Ok.
Let me try to express this properly.
It's a bit complicated and multi-faceted.
At this point I'm casting about for a new DAC.
I've been using a PCM63 based dac with a x-coupled Jfet/Mosfet buffer for years - it sounds very good. I listen via a very clean signal path, only a passive discrete"L" attenuator between it and the amp(s). Speakers can be Quad 57, Acoustat III (modified), wide range horns (~250Hz --->~14kHz) with high bass driver + BEAR Labs Quadripole subs (using Crown MacroTech 2400). Amps usually are my Symphony No.1 amp or BEAR Labs SE Mosfet amps (with the Quads mostly). Friends bring tube gear over, so I have that reference point. Easy to hear what "guest" amps, or other gear is doing... very. A rather relaxed and natural sound, not "hot", not "bright".
So, I'm looking at the Soekris 1102 thread, back when it started. Thinking, hmmm, interesting. Forgot about it. Until a few days ago. Then I realized that I had missed everything, and that there were now assembled boards for it. Also had come across the Schiit stuff, and some others that seem innovative.
I also see that Soekris is making a DAC product complete as well, same basic topology, all-in-one with a discrete output buffer (surprising).
I'm wanting to get a handle on both the sound and if it is worth the effort to DIY.
Frankly reading quite a lot of the thread(s) it seems like the Soekris DIY build is not simple and is complicated by things like thumps, outboard uproc controllers, displays, upgraded power supplies, etc, etc. That makes it a BIG project costing maybe 2-3x the board itself. That puts the hardware cost in range of some commercial DACs, before considering the time it might take to build it up.
Then, I'm thinking maybe I can slam the 1201 into a box with bare bones supply run it "naked" and see... but what will that sound like compared to the fully decked out builds?? How much difference? Any?
And sonics, what's the diff between V1,2 and now V4??
Does V4 not need the muting circuit??
But then too, I have no way that I know of to get an audition with the commercial DACs either.
Or the Soekris for that matter!
The other thing that I think I noticed is that the absolute values of the specs and the look of the FFTs are not utterly spectacular at all?? Seems to me that I recall seeing an order of magnitude lower THD values from previous generation chip DACs?? Am I imagining this?
So, I am hoping for a discussion regarding all of these factors. Especially, IF you're running a Soekris DAC, I'd really appreciate hearing your sonic impressions - hopefully in the context of your system, listening preferences/experiences, and possibly comparisons with other units.
Anyone have a built up unit they might wish to sell or ship for an audition?
(I'm in upstate NY, just in case you are not that far away (NYC <---> Boston)
Looking forward to hearing more...
_-_-
Let me try to express this properly.
It's a bit complicated and multi-faceted.
At this point I'm casting about for a new DAC.
I've been using a PCM63 based dac with a x-coupled Jfet/Mosfet buffer for years - it sounds very good. I listen via a very clean signal path, only a passive discrete"L" attenuator between it and the amp(s). Speakers can be Quad 57, Acoustat III (modified), wide range horns (~250Hz --->~14kHz) with high bass driver + BEAR Labs Quadripole subs (using Crown MacroTech 2400). Amps usually are my Symphony No.1 amp or BEAR Labs SE Mosfet amps (with the Quads mostly). Friends bring tube gear over, so I have that reference point. Easy to hear what "guest" amps, or other gear is doing... very. A rather relaxed and natural sound, not "hot", not "bright".
So, I'm looking at the Soekris 1102 thread, back when it started. Thinking, hmmm, interesting. Forgot about it. Until a few days ago. Then I realized that I had missed everything, and that there were now assembled boards for it. Also had come across the Schiit stuff, and some others that seem innovative.
I also see that Soekris is making a DAC product complete as well, same basic topology, all-in-one with a discrete output buffer (surprising).
I'm wanting to get a handle on both the sound and if it is worth the effort to DIY.
