Too late to edit my last post, so I will add a bit more here:
Some things I left out: How did I know not to quit? There are lots of times when it would be nice to think and easy to believe I must have this dac to 99% of the best it can be after all the work I have done on it.
The way I knew that I needed to keep trying things and keep working on improving it was by listening to DAC-3. And actually, I can tell I probably missed some other things besides only external DSP for interpolation. I have a suspicion that having proper power planes for everything would help. PSRR is really good in modern opamps, but when trying to get the very best performance possible down at -120dB and maybe even a little lower, nothing can be taken for granted. Keeping power supply impedance at opamp pins very low matters, just like keeping AVCC power impedance low at dac chip pins matters. I know these things because attending to them as much as I can has helped me get closer to DAC-3 sound quality.
Also, I keep listening through the modded LME49600 headphone amp I keep telling people they need, and using that to compare with DAC-3. It is a clean enough headphone amp to hear the changes to the dac. By the way, I went through similar steps to get the headphone amp working too which I wrote about in one of my posts back when. I didn't get it right the first time and had to keep making it better to by comparing it to another headphone amp, as I described in my posts at the time.
At least now people don't need to do all the testing and fine tuning I did to get to this point. There are posts with pictures and descriptions that show how to do everything I did including small SMD fabrication, without the need for a lot of experimenting and redoing things that didn't turn out to be good enough the first time around.
Still, although it is possible to do, it is a lot of work and pretty time consuming. If someone likes doing that or wants to do something tiny like building a ship inside a bottle but instead with electronics, then maybe this could be it. Probably a lot easier than a ship in bottle actually, but the small stuff can take time, especially at first.
Some things I left out: How did I know not to quit? There are lots of times when it would be nice to think and easy to believe I must have this dac to 99% of the best it can be after all the work I have done on it.
The way I knew that I needed to keep trying things and keep working on improving it was by listening to DAC-3. And actually, I can tell I probably missed some other things besides only external DSP for interpolation. I have a suspicion that having proper power planes for everything would help. PSRR is really good in modern opamps, but when trying to get the very best performance possible down at -120dB and maybe even a little lower, nothing can be taken for granted. Keeping power supply impedance at opamp pins very low matters, just like keeping AVCC power impedance low at dac chip pins matters. I know these things because attending to them as much as I can has helped me get closer to DAC-3 sound quality.
Also, I keep listening through the modded LME49600 headphone amp I keep telling people they need, and using that to compare with DAC-3. It is a clean enough headphone amp to hear the changes to the dac. By the way, I went through similar steps to get the headphone amp working too which I wrote about in one of my posts back when. I didn't get it right the first time and had to keep making it better to by comparing it to another headphone amp, as I described in my posts at the time.
At least now people don't need to do all the testing and fine tuning I did to get to this point. There are posts with pictures and descriptions that show how to do everything I did including small SMD fabrication, without the need for a lot of experimenting and redoing things that didn't turn out to be good enough the first time around.
Still, although it is possible to do, it is a lot of work and pretty time consuming. If someone likes doing that or wants to do something tiny like building a ship inside a bottle but instead with electronics, then maybe this could be it. Probably a lot easier than a ship in bottle actually, but the small stuff can take time, especially at first.
Last edited:
Do you think it could be done well on a two-layer board, or would it take a four-layer board? Need to wrap up a number of projects, but if it can be done well on two layers, I could probably make a halfway decent board. I could try my hand at a four-layer board, I suppose, there's a first for everything, right?
For best sound quality I can't see doing it will less than four layers. It should have power planes, and in particular a ground plane that is not obstructed by many traces or long traces running across it.
If you look at pictures of the original Chinese board they did a good job with a double sided board. The ground plane is pretty intact everywhere except where they intentionally divided it. Maybe such divisions could serve multiple purposes, I don't know.
Then if you look at the wiring I added for extra voltage regulators and opamps, routing those on the top in addition to signals would probably be virtually impossible. If there were power planes, +-15v where the opamps are, and maybe a 5v instead of one of the 15v at the very digital end, then only signal wires would have to be routed much.
With a top and two power planes, signals could probably be routed without causing much problems, and without having to cut into the ground plane.
Also, you should probably be aware that the AK4137 needs to be part of the same design. We need keep any jitter pickup along the way from it to the dac chip to be minimal. The cheap AK4137 board doesn't have a serial receiver chip, but the more expensive one does. AKM makes a companion chip for that purpose. Or, I guess the AK4137 could be omitted, but it is a big part of the best sound quality I can get. Upsampling to DSD helps a lot, IME.
Then there are power supplies. They could be on a separate board or integrated into one board. The power supplies I use currently are a probably a bit of overkill, but they need to at least be good enough.
One shared microcontroller should enough for all the chips that need it. But, I2C needs to be routed to AK4137 if on the same board.
That't my thinking so far, anyway.
Still looking not so bad?
If you look at pictures of the original Chinese board they did a good job with a double sided board. The ground plane is pretty intact everywhere except where they intentionally divided it. Maybe such divisions could serve multiple purposes, I don't know.
Then if you look at the wiring I added for extra voltage regulators and opamps, routing those on the top in addition to signals would probably be virtually impossible. If there were power planes, +-15v where the opamps are, and maybe a 5v instead of one of the 15v at the very digital end, then only signal wires would have to be routed much.
With a top and two power planes, signals could probably be routed without causing much problems, and without having to cut into the ground plane.
