Interestingly, while looking once again at some old Benchmark blogs or app notes I came across this which never fully made sense before:
"Benchmark ran a number of tests to determine the sweet spot for the oversampling ratio in the ES9018. We chose the oversampling ratio that delivered the lowest THD. PCM is delivered to the D/A elements in a format that closely resembles DSD-Wide."
Now I would suspect that they found the same thing I did, which is that Sabre dac chips sound better if fed DSD rather than PCM. If that is correct, the difference here is I am using the poor man's AK4137 to do it and looks like they are using a Spartan 6 chip for that task (among other things). If so, it would mean they are not stuck with the AK4137 filter choices.
"Benchmark ran a number of tests to determine the sweet spot for the oversampling ratio in the ES9018. We chose the oversampling ratio that delivered the lowest THD. PCM is delivered to the D/A elements in a format that closely resembles DSD-Wide."
Now I would suspect that they found the same thing I did, which is that Sabre dac chips sound better if fed DSD rather than PCM. If that is correct, the difference here is I am using the poor man's AK4137 to do it and looks like they are using a Spartan 6 chip for that task (among other things). If so, it would mean they are not stuck with the AK4137 filter choices.
Because, "Benchmark ran a number of tests to determine the sweet spot..."
Also, because I tried it and it just does.
I think that about summarizes what we know about it at this point.
Well, maybe add that the PCM reconstruction filters in the dac chip mostly don't get used for DSD, and it may handle DSD differently in other ways too that we don't know about.
On the other hand that may have been only for the earlier Sabre chip. Elsewhere Benchmark says they upsample to 211kHz to allow for a wider filter transition band (one really needs to study what information is available, no matter what form it is in). It could be they do both. There is an SRC4392 in DAC-3 that may upsample before PCM signals go into the Spartan 6. The exact details of what they do are their trade secrets, although it looks like Crane Song has figured it out. I probably could figure it out too if I wanted to poke around in my DAC-3 more and see what is going on. I am, however, definitely retired and not in the dac business at all. So, somebody else can do it if they want. I only came around this part of the forum several months ago to try to help out some DIYer's with modding Q2M dacs.
Also, because I tried it and it just does.
I think that about summarizes what we know about it at this point.
Well, maybe add that the PCM reconstruction filters in the dac chip mostly don't get used for DSD, and it may handle DSD differently in other ways too that we don't know about.
On the other hand that may have been only for the earlier Sabre chip. Elsewhere Benchmark says they upsample to 211kHz to allow for a wider filter transition band (one really needs to study what information is available, no matter what form it is in). It could be they do both. There is an SRC4392 in DAC-3 that may upsample before PCM signals go into the Spartan 6. The exact details of what they do are their trade secrets, although it looks like Crane Song has figured it out. I probably could figure it out too if I wanted to poke around in my DAC-3 more and see what is going on. I am, however, definitely retired and not in the dac business at all. So, somebody else can do it if they want. I only came around this part of the forum several months ago to try to help out some DIYer's with modding Q2M dacs.
Last edited:
After reading some of those links I have to wonder, how audible is the common mode distortion produced by the dac? I had an amp-shootout at friend of a friends house a few months ago, the guy wanted to hear an amp that I designed. My amp was the clear winner and it beat all of his expensive amps but his amps were all balanced and mine was SE. I have to wonder if maybe it's better to sum the balanced channels into SE channels instead for another 6db noise reduction?
When my ES9038 Buffalo dac was working, I didn't filter the output of my I/V stage and and there was no aliasing visible on the scope. Then again my I/V stage is a completely unique concept, maybe there's a HF filter in there somewhere I inadvertently created.
I only used a single output phase of the DAC and I've never heard anything sound more clean and realistic.
I have a difficult time believing common mode distortion can be so bad if the sound was that good.
I only used a single output phase of the DAC and I've never heard anything sound more clean and realistic.
I have a difficult time believing common mode distortion can be so bad if the sound was that good.
Scopes can't see below 1% distortion, or thereabouts. But, hey, some people like the sound of a little distortion. Seriously.
