There, just told you how to make a swiss army knife and mother of all r-2r dac's, and I would be first in line to pay ~$1000 for a board.
Nobody being offensive, I like Soren I really do, but the fact that I thought of that within 5 minutes, someone is going to pull it off eventually. I already seen what is happening first hand and involved in the MiSTer project, I too was skeptical at first, but the fact that such a board can run what it does, 100% accurate SNES simulation on SuperFX GSU2 with full RGB analog output to a CRT straight from the FPGA, already this dam1021 looks small fry.
Nobody being offensive, I like Soren I really do, but the fact that I thought of that within 5 minutes, someone is going to pull it off eventually. I already seen what is happening first hand and involved in the MiSTer project, I too was skeptical at first, but the fact that such a board can run what it does, 100% accurate SNES simulation on SuperFX GSU2 with full RGB analog output to a CRT straight from the FPGA, already this dam1021 looks small fry.
I think I’ve said all I wanted. And perhaps you can at least realize that it’s never about what a couple of people might think of the improvements. Like I said I’m not about to scrap my build that’s about 90% perfect for something that’s 100%, which is also not going come into existence in the short-term. It’s not my obsession talking here. Maybe try and think bigger.
Also I am no stranger to Soekris products, we spent a fortune on crypto acceleration boards when we had a large project that required them for offload, long before standard SoC's and CPU's even had instruction sets to do so, I know what the man is capable of and I am forever greatful of his knowledge and what he has achieved.
I think from the start many people had unrealistic expectations due to the way the first dam was presented. I know I did. Let's be honest and let's be clear: the original dam was not a "reference dac" as stated in the title of this original thread. It was a test platform with the early adopters serving as beta testers performing free debugging and development. Much was promised at the get go but alas proper implementation didn't occur until several revisions and models down the road refined and improved the product until the present version(s). Make no mistake, there were flaws in the original model, as admitted by the designer.
Most proper companies either do their beta testing in house and only release a product as "reference quality" once all the bugs are out and it truly performs at a standard worthy of the claims made in the sales copy, or, they call it a "beta" version and make it clear upfront that the product is still in the developmental stages, and, rather than charge full pricing, offer some sort of deep discount or trade in credit to those willing to debug and improve the product. Sadly, this was not done.
Attempts to return the original flawed model were pushed away and told that this was a DIY product and the purchasers should take care of the issues themselves. For those accustomed to the DIY products of Hypex and others, this was a bitter pill to swallow. Personally, I had to go on a negative publicity campaign to highlight this unusual business practice before I received an offer to upgrade the 4 units I purchased to version 2. While appreciative, I wonder how many others were able to get equitable treatment. Seeing where the product has gone over the last few years and the changes and improvements that have been made, it has become all the more clear how far from a "reference" the initial version really was....
Call it semantics, a gross difference in expectations, "marketing", naivete, or shoddy/shady business practices, or whatever. The original unit is not bad, but no doubt for many who thought they were buying a fully developed, "reference quality dac", it has been bitter disappointment. I stopped reading this thread a few years ago as I became sick of reading of the new and improved revisions and new models coming out on a regular basis. I fully realize all products are improved and new models released but in comparison to other similar products, the Hypex nc400 for example, and the major changes made, it has become obvious that the original version was far short of what was promised.
I learned a valuable, though costly, lesson. For those calling for changes, upgrades, additional features, etc., I suggest either waiting a year or so to purchase as perhaps yet another upgrade, revision, or new product will be released which comes yet another step closer to the "reference quality dac" which was originally promised. Or, perhaps simply look elsewhere.
Most proper companies either do their beta testing in house and only release a product as "reference quality" once all the bugs are out and it truly performs at a standard worthy of the claims made in the sales copy, or, they call it a "beta" version and make it clear upfront that the product is still in the developmental stages, and, rather than charge full pricing, offer some sort of deep discount or trade in credit to those willing to debug and improve the product. Sadly, this was not done.
Attempts to return the original flawed model were pushed away and told that this was a DIY product and the purchasers should take care of the issues themselves. For those accustomed to the DIY products of Hypex and others, this was a bitter pill to swallow. Personally, I had to go on a negative publicity campaign to highlight this unusual business practice before I received an offer to upgrade the 4 units I purchased to version 2. While appreciative, I wonder how many others were able to get equitable treatment. Seeing where the product has gone over the last few years and the changes and improvements that have been made, it has become all the more clear how far from a "reference" the initial version really was....
Call it semantics, a gross difference in expectations, "marketing", naivete, or shoddy/shady business practices, or whatever. The original unit is not bad, but no doubt for many who thought they were buying a fully developed, "reference quality dac", it has been bitter disappointment. I stopped reading this thread a few years ago as I became sick of reading of the new and improved revisions and new models coming out on a regular basis. I fully realize all products are improved and new models released but in comparison to other similar products, the Hypex nc400 for example, and the major changes made, it has become obvious that the original version was far short of what was promised.
I learned a valuable, though costly, lesson. For those calling for changes, upgrades, additional features, etc., I suggest either waiting a year or so to purchase as perhaps yet another upgrade, revision, or new product will be released which comes yet another step closer to the "reference quality dac" which was originally promised. Or, perhaps simply look elsewhere.
Last edited:
So Soren has just one of those attributes, both or none?Cab’s comment is sure to receive some straight up denials from the man himself.
If only we had a designer with respectable character AND good technical judgements. That’s my opinion at least.
This discussion is pointless and now it gets personal, what’s up? Sell your car and get a totaldac.
There is no reason to project too much into the dam1021. It is what it is..and that is a fine dac...for 250€.
Give the man a break, please.
You need to stop assuming that it’s about how my DAC sounds or will sound. I said many times I’m perfectly happy with how dam1021 performs. You might find it hard to understand or impossible to accept but my position doesn’t preclude me from saying everything I did.
P.S. I consider Soren to be a very competent engineer. The shortcomings of this DAC compared to TotalDac was not something that any engineer on this planet could’ve anticipated. TNT can attest to this as I too have believed last week that Soekris was perfect to the best of our knowledge. But I now have reason to believe that improving the Vref beyond what our measurement tests indicate is necessary will yield tremendous returns. The improvements are probably not difficult to implement on a new product, but not everything that should be done will be done.
Also, I even doubt that TotalDAC knew why they should do what they did. Their product, in my opinion, is more the result of a cost-no-object approach, which might’ve happened to be successful. MSB and many other delta-sigma manufacturers may not be so lucky. In comparison, Soren’s approach was much more technically reasonable and respectable. I have said many times I cannot stand the distasteful designs in audiogd etc, but I admire Soren’s work. However, I believe he was limited by what we know about audio and our best theories. I will be the first to admit that this is only reasonable speculation. Judge for yourself. But I’m sufficiently convinced and I’m looking forward to seeing some empirical proof soon.
Last edited:
I respect him as an engineer, but maybe not so much personally or as a business owner. I hope this is not beyond reason.
I've always operated under the principles that respect is earned, not something given freely, and that actions speak louder than words, generally speaking...
I'm not making a personal attack, I am just expressing disappointment at how this product was marketed originally and how the manufacturer's initial claims have proven to be inaccurate. A company is usually more than 1 individual so rather than blame any one person for this situation, I would rather frame it in terms of the company itself.
And, yes, it is what it is, it's just not what it was claimed to be, which wasn't simply "a good dac for $250".
Enough said.
My position does differ with you somewhat. You should to some extent respect the fact that Soren’s design, even dam1021 rev.1, is absolutely perfect in terms of our currently accepted theories on how measurements correlate to hearing, and the assumption that R2R avoids many pitfalls of delta-sigma, which I also believe now (validity yet to be confirmed by empirical results). I wouldn’t be too mad at Soren for not offering exchanges. The mods are really not that hard to perform, nor do they really matter based on what Soren believed at the time as a competent engineer. However, as a business maybe he should’ve done better.
P.S. there might also be inherent engineering standards to a product, e.g. PCB layout etc. I’m not familiar with electrical engineering professional standards so maybe the vref counted as a screwup not as an audio product but just as an electronic device. Maybe that’s why Soren backed down and admitted problems in rev.1. Otherwise he could’ve just insisted rev.1 is perfect and I would give the claim some credence, at least until my recent developments in the understanding of audio...
Last edited:
I would say, open a new thread to design a world class ladder DAC, and just post in the vendor thread, when you have something substantial to say....
And how can you deny personal respect for a person you never met is over my head.
But that is the internet I guess...
Respect is not binary is it? If only I were a professional designer, then I’ll just do it myself... and probably offer it at a very low price provided I probably do it only as a side gig. I believe that this has the potential to alter the entire DAC market, so the lower the price, the better it is. It’s an extraordinary claim, yet I cannot help but believe it. I’m happy to see evidence on either side.
Remember, there’s hardly anything innovative in this product or in the one I’m proposing, with the one small exception of getting two shunt regulated rails to accurately track each other, or find a workaround that still provides low enough output impedance on both rails. Please correct me if I’m wrong and all this isn’t a standard design task.
Last edited:
The gist of it is, and one can disagree, that I wouldn’t buy from anyone who can’t provide an explanation on why my speculations are incorrect or can’t implement very high quality Vref in R2R, to the limit of audibility, and the capability of our current technology, which I believe to allow for a near perfect solution. I also wouldn’t buy from those asking for a steep price, for that simply seems unacceptable, as is much of Hi-Fi, since the engineering required is nothing but ordinary. Perhaps one can give in a little on price but I would certainly stand fast on the first point.
Btw, I think the current Soekris price is probably reasonable, though I do lament the recent price hike in dam1021. It seems to me at first glance altogether deplorable when a business cannot see beyond its short-term profit or profit alone. Now this, is actually highly debateable, and it’s not a very refined position at all. But now you know where I stand - all through the convenience of internet.
Btw, I think the current Soekris price is probably reasonable, though I do lament the recent price hike in dam1021. It seems to me at first glance altogether deplorable when a business cannot see beyond its short-term profit or profit alone. Now this, is actually highly debateable, and it’s not a very refined position at all. But now you know where I stand - all through the convenience of internet.
Last edited:
One more thing, Soren said it himself when alecm first discovered the vref problem in dam1021 rev1 that he didn’t spend much time on the vref design. This is further evidence that he didn’t believe it to be important, since the ripples are beyond the monotonic ability of the DAC. But now it seems there might be more to THD, linearity and monotonicity (in number of bits) in determining what’s audible. I don’t think Soren is at fault for this, but maybe now is the time to re-evaluate our theories and run some proper experiments to test our speculations. There’s a lot of history in this thread and it can be too absorbing for my own good...
Sadly no one seems to be reading this... I can only hope that someone with the means might eventually find this to be interesting, if my speculations prove to be somewhat correct.
Also, it might be worth noting that all of our audibility experiments are probably done by Delta-Sigma DACs, including our knowledge on how many bits are audible. There is some real empirical exploration needed here, and I can’t help but think that diyAudio is one of the best places to make it happen
Sadly no one seems to be reading this... I can only hope that someone with the means might eventually find this to be interesting, if my speculations prove to be somewhat correct.
Also, it might be worth noting that all of our audibility experiments are probably done by Delta-Sigma DACs, including our knowledge on how many bits are audible. There is some real empirical exploration needed here, and I can’t help but think that diyAudio is one of the best places to make it happen
Last edited:
I made a mistake due to my lack of knowledge in electronics - a buffer or relay between shift registers and resistors is not so easy to make ideal. I sometimes joke that doing computer science makes you forget about the real world and its physics; this would be one of those times. My very rough guess now is that the shift register in 1121+ is probably much better than 1021 in terms of internal resistance, based on this page: Low RDS(on) MOSFETs in Ultra Small Pkg - Nexperia | DigiKey. Size might indeed correlate with Rds.
Now is the shift register on the 1121/1941 the best that human technology currently offers? Soren refuses to tell us despite many requests over the years. From what I've found, the lowest Rds on a transistor is about 2mR: IRFS7530 - Infineon Technologies
Not exactly sure if something like this will work with our application, but it might provide a lower bound on what we can achieve in Vref impedance. Anyways, 2mR is still much much better than the >300mR in the LVC595, right?
The other factor is probably rise/fall and delay time. They're on the order of ns and we don't want them to be out of sync which would cause jitter. One further complication is that speed seems to be inversely correlated to Rds based on the datasheets I read... Now would be a good time for me to stop and leave to the professionals...
Edit: The delay time only matters if there's a large variance, which I'm not sure if there would be theoretically. Slightly longer rise/fall time might be fine since op-amps are much slower and used everywhere in the audio chain...
Now is the shift register on the 1121/1941 the best that human technology currently offers? Soren refuses to tell us despite many requests over the years. From what I've found, the lowest Rds on a transistor is about 2mR: IRFS7530 - Infineon Technologies
Not exactly sure if something like this will work with our application, but it might provide a lower bound on what we can achieve in Vref impedance. Anyways, 2mR is still much much better than the >300mR in the LVC595, right?
The other factor is probably rise/fall and delay time. They're on the order of ns and we don't want them to be out of sync which would cause jitter. One further complication is that speed seems to be inversely correlated to Rds based on the datasheets I read... Now would be a good time for me to stop and leave to the professionals...Edit: The delay time only matters if there's a large variance, which I'm not sure if there would be theoretically. Slightly longer rise/fall time might be fine since op-amps are much slower and used everywhere in the audio chain...
Last edited:
Apparently ON makes some good MOSFETs that go as low as 0.4mR Rds, and with a little compromise we can get <50ns switching performance. I recall Soren saying something about current performance around 70ns. I'll leave it to more a more mature audience to decide what to do next...
https://www.onsemi.com/pub/Collateral/NVMFS4C302N-D.PDF
https://www.onsemi.com/pub/Collateral/FDMS8050-D.pdf
https://www.onsemi.com/pub/Collateral/FDD9409L_F085-D.PDF
https://www.onsemi.com/pub/Collateral/NVMFS4C05N-D.PDF
The last one has 4mR Rds, rise/fall 32/7ns, turn on/off delay 11/21ns at 4.5V. This is just a rough survey of what we can get nowadays...
Here's Paul's calculation again. He puts a good Vref supply impedance on the dam1021 at around 6mR. Vref musings: the final frontier? | moreDAMfilters. Maybe there's no need to replace the op-amp regulator yet.
My harebrained idea is that we try add-on boards premounted with SMT shift register + MOSFETs and somehow install it to the original shift register location...
https://www.onsemi.com/pub/Collateral/NVMFS4C302N-D.PDF
https://www.onsemi.com/pub/Collateral/FDMS8050-D.pdf
https://www.onsemi.com/pub/Collateral/FDD9409L_F085-D.PDF
https://www.onsemi.com/pub/Collateral/NVMFS4C05N-D.PDF
The last one has 4mR Rds, rise/fall 32/7ns, turn on/off delay 11/21ns at 4.5V. This is just a rough survey of what we can get nowadays...
Here's Paul's calculation again. He puts a good Vref supply impedance on the dam1021 at around 6mR. Vref musings: the final frontier? | moreDAMfilters. Maybe there's no need to replace the op-amp regulator yet.
My harebrained idea is that we try add-on boards premounted with SMT shift register + MOSFETs and somehow install it to the original shift register location...
Last edited:
Don't know if the HC574/HC597 combination that TotalDAC uses has extremely low impedance. It seems to be CMOS technology and the propagation delay is very low ~10ns but nowhere in spec sheet or catalog guides is impedance specified or mentioned. Or is the HC574 used for reclocking?
It seems there might be an unavoidable tradeoff between jitter and ripple in R2R. And I have no idea how bad the jitter in MOSFET can be. But so far it does seem like there's more difference from caps than from clocks. I'll be the first to contribute to a crowdfunded effort for a dam1021 test board and some custom add-on PCB. Maybe we should even ask Soren to sponsor the dam1021.
P.S. The MOSFET datasheet only specifies typical switching characteristic, and it seems pretty precise, down to 0.1ns for single digit figures. Might someone here with more electrical engineering knowledge know whether we should expect the MOSFET to match each other relatively well in switching performance, perhaps dependent on the manufacturing techniques?
It seems there might be an unavoidable tradeoff between jitter and ripple in R2R. And I have no idea how bad the jitter in MOSFET can be. But so far it does seem like there's more difference from caps than from clocks. I'll be the first to contribute to a crowdfunded effort for a dam1021 test board and some custom add-on PCB. Maybe we should even ask Soren to sponsor the dam1021.
P.S. The MOSFET datasheet only specifies typical switching characteristic, and it seems pretty precise, down to 0.1ns for single digit figures. Might someone here with more electrical engineering knowledge know whether we should expect the MOSFET to match each other relatively well in switching performance, perhaps dependent on the manufacturing techniques?
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.