Hi Mark,
>How does it sound now
I was not able to hear any defects, didn't compare side by side with a “reference” DAC yet.
(but an AudioGD 9038Pro I have shows higher harmonics )
So FFT at 1075 Hz (it’s my homemade notch filter frequency) before any changes made - before.png
After soldering 4 wires (2 wires per channel) and adding 2 resistors (1 per channel) to add amplification, I have - after.png
The notch FR (if you would like to check true harmonics levels) - notch.png
I would say it’s a good result of the repairs.
It’s time to play with the THD compensation registers.
>How does it sound now
I was not able to hear any defects, didn't compare side by side with a “reference” DAC yet.
(but an AudioGD 9038Pro I have shows higher harmonics )
So FFT at 1075 Hz (it’s my homemade notch filter frequency) before any changes made - before.png
After soldering 4 wires (2 wires per channel) and adding 2 resistors (1 per channel) to add amplification, I have - after.png
The notch FR (if you would like to check true harmonics levels) - notch.png
I would say it’s a good result of the repairs.
It’s time to play with the THD compensation registers.
Attachments
>Regarding the harmonic registers, perhaps settings also depend on sample rate, PCM, or DSD.
Maybe, I will try to check.
The curious current observation is THD on one channel depends on the second channel level (+2.5 db on C2). Probably it’s a power supply “leak”. The current schematics is primitive.
>Still curious about how it sounds.
You could do the same and check the sound. Unfortunately I can’t imagine a way to check the sound without having the device.
>IME one can't extrapolate from low level HD measurements if something will sound good.
Not at all for me, I measured a dozen of DACs and compared their sounds.
So for me there is a listening limit, where making THD lower has no audibly difference.
Blind comparison with ceteris paribus is a must though.
In your previous message
>opamps were heating due to RF mixed.
I have another hypothesis. DAC current is always positive as produced by DAC output in 0-AVCC range, So the inverting I/V is in range of -AVCC*k - 0 (where k is the amplification). It's not bipolar anyway. So if the I/V is loaded by a low impedance of the next stage, the I/V opapmp has to provide the current accordingly, so I may be not as cold as expected.
Maybe, I will try to check.
The curious current observation is THD on one channel depends on the second channel level (+2.5 db on C2). Probably it’s a power supply “leak”. The current schematics is primitive.
>Still curious about how it sounds.
You could do the same and check the sound. Unfortunately I can’t imagine a way to check the sound without having the device.
>IME one can't extrapolate from low level HD measurements if something will sound good.
Not at all for me, I measured a dozen of DACs and compared their sounds.
So for me there is a listening limit, where making THD lower has no audibly difference.
Blind comparison with ceteris paribus is a must though.
In your previous message
>opamps were heating due to RF mixed.
I have another hypothesis. DAC current is always positive as produced by DAC output in 0-AVCC range, So the inverting I/V is in range of -AVCC*k - 0 (where k is the amplification). It's not bipolar anyway. So if the I/V is loaded by a low impedance of the next stage, the I/V opapmp has to provide the current accordingly, so I may be not as cold as expected.
Regarding sound of opamps: One time I sorted unity gain noninverting opamp buffers (only difference was the particular opamp used) in order of distortion by ear and did it blind. It was not easy. It was so hard that I quit before sorting the last one. At that time my reproduction system for listening was not nearly as good as the one I have today. Lots of things become audible if the rest of the reproduction system is good enough. IME sometimes training of what to listen for is necessary, along with sufficient practice. Others may feel differently, of course.
Regarding no perceived benefit from reducing measured stationary harmonic distortion beyond some threshold level that might be judged as good enough, IME usually there are masking factors in the device and in the system that are the underlying cause of the perceived threshold. Others may disagree, of course.
Sometimes when I say things like the above I catch a lot of verbal flak for it. That's okay. I figure you are smart enough to think about it, listen carefully, and form your own conclusions.
Regarding no perceived benefit from reducing measured stationary harmonic distortion beyond some threshold level that might be judged as good enough, IME usually there are masking factors in the device and in the system that are the underlying cause of the perceived threshold. Others may disagree, of course.
Sometimes when I say things like the above I catch a lot of verbal flak for it. That's okay. I figure you are smart enough to think about it, listen carefully, and form your own conclusions.
Last edited:
>Regarding sound of opamps:
Could the difference be measured.
It’s hard to believe human ears is a more precise “instrument”
> Lots of things become audible if the rest of the reproduction system is good enough.
Obviously! If other components have more distortions a distortion of one stage in the chain can be “covered by the rest.
>IME sometimes training of what to listen for is necessary,
Oh yes. We should understand what/where to listen to be focused on. Otherwise, starting from a simple HiFi level, everything will sound equally.
I like to compare a single piano playing “on pedal”, a silent violin on a loud “bass” background.
violin ensemble soundstage depth etc…. The recording itself should be good enough at least.
>IME usually there are masking factors
Sure there are, DAC is just one component in a complex sound reproducing chain.
However the limit exists, its value may vary from system to system, from listener to listener though.
Anyway human ear is not almighty. It seems to be I’m already able to determine the limit to myself.
>I catch a lot of verbal flak for it.
I agree.
I guess there is only point we might have inconsistency. I’m not trying to make “the best of the best”, it’s an endless race distracting from listening and enjoying of music. Needless to say quality of modern recording is getting worse and worse, compressed (to be “nice” in noisy environment), intentionally distorted etc. So having a perfect playback system becomes a way to waste the “investments” sometimes.