Also, you should probably be aware that the AK4137 needs to be part of the same design. We need keep any jitter pickup along the way from it to the dac chip to be minimal. The cheap AK4137 board doesn't have a serial receiver chip, but the more expensive one does. AKM makes a companion chip for that purpose. Or, I guess the AK4137 could be omitted, but it is a big part of the best sound quality I can get. Upsampling to DSD helps a lot, IME.
Then there are power supplies. They could be on a separate board or integrated into one board. The power supplies I use currently are a probably a bit of overkill, but they need to at least be good enough.
One shared microcontroller should enough for all the chips that need it. But, I2C needs to be routed to AK4137 if on the same board.
That't my thinking so far, anyway.
Still looking not so bad?
Last edited:
David, The original headphone amp board was this one: 1PCS LME49720NA+LME49600 headphone amplifier KIT | eBay
My post on modding it is here: http://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-63.html#post5402398
EDIT: The headphone amp in it's original form didn't sound right. As I described in the post linked above, I covered the bottom with mylar tape and made a ground plane with copper foil. On top of the board where the opamp is I sanded off the traces and lifted the output pins of the LME49600's. I used some little pieces of PCB material and some very good quality 10k thin film resistors to make the feedback circuit and also connected the output jack there. That was to keep capacitance along the original output traces from coupling the outputs back into the power supply traces and causing potentially causing oscillations. Basic decoupling caps were added at all the IC power pins, then some more electrolytic and film caps were added because the stereo separation was not as good as the TP boards I was comparing to. I also put some thick red and black jumper wires to make a low(er) impedance path between the original filter caps. The additional power supply caps seemed to fix stereo separation okay. Also, it is good to download and read the LME49600 data sheet. There is a more involved headphone amp shown in it with DC offset servoing, but I didn't find any offset problem with my board so I left it DC coupled without a servo. Please let me know if there are any questions or clarifications I might be able to help with.
EDIT 2: By the way, a superb example of what is possible with a very carefully designed LME49600 HPA can be found here: HP-1: Ultra-High End Headphone Amplifier
To get anywhere close to that level of performance very careful layout and both power and ground planes are essential, other stuff too, but the point is the basic LME49600 buffer is extremely difficult to beat, and much like our dac chip so much depends on implementation details and not just on the part itself.
My post on modding it is here: http://www.diyaudio.com/forums/digital-line-level/314935-es9038q2m-board-63.html#post5402398
EDIT: The headphone amp in it's original form didn't sound right. As I described in the post linked above, I covered the bottom with mylar tape and made a ground plane with copper foil. On top of the board where the opamp is I sanded off the traces and lifted the output pins of the LME49600's. I used some little pieces of PCB material and some very good quality 10k thin film resistors to make the feedback circuit and also connected the output jack there. That was to keep capacitance along the original output traces from coupling the outputs back into the power supply traces and causing potentially causing oscillations. Basic decoupling caps were added at all the IC power pins, then some more electrolytic and film caps were added because the stereo separation was not as good as the TP boards I was comparing to. I also put some thick red and black jumper wires to make a low(er) impedance path between the original filter caps. The additional power supply caps seemed to fix stereo separation okay. Also, it is good to download and read the LME49600 data sheet. There is a more involved headphone amp shown in it with DC offset servoing, but I didn't find any offset problem with my board so I left it DC coupled without a servo. Please let me know if there are any questions or clarifications I might be able to help with.
EDIT 2: By the way, a superb example of what is possible with a very carefully designed LME49600 HPA can be found here: HP-1: Ultra-High End Headphone Amplifier
To get anywhere close to that level of performance very careful layout and both power and ground planes are essential, other stuff too, but the point is the basic LME49600 buffer is extremely difficult to beat, and much like our dac chip so much depends on implementation details and not just on the part itself.
Last edited:
David, You are quite welcome. Please see a few edits in my original reply to you. Enjoy!
EDIT: Ran out of time to edit my previous post one more time but wanted to say: For those who may not be familiar with the way pictures show in the forum, when you click on a picture to open it, it may not be shown at full size and level of detail. If not, there will be a white X that appears in the lower left corner of the picture when the mouse is hovered over that area. Clicking on the X will open the picture to full size. Pictures can also be downloaded if desired by right clicking and selecting from the menu.
EDIT: Ran out of time to edit my previous post one more time but wanted to say: For those who may not be familiar with the way pictures show in the forum, when you click on a picture to open it, it may not be shown at full size and level of detail. If not, there will be a white X that appears in the lower left corner of the picture when the mouse is hovered over that area. Clicking on the X will open the picture to full size. Pictures can also be downloaded if desired by right clicking and selecting from the menu.
Last edited:
Do you mean see with a scope? Or maybe, hear something possibly caused by AVCC and VCCA power? Hard to know what you mean sometimes if your description is too short and brief.
A bit above my current skill level, but staying in my comfort zone won't improve my skills. If nothing else, sounds quite time consuming, which would be my biggest problem.
Where can a ES9038Q2M chip be purchased? I did a quick check on Mouser and Digi-Key, they don't appear to carry any ESS chips.
I can see how that might be the case, with having additional layers to route on. My main thought is actually cost. I've been using Elecrow for all my PCB fabs since I started doing it. They are super cheap for 10x10cm (or less) and 2-layers. But their prices go up 10x if you exceed that size or add layers. Although I think I've seen others suggest there are fabs that will do four layer boards for a reasonable cost.