I spent weeks of spare time on this. Even adding an isolation transformer. Still the distortion increase was significant. I'm ready to go back at it but need a fresh take on the process.
Was it the injection signal that was disturbing the oscillator in a particular way that was causing distortion? I assume that the injection signal itself would at least have to be very low in distortion itself. Also, there might be a mechanism at work that causes distortion just as a result of pulling the oscillator frequency a bit. Might the distortion be a function of how much the oscillator was being pulled?
I would have guessed that the best approach to pulling the oscillator to the exact desired frequency would be to use a PLL. This is very straightforward with a state variable oscillator, but a little less straightforward with a Wein bridge oscillator. If only pulling it a little bit, a JFET VCR could probably be used to tweak one of the tuning resistors to pull the frequency a little bit.
Pulling the frequency of a state variable notch filter is the approach used for autotune in many analog THD analyzers.
Victor's oscillator has remarkably low distortion, and it would be interesting to see if a state variable oscillator could perform as well if it was designed and implemented with the same amount of care and expertise.
Cheers,
Bob
It does when powered from a Powerbank. Which is recommended for sensitive uses. I always have a few charged Powerbanks at hand ;-)
Jan
Thanks Demian. My experience is that where you inject the signal is critical and I expect it to be a critical issue between level and the amount of frequency pull required.
I suspect there is a very nonlinear cosine shaped relationship between pull amount and distortion.
My distortion reduction technique as mentioned before is to use an oscillator topology set just below oscillation as a filter for the same topology in critical-ish oscillation. Matching of the frequency networks is of course critical when the oscillator is analog. When it is digital I can get decent cleanup and frequency stability.
I have not done a direct comparison between frequency locking and filtering to compare distortion. I do know level stability can be significantly better with the filter approach as it allows a lot less critical level control or even fixed crystal like references.
The other advantage of the filter approach is that you can deliberately lock harmonic oscillators for additional wave shaping or distortion reduction.
I think the limit is still the active components and then the capacitors.
Using lock in amplifiers it is possible to measure harmonics 160 dB below the fundamental. Just to drive folks nuts the amplifiers can be DSP based!
I suspect there is a very nonlinear cosine shaped relationship between pull amount and distortion.
My distortion reduction technique as mentioned before is to use an oscillator topology set just below oscillation as a filter for the same topology in critical-ish oscillation. Matching of the frequency networks is of course critical when the oscillator is analog. When it is digital I can get decent cleanup and frequency stability.
I have not done a direct comparison between frequency locking and filtering to compare distortion. I do know level stability can be significantly better with the filter approach as it allows a lot less critical level control or even fixed crystal like references.
The other advantage of the filter approach is that you can deliberately lock harmonic oscillators for additional wave shaping or distortion reduction.
I think the limit is still the active components and then the capacitors.
Using lock in amplifiers it is possible to measure harmonics 160 dB below the fundamental. Just to drive folks nuts the amplifiers can be DSP based!
Last edited:
The VP7722A distributed under Panasonic brand uses 2 notch filters. Can be found on e-Bay etal for about 1200 USD. Same basic topology as ShibaSoku 725D
Dropbox - Panasonic VP-7722 service manual.pdf - Simplify your life
Either will help with ultra-low distortion generator development. Note -The VP7722 has built in generators (THD and IM). Easily beats its specs.
Both analyzers can be used with external FFT on the residual port provided and get much lower distortion readings.
THx-RNMarsh
Last edited:
here is KH 4402B with a passive notch at 2H... to cancel out the 2H
can do same for 3H. Just a crude proof of concept's effectiveness. Modest notch depth.
if you notch out 2H and 3H... pretty much for most gen = zero HD
Maybe a tracking notch to remove residual harmonics from a variable freq low distortion generator.??
2 years ago.
THx-RNMarsh
can do same for 3H. Just a crude proof of concept's effectiveness. Modest notch depth.
if you notch out 2H and 3H... pretty much for most gen = zero HD
Maybe a tracking notch to remove residual harmonics from a variable freq low distortion generator.??
2 years ago.
THx-RNMarsh
Last edited:
Gossip from the lab
I needed a good case for my oscillator. I found it and slaughtered a $1500 Nokia Crypto VPN router. Good enough for my little oscillator. The first screenshot shows the switched off oscillator (everything below approx. -150dB). The second screenshot shows the switched on oscillator (battery powered) with the 1st harmonic, which dances around -145dBc. I need a better ADC. As you know, you are never satisfied yourself. Any recommendations?
I needed a good case for my oscillator. I found it and slaughtered a $1500 Nokia Crypto VPN router. Good enough for my little oscillator. The first screenshot shows the switched off oscillator (everything below approx. -150dB). The second screenshot shows the switched on oscillator (battery powered) with the 1st harmonic, which dances around -145dBc. I need a better ADC. As you know, you are never satisfied yourself. Any recommendations?
Attachments
Very good question.
What is the amplitude of the fundamental frequency in this measurement? We can't see the full graph.
If you really can measure such low harmonics directly on the ADC, then this sound card is unique. I have never seen residual harmonics below -130 dB when the best ADC chips was used. Xonar DX uses CS5361 ... According to the specification, this is not the best ADC chip.