No! We got the ESS board some weeks ago, beacuse we wanted to try it.
If we were to decide, we would go for the CS4398, as we use it in our Reference DAC. Just seems somebody has a problem with just about any CS chip. And for some reason, no matter what Wolfson, AKM or other chip used, everybody seems happy no matter how bad implemented.
To me, this seems more like religion than real listening based.
Trust me... You can talk bad about the CS chips from until X-mas..... As long as nothing will outperform our DAC, it will stay in! Do feel free to come by and listen to it!
Last edited:
I had a chance to listen to your DAC.... 😱 Bit somehow it did not find it's way into my setup. Our reference DAC did.
The idea of this project is, to make something different from the rest. Something where everything is choosen by listening tests, and not by reading datasheets.
Here is a review of the reference. I would like other reviews to better place this "Reference DAC". Maybe comparisons with chinese 200$ DAC's based on the same chip.
Making a great DIY DAC at <200$ is a serious task and a laudable goal!
Keep up the good work and try to make everybody happy!
Cheers,
Nic
Hi,
Cannot get the drawing done on the quick, I need to re-design several commercial PCB's, including one with a 256-pin TQFP IC this week.
Have a look at this:
http://www.borbelyaudio.com/adobe/ae699bor.pdf
Figure 17 pretty well illustrates what I am proposing. There are simple ways to remove the coupling capacitors however as usually some form of Servo is needed to deal with circuit DC operating Drift I find that using a REALLY HIGH QUALITY (e.g. Audyn KP-SN or Mundorf M-Cap Zn) coupling capacitor sounds better than using Servos.
With Figure 17 as shown, the gain is 26.5 for balanced signals and 13.25 for unbalanced signals.
The Sabre DAC has around 4mA p-p current per output, so with 8 outputs parallel 32mA. We want 5.656V p-p output after a gain of 13, so at the input to the Borbeley line amp we want 0.427V p-p, so we need I/U conversion resistors of 13.34 Ohm for eaqch polarity.
I would suggest connecting 27 Ohm (nearest preferred value to 13.34R *2) between the two outputs with no reference to "ground" or "Vref". If we want an around 100KHz lowpass we connect a good quality (Wima FKP, polystyrene, silver mica) 47nF capacitor in parallel with the resistor.
Anyway, that would be my take on a nice non-tube analogue stage for the Sabre. And I would likely put a WM8805 upfront and include the MCU to control the Sabre DAC (volume control is pretty good on this one) and the WM8805 9176.4KHz & input selection).
The result would likely be the most desirable DIY DAC going right now. Cost would likely push to 300 Bucks, but I think it would be well worth it.
Ciao T
Cannot get the drawing done on the quick, I need to re-design several commercial PCB's, including one with a 256-pin TQFP IC this week.
Have a look at this:
http://www.borbelyaudio.com/adobe/ae699bor.pdf
Figure 17 pretty well illustrates what I am proposing. There are simple ways to remove the coupling capacitors however as usually some form of Servo is needed to deal with circuit DC operating Drift I find that using a REALLY HIGH QUALITY (e.g. Audyn KP-SN or Mundorf M-Cap Zn) coupling capacitor sounds better than using Servos.
With Figure 17 as shown, the gain is 26.5 for balanced signals and 13.25 for unbalanced signals.
The Sabre DAC has around 4mA p-p current per output, so with 8 outputs parallel 32mA. We want 5.656V p-p output after a gain of 13, so at the input to the Borbeley line amp we want 0.427V p-p, so we need I/U conversion resistors of 13.34 Ohm for eaqch polarity.
I would suggest connecting 27 Ohm (nearest preferred value to 13.34R *2) between the two outputs with no reference to "ground" or "Vref". If we want an around 100KHz lowpass we connect a good quality (Wima FKP, polystyrene, silver mica) 47nF capacitor in parallel with the resistor.