Frankly reading quite a lot of the thread(s) it seems like the Soekris DIY build is not simple and is complicated by things like thumps, outboard uproc controllers, displays, upgraded power supplies, etc, etc. That makes it a BIG project costing maybe 2-3x the board itself. That puts the hardware cost in range of some commercial DACs, before considering the time it might take to build it up.
Then, I'm thinking maybe I can slam the 1201 into a box with bare bones supply run it "naked" and see... but what will that sound like compared to the fully decked out builds?? How much difference? Any?
And sonics, what's the diff between V1,2 and now V4??
Does V4 not need the muting circuit??
But then too, I have no way that I know of to get an audition with the commercial DACs either.
Or the Soekris for that matter!The other thing that I think I noticed is that the absolute values of the specs and the look of the FFTs are not utterly spectacular at all?? Seems to me that I recall seeing an order of magnitude lower THD values from previous generation chip DACs?? Am I imagining this?
So, I am hoping for a discussion regarding all of these factors. Especially, IF you're running a Soekris DAC, I'd really appreciate hearing your sonic impressions - hopefully in the context of your system, listening preferences/experiences, and possibly comparisons with other units.
Anyone have a built up unit they might wish to sell or ship for an audition?
(I'm in upstate NY, just in case you are not that far away (NYC <---> Boston)
Looking forward to hearing more...
_-_-
Not sure it's likely to cost that much. Most of the major modding was done on v1 boards and incorporated into v2 and 3. I run v3 stock minus the opamps and it sounds fantastic. I don't recall much talk of upgraded power supplies being necessary, unless you get into battery-powered mods. Maybe you mean increased capacitance - but that has been addressed in the newer versions.
You just need a basic DC power supply, parts for input choice (caps, connectors and resistors, etc are cheap), and optional volume pot.
Most complicated part IMO is the upgrading firmware/filters. Thank goodness randy has such great directions.
As for the sound of the soekris, I would describe it as refined, relaxed, open and natural in its presentation. It REALLY needs warm up time, so best to keep it powered. I have a 2.1 channel setup with diy MTMs and Wiener TPA amp. I've been playing with sources and have heard the soekris via usb to spdif, an sd card player, and now a RedNet 3 (ethernet).
I'm curious to know more about your current PCM63-based DAC. For example is it being used NOS or in conjunction with a digital filter? Is there passive or active I/V?
I have yet to see any technical justification for why a discrete R2R DAC should beat an IC. IC DAC designers are in general far more knowledgeable about their art than the designers of the discrete DACs, who tend to rely, at least in part on appeals to use of boutique components (Vishay resistors in the case of the TotalDac). Resistor matching would seem to be much better when all the resistors are co-located (as on a die).
I have yet to see any technical justification for why a discrete R2R DAC should beat an IC. IC DAC designers are in general far more knowledgeable about their art than the designers of the discrete DACs, who tend to rely, at least in part on appeals to use of boutique components (Vishay resistors in the case of the TotalDac). Resistor matching would seem to be much better when all the resistors are co-located (as on a die).
sure, it's old school, with the standard PCM-63 and the Burr Brown filter chip that was supplied at the time, going to an I/V opamp converter then to that discrete buffer. Crystal 8414 (?) input receiver.
I'm in agreement that the laser trimmed resistors on a chip are doubtless almost perfect. But I think the justification for the R-2R in the Soekris design - from what I have gathered trying to read through some of the threads - is in terms of the 28bits of depth and the flexibility of being able to download various filters in firmware. Oh, and the ability to accept a variety of input types and formats...
There are other designs that use exotic AD chips (which one was that??) two of them with some sort of "glue" to make them extend to the extra bit depth...
Regardless, personally I only care about the resulting sound.
And here, part of my reservations come because there are multiple "mods" that allegedly "improve" the sound. A double edged sword. On one hand, IF the stock unit is on par with or maybe better in some areas and not in others with my old PCM63 dac, THEN I'm interested a whole lot more than if the stock unit is less good to my ears, and then I have to try to do "stuff" to bring it up, and hopefully beyond my old DAC.
But where does it stack up compared to current commercial DAC offerings - keeping in mind that I've not been impressed in the past by almost all commercial DAC offerings. In that I would tired quickly of their "sound".
