Thanks! will probably have to run the boards at ~-25db for headphones which seems to be okay...
Maybe due to the existence of distortion in the more significant MSB at 0db compared to, say, -48db? I just read earlier in this post that someone was worried and Soren never really explained (i think...) if and why it wouldn't matter, except saying that the -60db is better and -60db for other devices is usually worse... Not my area of expertise here so I'm just following others' lead
It will complicate things and "shift right" only work at -6, -12, -18 ... etc levels. And everything between needs to be done in an other way. Still, it would be a nice gesture
//
That makes sense
If what you said is true I would very much like to see it implemented! Or at least have Soren explain to us how to do it ourselvesBtw, do we have the extra 4bits (24db) in addition to the 24bits for digital volume attenuation? Either way, I guess Soren was absolutely right that at -24db we still have 20 (or 24) bits for signal processing, plenty for the human ear it appears: http://www.mojo-audio.com/blog/the-24bit-delusion/
Also.. for some reason I went to double check the resistor prices on digikey, if according to Soren the temp drift coefficient used is 5/10/25ppm for the 0.01/0.02/0.05% resistors, it might add up to $100 difference in resistor costs between 0.012 and 0.05 version of dam1021. Anyway, this would appear to be quite reasonable even at 100 profit for each board *2000 cts not including R&D costs. Just wanted to clarify/apologize...
Please feel free to correct me elsewhere too, as I know myself to be quite fallible...
Thanks!
Please feel free to correct me elsewhere too, as I know myself to be quite fallible...
Thanks!
------------- Build stuff -----------------
I finally got back to the build after a short New Years trip. Stacked the two dam1021's and paralleled the spdif/I2S inputs with extended (~21mm) headers. It couldn't get a lock at first, then I realized my homemade SPDIF input board is probably drawing too much current from the SK Lite 3.3v reg, so much as to cause the voltage to drop to 0.8v. I then removed the 3.3v power to the SPDIF board and the voltage came back to ~3.6v. I posted the schematics and pcb design of the SPDIF input board if anyone is interested in it. The board was designed to take 3.3v and regulate it onboard to 1.2v for the coax. The dimension is 50*25mm - a rather small footprint that was supposed to fit nicely in my chassis.
However, only the top board was able to get a lock in the first few minutes even after the power to SPDIF board was cut. I suppose it was because the S22 clone was still warming up and the voltage at the oscillator or whatever was causing the problem wasn't quite stable enough. After some time both boards were able to get signal lock without any problem.
My enclosure is 215*228*70mm so everything was really tight on paper. And it indeed took much thought just to fit everything together. Here's some pics to the setup. I'm using a 35w 12v toroidal and S22 clone (HP-X) at 9V DC output to power the 1021s. The USB board is an amanero clone. SK Lite is used for control and powering the ISO section (and the would-be SPDIF input board...) with an independent secondary.
Raw outputs are connected to 3-pin XLRs and RCAs just for convenience, and buffered outputs are wired to 6.3mm SE and 4-pin XLR on the front panel for headphone use. I had room on the back panel for two RS232 ports so I did, straight to the dam1021s - the most reliable and easiest way imo.
I had a small ground wire collector in my first build but there weren't really that many wires in the star ground (only 2 actually), so I got rid of it. Now the ground is just connected to the safety earth with a bridge rectifier in between.
I also attached my front and back panel drawings. I used fusion360 for the front panel but it seems STP format is more universal..
------------- Mod stuff ------------------
Many thanks to randytsuch and nige2000 for the additional reports on SQ gains with the battery mod. If only rev1 was tested to improve with battery supplies in vref, it might be explained away by the insufficient vref back then. I do have the ISO powered by an separate transformer and low noise regulator on the SK Lite so I guess I'm okay there.
Perhaps it's possible that the onboard regs and filters do indeed work well enough in the presence of "clean" DC supply, and that no additional mods are necessary or even could produce better subjective sound, as Soren seems to have always insisted. I'm not an EE person so I guess this is as much assertion as I could make with some confidence, and until Soren reveal more about the schematics or there is reliable information that a mod leads to better SQ, I'll just stick to the unmodded 1021s for now...
Much of the discussion on what to mod and why Soren did what he did seems to come down to the design goals and individual view on purchasing. Although Soren said outright that he is building the "best dac on the planet", such a claim cannot be interpreted to its extreme. In fact, I think maybe what he meant is something along the lines of "the best dac with a cost-effective approach in mind (as opposed to Totaldac/MSB's nonexistent budget
This is also reflected in Soren's design choices in his commercial products. Dam1541 is very compact, but at the cost of using switch-mode power supplies, which is not usually the to-go for high-end products. It corroborate the point that perhaps compactness is a meaningful dimension in the Soekris dacs, as well as cost-effectiveness as I'll discuss later. This effectively means that the performance will not be the "absolute" best, as I have admittedly sometimes suffered from imagining.
What most of us really care about is perhaps cost-effectiveness - based on our purchasing power and the goods available in the world, find a general threshold of cost-effectiveness, lower than which lies the numerous luxuries that we could not rationally purchase, as much as its positive but near non-existent marginal value to us maybe. We want to "eliminate the weakest link in the system", which is likely the most cost-effective approach, until perhaps it hits our budget, becomes too complex, time-consuming and tedious, or reaches a true engineering obstacle that cannot be solved with our current technology, no matter how much money we are willing to throw at it. The last item, I think, is the most interesting concept in this long section of my post. Although Soren did not explain his reasoning explicitly, there's another possible interpretation to his claim of "best dac on the planet" that might be meaningful to those of us pursuing the "cost-effective approach", which I did not take into account in my rebuttal of the "absolute best".
Again I’m not an insider in this area so please argue against if you think I’m wrong. The point here is that perhaps this discrete R-2R sign magnitude technology dac is most limited by how accurate and thermally stable we can make our resistors. In other words, without improvements in the resistors, anything else done to a rev4 board, and I’m speculating here, will have a much much lower cost-effectiveness – something that Soren is clearly against.
However, as I went back to the first pages of this thread, I became temporarily confused at how Soren seems to be aiming at a “low cost discrete dac” from the get-go. Soren did not provide the option of matching resistors or consider using better Z-foil resistors, when more accurate resistors is indeed the most important part of building a great discrete R-2R dac. After I slept on it, I realized the central position cost-effectiveness occupies in Soren’s designs, and perhaps should occupy in every good professional engineer. So far, my view has gone from brushing-off Soren’s pursuit of the best, to admitting that there maybe something in his claim in the insurmountable technical challenges that he may have faced. I now believe that Soren did indeed set out to build “the best”, but meaning something quite subtle – the best performance under a very specific cost-effectiveness threshold that he had in mind. Granted Totaldac probably beats the dam dac, but it has “the price of a car”. Granted good S-D dacs can be had for cheaper (or way more expensive…), but Soren believes that discrete R-2R tech is better than “any S-D dac”.
Therefore, the dam dac is supposed to hold a special place in the world of dacs in terms of cost-effectiveness – worse than the chips on your motherboard (not that it matters because we have a looser cost-effective threshold); worse than very cheap sigma-delta dacs (again, not that it matters because we have a looser cost-effective threshold); better than all S-D dacs priced higher than, say, 90% of the dam1021 solution, depending on how much you agree with Soren’s claim that R-2R when implemented well is superior to any D-S (I can at least agree that D-S made a lot of compromises to how the analog signal is produced, maybe too much to be considered a good candidate for the ultimate dac technology, as perfect as it might sound to the human ear.); better than all decent discrete R-2R dacs currently available because of its superior cost-effectiveness; and better than PCM1704 solutions because they no longer exist.
To me, the elegance in the engineering (R2R is a clear winner here) and objective measurements holds just as much if not more value than subjective enjoyment and the ability to be better immersed in the musical arts. I imagined earlier that the core r2r tech is the absolute best possible, which would hold much value for me as a source of inspiration to a perfectionist (I’d like to think of its positive connotations here
In conclusion, to consider the cost-effectiveness of an approach, we should perhaps first consider our own thresholds, then appraise the cost-effectiveness of the particular approach. In this case, rationally speaking, modding seems to be a way of saying that we have a looser constraint of cost-effectiveness than Soren, provided that he didn’t make any mistakes as appears to be the case in rev4. However, going back to how important the core R-2R part of the dac is, and how it consists of 216 0603 delicate resistors soldered in place already, perhaps modding will not be as useful as we hope it would be and we’ll just have to stick to Soren’s idea of the “most cost-effective approach”. That said, it is perfectly fine imo to mod just for learning or the fun of it.
Hope this is somewhat useful to other folks as it was to myself in clarifying how to mod and what cost-effectiveness means here. Matt_garman if you have read this far I wish you the best of luck in finding the perfect PSU that matches the dam1021 cost-effectiveness wise, and I’ll just happily stay with the S22 that you recommended to me as it seems to have good measurements, reviews and has a clone (HP-X by ZeroZone in China) that is just the right physical size for my current build.
---------------------- Sad stuff -------------------------
Unfortunately, in the process of moving the system, which didn’t make much use of connectors and has many soldered connections...., I broke the legs of a MOSFET on the PSU. When I tried to solder it back on, I must have burnt it as the PSU is sort of shorting and not working now… For many reasons I will have to hold off repair plans (as easy as it probably is..) until the end of year, but I’m glad to have gone this far. Many many thanks again to everyone here who helped clear things up for me and to those who contributed to this project, besides Soren of course, but he’s probably made enough money out of this “diy project” so I guess it’s fine to be a bit restrained in my gratitude
Also, I might not have as much free time as I’d like now to share my build in full detail which could be of help to those planning a build but don’t want to do everything all over again on his own. To be honest I don’t even know if my build is done right as everything is so tightly squeezed together as to perhaps cause EMI problems. Do I need to shield the toroidal transformers, or it is generally understood to be fine without, especially in a balanced setup? I would prefer to measure the EMI but good measurements can be prohibitively expensive…
Anyway, hope this helps!
P.S. The circle left open besides the OLED panel is supposed to have a black IR pass filter glued on top. Would've looked better if finished!
P.P.S The PcbDoc of the SPDIF board is 6mb and could not be uploaded, I'll just include a screenshot for now... would love to hear some feedbacks on it as it's not working well
I'd be worried about those transformers especially that one right beside where the digital signals go into the dam
The i2s lines are too long for my liking too between amanero and i2s in on dam would be more effective to use a usb extension
Doubt he's getting his resistors at digikey that's only for us Muppets
Parts in decent volume come at great discount
There's also around half the resistors on a rev 4 as a rev 1 that was less cost
Thanks for the feedback! I thought about extension cable too, when I was still worried about powering amanero properly... but it adds quite a bit of complexity to the build and take up a bit more space.
The current I2S cable is 15-20cm triple wires twisted and shielded, only the three data and clock signals are connected, gnd on amerno is used to ground the cable shield. Is the interference significant enough as to cause digital errors? Is there a way to detect subtle errors in the signal?... I’ve not had significant I2S signal locking problem besides the short moment after first dual-mono boot up
Transformer shielding wise, options might be fairly limited for me.. especially for the trafo that you said may be problematic... I found a 80*40 iron enclosure for the Talema toroidal that will likely work but for the blue trafo I’ll be limited to a small Mumetal or steel plate between the side of the trafo and dam1021. Would this plate be meaningful at all? Sometimes I wonder why transformer manufacturers can’t just shield at least some of their products for a small premium...
—-
I agree that Soren might have better sources for the components... but pricing based on Digikey 1000 reel costs seems reasonable to me. I initially thought the 0.01/0.02 resistors cost $30-40 on Digikey.... but you know, it would be great if Soren could lower the prices after the initial R&D costs are well compensated, even better if he could share with us his industry sourcing prices
I won’t be purchasing more boards at least until he releases his discrete design output stages, but hey, why not lower the costs for someone interested in a build if you can? It’s for diy anyways, and helping each other out is never a bad thing
Last edited:
The general rule of thumb for I2S lines is 10cm max. That doesn't mean longer than 10cm won't work, but you're "coloring outside the lines". In general, the shorter the better.
Transformer shielding wise, options might be fairly limited for me.. especially for the trafo that you said may be problematic... I found a 80*40 iron enclosure for the Talema toroidal that will likely work but for the blue trafo I’ll be limited to a small Mumetal or steel plate between the side of the trafo and dam1021. Would this plate be meaningful at all? Sometimes I wonder why transformer manufacturers can’t just shield at least some of their products for a small premium...
I'm all for small/compact builds, but if you can get a bigger chassis, that would be ideal. Besides shielding, distance and orientation can also help; a bigger chassis gets you more flexibility in both respects. Plus with a bigger chassis, you can make a simple shield via some angle iron or aluminum to corral the transformer(s) into their own little compartment.
What would be symptoms of I2S interference? I probably won’t be able to shorten the cable in this build as it’ll bring too many mechanical issues for me, and might have to live with that 7-8cm extra cable for now... one thing that makes me sleep better is that jitter on the isolated side won’t matter much here
I agree that my build is dangerously compact, but alas I went for it and it’ll be too cost prohibitive time-wise for me to rebuild any time soon. I’ll put my 35va trafo in a full shield no problem, but for the SK Lite PCB trafo the best I might be able to do is something like custom plates on the sides and top... is there a way for me to find out exactly how much trouble I’m facing EMI-wise without resorting to expensive instruments?
Thank!
But since the board has fifo reclocking and isolation, would that still be the case technically? If it still is and I have time later this year I’ll relocate the amanero...
Yes I agree that in an ideal free market prices will be determined by perceived utility on both sides, and things would most efficient. But there is no competitor here with even a similar product so us buyers will be at a great disadvantage - a reason for developers to keep proprietary tech proprietary or it ends up on sale on Taobao. Also, no one said high quality audio has to be less cost effective than it could be.
Surely one can say that the best is reserved only for those who can afford it, no matter how crazy the price. There’s something in that claim that I don’t like, but maybe that’s just what we have to live with given the best functioning economic and political systems that we have come up with.
And please don’t get me wrong, I think the price here is still far from exorbitant...