ES9038Q2M Board

wondering about ES9038Q2M the noise floor, how's it looks like?

Don't know, but I think it must be fairly low at least compared to many older dacs. The amount of filtering I decided to use was based on listening tests. Allo relied too much on measurements and had to go back and rework the first run of Katana dacs they made to add some additional output stage filtering. The problem was discovered when maybe 2-3 of the reviewers received the first units and heard problems. Most reviewers either did not notice, or assumed the dac was as good as Allo could make it so no point in sending them feedback about a problem with the sound.

Noise can cause distortion if it intermodulates with signal, and or to the extent it is deterministic noise. In some cases HF noise can cause distortion in opamps if it becomes demodulated in semiconductor junctions.

In addition, I would point out as I do from time to time that listening for distortion is mostly a learned skill, not innate. Low level higher order harmonic distortion (or perhaps more likely, the associated IMD) can cause listening fatigue in untrained listeners, so probably worth paying careful attention to. I wrote up some tips on starting to learn how to listen about half way though a post in another thread: https://www.diyaudio.com/forums/dig...proving-ak4490-thd-figures-5.html#post5666915 ...and a little more in yet another: https://www.diyaudio.com/forums/vendor-s-bazaar/294940-fifo-buffer-rpi-sbcs-302.html#post5669890
Knowing what to pay attention to when listening is one thing, but any skill is improved with prolonged practice. I have found that a year is enough time to get pretty darn good at listening, but people who want to develop skill have to start sometime and not worry about how long it will take to get much better at it.
 
Last edited:
How about for DSD? AKM has a recommendation regarding what to do in that case, as can be seen below.
 

Attachments

  • AK4490.jpg
    AK4490.jpg
    152.3 KB · Views: 386
A little bit more good dac modding news.

Update: For quite awhile I was pretty much exclusively trying to figure out a solution to the DPLL stability problem and didn't bother to work on any other unfinished tasks. Today I finally got around to adjusting harmonic distortion compensation. This time the magnitude of the adjustments were much larger than I had seen before, but a number of things about the dac are different, so not sure what to make of it. C2 came in at +1165, and C3 came in at -8985. After that found that different AK4137 filters seem to sound best. The modded dac now sounds more like Benchmark DAC-3 than it did a couple days ago, almost eerily so. AK filter 2 is a tad too bright to match DAC-3 and AK filter 3 is tad ever so slightly too dark. It seems remarkable that an ES9038Q2M dac built almost entirely on 2-layer boards can come so close in sound to an ES9028PRO dac built on a multilayer board. With a slight tweek of dac filters they possibly could be rather hard to tell apart, its getting that close.
 
In addition to the above, I will attach some pics below of a couple of Chinese low cost Spartan 6 boards, and an expansion board for experimenting. Although I don't enjoy programming much, it is something my son likes to do. We are trying out an idea which is that he will try to do a lot of the programming work if I will teach him about DSP. As always, no guarantee anything useful will come out of it, but at least it creates a possibility.
 

Attachments

  • Spartan6Boards2.jpg
    Spartan6Boards2.jpg
    935.4 KB · Views: 333
ESS claims -122dB noise and -120dB distortion for ES9038Q2M, the mobile chip we use in this thread. I think there is more noise in DSD mode though, but then again, noise is part of DSD.

The ES90x8PRO chips are rated for lower noise (than Q2M) which is associated with their higher output currents. If all eight channels of PRO chips are run in parallel groups of 4, noise can be reduced another 6dB, although most manufacturers don't seem to bother since the cost of additional output stage parts to get that lower noise doesn't usually seem worth it.

Then again, ESS tends to say -120dB THD+N for all the chips since distortion is about the same in all of them. Guess I am undecided about the merits of making noise lower than distortion. I think I might prefer to hear the noise if I have a choice.

Most of the output stage filtering is to keep low level HF junk out of subsequent preamps and or power amps that may not like it, so it is said.

EDIT: Looking at the AKM graphs, they seem to be focusing on noise from about 50kHz to 300kHz, then the graph stops. Don't know about that specifically for ESS dacs.
 
Last edited:
I could see how out of band noise could cause problems if output stages were not designed to attenuate it sufficiently. My output stage does use a lot of filtering, so wouldn't expect to see much after it, but I have not looked.

May I ask if you checked to verify that out of band noise is also very low for DSD mode? If only low for PCM then it still seems like output stage filtering might be prudent.
 
Okay, that would make sense. They do some more of the final filtering for you besides just the interpolation filtering. However, will that be the case for the new dac they have coming out?

EDIT: Still, odd that the AK4490 data sheet shows recommended output stage filtering for DSD.
 
Last edited:
If someone doesn't mention A-weighting, he has to specify the BW at least. Do you believe if ES9038Q2M has S/N=DR= -129db unweighted with 3.3VDC power and 25C ambient? Single one resistor 330ohm@25C should produce in 20-20k BW -129db(or about -132dbA) S/N regarding 1VRMS, if ES9038Q2M has -129db unweighted I'll be surprised. I saw this morning AKM mobile 3.3V DAC with -125dbA DR but with "Auto-mute gate" i.e. when the input data < threshold, DAC simply off and that state shows -125dbA. Higher class DAC, 4490 as well, has 5V power.