Soren, I hope you don't take my comments the wrong way. Without your expertise and experience in the field, Soekris DACs would never have existed or sounded as good as it does. This I'm sure you know.
No one knows everything about audio. But time gives us the chance to iterate and prototype. Maybe in 5 years we'll all have access to TotalDAC's potentially near-perfect DAC technology at the 1k euro price point!
P.S. I still have a feeling that you might find my comments offensive even if they are not actually. To clarify, I think Vref can use more work not because I'm sure of the effect, or why. Consider it an intuition that we need to do absolutely everything we can on Vref because it matters and we don't know how. At the same time, perfection doesn't seem so impossible with the availability of shift register buffers. Perhaps at the cost of oversampling capabilities, but there might be negligible difference between a equally good NOS and OS DAC, and our theories do seem to suggest so. Sales will keep going up, Soren.
No one knows everything about audio. But time gives us the chance to iterate and prototype. Maybe in 5 years we'll all have access to TotalDAC's potentially near-perfect DAC technology at the 1k euro price point!
P.S. I still have a feeling that you might find my comments offensive even if they are not actually. To clarify, I think Vref can use more work not because I'm sure of the effect, or why. Consider it an intuition that we need to do absolutely everything we can on Vref because it matters and we don't know how. At the same time, perfection doesn't seem so impossible with the availability of shift register buffers. Perhaps at the cost of oversampling capabilities, but there might be negligible difference between a equally good NOS and OS DAC, and our theories do seem to suggest so.
Sales will keep going up, Soren.
Last edited:
And hey, it even makes intuitive sense that Vref ripples screw with timbre, which is quite often just ripples of sorts, right? This theory would also suggest that you were totally right about small signals being important, even if they're not perfectly monotonic. Also probably right about delta-sigma being pretty much hopeless. We have some very smart researchers and engineers, but you can't artificially recreate music with all the tricks in the world when you don't know what to optimize for. Our intelligence is proven by our THD, but the path was hopeless to begin with, at least for quality music uses. Perhaps delta-sigma would fare better with our medical devices and military equipment; I don't think they have timbre problems.
This theory would also suggest that you were totally right about small signals being important, even if they're not perfectly monotonic. Also probably right about delta-sigma being pretty much hopeless. We have some very smart researchers and engineers, but you can't artificially recreate music with all the tricks in the world when you don't know what to optimize for. Our intelligence is proven by our THD, but the path was hopeless to begin with, at least for quality music uses. Perhaps delta-sigma would fare better with our medical devices and military equipment; I don't think they have timbre problems.
Last edited:
Your train of thoughtsis overloading me.
Sorry that part didn’t turn out to be practically useful.. (at least not yet
)All hope now is on Soren and those of us who can figure out the currently possible best tweaks to dam1021/1121!
Only because of the vultures in the distribution channels. But where is the fun in buying anyway?
I am more inclined to see you design your own dac.
As for getting the TD performance for less than 1k...dunno, do not underestimate the effect of the z-foils, those are indeed miracle parts. Besides, you already have a TD clone on the cheap and with the added bonus of bespoke filtering.
I still cannot see how clocking is done in the TD. Any luck with that looking at photos?
This is the whole point of being able to precisely adjust the offset for each bit so it matches perfectly, its not crazy out of the question as its just a firmware release. I have been messing with VHBL and MiSTer lately and to see these guys recreate in near perfection or perfection simulation of those old games, exactly how much harder would it be to replicate the FPGA function in Soren's dac if one were to design an ASIC to do the same?
I say this because the amount of reverse engineering required to add features one will take that further when they finally become tired of waiting on Soren and attempt it themselves and that will be the point where the product becomes cloned en-masse. Ignoring users on requested features is how things often get started. I doubt people are wanting things for free and if Soren needs some up front development funding for new features in his products than that's basically paying for what can make his dac's in the league of TotalDAC and that is only doing him favours.... the reality is its an FPGA, it's *not* merely an "old" product that simply no longer deserves dev time because of its age.
I don't understand why people are complaining I don't add features in firmware that require hardware that just aint there....
the dam1021 don't have hardware for synchronous clocking, and it don't have hardware for calibration, neither resistors nor timing....
If you're all that smart, go design a DAC with all those smart features you think you need, then start manufacturing and sell them at a price point where everybody want one....
the dam1021 don't have hardware for synchronous clocking, and it don't have hardware for calibration, neither resistors nor timing....
If you're all that smart, go design a DAC with all those smart features you think you need, then start manufacturing and sell them at a price point where everybody want one....
I have to agree, this discussion is a little bit from the"greedy boyz" point of view. Please remember the price region we are talking about, a Dam1121 is something like 250 bucks now. Add supply and case, you may end up at 500.
And now the total dac, which is... 20000?
So, gentlemen, get a grip. There is always room for improvement, in everything, but the question is, does it make sense.
And yes, ceramic caps, Vref and all that is part of the fun, it is DIY after all. I also wish some possibility to connect a shunt reg to Vref without major surgery, but OK.
If Soeren would put out a total Dac, in 3 weeks we would talk improvements here no doubt
And now the total dac, which is... 20000?
So, gentlemen, get a grip. There is always room for improvement, in everything, but the question is, does it make sense.
And yes, ceramic caps, Vref and all that is part of the fun, it is DIY after all. I also wish some possibility to connect a shunt reg to Vref without major surgery, but OK.
If Soeren would put out a total Dac, in 3 weeks we would talk improvements here no doubt
What about resistor calibration? We’re talking about using external ADC. I did not mention clocking at all because of its limited impact. I might try record the outputs to see how much delay there actually is in dual-mono since it may be more of an empirical question, but probably not anytime soon since it really doesn’t matter a ton for music uses.
I never dismissed your engineering contributions. It’s not trivial at all to design a product as performant and elegant as dam1021/dac1541, let alone all the logistics to get custom resistors trimmed. If I were more electronically-inclined (and have 40 years of experience in designing circuitry
), I think it would be really cool to make a sota DAC and democratize it. You can probably take over the entire market if you can make TotalDAC at $500 and provide a good enough theory on why it might be superior. There’s no reason that Hi-Fi needs to be more extravagant and elusive than everything else. I have to agree, this discussion is a little bit from the"greedy boyz" point of view. Please remember the price region we are talking about, a Dam1121 is something like 250 bucks now. Add supply and case, you may end up at 500.
And now the total dac, which is... 20000?
So, gentlemen, get a grip. There is always room for improvement, in everything, but the question is, does it make sense.
And yes, ceramic caps, Vref and all that is part of the fun, it is DIY after all. I also wish some possibility to connect a shunt reg to Vref without major surgery, but OK.
If Soeren would put out a total Dac, in 3 weeks we would talk improvements here no doubt
Total DAC is $4k euro. Another thing we can do is bypass 5V LDO on dam1021 and basically make the power supply identical to OEM products except for digital analog separation, which we probably can’t hear anyways. Dimdim and Salas found that a great shunt supply sounded better than linear supply. There’s probably some physical reality to it, even though the power goes through a 5V LDO and op-amp regulator.
TotalDAC is probably perfect given our current technology. Seriously what more do you need when you have a shunt supply like UltraBiB feeding the vref directly? The impedance is on the order of uR, the shift registers we have put us at 300mR+. And note that we’re hearing great changes just by adding caps. I would say resistor correction could be very impactful, contrary to what we previously believed, zfoil or not. So if TotalDAC hasn’t tried correction or doesn’t have the FPGA processing power to do it, I would at least suggest that they try... after that, it’s probably as perfect as it can ever get. The question would be to drive down cost and increase production scale, like every other consumer electronic. Allow me the indulgences of imagination
For a great engineer like Soren, maybe not that much
Talking about reverse engineering might be a bit offensive... part of the reason Soren reveals so little is to prevent something like that. The best patent is secrecy.
Anyway, I’m not a professional engineer and I’m perfectly happy to be of help to someone who can turn ideas into reality, even if they make all the money
P.S. Soren, I think you might want to work with Salas at some point if it is shown empirically that shunt supplies are better. TotalDAC had a university professor/lecturer who helped identify the problem and probably also with circuitry. I wouldn’t be surprised if Salas has even more experience. Numerous people benefitted from his shunt reg designs at virtually no cost. I really respect him and it might be a good thing to have some collaboration eventually. With efforts on an international scale in this project, it would make an even cooler story than Totaldac’s. Is that not worth trying for?
Leaving aside where Soekris might take us in the long term, from our DIY perspective, we should probably check a few things still:
(1) Vref circuitry: whether the low-res mod is beneficial both in measurement and listening. Whether the new Soekris vref circuit that seems to have replaced resistors with capacitors can be installed on rev4 and earlier, and whether it makes a difference.
(2) Power circuitry: check if removing 5V LDO and powering the 4V op-amp regulators directly with UltraBiB (or other good power supply) would help. If we get this far, it would be useful to do a quick comparison against 1121 to see if the shift register model does matter. Soren knows the theoretical predictions but we need experimental result to prove that such an improvement in Vref does matter in sound.
(3) Resistor correction: if Soren can give us a cost estimate and we manage to crowdfund him for this feature, we could then test a range of methods/algorithms to do the correction accurately and efficiently. I’m very inclined to believe that this would help.
There’s nothing we can do about jitter and the shift register used in dam1021, though it’s probably not that big of a deal. However, there are still things that we can try and if we succeed, I think it’s likely that most of us would find these mods to be more practical than scrapping our builds and buying a new set of 1941, which might have its own problems and is much harder to mod.
Our experiments would help Soren a long way in determining if our hypothesis is correct that better Vref equals much better sound. If so, he can have more confidence in setting up the next prototypes.
Please do try it. And you'll find lots of support on the UltraBiB thread, and from Salas himself when needed.Ignored the other gems but this one is just irresistible
-------------
Btw, one thing I forgot to mention is that there is probably also an audibility limit to timbre-related improvements in the time-domain. Now it's rather clear that we're far from it nor do we know what the limit is, since some polymer caps already made huge differences it seems. But it is only reasonable to believe that there is a limit just like THD. The TotalDAC approach, if combined with a good shunt regulator like UltraBiB but one that can be made to work in 4V (or whatever other Vref voltage that is preferable), seems to suggest perfection. It is likely that it might be overkill, and that at some point in our exploration we would find it more reasonable to stop and save costs. One such improvement that I can think of is trace width, which was among the improvements in the newer TotalDAC products. The last few uR may be less important than a smaller PCB footprint, and if Soren can find that perfect balance, then Soekris will be much more competitive than TotalDAC.
Soren, I think maybe the resistor compensation deserves much more attention than it has received so far. Since we're already saving a ton on Z-foils, I have a feeling that there is a lot to be gained from some decent compensation. The investment will also generate direct returns on all product lines.
Also, would you be able to share what the benefits of rev.5 op-amp Vref circuits are compared to rev.4? Could it be similar to choosing an optimal resistor value based on the capacitance in the system? Thanks so much for your insights.
Please do try it. And you'll find lots of support on the UltraBiB thread, and from Salas himself when needed.
.
I must have a total of at least 20 Salas and other shunt regs in various spots, both low and high voltage, so thanks for the invitation
So, nothing against the shunts, just admiring your use of language: a counterpoint between shunt and linear, priceless.
Also, if Soren does fail to live up to expectations, either because he doesn't agree with our speculations or because he doesn't have time or some other reason, I would love to see a community effort on developing such an ultimate DAC. I mean no disrespect to Soren, but I do understand that not every engineering company has to be interested in something that "fanatics" here are.
We have a lot of talented and interested professionals on this forum, it probably wouldn't be that hard to get a circuit design to get things started. Stocking and manufacturing might be harder since it's usually a centralized effort. If Soren already has customized resistors, maybe it wouldn't be impossible to find the same manufacturer and ask for the same stock. Heck, with resources from all over the world we might even be able to do things cheaper. Initial development can be crowdfunded, and the final product can be fully democratized, sold almost at cost perhaps with some compensation to the main contributing engineers if necessary. Probably need someone more familiar with the forum to take leadership roles, but it seems completely within the realm of possibility, and the potential benefits to the entire audio community could be immense. If our theories are proven to be somewhat accurate, it can also be a huge step to restoring sanity to the audiophile community. It's about time we take a stand against $20k femto-clocks that in almost all likelihood won't make a single audible difference in the sound.
It really is possible, but it's completely unnecessary at this point it seems...
I must have a total of at least 20 Salas and other shunt regs in various spots, both low and high voltage, so thanks for the invitation
So, nothing against the shunts, just admiring your use of language: a counterpoint between shunt and linear, priceless.
Sorry I meant series? I'm a cs person, so don't judge me for it
One last thought on the possible "community effort". Such a project invariably involves great uncertainties and unforeseeable challenges. Capitalism might have a special system of motivation to encourage individuals to engage in such efforts. If we are to take it outside of strictly capitalistic context, we need leaders who are not only technically competent to shoulder the technical challenges, but also very interested and strong-willed to handle the risks. Additionally, they must be specially situated to find such a project the best use of their time. It captures the imagination but to go through with it is non-trivial. However, the benefits are great. Some open-source projects are simply irreplaceable.
One more general comment. Soekris is very good for the price as it is, especially if we consider the possibility that it might be more transparent and natural than all delta-sigma systems. But the reason I've been so excited is that it seems there are only a few engineering hurdles to overcome in order to achieve possibly a perfect audio DAC. What's more exciting is that the solution doesn't seem to require any exotic or expensive components, only the careful application of quality parts. I believe that it's possible to manufacture such a DAC, especially for Soren, at not much more than the cost of dac1421. There will be a huge market for such a high fidelity device; even a "Beats Pro" sells for $400. If we're successful, it might mean that there will be no more need for various kinds of DACs from all sorts of vendors whose trustworthiness is more than questionable. At the same time, this will also drastically reduce the confusions in Hi-Fi, as well as the obsession that often naturally follows due to a number of reasons. In short, this could be a huge engineering achievement. If my speculations are not completely mistaken, I don't believe that my excitement is excessive.
I'm not an electrical engineer so I won't be able to play nearly as much of a role as others on this forum, and, without a doubt, Soren. Division of labor suggests that I should focus on my projects and studies, but I sincerely hope that those in a better position to take action will recognize the incredible opportunities that they have before themselves.
Calvin,
We can agree that the best Sigma-Delta DAC's are not bad at all, but I'm in pursuit of the absolute best, like so many other here
What happened to you over the last four years Soren? I can agree that "perfect" audio isn't always a worthwhile goal, but when it's easily within reach while still being very low-cost? I have to say I am very disappointed. All you do these days is get defensive at even the most well-meaning questions and requests. Maybe that's the new business strategy, but I would urge you to reconsider. You know you can do better than that.
Also I read that androa_mori is planning on releasing a discrete dac design. I usually prefer to support the pioneers, as I believe many do. But you would probably do well to keep your customers informed on your plans and progress, especially when the customers have been trying rather hard to help you develop your products.
I think I'm starting to share spikestabber's sentiments more and more. There is something very disagreeable going on here. Maybe it's miscommunication, and I can accept it if you acknowledge the problem. I'm not even personally invested in this situation since my initial build goals have been met and I'll probably continue to use my current configuration for quite a while, with no foreseeable need for upgrade or a second device. But I do care whether an idea that I believe is very promising will be implemented properly, especially when it might produce quite some benefits to many people. I guess it does matter quite a lot who we think deserves to be the leader in this field, and help make it happen by financial support and the less tangible sort including constructive feedback and spreading the word. As it is, I'm becoming inclined to change my vote. I hope I'm being reasonable and that my thoughts might also be of use to others.
On a slightly different note, just realized that TotalDAC isn't using sign-magnitude. So I'm inclined to believe that a proper update to the existing Soekris design might actually outclass TotalDAC while remaining a low-cost solution. The only semi-serious technical hurdle is getting the two Vref rails to track while maintaining the low output impedance of a shunt regulator. Whether it should be +/- input switched on a single ladder or passed directly to two ladders as it is now, or single rail input and switched somehow on a single ladder, I don't know. But it doesn't sound impossible. Maybe there's a way to track the two rails of a shunt regulator with some extra design? If only Salas can chime in here... So the real question, as I see it, is who will be leading this effort to build an affordable DAC that can very likely outclass TotalDAC. Would it be Soren or someone else? I'm curious to find out.
Seeing as the last few pages are full of back to back posts, could you have condensed your thoughts before clicking the reply button? It may be wise to stand back and deliberate before submitting a response.
Also, if you want to sanction development of a community dac, it may be better to start a thread dedicated to it rather than posting in a vendor thread.
Also, if you want to sanction development of a community dac, it may be better to start a thread dedicated to it rather than posting in a vendor thread.
Yes, yes, please! Perhaps a kind moderator can separate all the hundreds of pointless posts into their own dedicated thread. If financial contributions (a bribe) are needed i'd be the first to chip in
Why, you did most of the work, people just want you to continue to improve upon it, why should anyone else reap rewards after you put so much effort into it already? You talk about hardware limitations all the time yet not once have you ever talked about upgrading the dam architecture, larger FPGA, buffered signals to the resistors, etc. It is like you have no interest in taking it to the next level while still talking about hardware limitations... you can't explain why adding sufficiently large capacitors around the whole shift register network results in bass that will make anyone blush vs the face in the mud of the stock unit however you take care of all of that and then some in the 19xx...hmm. People just see potential and are voicing that. The FPGA is tiny and you've used the same one since 2014 and already we've seen that larger filters ~1m taps improve overall performance significantly. You could have a field day with a SoC based FPGA and rid the stupid microcontroller and USB interface entirely, and have full network audio support. The DE-10 Nano is dirt cheap for the power that FPGA has (110,000 logic elements) with an ARM bolted to the side of it, can't imagine the bare chips (5CSEBA6U23I7) being too cost prohibited for the quality such a product can bring, and certainly not $20,000 TotalDAC worth anyhow.... icing on the cake is the Linux running HPS can run anything the user wants.
Last edited:
Do tell me what if someone designs an R2-R interface board to the DE-10 that rivals the TotalDAC.... your project is as good as dead. I would certainly want that someone to be you if anyone due to what you have already achieved.
Such a board would have a voltage controlled femto clock generator (reclocking) and discrete buffer from shift register to the R-2R ladder, the difficult part for anyone would be the software and cooking up the required VHDL/Verilog to make it all work.
The DE-10 also has a 12bit 8 channel(!) ADC which can be very useful for realtime feedback adjustments to reduce clipping among other things that don't need full accuracy but nonetheless able to verify the output is within specs, among other monitoring such as the clock generator and vref voltage (might as well make vref range user adjustable and fully monitored too while we are at it, accuracy can be further had from decent large digit DMM where exact results are fed into the software)
Such a board would have a voltage controlled femto clock generator (reclocking) and discrete buffer from shift register to the R-2R ladder, the difficult part for anyone would be the software and cooking up the required VHDL/Verilog to make it all work.
The DE-10 also has a 12bit 8 channel(!) ADC which can be very useful for realtime feedback adjustments to reduce clipping among other things that don't need full accuracy but nonetheless able to verify the output is within specs, among other monitoring such as the clock generator and vref voltage (might as well make vref range user adjustable and fully monitored too while we are at it, accuracy can be further had from decent large digit DMM where exact results are fed into the software)
Guys, please stop offensive communication towards Soekris!
In my opinion, the Soekris DAM DAC offers a very respectable audio performance in an affordable way. And Soekris has spent a significant of time supporting people and answering questions on this thread. Yes, there are also some (minor) features I would like to be added to the DAM DAC. It would be nice if the platform keeps being maintained and improved, but we cannot insist on that. Note also, that we are dealing with a small company here. And it seems to be almost impossible to keep everybody happy here. If no further performance gains are achieved, people are unhappy. But if improvements are made, people will call the older revisions "broken" and demand upgrades. Oh well...
Fedde
In my opinion, the Soekris DAM DAC offers a very respectable audio performance in an affordable way. And Soekris has spent a significant of time supporting people and answering questions on this thread. Yes, there are also some (minor) features I would like to be added to the DAM DAC. It would be nice if the platform keeps being maintained and improved, but we cannot insist on that. Note also, that we are dealing with a small company here. And it seems to be almost impossible to keep everybody happy here. If no further performance gains are achieved, people are unhappy. But if improvements are made, people will call the older revisions "broken" and demand upgrades. Oh well...
Fedde
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.