Anyway, it is not possible to explain how something sounds to someone who hasn't heard it, or at least something very close to the same. I take a stab at trying to explain why in a post at the following link: http://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-234.html#post5519403
Regarding DAC-3, I think you just have to hear it. I do think it has a sound of it's own that sounds like the opamps and other parts in it, but it is still exceptionally good as DACs go. Most people probably wouldn't notice it at all, but a few would. Mikett is probably one who would, be my guess anyway. To me, it's a keeper, worth what it cost. That's how I feel about it, even though it was painfully expensive for audio gear, IMHO.
Anyway, it is not possible to explain how something sounds to someone who hasn't heard it, or at least something very close to the same. I take a stab at trying to explain why in a post at the following link: http://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-234.html#post5519403
Regarding DAC-3, I think you just have to hear it. I do think it has a sound of it's own that sounds like the opamps and other parts in it, but it is still exceptionally good as DACs go. Most people probably wouldn't notice it at all, but a few would. Mikett is probably one who would, be my guess anyway. To me, it's a keeper, worth what it cost. That's how I feel about it, even though it was painfully expensive for audio gear, IMHO.
Last edited:
Hello
I started a document for those who would like to embark on the adventure.
It's not a document for the "pro"
if it seems interesting to you I would complete it otherwise
I have started a thread ES9038Q2M production it's just for those who want to share there experience et realisation with details and pictures
Serge
Serge, Not sure if another thread is a good idea or not. This thread was at one time like the thread you are creating, a thread where people who don't know much about good mods for dacs become the advisers of people who know even less. The dacs don't end up sounding very good and everyone seems to have an opinion about what to try next, but it is rarely the right thing to fix the problems. In the final analysis, it would have been easier to do the right mods in the first place, and everybody would have really good sounding dacs too.
Hello Mark,
If I can also contribute with my modification, i think after any voltage/supply modification, the most important is I/V stage, i cut off simple opamp stage and made my own one HDAM with diamond buffer at the end, supplied by +-9V. Please see attached schema, from beginning i fight with some oscillation and the i had to reduce feedback capacitor to improve stability and also sound became. Overall sound with china board paired with kali reclocker is very good (cleaner than basic voltage opamp on china board). I will also try to use 2 separate power supply 3.3V for AVCC, dou you still reccomended increasing of AVCC capacitor from 100uf to somwhere around 220-330 ? Regards , Vita
If I can also contribute with my modification, i think after any voltage/supply modification, the most important is I/V stage, i cut off simple opamp stage and made my own one HDAM with diamond buffer at the end, supplied by +-9V. Please see attached schema, from beginning i fight with some oscillation and the i had to reduce feedback capacitor to improve stability and also sound became. Overall sound with china board paired with kali reclocker is very good (cleaner than basic voltage opamp on china board). I will also try to use 2 separate power supply 3.3V for AVCC, dou you still reccomended increasing of AVCC capacitor from 100uf to somwhere around 220-330 ? Regards , Vita
Attachments
Hello Blackfear, Wow, I am amazed that you took on something this complicated and got it to work, especially after having to fight oscillation problems. I really have to admire your work ethic and willingness to take on something like this. Don't know how much you know about oscillations and how to troubleshoot them, but it is possible we could have given you some help along the way if we knew what you were struggling with. So long as you figured it out, credit to you for doing so.
What I would say next I don't know if you will like hearing or not, but I feel like I should say something about it so you know, and so other people reading along can maybe learn about output stages at the same time.
It looks like you have made a discrete opamp and then configured it with some resistors and capacitors to work not exactly as a voltage output stage, and not exactly as a current output stage, but rather something in between. R19 sort of nearly matches the output R of the dac output, so that makes the non-inverting input impedance right in the middle between voltage and current modes. R14 and the opamap gain makes the inverting input see a similar load as the non-inverting input. Nowhere is the +1.65v reference voltage from AVCC used for anything, just another observation.
While I am still in awe that you would do all this, I am sorry to have to tell you that I think you could get the distortion down much lower and better SQ if you would not use this output stage, and instead change it to a normal ESS recommended type of 3-integrated-opamps-per-channel, I/V-and-differential-summing-with-filtering type of output stage. While I feel bad having to say that, I don't think it would be an option for me not to say it in case you don't already know about it, and in case you would like better sound quality. On the other hand, if you are happy with it as it is after all the work you have put into it, well, that's cool too, I guess.
