Hi Mark,
I will try more tonight. Have to go and get my Dac, standing at a friend. Before i just played around with my old PCB, not doing anything if it got more broken.
I will prepare PCB as we spoke about before. I will try to cut the right traces
I keep you updated if i’m ready to move forward.
BR //Daniel
I will try more tonight. Have to go and get my Dac, standing at a friend. Before i just played around with my old PCB, not doing anything if it got more broken.
I will prepare PCB as we spoke about before. I will try to cut the right traces
I keep you updated if i’m ready to move forward.
BR //Daniel
Last edited:
Since this thread has become one where from time to time I share thoughts on dac design and related issues, I would like to say that I have come to have some concern about the way I see audio design proceeding with some members.
What I mean is that there seems to be a mentality that design can be synthesized from theory and or simulations, a prototype built, and common measurements of the AP type taken to guide final tweaking of a design. One such measurements appear satisfactory, the job is done.
In another reality, what I have been doing recently is a lot of experimenting and listening to the learned wisdom of a few heretical high end audio designer friends. Sometimes I argue with them when they go against what I have learned elsewhere, but when and if I finally try what they recommend it almost always turns out they are right. They know what sounds good, and it isn't more distortion and noise.
However, sometimes more distortion and or noise is a byproduct of optimizing for things that we do no routinely measure. In that case some compromise of common distortion and noise measurements may suffer a bit.
Then it is a case of, "this is the best I know how to do." "If I optimize more for better commonly used measurement results, then I will deoptimize for what I find matters more."
Bruno Putzys seems to take a somewhat different approach which is that if an audible problem exists, it will be ignored until such time as he understands the math and the measurements to correct it. He will invent his own measurements and math if need be, but they may remain proprietary.
Whether its the guys who know electronics but optimize by experiments and careful listening or Bruno who will stick to conventional engineering methods all the way, neither type of designer believes that optimizing to AP measurements is all there is to it.
What I mean is that there seems to be a mentality that design can be synthesized from theory and or simulations, a prototype built, and common measurements of the AP type taken to guide final tweaking of a design. One such measurements appear satisfactory, the job is done.
In another reality, what I have been doing recently is a lot of experimenting and listening to the learned wisdom of a few heretical high end audio designer friends. Sometimes I argue with them when they go against what I have learned elsewhere, but when and if I finally try what they recommend it almost always turns out they are right. They know what sounds good, and it isn't more distortion and noise.
However, sometimes more distortion and or noise is a byproduct of optimizing for things that we do no routinely measure. In that case some compromise of common distortion and noise measurements may suffer a bit.
Then it is a case of, "this is the best I know how to do." "If I optimize more for better commonly used measurement results, then I will deoptimize for what I find matters more."
Bruno Putzys seems to take a somewhat different approach which is that if an audible problem exists, it will be ignored until such time as he understands the math and the measurements to correct it. He will invent his own measurements and math if need be, but they may remain proprietary.
Whether its the guys who know electronics but optimize by experiments and careful listening or Bruno who will stick to conventional engineering methods all the way, neither type of designer believes that optimizing to AP measurements is all there is to it.
Last edited:
It sounds reasonable, but let me rephrase
Do you mean there is something could be heard but couldn’t be measured at all?
Is there is just a measurement we don’t know or do ignore?
If so we should just follow the “magic” rules to get desired (audible) results and then confirm them with (potentially poor) measurements?
Do you mean there is something could be heard but couldn’t be measured at all?
Is there is just a measurement we don’t know or do ignore?
If so we should just follow the “magic” rules to get desired (audible) results and then confirm them with (potentially poor) measurements?
Everything is physical, so in principle everything should be measurable. Some effects may take quite a bit of work to track down, measure, and figure out how to mitigate.
Some things are complicated and or expensive to measure such as clock phase noise time domain behavior of high performance audio clocks. Maybe phase noise crest factor? Doesn't mean they don't matter, doesn't mean they do. Very few people are in a position to find out using measurements.
Some things are complicated and or expensive to measure such as clock phase noise time domain behavior of high performance audio clocks. Maybe phase noise crest factor? Doesn't mean they don't matter, doesn't mean they do. Very few people are in a position to find out using measurements.
