Regarding more "retrieval" with the CM circuit slowed, I have a suspicion about it nothing more. Its that there is some low level audio detail that is not reproduced exactly equal in magnitude and opposite in phase at the + and - dac outputs. There are component tolerances, etc., which means that audio information below 1% or .1% may be mistaken for CM noise. However, the brain may process the imperfectly rendered low level audio details as useful and desirable information if it is taken as SE only.
If so, maybe the remedy should be to focus on improving the accuracy of the dac conversion itself rather than to remove too much information later by treating it as CM noise.
EDIT: Actually I believe there is probably more evidence in support of the above conjecture than I am at liberty to disclose here. All I'm going to say about that.
If so, maybe the remedy should be to focus on improving the accuracy of the dac conversion itself rather than to remove too much information later by treating it as CM noise.
EDIT: Actually I believe there is probably more evidence in support of the above conjecture than I am at liberty to disclose here. All I'm going to say about that.
Last edited:
Mark, we had this discussion before.
Without CM in both differential signals, it does not matter whether the gain is exactly the same or even the phase of both signal is exactly the opposite as long as both have a stable relation between the two differential signals over the required frequency span.
For the receiver it is a linear process of an addition of two signals with the same frequency content, and adding those can only change level and phase but never create new content, for which you would need a non linear ingredient.
For uncorrelated noise in both channels, it will increase proportional to square of the summed differential signal increase, so S/N will be better as compared to using just a SE output.
However the attenuation of the CM content at the summing receiever will be strongly dependent on gain and phase between the preceeding differential channels
But the DM to CM ratio from a summed differential output will always be lower than from a SE output having no CM suppression at all.
So at the SE output you will have a worse S/N and added to that the full CM content.
When not using exactly the same signal paths, it will be difficult to tell If and Why a SE output will produce superior sound.
The only correct way to test this IMO is to short-circuit the output of one of the two Firdacs and keep everything else absolutely unmodified, apart from increasing the preamps gain by 6dB.
Hans
P.s. and of course disable the CM servo for audio frequencies.
Without CM in both differential signals, it does not matter whether the gain is exactly the same or even the phase of both signal is exactly the opposite as long as both have a stable relation between the two differential signals over the required frequency span.
For the receiver it is a linear process of an addition of two signals with the same frequency content, and adding those can only change level and phase but never create new content, for which you would need a non linear ingredient.
For uncorrelated noise in both channels, it will increase proportional to square of the summed differential signal increase, so S/N will be better as compared to using just a SE output.
However the attenuation of the CM content at the summing receiever will be strongly dependent on gain and phase between the preceeding differential channels
But the DM to CM ratio from a summed differential output will always be lower than from a SE output having no CM suppression at all.
So at the SE output you will have a worse S/N and added to that the full CM content.
When not using exactly the same signal paths, it will be difficult to tell If and Why a SE output will produce superior sound.
The only correct way to test this IMO is to short-circuit the output of one of the two Firdacs and keep everything else absolutely unmodified, apart from increasing the preamps gain by 6dB.
Hans
P.s. and of course disable the CM servo for audio frequencies.
Last edited:
Yes, you are right, DC offset should still be taken care by the CM servo, just for audio frequencies will it have to be disabled like Mark and Bohrok just did, but that means that the two Firdacs still must be active.
Instead of shortening the Firdac there must be found a better point to shorten, but the idea is that no signal is available at one differential output and that no intereference takes place between the two differential lines.
But sockets, interconnects and receivers should all stay the same to get the condition that only one variable has been changed when comparing.
Hans
Instead of shortening the Firdac there must be found a better point to shorten, but the idea is that no signal is available at one differential output and that no intereference takes place between the two differential lines.
But sockets, interconnects and receivers should all stay the same to get the condition that only one variable has been changed when comparing.
Hans
Last edited:
Think I already did that as one of my experiments with Marcel's dac. I used his output stage, but only took the positive phase into my line amp. At that point the CM servo was still working at full speed, so output I was using had already been differentially summed up to well above the audio band. The next experiment was to leave everything connected the same, but only to slow down the CM servo in way that was easily reversable so that the experiment could be repeated as often as desired. To me, the slowed servo version has superior "retrieval" as rendered through the remainder of my system. Of course, this was only after fixing/improving a number of other problems with power supplies, USB bus and other sources of conducted and or radiated EMI/RFI noise, clocking (better than a CPLD used for clock buffering), etc. IME once a lot of junk has has been cleaned up, then remaining problems/issues become less masked and or otherwise more easily audible.
