Ultra low distortion audio oscillator the neverending search for a pure sine

Ultra Low Harmonic Distortion 10kHz Oscillator with -160dBc

A do it yourself wien RC bridge sine generator with ultra low distortion in a range almost impossible to measure. An sine oscillator good for distortion measurment of operational amplifiers and audio equipment.

Generator - Frontansicht.jpg (18885 Byte)

Audio_Generator_Leiterplatte.JPG (35249 Byte) Generator - Unterseite.jpg (34451 Byte)

Frontview in a aluminium box with the "cake" batterybox 

Layout with amplifier, RC bridge, temperature compensation and voltage regulator

Bottomview, good to see the multilayer ground

Generator - Rückseite.jpg (28603 Byte) Stromversorgung durch - Batterie.jpg (29799 Byte) Erster Versuchsaufbau, fotographiert mit 16mm Objektiv - Fischauge.jpg (40816 Byte)

Backview with power supply, input and output of errorsignal

Batterybox with 2*20V NiCd accus in the cake EMC box. The cakes were delicious.

First running prototype, (foto done with 16mm fisheye objectiv)

" a search for a pure sine is a never ending story".

A step in the right way - a pure oscillation

This is a circiut I saw in a Linear Technology application note, a very interesting sine oscillator. Described in application note No. 67 on page 62 www.linear.com . In this note a designer developed a 10 kHz oscillator with a second harmonic distortion in the ppb parts per billion, 160 to 180dB, my full respect! I saw many sine oscillators, but this I build up at once. Specially the high open loop is very important for every control circiut. To get a high gain it is a main target for most controls. A small errorsignal is fine, but problem exists called stability. Ahuge open loop reduce distortion enormously. The designer was tricky enough for a super gain block of 180dB @ 10kHz. He took three 60dB amps in series - stable-! It is an frequency selective amplifier with an closed loop gain of 180dB at 10 kHz. A oscillator well suitable for distortion measurements on opams and many other audio applications. The designer was tricky enough to do a super gain block of 180dB @ 10kHz. He put three 60dB amp in series - stable! It is a selctive amp with 180dB at 10kHz. A oscillator well suitable for distortion measurement on opamps and many other audio application.

Important a precise builded up design

The ground connections must be state of the art to reach the possible low distortion at all. Wide groundplanes on different layers with low coupling under Ohm's law. It isa very important issue for maximum performance. The shielded battery power supply box leds to a good rejection of external sources, special for AC line transmitted by air.

HP3580A und Audio_Sinus_Generator.JPG (19159 Byte) Einstellungen_der_Messung_Spektrum__Audio_Generator.JPG (22329 Byte) 10kHz Audio Generator Signal mit 0dBVrms an HP3580A Analyzer Spektrum_Audio_Sinus_Generator.JPG (17004 Byte)

HP3580A low frequency spectrumanalyzer with the 10 kHz audio sine generator.

Settings on the spectrumanalyzer, for a better contrast on the photo I a dimmed room illumination.

A dream of a spectrum, 1Vrms on the input. Dynamic range of the analyzer 90dB. Measurement needs 2000 seconds.

Sauberes Spektrum bei 10 Volt rms und 1kOhm Belastung Hohe_Ausgangsspannung_10Volt.JPG (23264 Byte) Sinus_13V.JPG (14059 Byte) Sinus_100mV.JPG (9623 Byte)

A high amplitude of 10Vrms and a load of 1kOhm, the circiut is still within analyzers dynamic range.

Of course you see on the scope only a perfect oscillation.

With the lowest adjustable output amplitude of 100mV the beam gets thicker, indicates the working amplitude control.

"measure accurate the harmonics of this oscillator is almost impossible".

It is only possible to guess the harmonics: an open loop of 180dB leds to a theoretical harmonic distortion in a similar range.Possible only with an excellent build up and best available bridge parts and a well designed layout. On the test point "errorsignal" in the circiut, the 10kHz outputsignal is rejected by 60dB, but the errorsignal is unreduced. You can compare this point with a less distortion 10 kHz notch filter. At this test point there is a low second harmonic only. After doing this calculation not more than a guess can be made for the second harmonic distortion, 150 to 160 dB below output level.

What means 160dB? Almost nothing. A relation of one hundred millions to one. Imaginge the 20kHz second harmonic amplitude would have a length of only 1 inch, the basic 10 kHz would have a length of 1000 kilo inches. Twice the lenth of Germany compared to an inch.

1000 kilo inch output and only 1 inch harmonic distortion

"To nice if this are the only distortions".

The pictures are made with a 24bit ADC PCI soundcard and a program, which I wrote in LabView. Unfortunately LabView supports soundcard measurements only in the 16 bit mode, that means the the remaining bits are unmasked. If somebody knows how to solve this problem, I'am happy about every help. For all measurements, the used converters range 89%, a good compromise between dynamic range and nonlinearity of the converter. The AD converter has an integrated high order anti aliasing filter. The AD converter works with a sample rate of 44.1kHz, so unfortunately the third harmonic 30kHz can't be measured. The measured distortions came from the AD converter. The measured -115dBc are an excellent nonlinearity for an audio converter working in the 16bit mode. The measurement time was very long, enough to reduce random noise by averaging. With this method smallest periodical signals get visible.

