audioXpress.com

dehaan3067.indd 28 4/27/2009 3:42:45 PM is very clean, dynamic, and precise with setup. Musicality is enormous; they the modules I bought on eBay were a beautiful soundstage. The Dynaudio’s can rock and shake. Intimacy, imaging, likely taken from a Pragmatic SIWA 4Ω load and 84dB sensitivity are no soundstaging, and high resolution are in-wall configuration. However, the problem for these low power modules; its strong points. I guess these mod- Pragmatic Company sells modules as however, the Tannoy’s 94dB sensitivity ules are probably the best-kept secret in standard products. Check out their web- allows headbanging levels. audiophile circles. And remember: this site for the standalone amplifiers such Big Band sound, such as in Gerald class D design dates from 1996, which as the mono MBA-124 and the stereo Wilson’s In My Time CD, drives both is remarkable because class D back then SAA-224 amplifier (Photo 11), as well speakers to a very high level. Music is re- was not in the picture nor as developed as in-wall configurations such as the produced with great authority and ease, as today. SIWA-224; the first two are beautifully rhythm and pace are excellent, and it is Just before publishing deadline, I made standalone units with external also free of any distortion artifacts. Also, found out that the Pragmatic Company power supplies. A 12-channel version is at low and moderate levels all music can never sold these modules separately, so also available. aX be felt and heard; its dynamic behavior is excellent. Chet Baker’s Great Last Concert ABOUT THE DESIGNER cial applications. The success of these (Enja) sounds impressive, with very Prasanna Shah is the Chief Pragma- products has led to a strong and dy- good tonal balance and great atmo- tist and Thinker at Pragmatic Com- namic technology base, which has sphere/ambience. Piano solos such as munications Systems, Inc. He earned been applied to a new generation of classical pianist Arcadi Volodos’ Live in his M.S. degree in Electrical En- products such as CATSTM and others Carnegie Hall (Sony) are wonderfully re- gineering from Stanford University such as the DMSTM, TrueMusicTM produced; the Rachmaninoff pieces are and B.S. degree in Electrical Engi- wireless hi-fi system and TrueViewTM extraordinary and captured by this amp neering from Santa Clara University, wireless video system. The company’s in the best way. Audience, the bench, Santa Clara, Calif. He is a member philosophy is to use its technologies recording/hall noises/ambience—this of Tau Beta Pi and Eta Kappa Nu to solve problems that have gone un- amp offers high-resolution sound with Electrical Engineering Honor So- resolved for decades. impact and drive. cieties. He has almost three decades Further products include high- Another great piano solo recording is of experience working on semicon- performance, low-profile line-array Marc-André Hamelin’s Nikolaij Medt- ductors for analog and digital audio, speakers, high-efficiency audio am- ner four-CD box (Hyperion). This won- video, cellular telephones, wireless plifiers (monoblock, stereo, multi- derful recording by Tony Faulkner is communications, fiber optic commu- channel, in-wall), high-performance real, and, with these Pragmatic ampli- nications, data communications, data TV/CATV/FM tuners, category 5 fiers, the Dynaudios fade away. The low acquisition, and industrial controls (CAT5) audio/video transmission and and high units melt together as the left in the Silicon Valley. He holds one cross-point matrix switchers, hard disk and right speaker and results in a three- patent on power line communica- music servers, ambient audio levelling dimensional sound. tions, several patents on audio, video, systems, microphones, and IR detec- The music of Russian composer Niko- and wireless multimedia systems, and tors. laij Medtner results in a wonderful jour- many more patents pending in the ney. Natalie Cole’s voice on Take a Look area of industrial automation, health PADSTM—Multi-room audio distri- is full-bodied with lots of air around it; care patient interaction, and power bution system on a single category 5 the differences in recording of the differ- generation/energy storage technolo- wire. ent tracks can easily be detected. Fusion gies. An avid audiophile and music DMSTM—Multi-source, multi-zone jazz such as Lee Ritenour’s Wes Bound lover, in his free time he likes to distributed music system on a single (GRP) will make you smile. Such big apply his pragmatic thinking to solve category 5 wire. sound from such a tiny amplifier. Also complex problems. DAVESTM 4.43—Distributed Audio noticeable is that the higher frequencies Video Entertainment System. don’t show any rolloff. The Keith Jarrett ABOUT THE COMPANY CATSTM—Multi-source, multi-zone Trio Live at the Blue Note six-CD set is Pragmatic Communications Systems, DVD quality audio/video signal dis- a real treasure; I guess these Pragmatic Inc. (www.wireless-experts.com and tribution and IR control on a single modules already knew this. www.pragmatic1.com) is a Silicon category 5 wire. Yes, I am aware that I am overly en- Valley company established in 1994 CATSTM HD—High definition thusiastic. These modules are really fun to design, develop, and produce a vari- audio/video on category 5 wire. and capable of reproducing music in all ety of innovative products for wireless The Class D tiny power amplifiers are its pros and cons. audio, video, and data communica- part of the so-called Hide-AwayTM tions. The majority of the company’s Series (SIWA). EPILOGUE products until 1997 were primarily All are trademarks of the Prag- These interesting Pragmatic modules designed for industrial and commer- matic Company. ■ are very quiet, even in my simple test audioXpress June 2009 29

dehaan3067.indd 29 4/27/2009 3:42:45 PM speakers By Chi C. Wong Time-Adjusted Passive Crossover Systems This Australian study on passive crossovers is an interesting adjunct to Koonce’s work published last summer.

would like to thank G.R. Koonce the same acoustic center, and each identical to those in reference 2. for his very informative articles1 acts like a perfect point source. • PSpice does not use a true optimiza- on linear phase passive crossovers. • In frequency-domain analysis both tion algorithm. However, worst-case IIn the past 18 months, John2 and I the magnitude and phase responses simulation results should show that have spent some time analyzing passive are considered. The worst case is de- the actual worst-case, if different, time-adjusted crossovers proposed by rived when the phase plot produc- could only be worse. 3 Yamanaka and Baekgaard ; both require es the biggest discrepancy from the a “filler” speaker. Koonce’s investigation nominal. RESULTS concentrates on the acoustic character- • In time-domain analysis a 500Hz Figures 1A-1C show the (worst-case) istics of loudspeaker systems with linear square wave is used as the test signal. frequency response (Bode plots) of the phase crossovers. In contrast, we study The worst-case is derived when the second-order Yamanaka system with var- the non-ideal electrical properties of magnitude produces the biggest dis- ious component tolerances. In an ideal filter components and speaker imped- crepancy from the nominal. world the magnitude plot is a horizontal ance, using PSpice (student version 9.1) • Component values are similar but not line of 0dB and the phase plot is another to evaluate the worst-case conditions of various time-adjusted crossovers. This study appears to stand as a supplement to Koonce’s excellent articles. We use formulas by Yamanaka and Baekgaard to determine crossover components, which are accurate to three dig- its, and allow them to have tolerance. We assume that speakers are matched with zobel and/or notch networks such that the impedance is 8Ω and resistive. Again, we allow a tolerance to be added to individual speakers. As a comparison, we also include a first-order system, using worst-case and FIGURE 1A: Frequency response of the Yamanaka system with component tolerance: 1%, sensitivity analysis to evaluate the per- speaker tolerance: 10%. Upper plot is the magnitude plot and lower plot is the phase plot. formance of the different designs in both the time domain and frequency domain. In frequency analysis we use the phase of the summed output as the output variable. In time-domain analysis the magnitude is the output variable. We use the following conditions/assumptions: • Nominal crossover component values are calculated to the third significant figure with all speakers made equiva- lent to an 8Ω resistor. • Component tolerance and speaker tolerance are assigned to be +1% or +10%. • Crossover frequency is fixed at 2kHz in all cases. FIGURE 1B: Frequency response of the Yamanaka system with component tolerance: 10%, • Acoustic properties of speakers are speaker tolerance: 1%. Upper plot is the magnitude plot and lower plot is the phase plot. ideal; for example, all speakers have 30 audioXpress 6/09 www.audioXpress.com

