I think at that time people didn't have distortion free op amps like today. Historically interesting, but boring to build.
Jan, Here is a link to the Elektor May 1987 issue that was referred to by xavirom. The article starts on page 28. https://worldradiohistory.com/UK/Elektor/80s/Elektor-1987-05.pdf
0.008 3rd is respectable for an xtal oscillator. But I agree, nowadays the Viktor oscillator is orders of magnitude better.
But it comes down to what your target application is.
Jan
But it comes down to what your target application is.
Jan
And not to forget that an alternative technique using a soundcard style interface, and even with a following booster amplifier if needed, can provide a fixed frequency oscillator configured to null final output harmonics using REW capability of adding in digital nulling harmonic levels/phases.
Seems to me a typo error, likely 0.08 not 0.008%. Starting from rectangular wave, 3-d harmonics 1/3 (-9.54 dB) , there are RC 1-st order LPF + 8-th Butterworth total 9, with theoretical attenuation 9x6 = 54 dB. So , 3-d harmonics has to be -63.54 dBc, or 0.000665273 = 0.067%
Yep. Or if you only need a discrete set of test frequencies you could design an elliptical filter with notches at the harmonics you'd like to get rid of and clean up the sound card output considerably.
Tom
Tom, my post was trying to indicate an alternative technique that didn't rely on extra analog hardware to suppress distortion (ie. taking a digital path that has only just recently become available).
A long time ago, I built a turntable power supply using a crystal oscillator, dividers, and a 4015 to generate the quadrature sine waves. The "D-A" was as shown in Figure 9.93 of "Art of Electronics" - Second Edition. (It was in the first edition, too, which I gave away. Not sure about the third edition.)
The output was actually very clean and filtering made it even better. According to the authors, "The weighting shown generates an 8-level approximation to a sine wave, as shown, with a frequency 1/16th that of the clock, and with the first nonzero distortion term (assuming perfect resistor values) being the 15th harmonic, which is down by 24 dB."
It was so long ago that I don't remember the actual results I got, but it was pretty much as described. I would remember if the idea fell short.
(For the turntable application, you just need to add some inverters and a second string of resistors to get a quadrature signal. Plus the amplification.)
An approach like this does give a very low phase noise sine wave, essentially limited in phase noise by the noise floor of the CMOS logic parts and the power supplies.
I guess that if you built two of these, ran one through a notch at the desired frequency, and took the difference of the two, you could null out the harmonics. A bridged T notch filter has almost no delay out at the 15th harmonic and beyond, so getting the proper phase relationship shouldn't be insurmountable. Or, just use a low pass filter, if you care harmonics that high in frequency.
Something similar in concept is shown here: ZL2PD Audio Signal Generator
The output was actually very clean and filtering made it even better. According to the authors, "The weighting shown generates an 8-level approximation to a sine wave, as shown, with a frequency 1/16th that of the clock, and with the first nonzero distortion term (assuming perfect resistor values) being the 15th harmonic, which is down by 24 dB."
It was so long ago that I don't remember the actual results I got, but it was pretty much as described. I would remember if the idea fell short.
(For the turntable application, you just need to add some inverters and a second string of resistors to get a quadrature signal. Plus the amplification.)
An approach like this does give a very low phase noise sine wave, essentially limited in phase noise by the noise floor of the CMOS logic parts and the power supplies.
I guess that if you built two of these, ran one through a notch at the desired frequency, and took the difference of the two, you could null out the harmonics. A bridged T notch filter has almost no delay out at the 15th harmonic and beyond, so getting the proper phase relationship shouldn't be insurmountable. Or, just use a low pass filter, if you care harmonics that high in frequency.
Something similar in concept is shown here: ZL2PD Audio Signal Generator
Tom, there are some posts and my results in a thread started by jan.didden:
Distortion analysis - a thought experiment
Ciao, Tim
Distortion analysis - a thought experiment
Ciao, Tim
1. This signal generator is much like the historic HP200A Audio generator that launched Hewlett Packard ~80 years ago.
HP200A - Wikipedia
2. Signal generators from the ~1980s used filtered square and triangle waves. There was a chip that integrated those functions. Results were never very good wrt THD.
3. Modern signal generators use DDS (direct digital synthesis). A number of DDS chips are available and provide signals from a fraction of 1Hz to many MHz. The waveform comes from a (sine) table and can be any arbitrary waveform.
4. Software DDS is available for sound card signal generators. A 16 bit DDS provides a raw signal with about 0.0015% THD before the analog filters so no filter are required in many applications.
5. An Arduino or RPi are a good choice to base a DDS signal generator. They can be made with "shield" boards or from the ~Arduino alone using software.
Signal Generator with Arduino Using DDS and Pico - Arduino Project Hub
A DDS signal generator is rock solid with no bounce or seeking, or amplitude uncertainty.
HP200A - Wikipedia
2. Signal generators from the ~1980s used filtered square and triangle waves. There was a chip that integrated those functions. Results were never very good wrt THD.
3. Modern signal generators use DDS (direct digital synthesis). A number of DDS chips are available and provide signals from a fraction of 1Hz to many MHz. The waveform comes from a (sine) table and can be any arbitrary waveform.
4. Software DDS is available for sound card signal generators. A 16 bit DDS provides a raw signal with about 0.0015% THD before the analog filters so no filter are required in many applications.
5. An Arduino or RPi are a good choice to base a DDS signal generator. They can be made with "shield" boards or from the ~Arduino alone using software.
Signal Generator with Arduino Using DDS and Pico - Arduino Project Hub
A DDS signal generator is rock solid with no bounce or seeking, or amplitude uncertainty.
Some were analog oscillators. I'm thinking specifically of the oscillator in the HP8903. It was a state-variable oscillator with three opamps. Pretty nifty circuit.
I seem to recall the oscillator in the HP3581 having pretty good THD as well. "Pretty good" for the time was -80 dBc spec, -90ish dBc actual performance.
The oscillator in my HP8903 hit 0.002 % (-94 dBc) at 2 V RMS, 1 kHz, 20 kHz BW.
Tom
I realized after that THD is as power measurement so a 16 bit DAC is more like 0.000000023% THD but it also depends on the wave table length which is probably not 64K samples long.
I should mention that as few as 3 bits plus sign bit makes a fairly good sine wave and with a tracking filter almost perfect. That would be a simple /16 divider and x16 oscillator.
The chip I was thinking of is the XR2206.
https://www.electroschematics.com/function-generator/
I should mention that as few as 3 bits plus sign bit makes a fairly good sine wave and with a tracking filter almost perfect. That would be a simple /16 divider and x16 oscillator.
The chip I was thinking of is the XR2206.
https://www.electroschematics.com/function-generator/
With percentages there are too many zeros making for lots of potential errors so I much prefer dBC when talking about distortion products.
Digitally derived sine waves are really limited by the digital processing when you get below -100 dBC. The best seem to get to -120 dBC but its not easy. Very small errors from any of a number of sources will limit the performance.
The best analog sources are at -150 to -170 dBC. Its easy to blur function generators, RC oscillators as well as Digital synthesizers and arbitrary function generators which are similar but not the same. Function generators like the XR chip start by generating a triangle wave using current sources and caps then either squaring them or using various techniques 'rounding' the triangle into a sine wave. The best get to .1% THD (-60 dBC) right after careful tweaking. RC oscillators using clean amps and good agc can get to -150 dBC THD (e.g. Victor's oscillator) but every component will matter at that level of performance.
Synthesizers use digital dividers and sometimes VCO's to derive the output frequency. Challenges are distortion since the output starts digitally and spurious outputs since the fractional divides can have 'residue' which translate into small nonharmonic outputs. Arbitrary generators use DACs and clock frequency control to play stored bit sequences which can include sine waves. Their quite useful but are limited by the DAC's performance (16 bit at best) and output filters. The goal on most ARB's is not ultra low distortion but regenerating specific waveforms.
Each type of source has benefits and specific applications. The digital sources can have really precise frequency. The RC oscillators a freedom from harmonics and spurious outputs. Function generators can change frequency really quickly and have really good frequency flatness across a broad frequency range. it depends on what you need to test for. I have all these variations on my bench.
Digitally derived sine waves are really limited by the digital processing when you get below -100 dBC. The best seem to get to -120 dBC but its not easy. Very small errors from any of a number of sources will limit the performance.
The best analog sources are at -150 to -170 dBC. Its easy to blur function generators, RC oscillators as well as Digital synthesizers and arbitrary function generators which are similar but not the same. Function generators like the XR chip start by generating a triangle wave using current sources and caps then either squaring them or using various techniques 'rounding' the triangle into a sine wave. The best get to .1% THD (-60 dBC) right after careful tweaking. RC oscillators using clean amps and good agc can get to -150 dBC THD (e.g. Victor's oscillator) but every component will matter at that level of performance.
Synthesizers use digital dividers and sometimes VCO's to derive the output frequency. Challenges are distortion since the output starts digitally and spurious outputs since the fractional divides can have 'residue' which translate into small nonharmonic outputs. Arbitrary generators use DACs and clock frequency control to play stored bit sequences which can include sine waves. Their quite useful but are limited by the DAC's performance (16 bit at best) and output filters. The goal on most ARB's is not ultra low distortion but regenerating specific waveforms.
Each type of source has benefits and specific applications. The digital sources can have really precise frequency. The RC oscillators a freedom from harmonics and spurious outputs. Function generators can change frequency really quickly and have really good frequency flatness across a broad frequency range. it depends on what you need to test for. I have all these variations on my bench.
The 'modern' XR2206 used in that kit is not the same as the 'vintage' version - it has known power supply restrictions, and possibly other performance issues that may not meet 'original' datasheet specs.
I was told (being in Silicon valley in the 1970's) that the original XR stuff was built on a generic silicon base that could be inexpensively programmed by customers to make things like the XR function generator. This was from a guy who used the process for something else. Since that is all long gone and the chip on a completely different process it would not be a surprise that its not really similar.
I posted here QuantAsylum QA400 and QA401 some comparisons of oscillators using the QA480 notch. It works quite well with the QA401 as a system.