I am guessing that NOS is best with "hi-res" source, maybe the highest res??
_-_-
I'm in agreement that the laser trimmed resistors on a chip are doubtless almost perfect. But I think the justification for the R-2R in the Soekris design - from what I have gathered trying to read through some of the threads - is in terms of the 28bits of depth and the flexibility of being able to download various filters in firmware. Oh, and the ability to accept a variety of input types and formats...
There are other designs that use exotic AD chips (which one was that??) two of them with some sort of "glue" to make them extend to the extra bit depth...
Regardless, personally I only care about the resulting sound.
And here, part of my reservations come because there are multiple "mods" that allegedly "improve" the sound. A double edged sword. On one hand, IF the stock unit is on par with or maybe better in some areas and not in others with my old PCM63 dac, THEN I'm interested a whole lot more than if the stock unit is less good to my ears, and then I have to try to do "stuff" to bring it up, and hopefully beyond my old DAC.
But where does it stack up compared to current commercial DAC offerings - keeping in mind that I've not been impressed in the past by almost all commercial DAC offerings. In that I would tired quickly of their "sound".
I am guessing that NOS is best with "hi-res" source, maybe the highest res??
_-_-
There are areas of weakness that you can address - deleting the BB filter chip and the opamp-as-I/V which will enhance your listening experience, based on my own experience with DACs over the past few years.
Audio-gd has recently released a 'NOS' version of its multibit DACs which is getting subjective praise over on Head-Fi. The chips used are close cousins (PCM1704 or 1702, not exactly sure) of your PCM63. Indeed Mike Moffat of Schitt I understand considers PCM63 to be the pinnacle of the multibits, subjectively. So I for one wouldn't ditch the '63 in favour of a discrete R2R until I'd been convinced it would be superior.
Audio-gd has recently released a 'NOS' version of its multibit DACs which is getting subjective praise over on Head-Fi. The chips used are close cousins (PCM1704 or 1702, not exactly sure) of your PCM63. Indeed Mike Moffat of Schitt I understand considers PCM63 to be the pinnacle of the multibits, subjectively. So I for one wouldn't ditch the '63 in favour of a discrete R2R until I'd been convinced it would be superior.
I agreed on that, unfortunately I don't have chip making equipment and a laser trimmer, and all that do have discontinued making those R-2R DAC chips as they're not cost effective anymore at the low quantities nowadays....
Randy,
Great topic that is well aligned with things I've been thinking about lately. Over the years, I've built lots of DACs - TDA1541a non-oversampling, ES9018 based on the Buffalo III kit, inexpensive ES9023 (Subbu/JP based), and an old PCM63 DAC with asynch sample rate converter, NPC5842 digital filter, AD797 IV, and JFET/MOSFET discrete output buffer. Surprisingly, for redbook, nothing beats the PCM63 DAC. So I'm wondering if any of the newer DACs out there can knock the PCM63 off of its perch. I'll be watching this thread with great interest.
---Gary
As long as Søren is looking on, may I ask a few questions?
Briefly, in terms of the practical changes what has changed V1 --> V4?
Or another way, why would one want a V4 if one had a V2?
Are there "mods" to the circuit that were suggested that became incorporated, that if one had the earlier version one would likely want to do?
Briefly, in terms of the practical changes what has changed V1 --> V4?
Or another way, why would one want a V4 if one had a V2?
Are there "mods" to the circuit that were suggested that became incorporated, that if one had the earlier version one would likely want to do?
Digital filtering is best done on a PC which has vastly more computing resources than a FPGA or other DSP chip. Also, the extra 4-bits don't do any good unless the preceding 24 are perfect. AFAIK, the Soekris R2R does not use a single, precision current source but each segment of the resistor network is powered by the output of a different CMOS latch chip.
If you want 28 bits with laser trimmed resistors, look at the high sample rate PCM project on my blog.