Spektrum_FFT_6dBV_84mHz.gif (54140 Byte) Spektrum_FFT_8dBV_10k_zoom.gif (54529 Byte) Spektrum_FFT_6dBV_20k_84mHz.gif (52308 Byte)

Input signal 10kHz 2V bandwidth 84mHz, the measurement shows a harmonic of -115dBV below the fundamental. The nonlinearity of the AD converter cause this harmonic. Through the long mesurement time and the low adjusted bandwidth random noise signals are reduced

Input signal 10kHz 2V bandwidth 84mHz. A zoom within the frequency axe, shows two small mixing products close to the carrier. Most analyzers are not able to show very small signals, they don't have a high dynamic range plus low resolution bandwidth to measure close to the carrier.

Input signal 10kHz 2V 84mHz. A zoom within the frecuency axe shows two mixing products close to the second harmonic.Guess it is a modulation of the second harmonic with a AC harmonic (Europe use 50 Hertz line power).

Spektrum_FFT_6dBV_40mHz.gif (57993 Byte) Spektrum_FFT_6dBV_10k_zoom_42mHz.gif (53558 Byte) Spektrum_FFT_6dBV_DC_zoom_42mHz.gif (57430 Byte)

A wonderful nice spectrum, input signal 10kHz  2V  bandwith 42mHz. Measurement conditions like in the pictures above, only with halfed bandwidth. Averaging reduce the periodic noise to -146dB below center frequency.

Input signal 10kHz 2V bandwith 42mHz. Thru the halfed bandwidth compared to above it's more easy to see the both mixing products close to the carrier.

Input signal 10kHz 2V bandwidth 42mHz. Low frequency zoom. A small distortion caused by the AC line, (remember the soundcard is mounted in a personal computer with a dirty EMC environmental). Noise near DC rises up.

Spektrum_FFT_am_Fehlersignal_8dBV.gif (54523 Byte) HP3325A_Spektrum.gif (47070 Byte) HP3325A_Spektrum_2k.gif (54962 Byte)

Another measurement. Input signal 10kHz 7.5dB and bandwidth 672mHz Measuremnet is done at the errorsignal on oscillator backside. Using this signal a guess were made for the THD.

Example: spectrum of a Hewlett Packard 3325A frequency generator. This excellent general purpose generator was never designed by the manufactorer for distortion measurements. Easy to see the distortions are too high. Input signal 1kHz 2V bandwith 42mHz.

Spectrum of Hewlett Packard HP3325A in frequency zoom mode. Easy to see the parade of mixing products by the carrier and the high AC distortions. Higher noise. Inpu signal 1kHz 2V bandwidth 42mHz.

"what can you do with this thing?"

For a measurement THD + noise + other distortion, the source contains still some little mixing products and is a little noisy near DC. But an excellent source to measure harmonic distortion, it is curently the best source I use. Very good to measure nonlinearities of digital to analog converters or amplifiers. Top application is distortion measurement with audio amplifiers. Which audio amplifier do not like a pure sine like this.

"high open loop cause also problems"

A circiut with such a high open loop makes it more and more difficult to control the amplitude, the designer did already an excellent work with the amplitude control circiut. Only a very small uncontrolled errorsignal could be enough and the outputs responds with clipping.

A amplitude modulation e.g. caused by the amplitude control seen on the analyzer as two little lines close to the carrier and close to DC. The modulation frequency of the amplitude control circiut seen as delta frequency between mixing products and the carrier or DC. For most analyzers and also many FFT analyzers it is difficult to show such details. Many distortion measurement brigdes and audio analyzers, their build in tuneable filters are not sharp enough measuring such details. They can't measure these distortions very close to the carrier, due too high filterbandwidth, the instruments "think" these distortions are part of the signal. As result their displayed THD + noise + other distortion measurement result is lower than in reality. The high open loop gain leads to a high sensitivity for external signals. This super gain block amplifier takes his job very seriously, every signal on the inputs will be powerfull amplificated.

Now you understand, why I have done such enormous investigations in shielding and powersupply, it is necessary to keep the noise + other distortions as low as possible.

"what's going on?"

Of course it would be very nice to push more gain and performance to the circiut, but for what? You really want to reduce the harmonics more? Can you measure it? It makes no sense, no audio amplifier reaches such a low distortion level. Ok, perhaps some special measurement amplifiers. Yes, to design a better amplitude control, a good idea. I've never try it, but I think it is not an easy task and takes much time to get a better result, the current amplitude control is already very good. If somebody have a good idea, please contact me.

One issue could be getting a better stability frequency vs. temperature. I tried to compensate the time constant of the Wien Bridge. I reached a stability of 10,000 Hz +/-3Hz over a temperatur delta of 30 Fahrenheit, that was enough for me, I will not use the generator in a desert, arctic or in a nordic sauna. The visible wires wrapped around the big brown capacitor are resistor wires as heater. I designed a phase lock loop with heater and a controlled fan above the capacitor for cooling. I used a lot of time to run the control. A work with many parts and investigations. Another bad aspect, I saw the fan's influence on the spectrum. I closed the project. Another method could be a oven or a peltier element. But really no - for what - you really think it makes a different to measure the distortion at 9995 Hz or 10005 Hz? Stop this nonsens, don't try it, go and make better a nice amplifier, more gain for you, and the amp don't make always the same sound.

"the search for a pure sine is a never ending story"

Contact and Disclaimer

www.amplifier.cd english         Applications Notes