ES9038Q2M Board

[syn08]

My take on LM317/LM337 is that they are pretty good, but maybe not good enough with only one stage of them to necessarily eliminate power supply spurs on an FFT. Two of them in series, a two-stage regulator, is one possible solution but not the only one. There is also a noise cleaner add-on circuit that can be used with the regulators. Also,LM317 should be okay for dropping some higher voltage down to maybe 5v or so, which can then be regulated down to 3.3v for particular loads.

Regarding TL431, one of my professional audio designer friends doesn't like them, he says there are better sounding solutions. Just one person's opinion. They do seem to be popular though. Haven't used them myself, so no personal opinion.

I would say that two initial mods that make a lot of difference are the output stage, and the AVCC power supply. They both very directly affect the analog output of the dac. Also, the AVCC current draw is not entirely constant. It means the sound of the voltage regulator error amplifier could potentially be more or less imprinted on the sound of the dac. All the more so since AVCC has zero PSRR. That may be why a regulator made out of a good audio opamp has been found to be a good solution (not that its the only solution).

EDIT: The TPS7A47/TPS7A33 referenced in the preceding post by Mr. 08 above are what Topping used for the +-11v opamp supplies in their D90 dac. It sounds pretty good. However, for Vref (about the same as AVCC) they used LM317 as a first stage regulator.

Once again, hearsay, unsubstantiated claims and misinformations:

- "Two of them [LM317] in series, a two-stage regulator, is one possible solution but not the only one."

Such will only (and marginally, at low frequencies) improve the input PSRR. It will not improve the output noise, output regulation, spurious, etc...

- "Regarding TL431, one of my professional audio designer friends doesn't like them, he says there are better sounding solutions."

Likely Jam's kool aid. There are good real reasons to avoid the TL431, "better sounding solutions" is the usual BS.

- "All the more so since AVCC has zero PSRR. That may be why a regulator made out of a good audio op amp has been found to be a good solution"

Really, for all DACs? Always? Quotes please. And what would be the difference between a high quality DC regulator and a "good audio op amp" in the AVCC place? Hint, none at all. In fact, there are good reasons to believe that a good DC regulator implemented properly (layout, general and local bypassing (yes, with X7R ceramics), force/sense output connections) would be better by any metric.

 

[syn08]

I wold vote for LT3045/LT3094. But is is kind of overkill for op amp PS, IMHO

Agreed, op amps have usually very high PSRR, rejecting any noise or spurios from the supply rails. To add insult to injury, LT3045/LT3094 regulators are not in particular cheap.

But the topic here is the AVCC in a DAC, which is internally not feeding op amps only, and the AVCC PSRR, frequency response and sensitivity are usually not specified to allow designing based on the DAC datasheet.

 

[kozard]

I wold vote for LT3045/LT3094. But is is kind of overkill for op amp PS, IMHO

There is a similar LT3045 LT3094 module. (Pictured.) I don't like the low absolute maximum input voltages but I could power it from my LM317/LM337 supply set to something like +/- 20V to supply the ES9038Q2M board at +/- 15V.

It might be overkill but I suppose with a couple LT3045 I could be done with it and not question the DAC supply.

For improving the output stage I will likely wait until I have the time to use either KiCad or Eagle to do it neatly.

 

[syn08]

^^^
If that's the board available on EBay, you can easily dump the heatsink. ES9038 needs only a few mA for AVCC and the regulator doesn't need any extra cooling beyond the board ground plane. Also, it may be a good idea to replace the voltage adjustment pots with fixed resistors (for example 33kohm for 3.3V output. I would not trust the chinese pots for such a regulator.

 

[eziitis]

Agreed, op amps have usually very high PSRR, rejecting any noise or spurios from the supply rails. To add insult to injury, LT3045/LT3094 regulators are not in particular cheap.

But the topic here is the AVCC in a DAC, which is internally not feeding op amps only, and the AVCC PSRR, frequency response and sensitivity are usually not specified to allow designing based on the DAC datasheet.

what, AVCC again? so far I remember it was discussed in this very topic several years ago I can just say, no integral LDO is up to the task, believe or not. Even ESS own part ES9311. Vref-LPF-opamp (like in ESS demo boards) or a discrete shunt type one.

 

[syn08]

^^^
Based on what? Let me guess, subjective impressions coming from uncontrolled listening tests, correct?

Make no mistake, I'm not here to tell you what you hear, only that your subjective opinions may not be good for your next reader. Unless you could substantiate (aka "prove") your results, which I am pretty sure you can not.

 

[eziitis]

There is a similar LT3045 LT3094 module. (Pictured.) I don't like the low absolute maximum input voltages but I could power it from my LM317/LM337 supply set to something like +/- 20V to supply the ES9038Q2M board at +/- 15V.

It might be overkill but I suppose with a couple LT3045 I could be done with it and not question the DAC supply.

