Right. In addition it seems infiniteness of FM spectrum is exagerated. I tried LTSpice simulation, and the spectrum drops to below -140db very fast within audio band.PMA said:
IIRC, the gist of infinite spectrum of FM is related to linear change of frequency, where delta-f and delta-t are approaching zero. To express that, you'd sorta need infinite precision in both time and frequency. The infinite bandwidth is kinda expressing infinite precision.
With PC generated signal we aren't really even close to such precision, so I wonder if what we get is discrete frequency changes after discrete timesteps.
What is your interpretation of your opamp measurement results, in terms of what non-linearities are exposed? To me their spectrums look suspiciously like exact clones, and I'm wondering if that is really exposing any nonlinearities of amps or rather some measurement issues..
I wonder what effect tested opamp propagation delay has on the re-sampled results. The same audio card outputs the test signal and records the delayed result. Could the delay cause changes to measured spectrum? Loopback test has no delay..
Well it should be added that they are not the spectra of opamps itself, but poweramp with opamp as a driving stage (schematic can be seen on the web page). For opamps itself (627, 134), exactly said in the audio buffer circuit, there was almost no difference between test signal spectrum and output of the audio buffer circuit spectrum. For the power amp tested, it seems to me that it is a copy around the 2 x Fc frequency with amplitudes similar to harmonic distortion. The THD of the amp tested at 10kHz is higher to that of 1kHz.
Regarding the mess at the low frequency end, it is a problem of ground loop caused by card-out card-in measurement with amplifier inserted inside the measurement loop. With different signal source (like CD) this mess disappears.
Regarding the mess at the low frequency end, it is a problem of ground loop caused by card-out card-in measurement with amplifier inserted inside the measurement loop. With different signal source (like CD) this mess disappears.
Simple simulation of various distortion mechanisms will quickly reveal what the spectrum of the test waveform will look like.
A simple 2nd order distortion adds 2nd harmonics to each and every spectral line in the original. Same for third order and third harmonic. This is according to the usual theory and according to simulation results. The "cross-term" are down a fair bit if distortion is very low to start with. Nothing unusual, and nothing is particularly learned about the device that can't be obtained by standard harmonic analysis using sinusoids.
For example, using matlab (or octave) the simulation of these distortion sources are, for signal s:
s_distorted = s + A1 * s.^2 - A2 * s.^3
In the above formula the A1 and A2 are small, much less than 1.
If the device under test is behaving differently that these simulations would indicate, then there is some other distortion mechanism taking place. I would suggest that a test sequence be performed, one data collection with the test signal under discussion, the next with a sine wave. If these tests are performed in rapid succession, thermal drift will be less of a consideration. The resulting spectra can then be compared the a distortion model as above, and the consistency between the two types of test waveform can be compared (do they obey the same distortion model?).
To get to the question: what distortion mechanisms seem to be present?, it appears that 2nd and third harmonics are present, indicating that the 2nd and 3rd order distortion mechanisms represented in the formula above are present.
wimms said:
Maybe. To be sure, I am planning to build high performance differential testing amplifier to eliminate hum. But the results differ for different opamps in the same amplifier. For AD844 and LT1122 one can see 2nd a 3rd harmonic content, for AD797, OPA134 and OPA627 only 2nd harmonic.
wimms said:
wimms,
I have just built a differential measuring amplifier based on INA217 circuit. I am placing it at the card input. This is a substantial improvement to loop-like card-out card-in measurements, removing hum related problems. I will have to remake some of my measurements.
Pavel
The completely new results obtained by use of differential measuring amplifier and 6 different FM test signals are shown here:
http://web.telecom.cz/macura/fmanalysis/freqdist.html
http://web.telecom.cz/macura/fmanalysis/freqdist.html
Pavel
Agreed. I did see that. I'm wondering though what this type of test means now. Please let me clarify. The differences between the class a and the error correction amps are significant. So from a listening standpoint I would assume based on your graphs that the error correction amp sounds "better". For whatever that's worth.
I'm taking from previous threads that the point of this work was to attempt to characterize an amp when the THD is very low but the amps sound different.
I've never seen this type of test conducted before but it looks like it has merit, much merit.
I guess the hard part now begins. Investigate the data and quantify it.
Please keep up the work. I am learning much. Thanks
Mike
Agreed. I did see that. I'm wondering though what this type of test means now. Please let me clarify. The differences between the class a and the error correction amps are significant. So from a listening standpoint I would assume based on your graphs that the error correction amp sounds "better". For whatever that's worth.
I'm taking from previous threads that the point of this work was to attempt to characterize an amp when the THD is very low but the amps sound different.
I've never seen this type of test conducted before but it looks like it has merit, much merit.
I guess the hard part now begins. Investigate the data and quantify it.
Please keep up the work. I am learning much. Thanks
Mike
Nice, PMA,
Why do you make 1kHz conventional THD measures, not 10kHz? Is it fair for comparison with 10k FM?
How do you wire your diff-amp to your input now?
Have you tried IMD kind test with say 3 components, to compare with FM? If such test exposes same kind of HD, then benefit of FM is questionable. It would make sense to try to debunk it yourself, as it would give best insight into its merit.
Try Fc sinewave summed with -2*Fc sine at -6db from Fc.
Try FM with carrier 10.2kHz, deviation 10kHz and modulation 10Hz. Would it look familiar?
Why do you make 1kHz conventional THD measures, not 10kHz? Is it fair for comparison with 10k FM?
How do you wire your diff-amp to your input now?
Have you tried IMD kind test with say 3 components, to compare with FM? If such test exposes same kind of HD, then benefit of FM is questionable. It would make sense to try to debunk it yourself, as it would give best insight into its merit.
Try Fc sinewave summed with -2*Fc sine at -6db from Fc.
Try FM with carrier 10.2kHz, deviation 10kHz and modulation 10Hz. Would it look familiar?
wimms said:
wimms,
I have tried convetional 10kHz, but unable to generate enough "pure" 10kHz. You are right, the 10 kHz results are similar to that of 10kHz/100Hz/100Hz.
The diff amp is used in 2 ways:
1) to measure card own distortion ("test signal") in a card-out card-in method. It does not increase visible distortion compared to wire connection. The diff amp is based on INA217 and should have THD about 0.0003% for G=1.
2) it is connected at the power amp output to avoid ground loop problem (you have seen low frequency mess before). There is a voltage divider at the diff amp output to adjust input level for the soundcard.
Both tested amps perform quite well in a conventional 1kHz THD test. I assume that the FM method should be able to discover not only HD, but IMD related problems, especially for the lower quality amps. That's why I have generated different FM testing signals with index of modulation from 0.1 to 40.
Hi, PMA,
I built MrEvil's distortion eliminator feedback cct. Someone said it is quite like Error Correction. I posted the result here http://www.diyaudio.com/forums/showthread.php?postid=518453#post518453
Why is the result like that? Does Error Correction gives similiar result?
I built MrEvil's distortion eliminator feedback cct. Someone said it is quite like Error Correction. I posted the result here http://www.diyaudio.com/forums/showthread.php?postid=518453#post518453
Why is the result like that? Does Error Correction gives similiar result?
I am following with deep attention, and i a lot of consistences.
Despite having a simpler idea about that subject, i am following, as some very good guys are discussing ideas with a lot of care.
I am learning some, and beeing shocked with other aspects, but have not courage to discuss with Curl and others.
Have to respect those man.
Carlos
Despite having a simpler idea about that subject, i am following, as some very good guys are discussing ideas with a lot of care.
I am learning some, and beeing shocked with other aspects, but have not courage to discuss with Curl and others.
Have to respect those man.
Carlos
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.