wong3058.indd 30 4/27/2009 3:50:47 PM horizontal line with 0°. Figure 2 shows • Speaker units must be tamed to be a the response from a first-order system. constant resistance, probably within Figure 3 shows the results from a third- 10% tolerance or better. order Baekgaard system. All are affected • Crossover components must be se- by component tolerance with the third- lected (almost) exactly according to order Baekgaard system being the worst. the designed values and with the best Another set of worst-case results can tolerance. The importance of toler- be generated when the magnitude re- ance increases with system order. sponse dips instead of peaks. Figures • The third-order Baekgaard system is 4A and 4B show the (worst-case) time around two times worse than the sec- response of the Yamanaka system with ond-order Yamanaka system with the two different tolerance settings. In an same component/speaker tolerance. ideal case the original 1V 500Hz square • The second-order Yamanaka sys- wave can be reproduced. Figure 5 shows tem is close to the first-order system the relatively poor time-response from with component/speaker tolerance of the Baekgaard system. For the second- 10% if the former has 1% crossover order system, component and speaker components. tolerance must be kept within 10% if • The first-order system is not immune frequency response abnormities are kept to component and speaker tolerance. within ±2dB and ±20°. It is more tolerant. However, the shadow sloping of the filters is not CONCLUSION good for most speakers. From the above samples and John’s • The third-order Baekgaard system de- work2, I can conclude that mands very tight component tolerance

FIGURE 1C: Frequency response of the Yamanaka system with component tolerance: 10%, speaker tolerance: 10%. Upper plot is the magnitude plot and lower plot is the phase plot.

FIGURE 2: Frequency response of the first-order system with component tolerance: 10%, speaker tolerance: 10%. Upper plot is the magnitude plot and lower plot is the phase plot.

audioXpress June 2009 31

wong3058.indd 31 4/27/2009 3:50:49 PM and does not appear to be practical. • The second-order Yamanaka system seems to be the only viable solution. (This conclusion is quite different from Koonce’s listening experience.)

The performance of a complete speaker system is undoubtedly affected by non-ideal acoustic and electrical conditions, so it appears that a perfectly time-adjusted passive loudspeaker system is unlikely to be practical in the real world. Perhaps an active system, and possibly with the aid of digital tech- 4 FIGURE 3: Frequency response of the Baekgaard system with component tolerance: 10%, niques , is the only way to move closer speaker tolerance: 10%. Upper plot is the magnitude plot and lower plot is the phase plot. toward this objective. aX REFERENCES 1. G. R. Koonce, “Passive Crossover Linear Phase Speakers,” Part 1 and Part 2, aX 6/08 and aX 7/08. 2. John O’Connor, “Time Aligned [sic] Cross- overs for Audio”, M.Eng., Minor Thesis, RMIT University, 2008. (Unpublished) 3. Steve Stokes, “A Unique Crossover Design with Waveform Fidelity,” aX 1/07. 4. Iain McNeill, “Crossover Approaches,” aX FIGURE 4A: Time response of the Yamanaka system with component tolerance: 1%, 10/08. speaker tolerance: 10%.

FIGURE 4B: Time response of the Yamanaka system with component tolerance: 10%, speaker tolerance: 10%.

FIGURE 5: Time response of the Baekgaard system with component tolerance: 10%, speaker tolerance: 10%.

32 audioXpress 6/09 www.audioXpress.com

wong3058.indd 32 4/27/2009 3:50:50 PM Classified

Yard Sale Free Ads For VENDORS Subscribers Yardsale Guidelines: High Performance kits, Audiophile components 1. Submissions accepted from Custom designs, Custom Assembly subscribers to audioXpress www.borbelyaudio.com magazine only. You must include In North America: LBAudio, Les Bordelon, your subsciption account number [email protected] with each submission. 2. Submit your ad to Yard Sale, PO In Taiwan: TS audiolab, Tai-Shen Lee, Box 876, Peterborough, NH 03458. [email protected] Or fax to (603) 924-9467, or e-mail to [email protected]. AudioClassics.com Buys - Sells - Trades - 3. Please be sure your submission Repairs - Appraises McIntosh & other High is legible. We are not responsible for changing obvious mistakes End and Vintage Audio Equipment or misspellings or other errors 800-321-2834 contained in ads. 4. We will not handle submissions Vista-Audio, Radii, Audio Limits, Trafomatic. over the phone. Please do not Tubeamplifi ers, kits, custom transformers. call to verify acceptance or inquire about the status of your www.engineeringvista.com submission. 5. Each submission will be used until NEW PRODUCT Tuneable stereo rumble the next issue of the magazine fi lter, model 603A. 5Hz/step from 15 to 65 Hz. subscribed to is published. www.tdl-tech.com/data603a.htm 6. Maximum 50 words. No accompanying diagrams or illustrations or logos will be used. Class A Chassis! New ezPower Chassis® DIY power amplifi er chassis includes “Yard Sale” is published in heatsinks. www.designbuildlisten.com each issue of aX.

Ad Index

ADVERTISER PAGE Marchand Electronics Inc...... 32 ACO Paciﬁ c Inc ...... 21 Mouser Electronics ...... 24 Antique Radio Classiﬁ ed ...... 23 Mundorf EB GmbH ...... 9 Audience ...... 31 Parts Connexion ...... CV3 Audio Amateur Incorporated Parts Express Int’l., Inc...... CV4 Old Colony Sound Lab – Catalog ...... 17 PCBCart.com ...... 45 Audio Transformers ...... 25 Solen, Inc...... 15 Avel Lindberg ...... 35 Tang Band Industries Co.,Ltd...... 6 DH Labs Audio Cables ...... 5 E-Speakers.com ...... 7 CLASSFIEDS ETI – Eichmann Technologies ...... CV2 All Electronics ...... 33 Front Panel Express, LLC ...... 45 Audio Classics Ltd...... 33 Hammond Manufacturing ...... 3 Borbely Audio ...... 33 KAB Electro-Acoustics ...... 31 Design,Build,Listen ...... 33 Linear Integrated Systems ...... 43 ENG Vista, Inc...... 33 Madisound Loudspeakers ...... 41 TDL Technology, Inc...... 33

audioXpress June 2009 33

yrdsale-adindex-classy.indd 33 5/4/2009 9:59:38 AM sound solutions By Mark Williamsen Design and Measurement Of a Dipole Microphone Tone burst techniques, gated measurement, analog computing, and imaginary numbers.

’ve been interested in microphones rections. Sounds arriving simultaneously phone can distinguish front from back, for some time, especially those from various directions simply add up in the sense that sound from the front used to record music in concert to a single pressure value at the location gives a gradient signal which is in phase I and studio settings. More recently, of the microphone. Note, however, that with the omni response, while sound however, I’ve become especially inter- the addition is instantaneous, so phase from the back gives a gradient signal ested in how various microphones dif- matters when doing the summation. Two which is opposite in phase to the omni fer, how to measure and compare those sounds at the same frequency could add response. Assuming the omni and on- differences, and what design trade-offs up constructively or destructively, de- axis gradient responses have the same go into the development of a new mi- pending on their phase relationship. sensitivity, they may be added together to crophone. This article examines a par- Now suppose you want to design a obtain a response that is twice the omni ticular kind of microphone, the acoustic microphone that picks up sounds along response on-axis from the front, equal dipole or “gradient” microphone and its a particular axis, while ignoring sounds to the omni response from the side, and design parameters1. along some other axis. A simple tech- null (complete cancellation) along the First I’ll describe its operation math- nique is to place two pressure micro- dipole axis from the rear. ematically in enough detail to predict phones of equal sensitivity near each The dipole microphone is now giving frequency and polar response character- other, and then obtain the difference what is commonly called a “first-order istics in a spreadsheet. Next, I’ll discuss between the two signals. Sound waves cardioid” polar response pattern, prefer- the techniques needed to obtain accurate traveling along the axis of the dipole ring sounds from one direction while measurements of microphone frequency (that is, along a line through the acoustic rejecting sounds from the opposite di- and polar response, using a computer centers of both microphones) will arrive rection. This microphone type is widely sound card and some simple software at slightly different times. While the am- used—both in studios and in live re- programs. The spreadsheet and comput- plitudes will be roughly the same, there cording—because it allows direct sound er source code are posted online (www. will be a phase shift between the two from a performer or instrument to be audioXpress.com) for readers to down- signals so they don’t completely cancel. emphasized, while reducing ambient and load. Finally, I’ll show the construction This, then, is the “on-axis” signal of the reverberant sounds2. of a prototype dipole microphone using directional microphone. Most commercial microphones having two omnidirectional electret condenser Sound waves arriving from any di- a first-order cardioid response actually microphone cartridges. rection in a plane perpendicular to the have just one diaphragm and one electri- My goal here is not to produce a de- dipole axis will arrive at both capsules cal output. The gradient is obtained by sign that’s better or different from exist- at the same time, so there will be no having two ports to accept sound, which ing designs, but rather to gain a more phase shift between them, allowing for is then routed to the front and back of complete understanding of what goes complete cancellation of the signal. This the diaphragm with appropriate time de- into microphone design and measure- is the “off-axis” response, for what you lays and other acoustic elements. This is ment. This way, we can all be a little more now see is a bidirectional microphone, a form of analog computer, implemented confident about making the best use of having a figure-eight polar response pat- with air passages. The best way to see the microphones that are available to us. tern. The combined dipole microphone how this works is to take one apart, al- now responds to the gradient of pressure though you may ruin it in the process. MICROPHONE TYPES instead of pressure itself, and has a pre- For my work here, I’ll use two separate The simplest microphone is the omni- ferred axis for pickup. omnidirectional capsules and then de- directional pressure microphone, which But what if you want to pick up velop the various polar patterns using an senses instantaneous pressure at a point sounds arriving from one direction along analog computer implemented with gain in space. Pressure doesn’t have a direc- the dipole axis while rejecting sounds stages, adders, and an integrator. Clearly, tion, so an ideal pressure microphone arriving from the opposite direction? It the computations could also be imple- responds equally to sounds from all di- turns out that the dipole gradient micro- mented in a digital computer, but real- 34 audioXpress 6/09 www.audioXpress.com

williamsen3063.indd 34 4/27/2009 3:50:05 PM time digital signal processing is beyond a reference distance x0 at which pres- the scope of this article3. sure is arbitrarily set to unity. This will be the distance from the point source MATHEMATICAL DESCRIPTION to the acoustic center of the dipole. Be- Now let’s work out the math needed cause wavelength is just c/f, where c is to describe the polar and frequency re- the speed of sound and f is the frequency sponse for the dipole microphone. First in units of Hz, you can now write the you need some assumptions, so things wave equation as5: don’t get completely out of hand. I’ll assume the dipole is made of two ideal pressure microphones having identi- (2) cal sensitivity, and flat frequency and You’ve probably noticed that I’m using phase response over the range of inter- complex notation here, where i is the est. I’ll limit range of interest to 100Hz square root of -1. It may seem like an to 10kHz, for reasons that will become unnecessary burden, but it actually helps clear as we go along. I’ll assume free- to keep track of phase when you do the field conditions, considering only the calculations in Excel or other math pro- direct sound from an ideal point source, grams. You just need to be sure to let with no reflections or reverberations, and, Excel know that these are complex num- of course, no ambient noise. bers. Note that some math programs PHOTO 1: Auratone 5C Super Sound Sounds from the ideal point source expect “j” instead of “i” when entering Cube loudspeaker mounted on top of will propagate as spherical waves, so as- imaginary values. θ a tripod-style mike stand. Dipole mi- suming no transmission loss in the me- Now consider an angle between the crophone prototype is supported by a dium, acoustic pressure will fall as the in- axis of the dipole and a line from the mike stand sitting on a cheap turntable. verse of distance from the source. Acous- point source to the center of the dipole, Speaker and microphone are positioned tic power, of course, falls as the inverse where d is the acoustic distance between 1m apart, midway between floor and of distance squared, but the microphone the two microphone capsules in the di- ceiling, near the middle of the room. responds only to pressure, not power, pole (Fig. 1). By the law of cosines, the thus the 1/r dependence. The central distance from the source to each micro- axis of the dipole will, for my purposes, phone capsule is given as6: always remain in a plane that includes the point source. Now I write the wave equation giving instantaneous pressure p at distance x and time t, for a steady-state sound (3a) of frequency ω in units of radians per second4:

(1) (3b) Here e is the base of natural logarithms (2.71828. . . ), and k is the wave num- Here I’ve labeled the distance from λ λ ber 2π/ , where is the wavelength of source to microphone as xA for the front sound at the given frequency. I define capsule and xB for the rear capsule. For

FIGURE 1: Dipole microphone design features two omnidirectional capsules respond- ing here to spherical waves emanating from a point source. Distance from source to capsule A is xA, and from source to capsule B is xB. Acoustic center of the dipole is at distance x0. Length of the dipole d is the acoustic distance between capsules A and B.

audioXpress June 2009 35

williamsen3063.indd 35 4/27/2009 3:50:07 PM large x0 (that is, x0 >> d/2), this can be But now you notice that at 1kHz, tic of first-order cardioid microphones is ± θ ° ° simplified to x0 (d/2) cos , but let’s the difference (gradient) output level is that the response at 90 and 270 should keep the full form for now because I much lower than the sum (omni) output be exactly 50%, or -6dB when plotted will be considering small x0 (that is, x0 level, and as you change frequency you logarithmically. You now have a toy comparable to d/2) when I look into the find that the difference output level has model that reveals many of the charac- proximity effect. a large frequency dependence, increasing teristics of real microphones, as frequen- Now you can construct a spreadsheet along with frequency. What happened cy and distance from the source vary. If to calculate the omni, gradient, and was that when I calculated the differ- you are using Microsoft Excel, you can cardioid polar response patterns for an ence between the microphone A and download a fully populated spreadsheet acoustic sine wave. I’ll begin with some microphone B signals, I effectively took of this model (www.audioXpress.com). columns for given values, including the the differential ∆p, which caused a factor speed of sound (345m/sec assuming 22° of i(2πf/c)∆x to pop out, where ∆x is just MEASUREMENT TECHNIQUES C ambient temperature), distance from the dipole length d. This factor causes To test the model spreadsheet, I’ll de- source to dipole (1m), length of the di- the frequency dependence along with a scribe the construction of a prototype pole (0.01m), frequency of the incident 90° phase shift. dipole microphone. But first, you need a sound wave (1000Hz), and angle of di- To make the on-axis difference sig- set of measurement techniques in order pole axis with respect to the source (0°). nal look like the omni signal, you must to evaluate the prototype. According I’ve used the “Fill Down” command to divide out a factor of i(2πf/c)∆x. Luck- to the assumptions that went into the copy the given values into 72 consecu- ily, you can integrate the gradient signal spreadsheet, you want to measure the tive rows, except for angle, which I set to with respect to time to remove i2πf, and free-field steady-state response to spher- run from 0° to 355° in 5° steps. I placed ∆x/c is just a constant gain factor equal ical sound waves arriving from a point a column alongside the angle which con- to the dipole length divided by the speed source, in the frequency range from verts from degrees to radians using the of sound. So add a new column to the 100Hz to 10kHz. If you are already set Excel built-in function RADIANS(), spreadsheet where these factors have up to measure microphone polar and because many Excel functions assume been divided out of the complex differ- frequency response, you can skim over radians as input. Now you calculate the ence response. this part of the article. For readers who distance from source to microphone A With this change the gradient re- are interested in developing their own using equation (3a), and from source to sponse is now uniform over frequency, setup for microphone measurements, microphone B using equation (3b). but still off by a factor of 2, due to the however, I’ll give the complete details. All calculations up to this point have sum signal including the signal from To achieve free-field conditions is a been real numbers, but now I will switch two capsules. So you divide the com- bit of work, but well worth the effort. over to complex math to obtain the re- plex sum response by 2 (think of it as Building an anechoic room is one op- sponse to the source signal using equa- an average), and can now compare the tion, but way beyond reach for most of tion (2) for each capsule, replacing x amplitudes of the omni and gradient re- us. If you want to be able to stand up in- with the value of xA or xB as needed sponses on an equal basis. Alert readers side your anechoic chamber, you’ll need for each capsule. Because I’ve assumed will have noticed that x appears twice in 2m clearance vertically, plus 2m above steady-state conditions, the phase rela- equation (2), but when I integrated the and 2m below for the sound-absorbing tionship between the two capsules will difference signal with respect to time, I wedges, requiring a space of at least 6m be the same at any time t, so you can set accounted for only one of the x’s. The in height for construction. This would t = 0 to simplify the calculations. You presence of this second x is actually the be for a room with a cutoff frequency then take the complex sum and complex origin of the proximity effect in gradient difference of the signals from capsules microphones. A and B. The absolute value (radius in The final step is to show the cardioid the complex plane) is the output level response in the polar plot. This is done in units of pressure where the reference by taking the complex average of the level is unity, while the argument (phase omni and gradient responses and plot- angle in the complex plane) is the out- ting the absolute value. The result should put phase in radians. A radar plot of the be a familiar heart-shaped pattern, but difference output level should show a it’s pointed in the wrong direction! figure-eight pattern, while the sum out- It turns out that I overlooked a minus put level should be a circle. sign when obtaining the time integral of Use the Excel functions IMABS() the gradient response. Reinstating the FIGURE 2A: Calculated polar plot of and IMARGUMENT() to obtain real- minus sign will point the cardioid mi- 300Hz response with dipole length of valued numbers for plotting. If Excel crophone in the right direction. So you 10mm. Signals shown include omni, gives a “#NAME?” error when you try to can obtain both a front-facing and rear- gradient, and cardioid. Distance from use complex math functions, check to be facing cardioid response from the same source is set to 1m, so proximity effect sure that you have the Analysis ToolPak omni and gradient signals (Fig. 2). is minimal. Radial axis is linear. enabled, under “Tools Add-Ins. . .” By the way, an important characteris- 36 audioXpress 6/09 www.audioXpress.com

williamsen3063.indd 36 4/27/2009 3:50:07 PM around 86Hz, where 2m is a half-wave- the measurement. If loudspeaker and 10cm gives nulls at 1725Hz, 5175Hz, length. Below cutoff, the chamber re- microphone are moved closer, to 0.3m 8625Hz, and so on. I’ve used 10mm verts to pressure mode and is no longer separation, then I have 6.2msec in the diameter capsules for my prototype, so anechoic. Readers living in a mild cli- free-field regime, allowing good mea- the closest I can easily place the acoustic mate might consider using an open field surements down to 161Hz. Measure- centers of the capsules is 10mm, giving or empty parking lot, especially if they ments below cutoff don’t blow up, but the first null at 17.25kHz. This helps to can dig a small pit for a loudspeaker to rather begin to include more and more explain why, especially for large diameter be mounted so the baffle is even with sound from directions other than the microphones, it may be better to avoid the ground, pointing straight up in a 2π source with falling frequency. While it’s gradient behavior at high frequencies. (hemispherical) configuration. definitely not steady-state, you will see The basic requirement is to sample The microphone under test can then that this technique gives results that are pressure at two nearby points in space. be suspended directly above the loud- easily good enough to justify the effort To do this I used a pair of Panasonic speaker, pointing straight down. Note, in obtaining them. omnidirectional electret condenser mi- however, that you’ll need to accept what- All test signals were calculated and crophone cartridges, part #WM-034BY, ever ambient noise is present in the en- written directly to wave files using a sim- which I had on hand. It seems that this vironment. Also, the speed of sound has ple command-line utility program called part number has become obsolete, but a rather strong temperature dependence, “tbg.exe,” written for the purpose in C++. the cartridges themselves are completely varying almost 4% over the temperature This technique of using offline signal generic, and can be easily substituted range from 0 to 20° C. generation and analysis has worked out without affecting performance. I chose I actually chose to obtain free-field very well for me, and seems to have no to mount the microphone capsules fac- conditions in my living room with a real disadvantages when compared to ing upwards in a probe configuration tone-burst source and gated measure- more sophisticated graphical user inter- (Photo 2A). This allowed me to put the ment system7 using wave files and the face (GUI) programs that make their electronics beneath the microphone, sound hardware of a laptop computer. measurements in real time. Complete away from the line-of-sight for inci- My living room ceiling has a height of source code is available for readers to dent sounds, and made the microphone 2.43m. With loudspeaker and micro- download (www.audioxpress.com). nearly symmetric about the vertical axis phone suspended 1m apart midway be- for polar plots. tween floor and ceiling, the direct sound PROTOTYPE CONSTRUCTION You may ask why I didn’t face the arrives at 2.9msec after launch, while the Now you can build a prototype dipole capsules forwards, or perhaps one for- first echoes arrive at 7.6msec (Fig. 3). microphone, to see whether the charac- ward and the other one backwards. I That gives me 4.7msec of direct sound teristics predicted in the model spread- actually considered all of these configu- with no echoes, satisfying the require- sheet are meaningful. The first thing rations, but didn’t try them because I ments for a free-field measurement. I you must decide on is the length of the had in mind to make a bidirectional can fit at least one complete cycle of a dipole. As you can see from the spread- microphone that would allow me to de- free-field measurement at frequencies of sheet model, there are nulls (complete velop both front-facing and rear-facing 212Hz and above. cancellation) in the on-axis omni re- cardioid patterns from the same signals. When I measure frequencies below sponse whenever the dipole length is I also considered the fact that gradient 212Hz, at least some of the first echo equal to odd multiples of a half wave- microphones are notoriously sensitive from floor and ceiling is included in length. For instance, a dipole length of to handling noise and mechanical vibra-

FIGURE 3: Direct sound arrives at the microphone at time t = x0/c, where x0 is the distance from source to microphone, and c is the speed of sound in air. First echo arrives a bit later, after being reflected off floor and ceiling at height h. Free-field FIGURE 2B: Calculated frequency response of microphone with dipole length of conditions are obtained during the inter- 10mm. On-axis, 90°, and 180° plots are shown. Vertical axis is in dB. val between direct sound and first echo7.

