Ultra-Linear Operation of the Williamson Amplifier

DAVID HAFLER and HERBERT I. KEROES

The Famous “Williamson” can be improved simply by replacing the output transformer and making a few minor changes in other components. The results are well worth the effort and expense.

ollowing the original appearance of the preceding article on Ultra- F Linear operation of power output tubes, considerable interest has been evidenced in the application of this new circuit improvement to the famous Williamson amplifier. The William­ son circuit has been publicized in several arrangements including at least one commercial one, and the configur­ ation is undoubtedly the most popular high-quality audio circuit ever devel­ oped. For many people there is little necessity to attempt to improve this basic amplifier circuit. Its listening quality is excellent; it is easy to con­ struct; and it provides top quality at a cost comparable with units which cannot measure up to its capabilities. The one category in which the W il­ liamson amplifier is significantly de­ ficient is with regard to efficiency and power-output capabilities. Peak power output is less than 15 watts, and it takes a 450-volt supply at approxi­ mately 130 ma to achieve this power output. If this limitation can be over­ Proper location of the tap results in Ultra-Linear hook-up. In addition, KT- come without deterioration of quality, optimum input-output linearity simul­ 66’s can be used without deterioration a change in the original design is justi­ taneously with efficient operation, of quality as the slight mismatch is in fied. If simultaneously it is possible to power capabilities approximately double a favorable direction with respect to improve the amplifier both in measure- those of a connection, and low- characteristics. Therefore, able laspects and in listening quality, impedance output such as is offered by this transformer can be used with the then a change is not only justifiable, it . In short, it permits better per­ tube types normally used in William­ is mandatory. formance than either triode or son amplifiers without compromise of It is difficult to improve on some­ connection of the tubes, and this is sub­ characteristics. It is of interest to note thing which is really good. There are stantiated in comparative listening that the change in impedance to 6600 some audio enthusiasts who will scoff tests and distortion measurements. does not violate Mr. Williamson’s de­ at the idea that the Williamson circuit The unique merits of the Ultra-Linear sign considerations. The modified tube can be improved. However, it has been stage are particularly applicable to the characteristics of the Ultra-Linear con­ five years since Mr. Williamson pub­ Williamson circuit. The mating of the nection require this impedance if we lished his circuit; and in the course of two seems to have been inevitable. The wish to preserve operating conditions five years, there is little which can simple substitution of an output trans­ similar to those of the original amplifier. maintain supremacy without change or former with primary taps for Ultra- In other words, the tubes are still renovation. When a basic circuit im­ Linear operation and a few minor matched for minimum distortion rather provement—the Ultra-Linear output changes in circuitry, which will be than for maximum power output. The stage arrangement—came along, it was discussed below, combined the basic transformer, therefore, can be placed in natural to see how it could fit in with circuits into an amplifier which prac­ the circuit directly and the screens of the basic Williamson circuit. tically everybody agrees is an improved the output tubes connected to the ap­ The Ultra-Linear output stage is not version in all respects. Obviously, we propriate taps as shown in Fig. 1. This a triode stage as is used with the Wil­ must gain improvement if we substitute eliminates the two 100-ohm screen liamson circuit—nor is it a tetrode or a more linear output tube and use a stopper resistors of the original circuit. pentode stage. It combines the ad­ transformer which exceeds the origin­ The plate and screen leads of the trans­ vantages of both triode and tetrode by ator’s stipulations for performance. former are color coded to avoid phasing using an arrangement in which the The original Ultra-Linear circuit difficulties. screen grids of are energized utilizes a transformer, the Acrosound Several additional circuit changes from a tap on the primary of the out­ TO-300, which was designed for use have been found beneficial for optimum put transformer. This connection, on with tubes of the type. Its 6600 performance. One of these is the which patents are pending, modifies the ohms primary impedance therefore, is change in value of the feedback resistor operating characteristics of the tube. also correct for 5881’s and 807’s in the to 10,000 ohms in order to maintain 20 db of feedback. In the Ultra-Linear response, distortion, and transient char­ pairs of networks and increases the stage the gain of the stage is greater acteristics. In addition, it was con­ low-frequency stability margin at nom­ than for a triode stage. In addition, the sidered desirable to make certain other inal increase in cost. change in primary impedance changes slight changes which primarily increase The insertion of a 10,000-ohm para­ the proportion of voltage fed back. the stability under feedback conditions. sitic suppressor in the input grid and Thus the feedback is increased unless The low-frequency time constants of a 100-ppf across the feedback the feedback resistor is changed to the original circuit’s interstage coupling resistor adds to the high-frequency compensate. The readjustment of this networks were the same for both such stability margin and eliminates a slight resistor to the desired value then per­ networks. This is not particularly de­ ringing in the vicinity of 200 kc. mits the added gain of the Ultra-Linear sirable in a feedback amplifier since a One last optional difference from Mr. output stage to increase the amplifier given frequency loss is accompanied by Williamson’s original circuit lies in the sensitivity. It can now be driven with maximum phase shift. Separation of use of a bypass capacitor across the a little over 1 volt as compared to almost the time constants permits less phase cathodes of the output stage. This has 2 volts required for the original amp­ shift for the same frequency loss. In­ been found beneficial in both the Ultra lifier. creasing one pair of coupling Linear conversion and in the triode The feedback is taken from the 16- from .05 pf to .25 pf gives a five-to-one Williamson at high levels of operation ohm tap regardless of the speaker con­ ratio of time constants for the two as distortion at the overload point is nection. This tapped secondary arrange­ ment is extremely convenient when shifting to speakers of different im­ pedance as it does not require a change in the value of the feedback resistor. It is made possible by special transformer design (on which patents are pending) which permits equivalent response on all taps of a tapped secondary winding. The amplifier, as converted, now surpasses the original with respect to

