I recently built a simple low distortion audio oscillator based on an old design by Roger Rosens (Wireless World 1982). It uses the obsolete R53 thermistor which I just happened to have. It's a clever design as it cancells third harmonic distortion introduced by the thermistor. The original article suggests it should give THD around 0.0005% at 1kHz.
I don't have a spectrum analyser to measure distortion, so I thought I would try some of the free software that uses the ADC in the PC / sound card.
I tried using the Creative Sound Blaster Z PCIe card installed in my Win 10 Desktop.
The results looked much worse than I expected: THD 0.007%. Mostly 2nd harmonic.
At first I suspected the oscillator, so I spent some time trying different things but nothing I tried made any improvement.
Eventually I tried using a Creative Sound Blaster Audigy 2 PCI card installed in an old 32bit XP machine.
The results appeared considerably better: THD 0.001%
I now don't believe the measurements made using either system. I need something better.
It may be that the Sound Blaster Z is worse because it does not have a dedicated line input jack. Instead it shares the same jack with the mic input.
I am guessing there must be some gain adjustable amplification or switching between the jack and the ADC that is causing extra distortion.
The older Audigy has separate jacks for mic and line input. However, I dont know how much extra residual distortion is added by the card.
Can anyone recommend a better PCIe sound card with low input distortion to use with FFT analysis software?
I don't have a spectrum analyser to measure distortion, so I thought I would try some of the free software that uses the ADC in the PC / sound card.
I tried using the Creative Sound Blaster Z PCIe card installed in my Win 10 Desktop.
The results looked much worse than I expected: THD 0.007%. Mostly 2nd harmonic.
At first I suspected the oscillator, so I spent some time trying different things but nothing I tried made any improvement.
Eventually I tried using a Creative Sound Blaster Audigy 2 PCI card installed in an old 32bit XP machine.
The results appeared considerably better: THD 0.001%
I now don't believe the measurements made using either system. I need something better.
It may be that the Sound Blaster Z is worse because it does not have a dedicated line input jack. Instead it shares the same jack with the mic input.
I am guessing there must be some gain adjustable amplification or switching between the jack and the ADC that is causing extra distortion.
The older Audigy has separate jacks for mic and line input. However, I dont know how much extra residual distortion is added by the card.
Can anyone recommend a better PCIe sound card with low input distortion to use with FFT analysis software?
Last edited:
BTW, I do have an old HP 334a distortion analyser, but that instrument has a residual distortion reading of 0.03%, so it's nowhere near good enough.
You should test the oscillator at an input level around -10 to -20 dB from FS on the card. Most ADC's are at their best in that range. Building a passive notch can greatly enhance the ability to look at very low distortion levels. There is a lot about these techniques spread across some gigantic threads here.
be careful not to mix up THD+N and THD. The -N part is really significant at low distortions and makes it harder to know what you are reading. Are you using ARTA?
Can you share the schematic? That would be very low distortion for a thermistor based oscillator. I would like to see his techniques.
be careful not to mix up THD+N and THD. The -N part is really significant at low distortions and makes it harder to know what you are reading. Are you using ARTA?
Can you share the schematic? That would be very low distortion for a thermistor based oscillator. I would like to see his techniques.
Demian,
It's this one:http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1982/Phase-shifting%20oscillator%20-%20Roger%20Rosens.pdf
The osc that kicked off THE Monster thread was derived from this Rosen osc.
Simon B
It's this one:http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1982/Phase-shifting%20oscillator%20-%20Roger%20Rosens.pdf
The osc that kicked off THE Monster thread was derived from this Rosen osc.
Simon B
just google card name + rmaa
which will usually lead to reviews that use RMAA tests, give numbers for performance in loopback
and repeat the loopback tests yourself to see if your card, computer or setup is out of line
natually DAC could be limiting in loopback tests - but it should give a good idea, in fact the residuals you give are just about what I found for each card
which will usually lead to reviews that use RMAA tests, give numbers for performance in loopback
and repeat the loopback tests yourself to see if your card, computer or setup is out of line
natually DAC could be limiting in loopback tests - but it should give a good idea, in fact the residuals you give are just about what I found for each card
Today I tried adding an active twin tee filter between the oscillator and the Audigy sound card.
The filter attenuated the fundamental by about 54dB.
Although the second harmonic was reduced, the third harmonic initially appeared to increase by a few dB.
At first I thought this might be common mode distortion from the voltage followers in the active filter. However, I discovered that if I removed the coupling capacitor between the oscillator and filter, leaving it DC coupled, the third harmonic came down.
With the filter in place, the THD displayed by the FFT software indicates 0.2%.
To get the true THD figure, taking into account the 54dB drop in the fundamental, we have to multiply by a factor of 10^(-54/20) = 0.002.
That gives a figure of 0.0004% THD.
The second harmonic is 108 dB down.
Third harmonic is 125dB down.
Higher order harmonics are lost in the noise floor.
If I've got this right and the result is to be believed, it is quite remarkable for such a simple oscillator circuit.
It may not perform as well at low frequencies, but at 1kHz, it's performance is close to the more complex solutions using sample & hold detectors, FETs, opto couplers, multipliers etc.
It's certainly adequate for my own needs. I would struggle to measure anything much better with the limited equipment I have.
It's a shame the R53 thermistor is no longer in production.
The only difference between my implemention and Roger Rosen's original circuit, is a trimmer in parallel with the thermistor to further reduce the distortion. This is ok for a 1kHz spot frequency oscillator, but it could give stability problems for an oscillator that needs to cover the full audio band.
The filter attenuated the fundamental by about 54dB.
Although the second harmonic was reduced, the third harmonic initially appeared to increase by a few dB.
At first I thought this might be common mode distortion from the voltage followers in the active filter. However, I discovered that if I removed the coupling capacitor between the oscillator and filter, leaving it DC coupled, the third harmonic came down.
With the filter in place, the THD displayed by the FFT software indicates 0.2%.
To get the true THD figure, taking into account the 54dB drop in the fundamental, we have to multiply by a factor of 10^(-54/20) = 0.002.
That gives a figure of 0.0004% THD.
The second harmonic is 108 dB down.
Third harmonic is 125dB down.
Higher order harmonics are lost in the noise floor.
If I've got this right and the result is to be believed, it is quite remarkable for such a simple oscillator circuit.
It may not perform as well at low frequencies, but at 1kHz, it's performance is close to the more complex solutions using sample & hold detectors, FETs, opto couplers, multipliers etc.
It's certainly adequate for my own needs. I would struggle to measure anything much better with the limited equipment I have.
It's a shame the R53 thermistor is no longer in production.
The only difference between my implemention and Roger Rosen's original circuit, is a trimmer in parallel with the thermistor to further reduce the distortion. This is ok for a 1kHz spot frequency oscillator, but it could give stability problems for an oscillator that needs to cover the full audio band.
Actually a good PCIe card can reach residual THD levels as low as 0,0005% or better, used as an audio generator or as a THD meter/spectrum analyser.
If DIY a generator were not your main goal, you would use a good (and inexpensive or even free) software, as RMAA, and a good sound card, like Juli@, Asus Xonar, EMU etc to do that.
http://www.diyaudio.com/forums/equipment-and-tools/248720-emu-1212m-measurements-7.html#post5096961
Best regards,
If DIY a generator were not your main goal, you would use a good (and inexpensive or even free) software, as RMAA, and a good sound card, like Juli@, Asus Xonar, EMU etc to do that.
http://www.diyaudio.com/forums/equipment-and-tools/248720-emu-1212m-measurements-7.html#post5096961
Best regards,
- Status
- Not open for further replies.