Regarding AVCC, I am not sure what you are asking about with regard to increasing AVCC capacitors. My AVCC capacitors at the dac AVCC input pins are 47uf, but there are LME49720 opamps under the PCB that regulate voltage into the capacitors, so 47uf is a good value in that case. If I could see a schematic of what you are doing for AVCC, I could perhaps give better advice about capacitors for that case. If by chance you are using the original AVCC circuit that comes with the dac board, and you want to leave it that way except for capacitors, then the more capacitance the better. Victor said he used 3 x 1800uf per channel soldered to the AVCC cap pins on the bottom of the board, as one example of what you could do.
What I would say next I don't know if you will like hearing or not, but I feel like I should say something about it so you know, and so other people reading along can maybe learn about output stages at the same time.
It looks like you have made a discrete opamp and then configured it with some resistors and capacitors to work not exactly as a voltage output stage, and not exactly as a current output stage, but rather something in between. R19 sort of nearly matches the output R of the dac output, so that makes the non-inverting input impedance right in the middle between voltage and current modes. R14 and the opamap gain makes the inverting input see a similar load as the non-inverting input. Nowhere is the +1.65v reference voltage from AVCC used for anything, just another observation.
While I am still in awe that you would do all this, I am sorry to have to tell you that I think you could get the distortion down much lower and better SQ if you would not use this output stage, and instead change it to a normal ESS recommended type of 3-integrated-opamps-per-channel, I/V-and-differential-summing-with-filtering type of output stage. While I feel bad having to say that, I don't think it would be an option for me not to say it in case you don't already know about it, and in case you would like better sound quality. On the other hand, if you are happy with it as it is after all the work you have put into it, well, that's cool too, I guess.
Regarding AVCC, I am not sure what you are asking about with regard to increasing AVCC capacitors. My AVCC capacitors at the dac AVCC input pins are 47uf, but there are LME49720 opamps under the PCB that regulate voltage into the capacitors, so 47uf is a good value in that case. If I could see a schematic of what you are doing for AVCC, I could perhaps give better advice about capacitors for that case. If by chance you are using the original AVCC circuit that comes with the dac board, and you want to leave it that way except for capacitors, then the more capacitance the better. Victor said he used 3 x 1800uf per channel soldered to the AVCC cap pins on the bottom of the board, as one example of what you could do.
Last edited:
Mark, I follow this thread avidly and enjoy the discussions. As an beginner/intermediate amateur diyer much of it is over my head but I muddle through.
However it has not encouraged me to try modding my spare board as the gap between the hints and images lack direction.
Maybe the EE find it useful though I find the thread is more personal blog than guidance or recipe.
Not all of us are aiming for perfection and modding the chinese dac to compete with a Katana or Benchmark... some, like me ,simply want to try the upgrades we can competently achieve and afford in both time and $,,,, (oh to be retired like yourself and have the time and energy...
Serge's thread has ,overnight, given the recipe that bridges the gap between your technical descriptions and my humble skills and I feel that thread will be of great benefit in conjunction with this one and not as competition.
@DRONE7, Agreed without question that I have not tried to give any layout recipes. In some cases, not even schematic recipes. The historical reason would probably be because my dac project has been a work in progress, one that I am still not sure is completely finished. Think I pointed out before that we didn't even know basic things like what what architecture would end up working best. We still seem to be finding out more about how sensitive some things are to power quality, whether Q2M works better with some particular upsampling, digital filters, data formats, analog filters, oversampling ratio, etc. As time goes on some of those things seem to be continuing to solidify, and some things seem to still have more fluidity.
In this thread, my own dac project journey has had some better luck and some worse. Mostly, I have been fairly lucky in not having start over or make massive changes. Smaller changes have certainly happened, and my worst luck has come with blunders mispowering opamp polarity and having to replace a bunch of them. The good that has come out of it is that I have learned I could do and redo tiny work much more successfully than I ever imagined.
In trying to be supportive of other, perhaps less experienced thread participants I have always tried to answer any questions and explain how I would probably go about working on a particular problem. Sometimes there is one question and an initial answer or response, and then silence, perhaps premature silence as though there were nothing more to be gained by more discussion or possible additional back and forth clarification.
Changing the subject a little, I will admit my preference is not for writing cookbook approaches to modding, and that is true even for people pretty new to this stuff. I have suggested getting a SMD practice soldering kit from maybe ebay or Amazon, and recommended tools and supplies that can help make things easier. I have tried to talk about how to think about working on developing a circuit layout too. I guess what I am hoping for is that we can have helpful conversations to get through the hard parts, with hard being defined by whatever is new or hard for the person doing the work. I do want people to think about how they might layout a circuit and things like that because I am not here here to do all of other people's thinking for them. I think modding should be a learning experience and maybe a little bit challenging, again according to individual experience and abilities.
What is probably not so good is when somebody asks a question then comes back a few weeks later and says its finished and here it is, or I quit, or didn't start doing anything. The problem in the first case is that for many people there should have been more questions and more conversations along the way. That would be fine. Learning how to do electronics modding involves some thinking that is similar to learning math. People learn from doing as much of the homework as they can themselves, and then from talking over and focusing the next day on the problems that didn't yield to effort on the first try. There is some effort to learning where one has to think about things a lot, and ask questions maybe every day, but the good thing about is one actually learns so much more that way, and the satisfaction and accomplishment that comes with progress can be more than was ever expected.
For those that are adamant they want to follow a recipe or nothing, I guess we need to come up with a recipe for AVCC and maybe for output stages that isn't or aren't too intimidating on account of small parts or too much complexity. Unfortunately, drop in solutions for some things that are needed at prices that are in keeping with the idea of 'low-cost' is tough. I could probably come up with a recipe to make an AVCC supply with through-hole parts, but it would still be more work than many would want to do. The most popular wish seems to be maybe a handful of cheap ebay boards that can attach to the dac board and transform it into something actually better. That wish appears to be for things that unfortunately don't exist. At least so far not much luck finding cheap add-on boards that make a cheap dac board actually much better. Every once in awhile I find non-cheap boards, but they are too expensive for a cheap dac project, it tends to look like. That I can't fix. And that problem may make dac modding something not quite suited to some level of beginners.
By the way, retirement is not when one has time and energy to do lots of hobbies, not for me anyway. It is a time when there wasn't enough energy left to keep working any more. Yes, there is time but much of it is therefore not at all productive.
Okay then, having gotten through talking out loud while thinking about some future possibilities for this thread in response to the comments of DRONE7 and also in reference to the efforts of Serge to make things start happening for more people, looks like it is time for a response back to me to see if we can find a way to move forward at all. We may stumble and thrash a bit trying to work through hard organizational questions, but likely some improvements could come from it. That is, it could happen if we keep talking about it and don't give up too soon.
In this thread, my own dac project journey has had some better luck and some worse. Mostly, I have been fairly lucky in not having start over or make massive changes. Smaller changes have certainly happened, and my worst luck has come with blunders mispowering opamp polarity and having to replace a bunch of them. The good that has come out of it is that I have learned I could do and redo tiny work much more successfully than I ever imagined.
In trying to be supportive of other, perhaps less experienced thread participants I have always tried to answer any questions and explain how I would probably go about working on a particular problem. Sometimes there is one question and an initial answer or response, and then silence, perhaps premature silence as though there were nothing more to be gained by more discussion or possible additional back and forth clarification.
Changing the subject a little, I will admit my preference is not for writing cookbook approaches to modding, and that is true even for people pretty new to this stuff. I have suggested getting a SMD practice soldering kit from maybe ebay or Amazon, and recommended tools and supplies that can help make things easier. I have tried to talk about how to think about working on developing a circuit layout too. I guess what I am hoping for is that we can have helpful conversations to get through the hard parts, with hard being defined by whatever is new or hard for the person doing the work. I do want people to think about how they might layout a circuit and things like that because I am not here here to do all of other people's thinking for them. I think modding should be a learning experience and maybe a little bit challenging, again according to individual experience and abilities.
What is probably not so good is when somebody asks a question then comes back a few weeks later and says its finished and here it is, or I quit, or didn't start doing anything. The problem in the first case is that for many people there should have been more questions and more conversations along the way. That would be fine. Learning how to do electronics modding involves some thinking that is similar to learning math. People learn from doing as much of the homework as they can themselves, and then from talking over and focusing the next day on the problems that didn't yield to effort on the first try. There is some effort to learning where one has to think about things a lot, and ask questions maybe every day, but the good thing about is one actually learns so much more that way, and the satisfaction and accomplishment that comes with progress can be more than was ever expected.
For those that are adamant they want to follow a recipe or nothing, I guess we need to come up with a recipe for AVCC and maybe for output stages that isn't or aren't too intimidating on account of small parts or too much complexity. Unfortunately, drop in solutions for some things that are needed at prices that are in keeping with the idea of 'low-cost' is tough. I could probably come up with a recipe to make an AVCC supply with through-hole parts, but it would still be more work than many would want to do. The most popular wish seems to be maybe a handful of cheap ebay boards that can attach to the dac board and transform it into something actually better. That wish appears to be for things that unfortunately don't exist. At least so far not much luck finding cheap add-on boards that make a cheap dac board actually much better. Every once in awhile I find non-cheap boards, but they are too expensive for a cheap dac project, it tends to look like. That I can't fix. And that problem may make dac modding something not quite suited to some level of beginners.
By the way, retirement is not when one has time and energy to do lots of hobbies, not for me anyway. It is a time when there wasn't enough energy left to keep working any more. Yes, there is time but much of it is therefore not at all productive.
Okay then, having gotten through talking out loud while thinking about some future possibilities for this thread in response to the comments of DRONE7 and also in reference to the efforts of Serge to make things start happening for more people, looks like it is time for a response back to me to see if we can find a way to move forward at all. We may stumble and thrash a bit trying to work through hard organizational questions, but likely some improvements could come from it. That is, it could happen if we keep talking about it and don't give up too soon.
Mark, thank you for your response and I do look forward to your focusing our efforts to more practical outcomes.
Could I suggest the 'modular' approach taken by Serge.... perhaps starting with something requiring less smd work such as an expanded how-to for the I/V stage..?
I bought the Chinese ES9038Q2M ver 1.04 then, after reading yours and others comments re current output, I bought an ebay ES9028Q2M with I/V output and was very impressed at the improvement..
I would like to do the I/V mod to the first board but need a little more insight as to what and where.... I guess I'm more visual than text learner....
Could I suggest the 'modular' approach taken by Serge.... perhaps starting with something requiring less smd work such as an expanded how-to for the I/V stage..?
I bought the Chinese ES9038Q2M ver 1.04 then, after reading yours and others comments re current output, I bought an ebay ES9028Q2M with I/V output and was very impressed at the improvement..
I would like to do the I/V mod to the first board but need a little more insight as to what and where.... I guess I'm more visual than text learner....
Hello Mark, thanks for your response and kind words...this schema was designed by one person who reccomended to use discrete output stage instead opamp one, especially this basic one onto china board, what i can say abot this class A stage is that sound is very detailed open but not "hard" rather soft and smooth but also stay dynamic...basic opamp stage SQ was more transparent, unbalanced like with some distortion on high middles and highs...as i said from start i fight with some oscillation and reducing feedback capacitor helped and also softened sound...(seems that high oscillation caues high distortions)
But your idea to use simple output stage (recommedned by ESS) with 3 opamps is not bad and i could compare SQ against my stage... which schema shall i exactly use, and which opamps would you recommended ? regarding Avcc voltage, i was only asking if i use 2 separate power supply 3.3V-3042 for each channel of china board (remove regulator or interrupt feeding of 3.3v of each channelonto desc) so it should help with better SQ and also increasing capacitor directly behind china PCB parallel to original one (100uF are there if I am not wromg...)...thanks Vita
Last edited:
Yes It will be really great to get a step by step an visual tutorial for I/V stage.
Hi
I do not want to be in competition with anyone ,i just want to help those who want basic changes about this board .
After of course the Markw4 advices are very important to reach the limits of this card but for me it's hard to understand all the technical details.
The main thing is that some people find their happiness
ES9038Q2M Production
Serge
I do not want to be in competition with anyone ,i just want to help those who want basic changes about this board .
After of course the Markw4 advices are very important to reach the limits of this card but for me it's hard to understand all the technical details.
The main thing is that some people find their happiness
ES9038Q2M Production
Serge
Last edited: