Hi IVX,
Interesting project. However, I think you will have some real competition in that sort of form factor pretty soon (at around $150). Don't know for sure what they will be doing for stable power, but I would expect lots of filtering and probably some super capacitors will be used rather than going to SMPS, since it is very hard to make SMPS not hurt sound quality without some pretty sophisticated engineering and a fair amount of multilayer board space.
Also, not sure what you mean by a 'serious audio performer'? In what sort of price range do you think the dac you are working on would compete?
Might you have any interest in a rather high performance ES9038Q2M dac, too?
Interesting project. However, I think you will have some real competition in that sort of form factor pretty soon (at around $150). Don't know for sure what they will be doing for stable power, but I would expect lots of filtering and probably some super capacitors will be used rather than going to SMPS, since it is very hard to make SMPS not hurt sound quality without some pretty sophisticated engineering and a fair amount of multilayer board space.
Also, not sure what you mean by a 'serious audio performer'? In what sort of price range do you think the dac you are working on would compete?
Might you have any interest in a rather high performance ES9038Q2M dac, too?
Last edited:
Oh, agreed on that. I have a Benchmark DAC-3 here that uses SMPS and it sounds fine. However, most of the dacs on the market with SMPS I have seen have very audible switching noise on the output. Allo tried with their Katana dac, but couldn't make the SMPS inaudible no matter what lengths they went to. It is intended to be pretty serious dac, too. With linear supplies and enough film caps on the +-15v rails it sounds easily like it belongs in the $1,000+ sound quality class. Better than a lot of them up there until one gets a lot closer to $2,000.
Dual Mono ES9038Q2M XMOS DSD DXD 768khz USB Audio DAC with Bit-perfect volume control - DIYINHK
What you guys think about that website ?
Lots of interesting products (with a slightly premium pricing).
What you guys think about that website ?
Lots of interesting products (with a slightly premium pricing).
Many or most around here seem to know about diyinhk. I actually have one of their almost bare ES9028PRO boards here to play around with sometime: ES9028PRO ES9018 32bit Audio DAC PCB - DIYINHK
The output stage is textbook ESS. Strangely, they show resistor values for voltage dividers that would make Vref closer to AVCC/3 rather than the AVCC/2 it should be. Don't know what they were thinking.
There is no documentation for the board I have other than their recommended component values that are printed on it. The board only came with the ES9028PRO and the 1.2v VDD regulator, both of which were already installed (I removed the 1.2v regulator and replaced it with 1.3v as recommended by ESS for high sample rate operation). Diyink are not very good about answering questions regarding their products, IMHO, so I would only recommend the diy items to people that could reverse engineer them if necessary.
Last edited:
Ohh, sorry, I made mistake about opamp I gonna use there, it is ad8397. Quite an expensive beast, 2pcs cost as single SE9038Q2M i.e. $10. CM6642 $2, TPS7A8101 $.7, STM8S003F3 $.2, ok, $2 for all surrounding parts and connectors + PCB + $.5 for assembling . Total: about $26 for PCBa i.e. electronics BOM. I guess $60-70 retail(including mechanical, extrusion tolling etc) range would be ok for >1KU. I've no plan for max-performance DAC so far, just feel no challenge there(it is a trivial task if not limited size, price - no novelty just replicate the app-note), and I guess the market is full of it.
Thank you
So, for a "sure shot" DIY quality boards, what are my options ?
1. Soekris's R-R DAC
2. JL Sounds DAC
3. Abraxalito's Ling DAC
4. ????
Any recommendations ?
Hardly the case of no challenge to making high end dacs, but true the markets are full of app note designs. The app notes are only a starting point for really good dacs. It definitely gets audibly better. One obvious example is Benchmark DAC-3. Stereophile rates it as A+ recommended and SOA in terms of measurements. I assure you it is not app note dac, no way to get to that level of sound quality otherwise. In fact, I doubt ESS knows how to make their own dacs as good the best designers can do with ESS chips. Same type of situation for AKM.
I would suggest seeing if you can get a single board from someone who buys a bunch of Abraxalito's latest dac in the ling dac thread. They are cheap, Abraxalito says he thinks they sound as good as dacs probably costing a few hundred dollars, and you could get some experience under your belt for cheap.
Regarding JL Sounds, I have never seen one up close much less heard one, so can't say. They look reasonably competent from the webpage, but they should given the prices. Don't know exactly how good they are though without seeing and listening.
Soekris has its enthusiasts, but again I have not heard one. If one wants to go the discrete R-2R route I guess the prices are right, and there appear to be some happy customers. So long as you don't have to solder things or otherwise do work you are not already skilled it maybe they could be okay. Probably better to ask in the threads for that though.
They measure well for the price. Don't know much more than that. Would be skeptical the SMPS is totally inaudible until I heard it for myself, but would be happy to say so if they succeeded.
What I was trying to say in the above was that the evaluation board dac designs by ESS, and what they show in the app notes they have published, are useful information and lead to pretty good sounding dac board designs, but they are not sufficient to produce the best sounding dac boards that can be designed using ESS chips. To do the latter requires solving some additional technical problems, and different designers have taken different approaches to accomplish that.
It’s recommended that the multiple-quote function be used whenever possible.
Not sure it is clear to everyone how it is supposed to work. Is it assumed to be self-explanatory?
I see. That kind assumes I know what I want to respond to in advance. Often it takes all thinking I can muster to compose and edit a response to one post. Trying to do two or three at once would probably require working offline in a word processor, then combining responses together to make a nice looking multi-reply post. That's one reason why professional writers have editors, IMHO.
If it ends up being easy to read that means I did a good job of writing and editing, but not that it was really as off-the-cuff as it may look.
As an aside, it can get pretty challenging as it is to work in the narrow reply window we are given to use. I know it can be lengthened and I know how to do that, but it would probably help if it were wider too.
EDIT: Just figured out how to stretch it to the right.
If it ends up being easy to read that means I did a good job of writing and editing, but not that it was really as off-the-cuff as it may look.
As an aside, it can get pretty challenging as it is to work in the narrow reply window we are given to use. I know it can be lengthened and I know how to do that, but it would probably help if it were wider too.
EDIT: Just figured out how to stretch it to the right.
Last edited:
Clock Divider Update
Before moving along to trying the LTC6954 clock divider, decided to try adjusting delay timing with the LMK01000 divider. The two clock divider chips are rather different in terms of how they implement delays. For LTC6954, delays are determined by counting edges of the the input clock. LMK01000, on the other hand, apparently has some way of implementing delays in steps of 150ps without any reference to an external clock. Therefore I assume it must have something like a delay line inside it to provide timing for the delay increments.
After rewriting the LMK01000 software to make it easy for my to quickly switch between delay values, it turns out that the clock divider output delay matters a lot for ES9038Q2M sound quality when DSD DPLL = 1. The difference in sound quality, in particular in terms of what sounds like harmonic distortion, varies a lot from worse to best. The best sound quality as a function of delay is presumably about the same as what a carefully designed reclocker might be expected to do if located between AK4137 output and ES9038Q2M input. At the moment I am using the maximum delay setting of 15, which I sounded the best to me after a short trial of several delay values. I will need to go through it more carefully to verify which setting sounds best, and also to take a look with the scope to see how DSD clock timing is phased relative to MCLK.
Overall, I think this is an important result to be aware of. More importantly to me, I am liking the sound a lot. Some more listening tests over the next few days will be needed for a final subjective evaluation of what has been gained, but I am feeling pretty optimistic about it for now.
EDIT: As usual, after making adjustments for best sound quality with DPLL =1, subjective sound quality may be a little smoother and less distorted with DPLL =2. There may be a very slight, very nearly inaudible loss of timing accuracy with that adjustment, but seems like most people might prefer it.
Before moving along to trying the LTC6954 clock divider, decided to try adjusting delay timing with the LMK01000 divider. The two clock divider chips are rather different in terms of how they implement delays. For LTC6954, delays are determined by counting edges of the the input clock. LMK01000, on the other hand, apparently has some way of implementing delays in steps of 150ps without any reference to an external clock. Therefore I assume it must have something like a delay line inside it to provide timing for the delay increments.
After rewriting the LMK01000 software to make it easy for my to quickly switch between delay values, it turns out that the clock divider output delay matters a lot for ES9038Q2M sound quality when DSD DPLL = 1. The difference in sound quality, in particular in terms of what sounds like harmonic distortion, varies a lot from worse to best. The best sound quality as a function of delay is presumably about the same as what a carefully designed reclocker might be expected to do if located between AK4137 output and ES9038Q2M input. At the moment I am using the maximum delay setting of 15, which I sounded the best to me after a short trial of several delay values. I will need to go through it more carefully to verify which setting sounds best, and also to take a look with the scope to see how DSD clock timing is phased relative to MCLK.
Overall, I think this is an important result to be aware of. More importantly to me, I am liking the sound a lot. Some more listening tests over the next few days will be needed for a final subjective evaluation of what has been gained, but I am feeling pretty optimistic about it for now.
EDIT: As usual, after making adjustments for best sound quality with DPLL =1, subjective sound quality may be a little smoother and less distorted with DPLL =2. There may be a very slight, very nearly inaudible loss of timing accuracy with that adjustment, but seems like most people might prefer it.
Last edited:
WRT delay, very interesting. This DPLL is very sensitive. It must have a minimal memory buffer for the slow CF it can go down to.
I think you should try and disable it altogether and run the 4137 / DAC synchronously. This is absolutely possible as I know of others that have done it.
However, then you will not have any jitter reduction going on in the Sabre DAC.
If you want absolutely best possible (lowest) jitter, run a JLsounds USB bridge and clock everything off it synchronously.
End of jitter problems. They are isolated, reclocked and have provision for VHQ clocks. This should theoretically disable -both- DPLL's, the AK4137 AND Sabra DAC.