diminished. There are no changes required in the remaining stages nor in the power supply. Most of the publicized versions of the circuit utilize power transformers which furnish 400 volts at 200 ma. Since the drain of fhe circuit does not exceed 130 to 140 ma, the voltage ob­ tained out of a capacitor input filter and 5V4 is about 450 volts. This is the correct value for the circuit as con­ verted. Lower voltage will limit the power output capabilities. Figure 2 shows intermodulation dis­ made at a comparatively low level Williamson arrangement puts out close tortion versus power output. It can be which makes the test even more rig­ to 25 watts at 20 cps and at 30 kc with­ seen that the power output of the cir­ orous. At low excitation levels, the in­ out , attenuation, or other visual cuit is effectively doubled over that of ductance of an output transformer de­ distortion of the waveform as viewed the original circuit for a given distor­ creases, the phase shift increases and on a ’scope. By observing the transfer tion. At low levels, around 1 watt, the the tops of the square wave tilts. A high- characteristic, it is possible to detect by IM hits such phenomenal values as level square wave will appear better eye harmonic distortion of less than 2 .06 per cent. It is only 0.3 per cent at than a low-level one at low frequencies. per cent. The power curve of the amp­ 13 watts. This curve is based on equiv­ Similarly, high powers at high fre­ lifier thus deviates from flat by less than alent sine wave power in order to make quencies will clip any supersonic peaks 1 db over the range 20 cps to 30 kc. it comparable with all the other pub­ in the response and improve the ap­ As intimated above, the circuit has lished and advertised data on the Wil­ pearance of the square wave. The use excellent listening qualities. This is a liamson circuit. The values graphed of a high level of power can make a confirmation of the measurements. The in Fig. 2 can be divided by 1.47 for relatively poor amplifier appear better additional power available shows up in those who wish to have direct compara­ on square wave tests. cleaner and better articulated bass. The bility with the meter readings obtained The of the con­ overall effect is of greater smoothness, on the intermodulation test equipment. verted amplifier is flat ± 1 db from less more definition of detail in the sound, Figure 3 shows oscillograms of than 5 cps to 200 kc. Its phase shift and better transient response. Ultra- square-wave traces taken through the reaches 3 deg. at 20 cps and at 20 kc, Linear circuits seem to have a wider complete amplifier with repetition rates indicating symmetry of response with transient bandwidth—an audible benefit of 20 cps and SO kc. Traces at inter­ respect to the audio band. which is not readily susceptible to mediate frequencies approach theoreti­ The amplifier puts out 30 watts of measurement. The combined effect of cal perfection, and even such a rigorous power over a range greater than the the Williamson circuit configuration— test as the 50-kc wave shows up ex­ audio spectrum. However, this type of a wide-band, low-distortion arrange­ tremely well. The waveform has not power curve, as measured by response ment—plus an output stage of decreased "sined off,” and the extent of ringing is at high power levels, is not too meaning­ distortion and higher power capability, less than that exhibited by the 5000- ful. The important consideration is the a stage which exceeds the original cps wave of many good quality amp­ amount of undistorled power available specification and operating parameters, lifiers. These square-wave tests were at various frequencies. The Ultra Linear must be heard to be appreciated.