Could the difference be measured.
It’s hard to believe human ears is a more precise “instrument”
> Lots of things become audible if the rest of the reproduction system is good enough.
Obviously! If other components have more distortions a distortion of one stage in the chain can be “covered by the rest.
>IME sometimes training of what to listen for is necessary,
Oh yes. We should understand what/where to listen to be focused on. Otherwise, starting from a simple HiFi level, everything will sound equally.
I like to compare a single piano playing “on pedal”, a silent violin on a loud “bass” background.
violin ensemble soundstage depth etc…. The recording itself should be good enough at least.
>IME usually there are masking factors
Sure there are, DAC is just one component in a complex sound reproducing chain.
However the limit exists, its value may vary from system to system, from listener to listener though.
Anyway human ear is not almighty. It seems to be I’m already able to determine the limit to myself.
>I catch a lot of verbal flak for it.
I agree.
I guess there is only point we might have inconsistency. I’m not trying to make “the best of the best”, it’s an endless race distracting from listening and enjoying of music. Needless to say quality of modern recording is getting worse and worse, compressed (to be “nice” in noisy environment), intentionally distorted etc. So having a perfect playback system becomes a way to waste the “investments” sometimes.
Distortion was measured by the guy running the experiment. It was normal expected low level audio opamp distortion.
People have tried it, done it, etc. It works okay when measured performance is not too near SOA, and when the problems are bad enough to be heard despite some masking by other reproduction system imperfections.
When it comes to devices that measure at or close to SOA levels, then I personally don't think we typically measure everything that seems to be audible. Sometimes audible differences just don't show up very well in typical measurements, although they may show up a little. For example, a very small FFT spur may result in substantial sound quality improvement if the underlying cause is found is fixed. In other cases a bigger FFT spur may not produce much audible change if it is fixed. In some cases, particularly where RF is involved (always in dacs, maybe in other equipment when wires act like antennas), whatever audible distortion or noise that may be produced may be changing enough from moment to moment that its not stable enough to make a nice sharp tall spur on a typical audio FFT. The justification for using FFTs as we do is based the assumption that a source of nonlinear distortion is 'stationary.' That is, we assume it is unchanging as a function of time. We sometimes don't allow there could be any way a distortion could be changing and still audible. I personally don't like to rule out such possibilities without sufficient experimental justification in a case like we have here where we find measurements and perceived sound quality don't always correlate well.
When it comes to devices that measure at or close to SOA levels, then I personally don't think we typically measure everything that seems to be audible. Sometimes audible differences just don't show up very well in typical measurements, although they may show up a little. For example, a very small FFT spur may result in substantial sound quality improvement if the underlying cause is found is fixed. In other cases a bigger FFT spur may not produce much audible change if it is fixed. In some cases, particularly where RF is involved (always in dacs, maybe in other equipment when wires act like antennas), whatever audible distortion or noise that may be produced may be changing enough from moment to moment that its not stable enough to make a nice sharp tall spur on a typical audio FFT. The justification for using FFTs as we do is based the assumption that a source of nonlinear distortion is 'stationary.' That is, we assume it is unchanging as a function of time. We sometimes don't allow there could be any way a distortion could be changing and still audible. I personally don't like to rule out such possibilities without sufficient experimental justification in a case like we have here where we find measurements and perceived sound quality don't always correlate well.
Last edited:
Of course any audible difference can be measured. With today's instruments, even the differences that are too small for any human's hearing ability can be measured. Anyone who says otherwise is either uninformed on the subject or trying to sell you something and for that reason, you don't want to listen to them.
So the question is what/how to measure to have representative, correlated to audible sound quality results.
FFT of 1kHz 0db is definitely not enough.
For me it allows to surely separate, good from middle and from poor sound.
In site of separating in the “good” category by FFT is possible as well, it’s not too strict and sometimes can’t be formalized.
Of course I don’t mean an aggregated value of THD only, harmonics distribution over the spectrum is very meaningful
IMD generally correlates to FFT spectrum.
Varying testing signal level typically gives nothing, following my experience at least.
So what/how to measure?
I would be happy if you share your thoughts.
FFT of 1kHz 0db is definitely not enough.
For me it allows to surely separate, good from middle and from poor sound.
In site of separating in the “good” category by FFT is possible as well, it’s not too strict and sometimes can’t be formalized.
Of course I don’t mean an aggregated value of THD only, harmonics distribution over the spectrum is very meaningful
IMD generally correlates to FFT spectrum.
Varying testing signal level typically gives nothing, following my experience at least.
So what/how to measure?
I would be happy if you share your thoughts.
We are not salesmen here I believe, Salesmen (imho generally) sale a kind of "happiness" ;-)
Maybe some iterative process of listening, hypothesizing, measuring, and defining.
Sometimes studying something like nonlinear dynamic behavior of circuit topologies which are often claimed to sound good can provide clues. That may correlate better with perceived sound quality as verses the more easily measured static behavior.
Sometimes studying something like nonlinear dynamic behavior of circuit topologies which are often claimed to sound good can provide clues. That may correlate better with perceived sound quality as verses the more easily measured static behavior.
When it comes to sound replaying electronics such as DAC, it's the level of output fidelity to the input that judges the performance quality. The term "hi-fi" is the goal and the DAC technology has already achieved it, at least 20 years ago.
Compare the measurements you mentioned between different DACs and see what you find.
Human choice is emotional.
Is it basically emotional or just partially - is a personal psychology
I’ve been just trying to be rather practical/rational than emotional
Last edited:
