Wahab,
There might be some reasons why less peaks show up for the opamp, since all this is way beyond audio frequencies and your preamp might filter this out.
Would be interesting to see comparison with fundamentals at let's say 1K and 100Hz. For good measure, why not throw in some 10Hz as well?
vac
There might be some reasons why less peaks show up for the opamp, since all this is way beyond audio frequencies and your preamp might filter this out.
Would be interesting to see comparison with fundamentals at let's say 1K and 100Hz. For good measure, why not throw in some 10Hz as well?
vac
Last edited:
At lower frequency , the LME49860 has THD at 0.1ppm/1khz ,
wich is in the datasheet range , but still higher than the one
of the discrete design , although not as dramatically as at 20khz.
Attachments
There must be some analog dithering going on in the discrete ...
Who knows ?...
More simply , the discrete has better intrinsical linearity
despite its 143db OLG that would imply a low THD thanks only to GNFB.
Hi Wahab,
Thank you for putting up another measurement. I really like to look at stuff like this and try to figure out what is going on. There is a question though I have with regard to the two measurements in order to get a better understanding.
It has to do with the noise floor. The two measurements of the opamp seem to connect, give or take. However, for the discrete, the noisefloor appears to jump > two orders of magnitude.
Another interesting observation that could be made from the latest measurement is that the opamp creates inverse harmonics. That is, at harmonic intervals it lowers the noise floor.
vac
Thank you for putting up another measurement. I really like to look at stuff like this and try to figure out what is going on. There is a question though I have with regard to the two measurements in order to get a better understanding.
It has to do with the noise floor. The two measurements of the opamp seem to connect, give or take. However, for the discrete, the noisefloor appears to jump > two orders of magnitude.
Another interesting observation that could be made from the latest measurement is that the opamp creates inverse harmonics. That is, at harmonic intervals it lowers the noise floor.
vac
The apparant difference in noise floor is just a convergence matter,
as i simulated using same number of cycles and same step, but the discrete has more
nodes to compute , so the precision is less than with the LME.
Last edited:
kinda pointless without schematics, load conditions
you could be loading the op amp too heavily - composite structures with higher current output amp/buffer in the loop can greatly reduce distortion with op amp gain stages
for DAC I/V specifically the current is defined, and the output Z of the op amp is important so "low Z" buffers in the loop have been recommended in DAC datasheets for nearly a long as monlithic op amps, DAC have existed ~40 years now
THD20 can also be less useful if comparing a "high gain" op amp to a "flat loop gain" circuit - the 20KHz harmonic distortion products aren't themselves audible - only the IMD difference products with multiones, complex signals that "fold down" into the audio frequency range are directly audible
the "high gain" circuit can reduce the audio frequency IMD differences below 20 KHz by the excess loop gain
Ahhh. But Wahab, you need an apples to apples comparison. Drop you discrete design rails to the same as the opamp - maybe then it will not be as good.
I did a lot of simulation work on a CFA design and got to <<1ppm at 20KHz, but I used +-25V rails. As soon as I dropped them to +-15, it did not performa nearly as well. The high rails help to mitigate Early effect distortion.
There must be some analog dithering going on in the discrete ...
. . please don't go there . . that's been thrashed to death on the other thread
It s written on the graphs that load is 1K.
Scematics are basicals non inverting configurations.
Agree that 20khz THD is more or less meaningfull, 10KHZ being the
most relevant , still , with only one resisual in the audio band.
IMD should be within THD numbers, but will check neverless..
Hi Bonsai
Only a version i use as amp front end has high voltage rails, +-28V.
The line preamps in the sims , including the LME, are supplied with +-20V.
The discrete design being a symmetrical differential , this should
tame down the legend that single ended differential topologies,
aka blameless, are better in matter of THD distribution..
Thank you for your contribution. It could be a lot of things, or even all the same time together.
I want to focus again on overshoot, settling, ringing or ringing artifacts. I mentioned already the wikipedia where audio is also touched:
Ringing artifacts - Wikipedia, the free encyclopedia
I ask myself according overshoot, settling, ringing etc.
What does the phase do when the response is overshooting, settling, ringing etc.?
Hi Bonsai
Only a version i use as amp front end has high voltage rails, +-28V.
The line preamps in the sims , including the LME, are supplied with +-20V.
The discrete design being a symmetrical differential , this should
tame down the legend that single ended differential topologies,
aka blameless, are better in matter of THD distribution..
Yes, we agree on that point. The CFA I mentioned was also symmetrical (as the alsmost always are).
Symmetrical with mirror loaded front end + EF3 and nothing special in terms of the rest of the design easily gets 1ppm on sims.
However, I won't start locking horns with OStripper, BC or DS on which is best - I think its just a case of what you are comfortable with. For me its symmetrical designs.
Yes, 180deg phase shifted may not given to negative input with gain more than 1x (0dB+), or it will oscillate and almost 1x (less than 0dB) is ringing less than that is overshoot, and so. Also happened for less than 180deg PS with higher gain limit (while gain limit for 180deg is 0dB), and 90deg is safe phase lag.
Except there is extra bypass (most used is RC series or C) applied to get 3rd pole path feedback instead from 2nd pole in high frequency, then for high frequency the 2nd pole is shifted to 3rd pole that is higher (up to 40MHz for common fast amplifiers). Then solid stability may reached with faster response than not bypassing.
Except there is extra bypass (most used is RC series or C) applied to get 3rd pole path feedback instead from 2nd pole in high frequency, then for high frequency the 2nd pole is shifted to 3rd pole that is higher (up to 40MHz for common fast amplifiers). Then solid stability may reached with faster response than not bypassing.
Not directly related, but the "audio band" term always bothers me, in contexts like this. Can people hear the difference between a 22 kHz sine wave and an ideal square wave with a 22 kHz repetition frequency? According to everything most people seem to believe, the answer must be "no". And I guess a 44 kHz sampling rate to digitize will then reproduce a 22 kHz square wave as a 22 kHz sine wave. What about somewhat lower repetition frequency square waves? And even lower? Do we really not care about anything outside of "the audio band"? If not, then why is 44 kHz sampling frequency not nearly good enough, by all accounts?
Around 1977, one of my EE professors at Purdue, when he discussed the Nyquist frequency stuff with us for the very first time, said 2x the highest component frequency was required to theoretically be able to reconstruct the original sampled signal from the samples. But then he quickly added that in reality we had better be sampling AT LEAST at 10x. Did none of the CD or digital audio designers ever hear about anything like that, when they designed the first generation of digital audio stuff? And aren't all of our "audio band" assumptions similarly affected?
Tom
Good point, Tom. There is a lot of ancecodtal information about that some people can perceive bandlimiting beyond 20KHz.
I would think however that the 44.1KHz CD sampling frequency was chosen as much for reasons of space as anything else; 9.4Gb DVDs are relatively recent, certainly nothing like it around in 1982. In truth, not too many audiophiles seem really interested in 24 bit 192Khz sampling, or even 96Khz for that matter. Commercial interests perhaps?
Hugh
I would think however that the 44.1KHz CD sampling frequency was chosen as much for reasons of space as anything else; 9.4Gb DVDs are relatively recent, certainly nothing like it around in 1982. In truth, not too many audiophiles seem really interested in 24 bit 192Khz sampling, or even 96Khz for that matter. Commercial interests perhaps?
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.