Elektor•Post Project No. 53 Reference Sine Wave Generator How do you build a generator using only a few components while obtaining the lowest possible distortion? Here follows the account of the practical experience from many experiments, which resulted in a state- variable oscillator with a distortion well below 0.001%. When measuring, calibrating and testing of of a sound card is often the result of all kinds By audio (test) equipment the availability of a of interference products which appear in the Hein van den Heuvel test oscillator, which is capable of generating audio bandwidth. In any case, my hobby PCs (Netherlands) a sine wave (of 1 kHz, for example) that is as (a desktop and two old laptops) with built-in pure and low distortion as possible, is practi- sound cards were all unsuitable for generat- cally indispensable. These days, an obvious ing something that looked like a clean signal. solution would be to use the sound card of a PC as a sine wave generator. In conjunction Low distortion with few with one of the many available (freeware) components programs it is very easy to obtain a defined Would it be possible the build a sine wave gen- sine wave from the DAC in the sound card. If, erator, using only a few components, which however, we take a close look at the specifi- provides a much better performance? A gen- cations of a good sound card (even including erator that can produce a signal with a dis- those with 24-bit resolution) then the specified tortion in the region of 0.001% (–100 dB) or values for distortion will be somewhat disap- lower, that would be very nice. A sine wave pointing. Even a more expensive card or (even signal with such a low distortion would be better) an external sound adapter which is a very usable reference signal, certainly for connected via USB, generates signals with a the audio measurements that I had in mind. THD somewhere between 0.01% (–80 dB) and Because there are plenty of Elektor readers 0.003% (–90 dB). The aforementioned THD who are also keen to experiment with analog number gives the ratio between the voltage circuits, I would like to elaborate about the of the sine wave signal (the fundamental) and circuits and designs you are likely to encounter all other noise components. These noise com- during your quest. There is, after all, a certain ponents include harmonics, random noise and charm to assemble a few components, without non-harmonic whistles and beeps. In practice getting involved with firmware or software, it turns out that the distortion at the output and have something functional in your hands. elektor post | Project No. 53 | 1 Elektor•Post Project No. 53 And for test equipment it remains downright convenient to have a standalone device which 15V P1 will work by itself, without having to drag C1 along a PC as well. 100R C3 C5 MT The challenge for me was to build a sine wave 120n generator that could be easily copied by oth- R1 R3 100n 100u 25V ers, uses only a minimal number of compo- 1k33 2k05 IC1 nents and has a low cost. Oh yes, and also 1 requires no calibration. The aforementioned 3 7 1kHz/8Vrms 5 6 desire to obtain a distortion below 0.001% OPA627 (–100 dB) is, of course, also still standing. 2 R4 4 If our preference is for a design with a min- imum number of components, then a natu- 1k ral first attempt is to start with a classical R5 Wien bridge oscillator. Much useful informa- 3k3 R2 tion about this can be found in an application C2 La 1 note from the company Linear Technology, 1k33 120n written by Jim Williams [1]. This contains a 12V C4 C6 50mA few examples of oscillators with a Wien bridge. In its simplest form, in addition to the Wien 100n 100u 25V bridge, there is an opamp as amplifier, while an ordinary incandescent light bulb serves as 140424 - 11 15V amplitude stabilizer. A light bulb in a mod- ern electronic circuit? This is perhaps not so strange when you realize that an incandescent is mostly generated by the lamp. This is Figure 1. lamp effectively operates as a voltage-depen- because of the fact that the dissipation 1000-Hz sine wave dent resistor. By using the properties of the in the lamp varies during the period generator from Elektor filament we obtain a simple form of ampli- of the sine wave. This therefore varies 12-94 tude control. After all, the higher the volt- the temperature of the filament in time age across the lamp, the higher the dissipa- with the sine wave and therefore also tion, and that means a higher temperature the resistance of the wire. The result of the filament. Just like most other metals, is a distortion which predominantly the resistance of the tungsten from which contains a third-order component. The the filament is made has a positive tempera- amount of this distortion also depends ture coefficient: A higher temperature results on the thermal inertia (‘slowness’) in a higher resistance of the wire. The fil- of the filament. This is, among other ament therefore behaves like a PTC-resis- things, also dependent on the (average) tor. By including the incandescent lamp in a temperature of the filament. It is feedback circuit a controller can be realized difficult, if not impossible, to establish which will reduce the gain when the ampli- the exact amount of distortion that will tude of the sine wave signal increases. When result beforehand. This will therefore the closed-loop gain becomes less than 1, have to be established experimentally. then amplitude will reduce again. After some In practice this means an acceptable time, a steady-state will be reached at a cer- compromise has to be reached between tain amplitude. In this way the amplifier is the amount of distortion, speed of the prevented from limiting itself. This limiting control loop and amplitude stability. action would otherwise result in a distortion level that is unacceptably high. Sine wave generator If we prefer a low distortion at the expense In the story from Linear Technology of amplitude stability, then with a few more even the simplest Wien bridge oscillator components we can retrieve a previous already generates a signal with a Elektor design: A simple sine wave oscil- distortion (THD) of only 0.0025% lator was already described in Elektor some (–92 dB). Not bad at all! This distortion 20 (!) years ago [2]. The opamp used in elektor post | Project No. 53 | 2 Elektor•Post Project No. 53 R1 10k R2 10k C1 C2 R6 2 10n 27k R3 1 6 10n R5 IC1.A 16k R4 3 7 9 330R IC1.B 16k R7 5 8 IC1.C 470R 10 LA1 6V 30mA +15V C3 12 14 4 IC1.D 220u 25V 13 IC1 = TL074 IC1 C4 11 220u 25V Figure 2. –15V State-variable sine wave 140424 - 12 generator with a lamp as stabilization element. this design (see Figure 1) costs more than tude. There are various alternatives to the € 20 these days; that is why I first tried incandescent lamp as amplitude regulator to build the sine wave generator with a in the form of JFETs and opto-couplers, for cheaper opamp, as was already suggested example. A number of these can be found in the original article. in the aforementioned story from Linear The results were not disappointing: with Technology. Some — a trade-off with sim- one of the two opamps in a TL082 a THD of plicity. And if you then also have to use –96 dB was obtained and with a NE5532 I expensive or critical components then we even reached –105 dB. Perfectly suitable for will have missed our design goal. me. But it was, however, not all that straight- So, the circuit with the lamp is and remains forward: the low distortion value was only a great starting point. You could, for exam- obtained after some experimenting. The com- ple, start with the Wien bridge oscillator ponent selection appears to be very critical (with lamp) and follow it with a simple and also the calibration of the amplitude has band-pass filter or low-pass filter to sup- to be very precise. It is very difficult to ‘tame’ press the harmonics. This could be built the amplitude of the oscillator in this design. very easily using a dual opamp. When the ambient temperature increases the Continuing with the possibility of building amplitude also decreases quickly. This is to an oscillator followed by a harmonic filter, be expected, because the operating point of I arrived at an oscillator type that effec- the oscillator is such that the temperature tively has the filter built in: the ‘state-vari- of the filament is very low. This means that able’ oscillator. This type of oscillator was variations in the ambient temperature have often used, among other things, in profes- a significant effect on the output amplitude. sional audio test equipment before digital techniques became commonplace. Strange State-variable oscillator enough, I never came across such an oscil- In spite of all this I was keen to continue lator that uses a lamp as its stabilizing experimenting, with the objective of finding element. So it is of course interesting to a circuit that is less sensitive to component quickly try out such a lamp-circuit on a values and has a more stable output ampli- breadboard. With surprisingly good first elektor post | Project No.