audioXpress June 2009 37

williamsen3063.indd 37 4/27/2009 3:50:08 PM tion. Facing both capsules in the same package contained in an aluminum proj- technique with enameled magnet wire. direction should minimize the difference ect box. The tubes and project box are A set of jumpers JMP1 allows me to signal between them when both are sub- all tied to ground for proper shielding. I reconfigure the outputs between mea- jected to the same vibrations8. obtained a threaded mike stand adapter surements. As shown in Photo 2B, two brass tubes from Atlas Sound which I mounted on Instead of providing trimpots for bal- hold the capsules up above an electronics the bottom of the project box, centered ancing gains, I decided to mount fixed along the vertical axis of the microphone. resistors in parallel to accomplish the Although not shown, you can easily fit same thing. I used packaged resistor net- an inexpensive foam windscreen around works RN1 and RN2 for the 15kΩ resis- the capsules if needed. tors, because their values should track I provided two jacks at the bottom well over time and temperature. Cou- of the project box, one for a 9 to 12V pling capacitors C1, C2, C4, and C5 in DC power input, and one for a stereo the circuit path are all sized to put the line level output to match my comput- low-frequency cutoff well below the lower’s sound card input. The two capsules est frequency of interest. The integrator should be rigidly mounted with respect capacitor C3 is sized to roughly match PHOTO 2A: Two electret microphone to each other, and should have mechani- the gradient and omni signal levels. R5 cartridges are mounted side-by-side in a cal resonances well damped. In my pro- limits the DC gain of the integrator, probe configuration above an electronics totype, I found that the wires connecting and can be used to adjust low-frequency package. Thin wall brass tubes are tack- to the capsules had a strong resonance in response of the gradient signal, and thus soldered to each other, and rigidly mount- the low audio range, so I stuffed the brass tune the proximity effect in the cardioid ed to the perfboard circuit assembly. tubes with tissue paper to keep the wires outputs. quiet. I also experimented with using a pyramid, glued from foam-core board, to blend the surfaces of the project box to the probe tubes, but wound up not using it for the measurements in this article. The electronics package implements the analog computer I designed while coding the spreadsheet earlier in this article. It consists of gain stages, adders, and an integrator, as indicated in the PHOTO 3: Analog computer is as- block diagram of Fig. 5. A detailed sche- sembled onto a piece of perfboard with matic is given in Fig. 6. The completed PHOTO 2B: Threaded mike stand adapt- Photo 3 0.1″ hole grid. Power supply and ground er is aligned with the acoustic center of assembly is in . ″ bus wires are at top and bottom. Circuit the dipole. Power supply input and line I mounted all parts on a piece of 0.1 adjustments are made by soldering in level output jacks are also mounted on grid perfboard using DIP sockets and optional resistors. Jumper block at right the bottom end of the project box. 0.042″ perf-clips. Wiring was done allows various signals to be monitored. point-to-point using a “wrap and burn”

FIGURE 4: Simple sine tone burst shown above has discontin- uous slope at the start and end. Raised cosine has continuous slope, but gives a DC bump during each tone burst. Adding in the second harmonic with equal amplitude and opposite polar- FIGURE 5: Block diagram of dipole microphone with analog ity corrects the DC bump, and gives a continuous band-limited computer. Gradient and cardioid signals are derived using sum, function that can be repeated as necessary to fill a tone burst. difference, and integrator blocks1. Gain stage for each capsule Matched filter removes the harmonic during analysis. converts current to voltage.

38 audioXpress 6/09 www.audioXpress.com

williamsen3063.indd 38 4/27/2009 3:50:11 PM FIGURE 6: Detailed schematic diagram for dipole microphone and analog computer. Electret capsules are operated as current sinks at constant voltage. Jumper block JMP1 lets you assign the two output pins to a variety of internal signals. Vbias is set to 4.5V DC, to allow single-supply operation. Required filter and bypass capacitors are not shown.

I used two Burr-Brown OPA4134 relative phase response instead of the possible. I did this by first measuring the quad op amp packages at IC1 and IC2 absolute response, and thus subtract out amplitude difference versus frequency for for the analog computer, mainly so I the absolute distance from speaker to the two capsules when connected to the could become familiar with their char- microphone. The assumption here is that preamps as shown in Fig. 7. I then calcu- acteristics. I found them to have low I’ve precisely set the two capsules equi- lated the gain change required to match noise, moderate current consumption, distant from the loudspeaker. Note that the outputs at 1kHz and soldered in the and they are very fast. Layout and de- at 100Hz, 1 milli-radian (mrad) of phase appropriate resistors at optional locations coupling are important, as is an oscil- difference is just 0.55mm! R1 and R2. The gain and phase differ- loscope that can see oscillations well be- The next step is to match the sensi- ences between the two capsules after this yond 20MHz. Readers may prefer to use tivity of the two microphone cartridges adjustment are shown in Fig. 7, measured TL074 or other unity-gain-compensated over as broad a range of frequencies as using the harmonically shaped tone burst op amps in this application. I technique. chose an LM10CN integrated You can now measure the op amp and voltage reference omni and gradient responses, for IC3 to set the bias voltage both in frequency and polar for single-ended power supply domain. The results look en- operation. couraging (Fig. 8) with almost Instead of just buying two 30dB of side rejection in the electret microphone capsules gradient response at 300Hz. It and hoping they would match, seems that the integrator de- I actually bought ten capsules sign was correct, because the (they’re not expensive), mea- omni and gradient on-axis fre- sured the on-axis phase and quency responses are a good frequency response of each, match. To obtain the correct and then chose the two that cardioid response, you need had the best phase match at FIGURE 7: Measured gain and phase difference between to precisely match sensitiv- 100Hz, while maintaining a capsule A and capsule B in the dipole microphone prototype. ity of the omni and gradient similar frequency response all Best match is obtained at 100Hz where it matters most, due responses for sounds coming the way up to 10kHz. I actu- to integration of the gradient signal. Note the expanded verti- from the rear, the direction of ally measured the capsules in cal scale for both amplitude and phase. greatest rejection in the car- pairs, so I could look at their dioid pattern. audioXpress June 2009 39

williamsen3063.indd 39 4/27/2009 3:50:12 PM It turns out that the gradient polar sistors at optional R3, or in series with was about 1dB different from front to response can be asymmetrical due to a R4 to optimize the left cardioid pat- back, so I adjusted the gradient signal in small mismatch in phase between cap- tern. The goal is to obtain the best null the right cardioid by adding some resis- sules A and B, so you may need to ad- for rear rejection over the widest range tance in parallel with RN2A or RN2B just the “left” and “right” cardioid bal- of frequencies. around amplifier IC2D. Now I have two ance separately. Gross adjustments If your gradient polar response is sym- cardioid responses, one facing front and can be made by changing the value of metrical, then the right cardioid response the other facing back, with rear rejec- integrating capacitor C3. Fine adjust- should also be optimum at this point. In tion of 20dB or better in the range from ments can then be made by adding re- my case, however, the gradient response 200Hz to 10kHz (Fig. 9). You may be wondering whether the TABLE 1 Dipole Microphone Prototype Parts List sum and difference operations imple- Designator Part Number Description Manufacturer mented in the analog computer discard 0.1µF 50V ceramic disk bypass capacitors any any information. It turns out that they suitable project box 4 pcs. #6-32 ¾″ metal standoffs don’t, in the sense that given the left and 2 pcs. 137 7/16″ dia. brass tubing K&S Engnrng. right cardioid signals, you can recover 169P44 0.1″ grid perfboard Vector the original omni and gradient signals by AD11B 5/8″-27 mike stand flange Atlas Sound applying another matrix operation which T42 0.042″ perf-clip terminals Vector is the inverse of the sum and difference 2 pcs. 14-pin DIP socket any 8-pin DIP socket any operation. Expressed as a matrix, it turns C1,C2 220µF 35V electrolytic capacitor any out that the sum and difference operator C3 1nF mylar film capacitor any is self-inverse, times a factor of 2. C4,C5 100µF 35V electrolytic capacitor any C6 10µF 16V tantalum capacitor any Capsule A, Capsule B WM-034BY Omnidirectional Electret Condenser Microphone Cartridge Panasonic IC1,IC2 OPA4134PA High-performance quad op amp Burr-Brown (5) IC3 LM10CN Op amp and voltage reference National Semi JMP1 8-position 2-row header any This same matrix operator appears in line out 3.5mm 3-circuit phone jack any other places in audio, for instance, in ste- power in 2.5mm 2-circuit phone jack any R1,R2,R3 Optional gain trim resistors any reo modulation of broadcast signals, and R4 15kΩ ¼W resistor any in mid-side stereo microphones. R5 1.5MΩ ¼W resistor any At this point, I’ve barely scratched the R6,R7 100kΩ ¼W resistor any surface of what can be done with direc- R8,R9,R11 1kΩ ¼W resistor any tional microphones. I’ve assumed flat R10 22kΩ ¼W resistor any R12 10kΩ ¼W resistor any response everywhere, except for rolling RN1,RN2 761-3-R15K 15kΩ DIP resistor network CTS off the integrator on the low end. Obvi- Substitutions are encouraged, in the spirit of promoting research and development. ously, much more work can be done with equalization and phase matching here.