Last edited:
Serge,
I would like to reply separately to your question about 'magic' rules and listening. For me, there is no simple answer. Electrical engineering is a profession, there is a lot to learn. Another sort of profession might be being a mastering engineer like, say, for example, Bob Ludwig. The people I know who are high end audio designers have some mix of both types of training and skills. They know electronics, they know what circuits sound tend to sound good for audio, they know how to listen, etc. Its taken them decades to gain experience to be good at what they do.
Some people seem to have the idea that knowledge of feedback loop stability, frequency response, noise, output impedance verses frequency, simulation, etc., form a complete set of knowledge for being a good audio designer. IMO its good enough for run of the mill consumer stuff. It can also be good for competing in terms of who can produce the most impressive measurements. IMO, its not all there is to it though.
I would like to reply separately to your question about 'magic' rules and listening. For me, there is no simple answer. Electrical engineering is a profession, there is a lot to learn. Another sort of profession might be being a mastering engineer like, say, for example, Bob Ludwig. The people I know who are high end audio designers have some mix of both types of training and skills. They know electronics, they know what circuits sound tend to sound good for audio, they know how to listen, etc. Its taken them decades to gain experience to be good at what they do.
Some people seem to have the idea that knowledge of feedback loop stability, frequency response, noise, output impedance verses frequency, simulation, etc., form a complete set of knowledge for being a good audio designer. IMO its good enough for run of the mill consumer stuff. It can also be good for competing in terms of who can produce the most impressive measurements. IMO, its not all there is to it though.
Last edited:
Markw4,
>dac current output is very low impedance.
ES9038Q2M – 700 ohm (as declared)
ES9038Pro - 200 ohm
So not too low I believe.
I see the recommended I/V with 680 ohm negative feedback (for the opamp) drops the signal DAC output level by two.
Following my experiments with ES9038Q2M, the I/V conversion adds absolutely nothing to THD.
It may be of course only a chip I have. Anyway the DAC native distortions are too high on their own and I’m staring to realize why good quality ES9038Q2M based devices have 2 of them. There is just nothing to balance (in spite of whether I/V or not) on a single chip.
I’m even under strong impression there is no way to get THD noticeably above 90 db with a single chip.
>dac current output is very low impedance.
ES9038Q2M – 700 ohm (as declared)
ES9038Pro - 200 ohm
So not too low I believe.
I see the recommended I/V with 680 ohm negative feedback (for the opamp) drops the signal DAC output level by two.
Following my experiments with ES9038Q2M, the I/V conversion adds absolutely nothing to THD.
It may be of course only a chip I have. Anyway the DAC native distortions are too high on their own and I’m staring to realize why good quality ES9038Q2M based devices have 2 of them. There is just nothing to balance (in spite of whether I/V or not) on a single chip.
I’m even under strong impression there is no way to get THD noticeably above 90 db with a single chip.
Serge,
Two dac chips are not needed. That's not the problem at all. Distortion can be -120dB or better with only one chip.
First, there is RF coming out of the dac chip, that RF can create nonstationary distortion in an output stage that might not show up well or maybe not show up at all on a typical audio FFT. In addition, you need a good 3-opamp output stage more or less like those recommended by ESS. An example of that and an AVCC supply is attached to post #3003 of this thread.
2nd, the ASRC DPLL Bandwidth setting in the dac chip has a large effect on perceived sound quality. IME, it does not show up well an a typical audio FFT. Trying turning down the bandwidth as ESS recommends can see if you don't hear a change in sound quality.
3rd, the AVCC voltage regulation is critical to sound quality. It has zero PSRR, and it is directly multiplied by the dac audio output. That leads to audible distortion that may or may not show up well on an audio FFT.
4th, other things matter too. Clock jitter matters, clock voltage regulation too. Voltage regulators for the various dac power inputs, DVCC, VCCA, and any others that don't have internal regulation. This is true for high performance dacs in general. ESS recommends that the MCU not be on the same voltage regulator as the dac chip.
5th, ESS and AK4499 sound better if sent DSD256 or DSD512. They just do. Try it for free with a trial version of HQ Player, or the DSD conversion software of your choice. Basically, this is the best sounding way to play CDs if using the dac chips we are talking about.
In other words, you have lots and lots of work to do if want to get the best sound out of the dac chip. Using two dac chips will only help slightly after everything else is fixed. Have you carefully studied the document available on the ESS downloads page?: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
Two dac chips are not needed. That's not the problem at all. Distortion can be -120dB or better with only one chip.
First, there is RF coming out of the dac chip, that RF can create nonstationary distortion in an output stage that might not show up well or maybe not show up at all on a typical audio FFT. In addition, you need a good 3-opamp output stage more or less like those recommended by ESS. An example of that and an AVCC supply is attached to post #3003 of this thread.
2nd, the ASRC DPLL Bandwidth setting in the dac chip has a large effect on perceived sound quality. IME, it does not show up well an a typical audio FFT. Trying turning down the bandwidth as ESS recommends can see if you don't hear a change in sound quality.
3rd, the AVCC voltage regulation is critical to sound quality. It has zero PSRR, and it is directly multiplied by the dac audio output. That leads to audible distortion that may or may not show up well on an audio FFT.
4th, other things matter too. Clock jitter matters, clock voltage regulation too. Voltage regulators for the various dac power inputs, DVCC, VCCA, and any others that don't have internal regulation. This is true for high performance dacs in general. ESS recommends that the MCU not be on the same voltage regulator as the dac chip.
5th, ESS and AK4499 sound better if sent DSD256 or DSD512. They just do. Try it for free with a trial version of HQ Player, or the DSD conversion software of your choice. Basically, this is the best sounding way to play CDs if using the dac chips we are talking about.
In other words, you have lots and lots of work to do if want to get the best sound out of the dac chip. Using two dac chips will only help slightly after everything else is fixed. Have you carefully studied the document available on the ESS downloads page?: http://www.esstech.com/files/4514/4095/4306/Application_Note_Component_Selection_and_PCB_Layout.pdf
Last edited:
>Distortion can be -120dB or better with only one chip.
With a single 9038q2m? Hardly, I won’t believe till a I see a single chip based device with -120db. ;-)
On the other hand if -120db is reachable with a single chip, why the manufactures use 2 of them (even in cheap ChiFi)?
Regarding all the rest possibilities, theoretically you are obviously and absolutely right.
But my the current goal is completely different, I’m not making (designing from scratch) the best DAC. I’m rather trying to “repair” an existing DAC to make it just decent, just not as ugly as it is now. For now I even can’t reach -100db but have evidences that harmonics come from the chip itself. I have nearly exhausted ideas “what to try next” beyond giving up the board(s) and order new one(s).
I can cut a wire, solder 1 or 2 passive components, replace a cap, an opamp, etc (as I already did), but I won’t, for instance, separate 3.3v for DAC from 3.3v for MCU. It would be closer to a new board, but I’ve been trying to repair this one.
So I’m keen to have a simple idea for what to do.
Obviously RF on the output can distort, bad clock/power can do the same, but what to try to verify the practice, a right direction at least with a simple experiment. Isn’t it impossible?
I’m not in the position to make significant changes at all. I just need -100Ddb THD and nothing more.
When (if) I start to design from scratch I will definitely follow ALL the recommendations, but now my purpose is different. Hope you understand
With a single 9038q2m? Hardly, I won’t believe till a I see a single chip based device with -120db. ;-)
On the other hand if -120db is reachable with a single chip, why the manufactures use 2 of them (even in cheap ChiFi)?
Regarding all the rest possibilities, theoretically you are obviously and absolutely right.
But my the current goal is completely different, I’m not making (designing from scratch) the best DAC. I’m rather trying to “repair” an existing DAC to make it just decent, just not as ugly as it is now. For now I even can’t reach -100db but have evidences that harmonics come from the chip itself. I have nearly exhausted ideas “what to try next” beyond giving up the board(s) and order new one(s).
I can cut a wire, solder 1 or 2 passive components, replace a cap, an opamp, etc (as I already did), but I won’t, for instance, separate 3.3v for DAC from 3.3v for MCU. It would be closer to a new board, but I’ve been trying to repair this one.
So I’m keen to have a simple idea for what to do.
Obviously RF on the output can distort, bad clock/power can do the same, but what to try to verify the practice, a right direction at least with a simple experiment. Isn’t it impossible?
I’m not in the position to make significant changes at all. I just need -100Ddb THD and nothing more.
When (if) I start to design from scratch I will definitely follow ALL the recommendations, but now my purpose is different. Hope you understand