Terry
I think you should try and disable it altogether and run the 4137 / DAC synchronously. This is absolutely possible as I know of others that have done it.
However, then you will not have any jitter reduction going on in the Sabre DAC.
If you want absolutely best possible (lowest) jitter, run a JLsounds USB bridge and clock everything off it synchronously.
End of jitter problems. They are isolated, reclocked and have provision for VHQ clocks. This should theoretically disable -both- DPLL's, the AK4137 AND Sabra DAC.
Terry
Terry,
There is always some jitter, as there is no such thing as a perfect clock. What you are suggesting is not possible for what I want to do, but I know how that can sound. I have an Allo Katana here which basically works the way you are describing. Allo says it has the lowest jitter ever measured at ASR, a bit lower than even DAC-3. I can already do that well in terms of sound quality, I am trying to do better. It gets really tricky down at the low levels of jitter we are talking about, but the effects are still audible, just like other forms of distortion are still audible. I'm simply trying to figure out what the best I can do is. Give it some time and let me keep at it a bit longer, and we'll see how it turns out. I haven't given up yet, but I know am getting pretty close to as good as I can do this way.
There is always some jitter, as there is no such thing as a perfect clock. What you are suggesting is not possible for what I want to do, but I know how that can sound. I have an Allo Katana here which basically works the way you are describing. Allo says it has the lowest jitter ever measured at ASR, a bit lower than even DAC-3. I can already do that well in terms of sound quality, I am trying to do better. It gets really tricky down at the low levels of jitter we are talking about, but the effects are still audible, just like other forms of distortion are still audible. I'm simply trying to figure out what the best I can do is. Give it some time and let me keep at it a bit longer, and we'll see how it turns out. I haven't given up yet, but I know am getting pretty close to as good as I can do this way.
SYNCHRONOUS CLOCK - the magic words!
A couple of hours ago I have listened music for the the first time from my last dual ES9018K2M DAC.
On short:
- a STM32 controller read F0-F3 pins from XMOS USB board (I built it on the Amanero like PCB with the same pinout), so it writes the right values in Si549 registry for 98.xxxx/90.xxxx frequencies for DAC's MCLK
- this frequency is divided by 2 and sent to the XMOS board via 150MHz isolator; it is necessary to be divided by 2 because 50MHz is the maximum recommended clock input for XMOS.
- the same divided by 2 frequency is used for a rudimentary re-clock for the I2S lines who come out from XMOS via 150MHz isolator; 6 pieces of 74AUP1G79 chips are used for this purpose (3 for DAC left chip and 3 for the other).
- nothing to say about the IV converter and the output filter; they are as simple as the version ESS recommended in their reference design document.
- it is probably important to mention that I have paid a very big attention to ALL power supplies, including +/-15VCC for AD797s.
Well, I am impressed!
I hear the sound clear, detailed, not so smooth as on AKM DAC, but very precise. Every time I build a new DAC, I listen some Nightwish songs because they have some passages which the audio equipment could easily mix them and the instruments sound. I can say now that this K2M DAC is the best Nightwish player I have heard.
pcmDPLL and dsdDPLL are set to 1 or even to OFF, but I cannot hear any difference between these settings (on the ebay 9USD K2M board, dsdDPLL cannot be smaller than 7 because the sound stops at least once in a second).
A couple of hours ago I have listened music for the the first time from my last dual ES9018K2M DAC.
On short:
- a STM32 controller read F0-F3 pins from XMOS USB board (I built it on the Amanero like PCB with the same pinout), so it writes the right values in Si549 registry for 98.xxxx/90.xxxx frequencies for DAC's MCLK
- this frequency is divided by 2 and sent to the XMOS board via 150MHz isolator; it is necessary to be divided by 2 because 50MHz is the maximum recommended clock input for XMOS.
- the same divided by 2 frequency is used for a rudimentary re-clock for the I2S lines who come out from XMOS via 150MHz isolator; 6 pieces of 74AUP1G79 chips are used for this purpose (3 for DAC left chip and 3 for the other).
- nothing to say about the IV converter and the output filter; they are as simple as the version ESS recommended in their reference design document.
- it is probably important to mention that I have paid a very big attention to ALL power supplies, including +/-15VCC for AD797s.
Well, I am impressed!
I hear the sound clear, detailed, not so smooth as on AKM DAC, but very precise. Every time I build a new DAC, I listen some Nightwish songs because they have some passages which the audio equipment could easily mix them and the instruments sound. I can say now that this K2M DAC is the best Nightwish player I have heard.
pcmDPLL and dsdDPLL are set to 1 or even to OFF, but I cannot hear any difference between these settings (on the ebay 9USD K2M board, dsdDPLL cannot be smaller than 7 because the sound stops at least once in a second).
This makes perfect theoretical sense and everyone who tries it agrees, it is a step up in sound.
My only recommendation is to replace the Si clock, it is not super low jitter and configure ES DAC to run of 45 / 49MHz clock.
WRT DPLL, yes, I believe it drops out (isn't used) when running synchronous.
T
Last edited: