There are various choices, you want to try something. At one time I made a through-hole output stage project which can be downloaded at: https://www.dropbox.com/s/gby3v49oxoqi41q/Output Stage Instructions.zip?dl=0 Some people got a warning about the .rtf file being unsafe, but it only contains images and text. Anti-virus checks show no problem, so don't know why some people get the warning.
Then there is a simpler way to make some improvement posted by forum member Victor (aka vicnic ?) that used two opamps per channel to make an output stage:
However, LME49720 is probably not the best opamp for the above since it appears to sensitive to EMI/RFI present at the dac chip outputs. You might try OPA1612 instead.
Also IIRC, Victor attached three 1800uf caps in parallel on the bottom of the dac board per AVCC channel (from each AVCC pin to ground) to help stabilize and clean up AVCC without making a complete regulator.
Otherwise the opamp buffer AVCC supply in the ESS application note posted previously can work to power AVCC. A low noise reference can also be used for the opamp buffer reference inputs. The circuit can be mounted on the ground plane side of the dac board. The AVCC buffer opamps can be run from +12 or +15 volts, and with ground used for the negative opamp rail.
Then there is a simpler way to make some improvement posted by forum member Victor (aka vicnic ?) that used two opamps per channel to make an output stage:
However, LME49720 is probably not the best opamp for the above since it appears to sensitive to EMI/RFI present at the dac chip outputs. You might try OPA1612 instead.
Also IIRC, Victor attached three 1800uf caps in parallel on the bottom of the dac board per AVCC channel (from each AVCC pin to ground) to help stabilize and clean up AVCC without making a complete regulator.
Otherwise the opamp buffer AVCC supply in the ESS application note posted previously can work to power AVCC. A low noise reference can also be used for the opamp buffer reference inputs. The circuit can be mounted on the ground plane side of the dac board. The AVCC buffer opamps can be run from +12 or +15 volts, and with ground used for the negative opamp rail.
Perhaps it is worth emphasizing that this DAC is very crippled without a I/V proper output stage. We are not talking about a few dB here and there. We are talking about 40+ dB. It is perhaps worth searching through the whole thread and finding the measurements done on this board. In the crippled state I am not sure what can be hoped to be gained with changes like FKP caps. The output stage is a very severe flaw. When this board is seen on Ebay or AliExpress (or other site) the idea of an ES9038Q2M suggests a certain level of performance. With the crippled output stage the performance is nothing of the kind.
In other words with the present output stage this is not a suitable board to mod with high quality capacitors and listen for improvements.
In other words with the present output stage this is not a suitable board to mod with high quality capacitors and listen for improvements.
No. There is a bass and quite dense. I listen with my Hifiman 560. Tomorrow I want listen trough my stereo setup. I also change fake 8920 on LM6172. I don't have other opamps (need to buy).
Like the guys are trying to tell you. Its not just opamps, not just DC blocking caps. There are significant problems with this board. Like I told you, output stage and AVCC supplies help the most. OTOH, Wima FPKs in large values are a waste for something like this. Large size film caps may sound good in speaker crossovers, but not for DC blocking in a DAC. Doesn't work like that. Sorry.
Turns out its exactly 1/2 of +3.3v. Historically it was taken from each AVCC regulator output, then reduced to 1.65 using two 10k resistors in series to make a voltage divider to ground. The output of the voltage divider had an electrolytic filter cap to clean up that I/V opamp common mode voltage. Later it turned that the voltage didn't have to be half of AVCC, sometimes less voltage was used for two possible purposes: (1) to reduce hump distortion under the conditions it was being measured, and or (2) to remove the dac chip DC offset using the I/V opamps. Two things about that, (1) changing the measurement conditions can merely move the hump to a different frequency and or level. IOW, lowering I/V common mode voltage can hide the hump, but not necessarily completely eliminate the problem, and (2) reducing the I/V common mode voltage causes a DC current to flow out of the dac output and that current is supplied by the I/V opamp. If you want to try it you could use a 20k pot (instead of the two 10k fixed voltage divider resistors) and dial down the common mode voltage to reduce or eliminate the DC offset at the opamp outputs. However, you might need to change the filter cap and or change how the filtering is done since using other than two 10k resistors would reduce the time constant depending on the exact pot setting.
All resistors should be good quality metal film or metal thin film. The better quality Yageo resistors are what I would suggest to try.
Depends. If you remove the DC offset in the I/V opamps then maybe you don't need DC blocking caps. Otherwise you could try may Pansonic FC, or similar, maybe with a small Wima mkp cap in parallel. In that case the film cap is a lot smaller than most people think if you don't want the sound to be too bright. You may find .1uf to be too much, and maybe find 4.7nf to be too small. Have to try it and see what you think. Also, when electrolytics are used for DC blocking they should be large enough value so that the AC voltage across them at the lowest frequency of interest is very small. Less than 67mV is a number that comes to mind. In addition, it can take a week or three for sound of an electrolytic in the sound path to audibly stabilize. You may have to burn it in for awhile before deciding if you like a particular choice of cap.
They would be from the AVCC pins to ground (3x 1800uf per for each AVCC channel). That was as an alternative to using dedicated low noise regulators for each AVCC channel, such as, for example, opamp buffers.
Last edited:
Hi Mark,
Apologies if this has been asked before, but I want to ask if you have seen (or even better, use) this particular ES9038Q2M board. It advertises itself as a Raspberry Pi DAC of some kind, and the features correspond (it has tiny buttons for volume control, input control...). The design reminds me of the green board of the first link, with the arrangement of the op-amp and surrounding caps (though it might just be because it's copied from a reference design).
Here's the Aliexpress link.
I would like to ask if you think something's off with the board, and whether or not it's suitable as a "first DAC". I have read through this thread somewhat and understand your sentiment towards these budget DACs. But my quality standards aren't particularly high (an improvement from integrated computer audio) . Maybe when I have some more spare cash I can throw in a few mods. The question is, is it worth it, for the $25-30 price tag?
Thank you.
Obligatory pictures:
Its about the same as the other dac just under discussion. Same problems with output stage and AVCC power.
Plus this one for RPi may be worse in a few ways:
1. Looks like this one doesn't have crystal clock, rather its one that produces probably worse close-in phase noise.
2. This one will get I2S from RPi which is typically of lesser quality than you would get from a USB board.
3. This one is normally located in very close proximity to an RPi, which is a source of conducted and radiated EMI/RFI which tends to cause SQ problems in nearby dacs. That's why if someone wants to use RPi as a streamer, I would recommend to connect the dac to RPi via USB, with substantial steps taken to isolate the dac from RPi electrical noise.
Thus IMHO this dac is going to be more work to get sounding better than other non-RPi dac we were just talking about a few posts ago. BTW, there is a long history of using RPi hat dacs, including all the work that Iancanada has done in that area. You can spend a lot a money trying to fix one of these things to sound decent, or else you could spend less money in the end and get better sound buying a Topping dac or equivalent at whatever level you can afford. Also, anyone wanting to work on dacs needs a scope. 2-channels, 100MHz bandwidth, is about the minimum for dac work. Without that, there are some people here who won't even try to help you if you have problems.
The above having been said, if you want to start learning about dacs, their problems, proper dac design, or just play around with a dac for a hobby, then one of these low-cost dac boards can be one way to get started. However, please don't hold out hope that the guys who design these cheap dacs designed a really good dac, but they made one stupid, easy to fix, mistake, and if you just replace a few easy parts you will have a great dac for a great price. Actually, the truth is more like the guys who designed these dacs optimized them in every way to be as cheap to make as possible by using the fewest and cheapest, worst parts they could find, except for only the dac chip is real. Why? Because that's the only way they can make any money at the prices these dac boards sell for.
Last edited:
As an example, if you consider datasheets only the ES9038Q2M is much greater than a PCM5102. However if you consider the terrible implementation of the ES9038Q2M in the last two boards in this thread you might be surprised to find that the PCM5102 performs better because it does not need a complex output stage/application circuit. So if you have a cheap board with a disaster output stage the ES9038Q2M will perform terribly. If you want something very cheap for the Raspberry Pi the PCM5102 is probably better because the cheapness in the application circuit and implementation will do less damage to it. (It is designed for fewer and simpler external components/application circuits. The ES9038Q2M is designed with far most stringent and costly external components and application support circuitry.) Of course this is a bit of a quick illustration/comparison that should not be pushed too far. But I hope that you get my point.
It becomes necessary to get a board that is implemented correctly with good quality parts, proper support circuitry and good layout if you want really good performance.
So either you need to find a better implementation or be satisfied with what you can get. However I don't recommend a poor ES9038Q2M implementation. After I tried this ES9038Q2M (and made a prototype proper output stage and power supplies) I built the Ti recommended circuit for the PCM1794 on a four layer PCB with quality parts. Again boards were available for PCM1794 (cheap) but the components and/or layout were poor. Quite a bit can be done with quality 1/4W metal film resistors, quality PP film caps and quality (such as Panasonic) electrolytics. Unfortunately many (most?) of the cheap Ebay or AliExpress boards are significantly compromised one way or another. [Aside: Finally beware that many/most of the good quality components are faked online. That includes op-amps, transistors, capacitors, etc. If it is a desired and quality audio component you can be sure that it has been faked if possible. So I suggest sticking with Tier 1 real distributors for your parts. I would recommend the NE5534 and NE5532 from a genuine Tier 1 distributor over just about any cheap online market place part number because most online are fake. I eventually learnt to stop buying op-amps online...]
It becomes necessary to get a board that is implemented correctly with good quality parts, proper support circuitry and good layout if you want really good performance.
So either you need to find a better implementation or be satisfied with what you can get. However I don't recommend a poor ES9038Q2M implementation. After I tried this ES9038Q2M (and made a prototype proper output stage and power supplies) I built the Ti recommended circuit for the PCM1794 on a four layer PCB with quality parts. Again boards were available for PCM1794 (cheap) but the components and/or layout were poor. Quite a bit can be done with quality 1/4W metal film resistors, quality PP film caps and quality (such as Panasonic) electrolytics. Unfortunately many (most?) of the cheap Ebay or AliExpress boards are significantly compromised one way or another. [Aside: Finally beware that many/most of the good quality components are faked online. That includes op-amps, transistors, capacitors, etc. If it is a desired and quality audio component you can be sure that it has been faked if possible. So I suggest sticking with Tier 1 real distributors for your parts. I would recommend the NE5534 and NE5532 from a genuine Tier 1 distributor over just about any cheap online market place part number because most online are fake. I eventually learnt to stop buying op-amps online...]
One can only imagine what a listening session to find the worst parts must be like and to think some actually think they failed.
That resistor after the 220uf cap is just to discharge the cap if the output of the dac is not connected to anything else. We probably want to keep the impedance high after the cap so as to keep any voltage drop more across the load, rather than across the cap. Otherwise we might try a bigger value cap so its impedance is low compared to the load impedance. The resistor value depends on how long a period of time we are willing to allow for the cap to discharge, and how much current we want to go through the resistor in normal operation. You might try making the resistor something like 100k as an initial estimate.
EDIT: Not sure what you mean by "tune the opamp from both side?"
EDIT: Not sure what you mean by "tune the opamp from both side?"