FIGURE 8A: Measured polar response of FIGURE 8B: Measured polar response of FIGURE 8C: Measured polar response prototype dipole microphone, showing left and right cardioid signals at 300Hz, 1m of left and right cardioid signals at omni and gradient signals at 300Hz, 1m distance, averaging of 4. Rear rejection is 300Hz, 10cm distance, averaging of 4. distance, with averaging equal to 4. Side greater than 20dB in both directions. Prox- Proximity effect causes a loss of rear rejection is almost 30dB. Plot is normal- imity effect is minimal at this distance. Plot rejection at this frequency. Plot normalized to on-axis omni response. normalized to on-axis response. ized to on-axis response.

40 audioXpress 6/09 www.audioXpress.com

williamsen3063.indd 40 4/27/2009 3:50:15 PM Fountek ribbon tweeters are an excellent choice for any high end speaker system. Ribbon tweeters are known for their light and effortless sound, described as being open and spacious. The Fountek ribbons are a perfect choice for new speaker projects and add on super tweeters. • Sandwich alloy diaphragm • Strong neodymium magnets • Black aluminum face plate The new sandwich aluminum alloy FIGURE 9: Measured frequency response of left and right cardioid signals measured at ribbons have higher power handling 1m with averaging equal to 4. Left and right curves are superimposed, with dipole axis and won't stretch or sag. at 0°, 90°, and 180° with respect to the source. All curves are normalized to the 90° re- NeoCd3.0M 3" Ribbon sponse of the rear capsule. Side response is down 6dB, as it should be. Rear rejection is 20dB or better from 200Hz to 10kHz from either side, which is quite respectable.

You could add more capsules to obtain Proximity effect is less than I’ve experi- higher order polar patterns or to simulate enced with other microphones, and may larger diaphragm microphone designs. take some getting used to. Now I can The electret capsules weren’t modified begin the real work of correlating mea- for improved headroom, as reported by sured results with listening performance, some authors. I haven’t used any acousti- for both purchased microphones, and cal elements in this design, but you could those of my own design. For more on $89 ea certainly do that as well. the test methods used here, go to www. 95dB, 2.5kHz to 40kH CONCLUSION audioXpress.com. Ribbon: 8 x 60 x 0.02mm, 18mg So what does a dipole microphone sound REFERENCES NeoCd2.0M 5" Ribbon like in actual use? To find out, I recorded 1. A.D. Blumlein, “Sound-Transmission, a rehearsal of some musician friends. Sound-Recording, and Sound-Reproducing Sys- The left- and right-facing cardioid pat- tem,” U.S. Patent #2,093,540 (Sept. 1937). terns cleanly divided the group into two 2. J. Eargle, The Microphone Book, Second Edi- bunches, with a few stragglers in be- tion (Focal Press, Oxford, 2004). tween. The bidirectional cardioid pattern 3. G.W. Elko and A.N. Pong, “A Steerable and could be useful in cases where the only Variable First-Order Differential Microphone place you can put a microphone is in the Array,” 1997 IEEE International Conference on middle of the stage. Note that the sum of Acoustics, Speech, and Signal Processing, Vol. 1, the left and right signals is just the omni pp. 223-226 (1997). response. This really is a special-purpose 4. L.L. Beranek, Acoustics (Acoustical Society microphone as designed, but it holds out of America, Woodbury, 1993). lots of promise for future development. 5. H.F. Olson, Acoustical Engineering (Profes- Mechanical noise, coupled from the sional Audio Journals, Philadelphia, 1991). floor through the mike stand, is tolerable 6. G.B. Arfken and H.J. Weber, Mathematical but needs improvement. Signal ampli- Methods for Physicists, Fifth Edition (Academic $126 ea tude is awkwardly positioned between Press, San Diego, 2001). 98dB, 2kHz to 40kHz microphone level and line level. Some of 7. Henning Møller and Carsten Thomsen, Ribbon: 8 x 120 x 0.02mm, 34mg the outputs are inverted in terms of ab- “Electroacoustic free-field measurements in ordi- solute phase. Self-noise out of the analog nary rooms—using gating techniques,” Brüel & computer is acceptable but needs work. Kjær Application Note 17-196 (1975). Wind and pop noise sensitivity are high 8. Roger Sanders, “A Simple Shock Mount for as expected, and readily improved with Your Microphone,” The Audio Amateur, #1, 1972, aX a generic open-cell foam windscreen. p. 1. audioXpress June 2009 41

williamsen3063.indd 41 4/27/2009 3:50:16 PM XPRESSMail

CORRECTION not been able to agree on his statement under test. More legible figures for the published that the total impedances at the phase Next, a short circuit to ground at the article T-reg: a High-Voltage Regulator splitter’s (PS) plate (P) and cathode (K) frequency of the test signal is applied for Tube Amps are available on the au- are identical if the P and K load imped- to that node, and the current into the thor’s website: www.linearaudio.nl. The ances are also identical. short is measured. The output imped- figures as published in the magazine Mr. Yaniger’s proposed test of output ance is defined as the voltage divided by (April '09, p. 24) are correct.—Eds. impedance requires that identical test the current. Never is any mention made loads be present simultaneously at both of placing additional test loads at any TESTING IMPEDANCE the P and K. No justification is given other node in the C.U.T. Nor is there I thank Mr. Yaniger for a cordial ex- for this requirement, which is not found any discussion of testing certain kinds change of e-mails and especially for his in any engineering textbook that I have of circuits differently, whether those cir- willingness to test my 12AX7 sugges- ever seen. The standard prescription for cuits are balanced, symmetrical, have tion, which did not pan out (Xpress testing output impedance involves driv- differential outputs, or have any other Mail, April '09, “The Impasse Preamp,” ing the input of the circuit under test particular characteristic. p. 42). His original 6SN7 was the bet- (C.U.T.) with a signal. The unloaded Certainly, shorting an output will lead ter gain stage. Unfortunately, we have voltage is measured at the output node to operation in a manner quite different

42 audioXpress 6/09 www.audioXpress.com

XpressMail609.indd 42 4/27/2009 3:51:16 PM from what was intended by the designer, The total impedance of these circuits fier inputs with finite impedances. I was but this is clearly a legitimate part of an at the load is the parallel combination remiss in not having pointed out that impedance measurement. Clearly, add- of the source impedance and Z. Because there would need to be a compensatory ing test loads to multiple circuit nodes the source impedances differ in the two change in the rest of the circuit to ad- will yield results different from the clas- circuits, the total impedances differ even dress that imbalance. sical impedance test. Such a change though the same voltage arises across It turns out that you can modify the in an established definition requires a each Z. So equal voltages across identi- PS to have both balanced output signals hearty justification, especially since the cal Zs in the PS clearly do not imply and balanced drive impedances for a se- results are at odds with the standard equal total impedances. Equal voltages lected power amplifier input resistance. measurement. are a simple consequence of the facts You need to increase slightly the value Mr. Yaniger mentions the Jones text, that the same current flows through of the cathode bias resistor when you which offers two arguments in favor the P load, the triode P and K and the add the series resistor. Unfortunately, the of his position. The first argument K load, and that the P and K loads are gain balance then becomes worse than states without proof or derivation that identical, as noted in the RDH4 refer- that of the standard PS the moment feedback in the PS reduces the P im- ence. Ohm’s law takes care of the rest. you add a common value of capacitance pedance. It is worthwhile noting that A straightforward, classical, by-the- across the power amplifier inputs. Not pages 312-313 (and 329-330 in more book measurement or derivation of exactly a step forward. Again, if anyone detail) of the 4th edition of the Ra- the impedances at the P and K in the is interested, I’d be happy to share the diotron Designer’s Handbook (RDH4) PS with identical P and K loads shows details. provide a derivation which shows that those impedances in general to be quite the feedback through the cathode load different—not identical. Christopher Paul increases the P impedance. There are I stated in my prior letter that it [email protected] numerous web references (check out, for would be possible to eliminate the im- instance, www.rocw.raifoundation.org/ pedance imbalance of the PS by add- STRANGE MIDPOINT computing/BCA/electroniccircuit/ ing a resistor in series with the cathode I studied the solid-state piece by Ian lecture-notes/lecture-26.pdf ) which output. Mr. Yaniger correctly pointed Hilton in the April 2009 issue of agree with this conclusion as do my own out that this change alone would cause audioXpress, “Keeping the Hobby Alive” derivations, which I’d be happy to share a voltage imbalance at power ampli- (p. 38), with interest. I am sure that with anyone interested. This wealth of contradictory derivations should at least make the text’s claim a contentious one requiring a justification. Strangely enough, the text itself on page 404 states that the impedance looking into the plate of a triode is rp + (1 + µ)Rk. This would clearly be reduced rather than increased (as later claimed) if Rk were bypassed and cathode feedback eliminated. It is unclear to me how these contradictory statements in the text itself can be reconciled. The second claim of the text is that since the frequency responses are identical at the P and K in a triode with identical P and K loads, “the output re- sistances must also be equal.” But while this observation is true, the conclusion is demonstrably false. Suppose I drove a load Z from a voltage source with a low impedance r. A current i(t) would flow from the source through r and Z, and a voltage v(t) = i(t)Z would arise across Z. Now suppose I drove the same load Z with a current source of a high impedance R. If I wished v(t) to arise across Z again, I’d set my current source to supply i(t) + v(t)/R. audioXpress June 2009 43

XpressMail609.indd 43 4/27/2009 3:51:18 PM FIGURE 1: Modified headphone amp circuit.

4HE3OUND3TROBEÛ 30%!+%2$)!'./34)#4//,

4HISASSEMBLEDUNIT COMESINADURABLE PLASTICCASE

Produces six selectable spectrum shape pulse To ﬁ nd out more about The Sound Strobe™contact customer service by signals to help you identify problems with your calling 888-924-9465 or e-mail [email protected]. speakers and their interactions with your listen- The Sound Strobe™ was designed by Dennis Colin, currently working as ing room. an Analog Circuit Design Consultant for microwave radios and a frequent Quickly hear your speaker’s degree of image contributor to audioXpress magazine. focus, transient precision, and tonal neutrality. !VAILABLEINKITORASSEMBLEDVERSIONS A great product for DIY speaker builders who +4 ! !SSEMBLED3OUND3TROBE 3H7TLBS want to tweak their designs and a must-have for +4 3OUND3TROBE+IT 3H7TLBS anyone installing loudspeakers professionally. 0#"4 3OUND3TROBE0# BOARDONLY 3H7TLB #HECKOUT%D3IMONSREVIEWAT Old Colony Sound Laboratory, PO Box 876, Peterborough NH 03458-0876 USA Toll-free: 888-924-9465 Phone: 603-924-9464 Fax: 603-924-9467 WWWAUDIO8PRESSCOM E-mail: [email protected] www.audioXpress.com

44 audioXpress 6/09 www.audioXpress.com

XpressMail609.indd 44 4/27/2009 3:51:20 PM the solution for this headphone-circuit Ian Hilton responds: sounds great, but, in my opinion, the Thanks for pointing that out, Hans. You emitter-follower midpoint circuit does are right, of course. The emitter-follower not work. First of all, there is nothing scheme will only work if the majority of to define the midpoint voltage. I sug- current is drawn from the pseudo -4.5V gest that what really does the trick is the rail. In my defense, it was just a quick common mode and the power-supply lash-up and not fully thought out, but it rejection ratios of the op amps. I believe still sounds good. the circuit would work better if you en- A better scheme is to have a push- tirely removed the 2N3904 Q1 transistor pull NPN-PNP pair driving the pseudo and substituted a connection between ground. Incidentally, Quad uses such a the junction of R1 and R2 and the junc- scheme (push-pull pair) to generate a tion of C1 and C2, and call that GND. pseudo ground in their Quad 306 power As it is, the midpoint is not a mid- amp. point at all; it is some voltage in which Further to the constructive com- Q1 just barely—perhaps—conducts. ments from Hans Weedon concerning There is nothing to define that voltage the pseudo ground circuit in the simple if the midpoint is more positive than headphone amp, I have revisited the de- -0.6V. In other words, if the midpoint is sign and amended it (Fig. 1). The pseu- anywhere in the positive direction above do ground is now generated using U2, a -0.6V of the middle, Q1 is completely TLE2426, a device specifically designed off and the midpoint is undefined. for that purpose. Further amendments I suggest that what really does the are the inclusion of 470R input resis- job of defining an AC ground are the tors. These were required with a spe- two 100µF capacitors; the transistor cific MP3 player that did not like a high does nothing. impedance load and commenced oscil- Hans J. Weedon lating. aX [email protected]

CONTRIBUTORS FRONT PANELS Luciano Peccerini, Jr. (“Beyond G4OEP: Improving a Flea-Powered Headphone Amp” p. 8) is an Electronic Technician from “ETI Lauro Gomes Technical School,” Electronic Engineer BSc from “FESP University,” and Industrial Automation & ENCLOSURES MSc from “FEI University” (all in Brazil). He has been involved with audio gear since his teens, fixing tube amplifiers and old radios. Nowadays he spends his time between “Business Intelligence Consulting” for industry, sci-fi scratch-built Customized front panels can modelling, and designing audio stuff, mostly tube amplifiers. be easily designed with our free software Front Panel Designer Ed Simon (“Removing the Six “Straight Wire” Kinks: Prologue to a Better Preamp,” p. 14) received his B.S.E.E. at Carnegie-Mellon University. He has installed over 500 sound systems at venues including Jacob’s Field, Cleveland, Ohio; MCI Center, Washington D.C.; Museum of Modern Art Restaurants, New York; The Coliseum, Nashville, Tenn.; The Forum, • Cost-effective prototypes Los Angeles; Fisher Cats Stadium, Manchester, N.H. and production runs • Wide range of materials or customization of provided Ari Polisois (“Three Novel Transformers for SE Amps,” p. 20) resides in France. material • Automatic price calculation Wim de Haan (“The Pragmatic Class D Amplifier,” p. 26) is a medical engineer from the Netherlands. His interest • Fabrication in 1, 3 or 7 days in audio electronics began at age 14 and his interest in music sometime before that. He is webmaster of the website of legendary Hammond jazz organist Jackie Davis. He also participated in a CD production of classical pianist Evelina Vorontsova which had very good reviews in Piano (UK magazine) and Luister (NL magazine).

Chi C. Wong (“Time-Adjusted Passive Crossover Systems,” p. 30) resides in Melbourne, Australia. Sample price: $ 43.78 plus S&H Mark Williamsen (“Balanced Cable Hum: Cause and Cure,” p. 34) is a graduate student in Physics at the University of Wisconsin in Milwaukee. He is active in Milwaukee’s busy music festival scene, and has recorded several album projects for local musicians. More info is available at his home page, http://my.execpc.com/~williamm/.

John Shand (“Jazz Track,” p. 46) is a CD reviewer for Australian HiFi and Home Theatre Technology. www.frontpanelexpress.com (206) 768 - 0602

audioXpress June 2009 45

XpressMail609.indd 45 4/27/2009 3:51:23 PM Jazz Track By John Shand Melbourne www.audaud.com BursonAudio #VˌPOTPVOETHPPE MJLFBTPMJETUBUFTIPVME Manu Katché|Playground|ECM 2016 Manu Katché has drummed with Jan Garbarek, Peter Gabriel, Sting, Joni Mitchell, Joan Armatrading, Youssou N’Dour, and scores more on more than 200 albums. His “Neighbourhood” album was a key 2005 release, and here he has retained pianist Marcin Wasilewski and bassist Slawomir Kurkiewicz (from Tomasz Stanko’s band). But where “Neighbourhood” gloried in the combined brilliance of Stanko and Garbarek, trumpeter Mathias Eick and saxophonist Trygve Seim don’t put such a rm stamp of individuality on the music. Nor are Katché’s compositions as strong this time, although they do gradually insinuate themselves. Understated grooves and textural subtlety are the mainstays, and given that the man with one of the planet’s most beautiful- sounding drum-kits has been well recorded, the aesthetics are not in question. You just feel there was another level that Playground could have reached. . . Mike Nock Project|Meeting of the Waters|Jazzhead HEAD086 . Despite being made by a marvelous nonet, this CD’s ultimate glory is three minutes of more or less solo piano on a heart-rending . little lullaby called Sho’s Cradle Song. e spiky styles of trumpeter Eugene Ball and alto saxophonist Tim Wilson contrast with the more eshy approaches of trombonist James Greening and tenor saxophonist Julien Wilson. Meanwhile, Dale Gor nkel’s vibraphone and Stephen Magnusson’s guitar blur the line between horns and rhythm section. Bassist Brett Hirst solos pensively over the gentle wash of Cyboreal (as does Nock), and robustly on the title track. Mysteries has the vibraphone and guitar clouding the AD712, AD827, C4570, JRC4580, JRC5532, JRC5532D, JRC5534, LF353, ensemble in a shimmering heat-haze, after which a woolly, languid tenor solo occasionally arches up to a plaintive high note, around LM833N, NE5532, NEC4520, NEC4570, NJM2068D, NJM2114, NJM2214D, which Magnusson instantly paints a tiny aura. NJM4558, NJM4558D, NJM4560, NJM5532, NJM4558P, OP275, OPA2132, Wholesale OPA2134, OPA2604, RC4558D, RC4558P, TL052, TL072, etc. Discounts Available NE5534, LT1122, OPA627, AD811, AD797, AD844, etc. for Modifiers & Judy Bailey Trio|Pendulum|Birdland Maria Schneider Orchestra|Sky OEM’s BL017 Blue|ArtistShare AS0065 ose who still dabble in “generic post-bop Maria Schneider eschews the tired cliché of jazz” often are merely echoing a style, rather big-band writing. She treats her orchestra as than molding substance. Pianist Judy Bailey, exactly that, so the trumpet section, for in- however, has been steeped in this music for stance, is not a block of sound, but a set of half a century, and deploys an e ervescent possibilities. Nor does she have a chain of so- lightness of heart with a commensurate los over a given backdrop. Rather, each solo is lightness of touch (especially notable here a true extension of the composition, speci - on Night and Day). is album, with bass- cally colored by the orchestrations. e Pretty ist Craig Scott and drummer Tim Firth, is centered around breezy stan- Road has Ingrid Jensen’s digital delay-treated trumpet soloing elegiacally dards. After an overly fussy arrangement of Love Walked In, Bailey deftly against ever-changing water-color washes from the orchestra. It’s such a The Authority on Hi-Fi DIY! slides into an improvisation encrusted with glistening, gem-like ideas. She glorious piece of music that the tune it resolves into seems bland by com- weights the notes with aching precision on the delicate Flamingo, and lilts parison. Any big band is only as good as its rhythm section, and guitarist and skips on the ¾ title track. Scott develops a compelling narrative in his Ben Monder, pianist Frank Kimbrough, bassist Jay Anderson, and drum- solo on You and the Night and the Music, while Firth swings throughout mer Clarence Penn provide a mighty foundation for this kaleidoscope of No matter what your requirements are, pcX has what you need. and solos melodically. beauty. Whether it be vacuum tubes (NEW, New Old Stock or OS), sockets, transformers, capacitors, resistors, connectors, hook-up wire, etc... We’ve got a world class selection of all the best brands Visit our Enrico Rava/Stefano Bollani| e ird Man|ECM 2020 Although the expectations of rapturous lyricism from trumpeter Enrico Rava and pianist Stefano Bollani were not disturbed, ...and more arriving every month! website and sign up for our this album still surprised, as though hearing music for the rst time, so each note glistened, newmade and just thought of. e eConneXion Newsletter! source of this surprise is not how brilliantly the pair play, but where the music takes these old colleagues (a generation apart), between whom the music falls as naturally as a waterfall. Perhaps that is the source of the album title (and an enthralling free improvisation), e ird Man: the phantom player who emerges as a combination of the two. e piano is supremely well- www.partsconneXion.com recorded, and needs to be to capture all the subtlety of Bollani’s playing as he delicately cross-hatches the sadness and exultation o r d e r @ p a r t s c o n n e X i o n . c o m spilling from that silvertoned trumpet. Toll free order line 1-866-681-9602 Tel: 905-681-9602 Fax: 905-631-5777

Paypal, US money order, bank draft, cashier’s check Albert Mangelsdor Quintett|Folk Mond & Flower Dream|Nekta/Planet 68.861 A Z U M A Inside a few notes a brilliant wind player tells you just as much about himself as a good singer. e late Albert Mangelsdor International orders was already one of the world’s great trombonists when this was recorded in 1967. e German virtuoso had not yet developed welcome! multiphonics (the art of playing more than one note at once) and was ve years away from launching his famed unaccompanied concerts. Already in place, however, was a powerful sense of the warmth of the man just bursting out of the bell of the trombone. His sound is as plump and inviting as a ripe mango, and nds perfect foils in his Quintett’s two slicing saxophonists (the other instruments being bass and drums). His lively original tunes are adventurous, but never anarchic: Mangelsdor would rather charm than bully you.

46 audioXpress 6/09 www.audioXpress.com

shand2978.indd 46 4/27/2009 3:46:42 PM