Also, if you would like measurements of that at this point you would have to speak to Acko. The dac, along with the pluggable servo slow-down components are now in his possession.
However, I don't doubt that CM noise was made worse. I only doubt that laser-focus on minimizing CM noise is not the only DM/CM tradeoff strategy that matters for a dac intended for human listening (and I believe the experimental result suggests such a tradeoff may exist).
Last edited:
As you were also able to hear low-level distortions with bit-perfect Foobar2000 there are probably other reasons for hearing more detail with crippled CM servo than the dac itself.
saboteur
a person who deliberately damages or destroys things
https://www.google.com/search?q=sab...CAIQARgTkgcEMTIuNKAH95QB&sclient=gws-wiz-serp
It's actually a leftover from my failed attempts to make a common-mode loop with second-order compensation in my earlier solid-state DAC (the logic gate DAC). It can probably be removed when the compensation capacitor is increased accordingly, although I have to think a bit more carefully about whether anything can go wrong during start-up.
Taking it literally, I think this applies when you intentionally make a common-mode loop inoperative for an experiment: you destroy the common-mode loop and you do it deliberately. Maybe the connotation is incorrect, though. If I have insulted anyone by my literal interpretation of sabotage, that is completely unintended.
I made an overview of the distortion measurements, see the attachment. I ignored the simulations, as I don't trust them.
As far as I'm concerned, all measured values are perfectly adequate, but anyway, they make me wonder whether the distortion isn't dominated by the output stages of the op-amps of the second filter stage. The only two cases where the 10 kHz distortion dropped substantially were a case where the op-amps of the second stage were replaced with an OPA1632 (and the feedback network of the second stage was different) and a case where the common-mode biasing was disturbed such that the second stage was forced to work in class A. The frequency-dependence appears to be just a bit stronger than first order when I look at the various products on bohrok2610's intermodulation plot, which is also more consistent with op-amp output stage than with op-amp input stage distortion.
Changes to the first stage that improve its loop gain or slew rate don't do much, and one of the cases with low distortion much degrades its common-mode suppression.
As far as I'm concerned, all measured values are perfectly adequate, but anyway, they make me wonder whether the distortion isn't dominated by the output stages of the op-amps of the second filter stage. The only two cases where the 10 kHz distortion dropped substantially were a case where the op-amps of the second stage were replaced with an OPA1632 (and the feedback network of the second stage was different) and a case where the common-mode biasing was disturbed such that the second stage was forced to work in class A. The frequency-dependence appears to be just a bit stronger than first order when I look at the various products on bohrok2610's intermodulation plot, which is also more consistent with op-amp output stage than with op-amp input stage distortion.
Changes to the first stage that improve its loop gain or slew rate don't do much, and one of the cases with low distortion much degrades its common-mode suppression.
Last edited:
Disable and destroy are not exactly the same. To Disable something may be a temporary setting selected by a switch or some jumpers. To destroy something suggests a more irreversible condition, say, perhaps involving kinetic munitions.
Similarly, to sabotage something would seem to imply some ulterior motive, say, perhaps in order to weaken a perceived enemy, or maybe just to cause chaos.
https://royalsocietypublishing.org/doi/10.1098/rstb.2020.0147
Doesn't seem like we have anything quite that intense going on here, despite the little bickering that sometimes occurs.
Just sayin'
Last edited:
As a third party reader, I thought it was appropriate diction and entirely inoffensive, colloquial use of the term. We're not talking war and peace here...
Actually, bohrok2610 has in possession his own version that can be modified accordingly and check if these ultrasonic distortions can be heard or affect elsewhere in the audible range. So far only you have reported as such. Interesting to know listening experiences from others as well
That's what I thought, but I'm neither a native speaker of English nor neurotypical, so I could very well be wrong.
Off topic: at university I once wrote an article about translinear circuits that upset the Australian lady who corrected the articles. The title "Square rooting circuit based on a novel, back-gate-using multiplier" gave her the impression it was about anal sexual intercourse, as "rooting" means something very different from taking the square root in Australian slang.