Yes, the whole point of doing my discrete R-2R design is to get an updated version as development of R-2R chips for audio stopped 20 years ago. With support for higher clocks, more bits, DSD and programmable filters.
Yes, you can find old stock, or you can use expensive industrial DAC chips, or you can parallel multiple slower chips. But with discrete 0.01% smd resistors now priced more reasonable than they used to it's much more cost effective to do a design like mine, especially when using modern automated manufacturing.
I'm interested in being able to play redbook CDs and have the ability to play ripped files at hi-res from a computer.
Given that all the CMOS chips are powered from the same source, I'm not clear if in practice there is any non-linearity introduced beyond the tolerance already present in the resistors. Or if that actually matter much in terms of what is heard.
These technical points are interesting, but only to the extent that there is a bearing on the resulting performance and so to the way the dac sounds.
I'm mostly interested in what is being heard and in what system context.
Given that all the CMOS chips are powered from the same source, I'm not clear if in practice there is any non-linearity introduced beyond the tolerance already present in the resistors. Or if that actually matter much in terms of what is heard.
These technical points are interesting, but only to the extent that there is a bearing on the resulting performance and so to the way the dac sounds.
I'm mostly interested in what is being heard and in what system context.
None that I'm aware of, you'll just have to search the main thread. Once you do, you'll see that most of the major mods you think might be needed are no longer necessary.
Sounds to me like you should build one and compare yourself. Given its popularity you'll have no problem reselling it if you are not satisfied.
One problem with asking this question about DIY stuff in general, and this DAC in particular, is that everyone is going to have a different build. So when someone makes a comment about the DAC (good or bad), is it due to a fundamental feature/problem with the unit itself, or the user's particular build and/or any attempted mods?
And I think out of the box the Soekris DAC has immense flexibility, so even if you take modding out of the picture, simple build variation is huge. You can direct-connect an appropriate transformer, get a cheap I2S or or SPDIF device off ebay and be done. Or you can get fancy with ultra-regulated power supplies and super-duper inputs, or even parallel two boards as I'm attempting to do... and don't forget the custom filters!
That said: here's a brief writeup I did on a different DAC (tda1387 x8). I'm still not done with my Soekris DAC (meaning I didn't spend as much listening time with the initial Soekris build as the tda1387x8), but with a little salt, the comments I made about the tda1387 x8 also apply to the Soekris. My initial dam1021 build was fairly simple: v3 board (no mods), Amanero USB intput, R-core transformer, DIYINHK +/-12V regulator.
I've already acquired a second dam1021 board and am (slowly) working on a dual mono balanced build.
And I think out of the box the Soekris DAC has immense flexibility, so even if you take modding out of the picture, simple build variation is huge. You can direct-connect an appropriate transformer, get a cheap I2S or or SPDIF device off ebay and be done. Or you can get fancy with ultra-regulated power supplies and super-duper inputs, or even parallel two boards as I'm attempting to do... and don't forget the custom filters!
That said: here's a brief writeup I did on a different DAC (tda1387 x8). I'm still not done with my Soekris DAC (meaning I didn't spend as much listening time with the initial Soekris build as the tda1387x8), but with a little salt, the comments I made about the tda1387 x8 also apply to the Soekris. My initial dam1021 build was fairly simple: v3 board (no mods), Amanero USB intput, R-core transformer, DIYINHK +/-12V regulator.
I've already acquired a second dam1021 board and am (slowly) working on a dual mono balanced build.
You might give some consideration to having two DACs, each optimized for those tasks. AMR has applied this within their digital products, there are two DACs within - one optimized for RBCD and the other for 'everything else'.
There's at least some anecdotal evidence that linearity in the classic (DNL, INL) sense doesn't matter as much as is generally supposed. For example the TDA1543 has desperately poor linearity but is relatively highly rated subjectively.