For improving the output stage I will likely wait until I have the time to use either KiCad or Eagle to do it neatly.

I would be more concerned to have dedicated regs for each of the loads, and separate secondary windings for digital and analog supplies.

 

[John Luckins]

^^^
I totally agree, and base your findings on extensive listening

 

[cogitech]

To respond to some previous posts in a different context, I would agree that trying a bunch of cheap dacs trying to find a good one out of the bunch is a good way to waste money. The same amount of money spent on a D50, or another another dac that is a good value for the money, is probably going to give better bang for the buck than several cheap dacs could.

I would agree with all of this, however I spent money on cheap DACs and had fun building and modifying them. It is a hobby, after all.

I would add that the Khadas Tone Board is in the same league as the Toppings being discussed, and it is 1/2 to 1/3 the price. If someone just wants ones and zeroes to be effectively converted to analog with minimal jitter, noise and colouration then $100 can get you there.

I have come to the conclusion that all really good DACs should sound nearly identical if they are doing what they are supposed to do.

 

[eziitis]

^^^
Based on what? Let me guess, subjective impressions coming from uncontrolled listening tests, correct?

Make no mistake, I'm not here to tell you what you hear, only that your subjective opinions may not be good for your next reader. Unless you could substantiate (aka "prove") your results, which I am pretty sure you can not.

You are wrong, pal. I already posted this a while ago in a different topic, but once you insist ...
I would recommend you to read the whole current topic from very beginning, even you may learn something Regarding listening vs measurements, I believe the situation is a bit more complex, than you just said, because at the end of the day "Directly or indirectly, all questions connected with this subject must come for decision to the ear, as the organ of hearing; and from it there can be no appeal" (J.W.S. Rayleigh)

 

[syn08]

^^^
I totally agree, and base your findings on extensive listening

...then sell those results to the gullible. A safe way to success/recognition as an audio guru, doesn't require any expenses or skills, other than putting together a nice attractive story line, that must include a set of predefined keywords.

This approach to designing audio electronics doesn't bother me at all; what bothers me is attempting to find, or invent, engineering arguments to support these uncontrolled listening impressions. When everything else fails, going as far as creating an own, private physics.

So good luck with it, no need to re-iterate the "music is more than a collection of sines" line of reasoning.

 

[kozard]

I would be more concerned to have dedicated regs for each of the loads, and separate secondary windings for digital and analog supplies.

Right now (on my AK4396 board) I use a separate external regulator to power the on-board digital section power supply. Then I use another external LM317/LM337 board to pre-regulate (to +/- 20V) to feed the on-board analog section power supply. So completely separate external transformers.

When my ES9038Q2M board arrives I will likely start with feeding +/- 15V from one of my existing supplies (need to determine which is the best to start with) before determining how to best modify the board with multiple external supplies (each with their own transformer).

 

[syn08]

You are wrong, pal. I already posted this a while ago in a different topic, but once you insist ...
I would recommend you to read the whole current topic from very beginning, even you may learn something Regarding listening vs measurements, I believe the situation is a bit more complex, than you just said, because at the end of the day "Directly or indirectly, all questions connected with this subject must come for decision to the ear, as the organ of hearing; and from it there can be no appeal" (J.W.S. Rayleigh)

Understood. From what I can gather from the photo, you can hear -130dB spectral components? Good for you (or not so much).

BTW, since "The Theory of Sound" in 1945 we successfully measured the gravitational waves and built receivers with about five times the quantum noise hf/k. No ears were used in the process.

I'm out of here for the time being.

 

[Markw4]

kozard,
Multiple transformers may not always be a good thing. Some types of transformer construction offer more isolation of common mode noise between primary and secondary. R-core type is what I usually recommend for audio. One with, say, 4-individual secondary windings.

 

[kozard]

kozard,
Multiple transformers may not always be a good thing. Some types of transformer construction offer more isolation of common mode noise between primary and secondary. R-core type is what I usually recommend for audio. One with, say, 4-individual secondary windings.

I see. I don't have one of those but I could put it on the purchase priority list. How would you rank the priority of that versus the LT3045/LT3094 or the output/operational amplifier upgrades?

Perhaps you can offer some advice that will help with my ES9038Q2M as well as my AK4396 and noise measurements. I have been working on a Douglas Self type MC LNA (60 dB) so that I can make noise measurements on my DAC boards as I modify them. (Both the AK4396 and later the ES9038Q2M which is on order.)

The LNA has LM317/LM337 onboard with the denoiser (BC337/BC327) and in addition I use another LM317/LM337 supply externally to preregulate +/- 20V to supply the LNA board.



Unfortunately I have a stubborn 8nV 60 Hz peak (8uV an the output of the LNA on the 60 dB gain setting) that I have not been able to get rid of yet. I have attached pictures of the LNA, the LM317/LM337 pre-regulator as well as the noise measurements with the LNA input shorted (60 dB gain setting) and with the LNA powered off (to see what is not coming from the LNA).

The spectrum show what is at the output of the LNA. So divide the voltages by 1000 for the level at the input of the LNA.


Any suggestions on how to find and eradicate this 60 Hz (and 180 Hz) peak? I have unplugged all other equipment in the area and the laptop is running on battery with the WiFi disabled.

 

[Markw4]

Personally, I haven't found that I need LT3045/LT3094 to get excellent audio sound quality. The recommended circuits attached to post #3003 are pretty good as far as schematics go. Everyone I know of who tried LT3042/LT3045 for AVCC eventually found an opamp based regulator of some type sounded much better to them. Why? I don't know. Don't have time to study it.

The ebay voltage regulator boards I have recommended are more than good enough for a quite/exceptionally good, few-thousand dollar dac (although one of them does use LT4042, its better than needed). Specialized regulation beyond that is only needed for AVCC, at least for these ES9038Q2M dacs. The rest of getting it to work well is all in the implementation details.

Regarding the power line frequency spurs, how are you powering what you are measuring? Is the circuitry in a shielded metal box? How much care are you taking with interconnection wiring? With grounding? Layout?

In other words, it might help if we had more info and pics of exactly what the setup is where you are doing the measuring.

 

[syn08]

Unfortunately I have a stubborn 8nV 60 Hz peak (8uV an the output of the LNA on the 60 dB gain setting) that I have not been able to get rid of yet. I have attached pictures of the LNA, the LM317/LM337 pre-regulator as well as the noise measurements with the LNA input shorted (60 dB gain setting) and with the LNA powered off (to see what is not coming from the LNA).

Learn to make noise measurements. Your representation in V-f coordinates for noise measurement makes no sense, since the frequency bin size is not considered, you need to represent noise measurements in V/rtHz or V^2/Hz (in dB) vs. frequency (log scale), all PC based software have somewhere the option to correctly represent noise.

60Hz and harmonics cannot be completely avoided, unless you shield the LNA in 1/2 inch of aluminum or a 1/8 steel enclosure. Double shielding in thin steel cans can work too.

See https://www.diyaudio.com/forums/ana...ra-low-noise-mc-head-amp-172.html#post5977292

 

[1z2a3y]

SUPERB!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

du you use opamps as simple regulator like in the flea ? PFM Flea regulator - group buy (cancelled, but do read the thread) | pink fish media
.

Yes .... I do and it's ( quoting Donald Trump) "...because I'm RICH"!!!!!!!!!!

Lotta goodies ..... OHHHHHHHHHHHHHHHHH ............ so SEUUUUUUUUUPUHHHBBBB......

 

[kozard]

Regarding the power line frequency spurs, how are you powering what you are measuring? Is the circuitry in a shielded metal box? How much care are you taking with interconnection wiring? With grounding? Layout?

In other words, it might help if we had more info and pics of exactly what the setup is where you are doing the measuring.

Right now I am just measuring the 60 dB LNA with the LNA input shorted. (Not measuring the DAC yet.) So I am still debugging and verifying the LNA.

The LNA is in a recycled steel DVD chassis. I have tried various grounding with the best seemingly with the ground of the power supply section (LM317/LM337 on the LNA board) connected to the steel DVD chassis. But the results are similar to connecting the steel DVD chassis to the ground side of the LNA input.

The LNA has LM317/LM337 onboard with denoiser. That LNA board is being powered by the separate external LM317/LM337 preregulator. The preregulator supplies +/- 20VDC to the LNA that is inside the DVD chassis via twisted pair.

I can get additional photos later. Right now it is just the LNA inside the steel DVD chassis with the twisted pair feeding the +/- 20VDC from external.

 

[kozard]

Learn to make noise measurements. Your representation in V-f coordinates for noise measurement makes no sense, since the frequency bin size is not considered, you need to represent noise measurements in V/rtHz or V^2/Hz (in dB) vs. frequency (log scale), all PC based software have somewhere the option to correctly represent noise.

60Hz and harmonics cannot be completely avoided, unless you shield the LNA in 1/2 inch of aluminum or a 1/8 steel enclosure. Double shielding in thin steel cans can work too.

See https://www.diyaudio.com/forums/ana...ra-low-noise-mc-head-amp-172.html#post5977292

Yes, I do have other options in REW. I will do that. However I was checking the LNA gain by feeding in known peak to peak input voltages and left the RTA display as V so that it matched what my test equipment was reading directly. (Just an easy double check.) I was generating known peak to peak voltages, feeding them to a 60 dB attenuator and then into the 60 dB LNA.

I just built the LNA so I am doing various tests on it.

Perhaps I can try some sort of double enclosure. Perhaps I can put the LNA first into an aluminum project case (die cast?) then then put that die cast case into a larger steel chassis.