Anyway, that would be my take on a nice non-tube analogue stage for the Sabre. And I would likely put a WM8805 upfront and include the MCU to control the Sabre DAC (volume control is pretty good on this one) and the WM8805 9176.4KHz & input selection).
The result would likely be the most desirable DIY DAC going right now. Cost would likely push to 300 Bucks, but I think it would be well worth it.
Ciao T
Hi,
Funny, I would have said this instead of the blind support of CS Chips in general. I have listened to them and many others (I should, after all I have serious influence on what ends up in commercial high end gear) and find the results pretty much the worst among the available choices. Their technical data is also much worse than what one could consider "par for the course", not that I am particularly concerned about technical data.
Well, your call.
From where I stand I still find nothing better for CD-replay (strictly CD though) than a well implemented TDA1541, so WTFDIK? The Sabre was rather close though.
Ciao T
Funny, I would have said this instead of the blind support of CS Chips in general. I have listened to them and many others (I should, after all I have serious influence on what ends up in commercial high end gear) and find the results pretty much the worst among the available choices. Their technical data is also much worse than what one could consider "par for the course", not that I am particularly concerned about technical data.
Well, your call.
From where I stand I still find nothing better for CD-replay (strictly CD though) than a well implemented TDA1541, so WTFDIK? The Sabre was rather close though.
Ciao T
I'd bet once you disable the sabre's ESRC and rely on the wolfson receiver instead (that has no elastic buffer inside, despite the marketing fluff), you suddenly burn away a big enough chunk of DNR ( 15+, 20+? ) that digital volume control becomes less and less desirable ...Why would you choose ESS then?
Anyway, I saw here people using digital volume control and nos tda1543 , so it goes..
Anyway, I saw here people using digital volume control and nos tda1543 , so it goes..
Last edited:
Hi T,
Long time... Nice to see you still like to hang out.
You mention programmable oscillators. I have been involved in a project, where we do a 3Gbs SerDes link, and here we have used a SI570, from Silicon Labs. Not to expensive, but arguably not the cheapest, and it needs MCU programming. Any thoughts on that?
Cheers,
mkc
Long time... Nice to see you still like to hang out.
You mention programmable oscillators. I have been involved in a project, where we do a 3Gbs SerDes link, and here we have used a SI570, from Silicon Labs. Not to expensive, but arguably not the cheapest, and it needs MCU programming. Any thoughts on that?
Cheers,
mkc
Welcome back Nicmac.
If you want to do any further posts in this thread, then let it be of technical value to everybody.
For your info this project is to be done in the way most high end design is done, which means a lot of input is considered, and the the most attractive solution are chosen, when done the whole thing is build, and if it works finito.
The old project, which you happen to dislike so much without ever hearing it, was build in a completely different way which of course never was intended in this project.
This new project is for everyone finding pleasure in building something for themselves with properties mostly found in pretty expensive gear, but with both personal influence on solutions and budget.
My original sugestions for this project from the beginning was CS8416, CS8421, AKM 4399 or ESS sabre DAC and then an op-amp based I/V and analog stage, all put together, and if it would work finito.
Now the project is open again, due to different sugestions from others, yours could be welcome to.
But still building this new project is on a completely different level than the old one was.
The earlier one was for own use primarily, this one is for everyone else.
Well it is now easy for me to see, that your sound preferences must be completely different from ours.
We will not start using FETs in any linelevel design.
We also do not want coupling caps in any part of the design.
These points are crusial to us, no FETs, no caps where avoidable, and analog decoupling caps are to be PFS placed close to the supplied circuitry.
Servo´s could be onepoint types.
Originally we thought of a high quality op-amp, maybe succeded by a NFB, DC coupled bipolar buffer, but discrete design is OK with us, as long as it is NFB, bipolar and DC coupled.
We could use the earlier design without the shunts, maybe even in a differential configuration if wanted - God knows why.
Only things to be considered, are gain and the need for postfiltration.
Hi Bluesky
I know that most vendors recommend both ceramics and tantalums for decoupling purposes almost anywhere throughout their application notes.
The best thing to do, is to try to listen to their evaluation boards, how does their own implementation sound.
In our design we did start out with ceramics, and ended up using them for digital circuitry only. They are the best in that position IMHO.
For analog design they probably are the worst I´ve been able to find. We also tried tantalums, but they really screwed up the whole design. I think they might work with up-amps or so, which is not that critical to PSU, but in discrete design with almost no PSRR, you have to use low esr types with as linear impedance as possible, or at best completely avoid caps.
We chose the latter, because finding good caps is not possible, they are all very bad, but some worse than others, IMHO PFS does the least harm whereever they are used.
But in digital circuitry there is no way around ceramics these days.
Regarding the Crystal DAC, this need decoupling several places, both VD, VA, FILT+, VQ, REF GND and VREF, we did do as sugested by Crystal (electrolytics and ceramics), but with very disapointing results. Then gradually we tried to change the ceramics at VA, FILT+ and a few places more as I recal, and voila, it suddenly began to sound like something.
Later on we changed the PSU´s to discrete fast ones also, but especially FILT+ is very important for this particular chip.
Also the recievers analog supply needs care, the gain you get is both cheap and natural.
Changing the postfilter caps from ceramics to film though may be the biggest improvement regarding caps.
Once you hear what ceramics do to sound, you´ll be very carefull where you put them.
Last edited:
Hi,
I did not say disable the ASRC, merely feed it a low jitter signal. This goes BTW for any ASRC, if you have to use one.
The Sabre Architecture is sufficiently unique to mean it is best used as designed, but it will still sound better fed less jitter.
As for the WM Receivers, I do not care exactly what they do inside, whatever they do, it does result in dramatic jitter attenuation when compared to the "market leading" Cirrus Logic parts.
Ciao T
I did not say disable the ASRC, merely feed it a low jitter signal. This goes BTW for any ASRC, if you have to use one.
The Sabre Architecture is sufficiently unique to mean it is best used as designed, but it will still sound better fed less jitter.
As for the WM Receivers, I do not care exactly what they do inside, whatever they do, it does result in dramatic jitter attenuation when compared to the "market leading" Cirrus Logic parts.
Ciao T
Hi Kurt,
The funny thing is that "decoupling caps" are also coupling caps. And the cap's in the DC Servo to keep your output at 0V DC are in effect AMPLIFIED coupling caps.
Of course, next thing you tell me that you also believe in "ground", just as you believe that "coupling caps" only are in line with a signal line...
There are reasons why I like J-Fets, but the same circuit almost to the dot can be made using BJT's. Both J-Fets and BJT's have drawbacks and advantages. In this case I'm certainly leaning towards J-Fets.
Most modern DAC's BTW manage with minimal post filtering, then again, I use non-oversampling DAC's with minimal post filtering and it works fine too.
Ciao T
The funny thing is that "decoupling caps" are also coupling caps. And the cap's in the DC Servo to keep your output at 0V DC are in effect AMPLIFIED coupling caps.
Of course, next thing you tell me that you also believe in "ground", just as you believe that "coupling caps" only are in line with a signal line...
There are reasons why I like J-Fets, but the same circuit almost to the dot can be made using BJT's. Both J-Fets and BJT's have drawbacks and advantages. In this case I'm certainly leaning towards J-Fets.
Most modern DAC's BTW manage with minimal post filtering, then again, I use non-oversampling DAC's with minimal post filtering and it works fine too.
Ciao T
That explains a lot, and I guess you will newer end up with this DAC. You see...: We stared our last DAC project, in pure desperation, over the fact that we could not seem to find any commercial DAC, that really do perform.
As I wrote earlier, a Mark Levinson owner tested our first prototype, and was ready to change his ML DAC right away...
Even though I am not an ML fan, that must count for something 😉
I feel like asking you this: Do you agree, that making "State Of Art Audio Gear", is really about synergys through the whole product, by tuning each and every part with each and every other??
Or do you believe, that it is sufficient to take the best parts, and then expect the best result??
Maybe it would be an idea to use CS8416 followed by CS8406 and then by TI or WM recievers and then ASRC, DAC?
The CS is known to lock on almost anything with a pulse😀, which TI is not.
Then best of both worlds is incorporated.
Believe me, I know the impact of caps where ever they are placed, thus my discontend with them.
I won´t tell you anything about ground.
There are sonic reasons why I dislike FETs of any kind in my signal path, and we cannot introduce them in this design.
NOS DACs is out of the scope for this project, so I will not comment on that.
This project is basicly to be non idiosyncratic in all stages.
Hello all,
Very nice initiative, lots to learn about! Also, welcome back Mr. Loesch, been missing your posts 🙂
Regarding the selection of the DAC chip, although I feel this is preety much already decided upon, you can check the page below. Is the only DAC comparison I know of in a relatively controlled enviroment.
AD, BB, CS, Wolfson and AKM parts are included.
LINK
Very nice initiative, lots to learn about! Also, welcome back Mr. Loesch, been missing your posts 🙂
Regarding the selection of the DAC chip, although I feel this is preety much already decided upon, you can check the page below. Is the only DAC comparison I know of in a relatively controlled enviroment.
AD, BB, CS, Wolfson and AKM parts are included.
LINK
@ TorstenL:
I just discussed with KvK, and we would like to invite you to a DAC listening session here i DK. It's not that long a trip from London, and flight ticket are really cheap!
If you do design High End audio gear, I am sure that visiting KvK will be of great interest to you. Most people are amazed!
Ready or not???
I just discussed with KvK, and we would like to invite you to a DAC listening session here i DK. It's not that long a trip from London, and flight ticket are really cheap!
If you do design High End audio gear, I am sure that visiting KvK will be of great interest to you. Most people are amazed!
Ready or not???
Hi,
Well, given that you are aiming that low, I could counter that the commercial DAC (which does not use CS4398 I may add) that I helped to equip with a WM8804 receiver has actually since replaced in the real world with cash on the barrel head several ML DAC's.
BTW, this DAC also uses pretty generic Op-Amps and 3-Pin regulators and I personally think it sounds okayish, but far from great.
Neither, actually.
You present the Heglian Angel Thesis and his Oponent the Daemon Antithesis as if there was any need to choose.
I find true state of the art comes from combining the two into synthesis. I like to start from the best fundamental (unchangeble) items, such as the basic design, choice of actives etc. and then follow up by tuning things until I get the best synergy and synthesis from the best.
The results tend to speak for themselves.
Of course you are not going to believe this, as I prefer to use Tubes and other design features you seem to dislike intensely.
Ciao T
PS, thanks for the offer to visit, I'll take it up when I am back in the UK/Europe, currently I am in Hong Kong and will be for a while longer, I do not like travel as much as I did when younger.
Well, given that you are aiming that low, I could counter that the commercial DAC (which does not use CS4398 I may add) that I helped to equip with a WM8804 receiver has actually since replaced in the real world with cash on the barrel head several ML DAC's.
BTW, this DAC also uses pretty generic Op-Amps and 3-Pin regulators and I personally think it sounds okayish, but far from great.
Neither, actually.
You present the Heglian Angel Thesis and his Oponent the Daemon Antithesis as if there was any need to choose.
I find true state of the art comes from combining the two into synthesis. I like to start from the best fundamental (unchangeble) items, such as the basic design, choice of actives etc. and then follow up by tuning things until I get the best synergy and synthesis from the best.
The results tend to speak for themselves.
Of course you are not going to believe this, as I prefer to use Tubes and other design features you seem to dislike intensely.
Ciao T
PS, thanks for the offer to visit, I'll take it up when I am back in the UK/Europe, currently I am in Hong Kong and will be for a while longer, I do not like travel as much as I did when younger.
- Status
- Not open for further replies.