It seems from my own research that there's a blind spot shared amongst DAC developers, no-one that I've seen has given much importance to dynamic performance. Not even Mike Moffat who's arguing strongly that a DAC requires excellent INL and DNL (his flagship using an AD device which is probably the best out their in these terms) has talked about dynamic performance other than 'a deglitcher is required' with that DAC. The popularity of NOS DACs to me indicates dynamic performance is an issue, the importance of which isn't regarded as highly as it should be.
dam1021 v1 - The first one, really need the vref mod
dam1021 v2 - vref fixed, improved power on/off mute
dam1021 v3 - custom resistors, more vref capacitors
dam1021 v4 - new output buffers
v2 might improve with a little more vref capacitance, otherwise no mods are needed on v3 & v4. You should of course always use latest firmware.
Ok, thanks Søren. Good info.
Why the change in the buffers??
-------------------
As I said at the top I am more interested in what the sound "is".
Time, time, time is not on my side.
So, committing to a build has to be a seriously considered and a pretty certain and clear path.
I'd prefer to buy someone else's if only to see what it is/was, at least to get started.
have yet to see one for sale!
gave a PM to someone who has a post up now, but they seem to have not checked back in, it's been a week or more.
The key is to make some sort of determination, which is the aim of this thread anyhow.
There seems to be an awful lot of raspberry/arduino control work, input boards, muting boards (not needed any longer???), power supply mods. No desire at this point to need to become proficient with the uproc boards! Too steep of a learning curve just to make a DAC do things...
Here's an example - from another forum's post regarding his build and you can see the comments and their significance (yes I gather the shift register PS issue is more or less solved now).
"- Shift reg power bypasses (added rather ocd level of capacitance for one per each shift reg + some additional ceramics). This should be sort of fixed on the newer boards, still more further bypasses usually help (confirmed in some posts by fellow DIYA modders). This one gave most impact on dynamics and 'clean slam', background feels darker - 10 to 15% better
- 1x Salas Ref-D powers clean side of USB-I2S Better imaging and better defined transients - 2% better
- (1x Salas Ref-D powers onboard osc)* Better imaging and better defined transients - 1% better
- Salas BiB does bipolar main dc rails at 7 V or something Smoother, cleaner, more 'refined' - 5 to 10% better
- Removed main psu side from the board (not needed now) Not much difference in sound
- Removed output buffers, can't say if it made a difference, but I'm not using those
- Latest Filters by Spzzzkt Stock filter should not be used at all, it's lo fi. "
Sure wish those folks who have built them up would show up and say what they found and what they think...
Why the change in the buffers??
-------------------
As I said at the top I am more interested in what the sound "is".
Time, time, time is not on my side.
So, committing to a build has to be a seriously considered and a pretty certain and clear path.
I'd prefer to buy someone else's if only to see what it is/was, at least to get started.
have yet to see one for sale!
gave a PM to someone who has a post up now, but they seem to have not checked back in, it's been a week or more.
The key is to make some sort of determination, which is the aim of this thread anyhow.
There seems to be an awful lot of raspberry/arduino control work, input boards, muting boards (not needed any longer???), power supply mods. No desire at this point to need to become proficient with the uproc boards! Too steep of a learning curve just to make a DAC do things...
Here's an example - from another forum's post regarding his build and you can see the comments and their significance (yes I gather the shift register PS issue is more or less solved now).
"- Shift reg power bypasses (added rather ocd level of capacitance for one per each shift reg + some additional ceramics). This should be sort of fixed on the newer boards, still more further bypasses usually help (confirmed in some posts by fellow DIYA modders). This one gave most impact on dynamics and 'clean slam', background feels darker - 10 to 15% better
- 1x Salas Ref-D powers clean side of USB-I2S Better imaging and better defined transients - 2% better
- (1x Salas Ref-D powers onboard osc)* Better imaging and better defined transients - 1% better
- Salas BiB does bipolar main dc rails at 7 V or something Smoother, cleaner, more 'refined' - 5 to 10% better
- Removed main psu side from the board (not needed now) Not much difference in sound
- Removed output buffers, can't say if it made a difference, but I'm not using those
- Latest Filters by Spzzzkt Stock filter should not be used at all, it's lo fi. "
Sure wish those folks who have built them up would show up and say what they found and what they think...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.