As far as I can find there's precious little discussion about the TDA1541 around here. @Purrin mentions it, but he seems to get it mixed up with the 1543, and won't elaborate; I've asked.
To me, this seems an omission; the 1541 has such a following. Whether this attention is truly justified, I'm not sure; that's what I've been hoping the discussion, on this thread, might go some way to answering!
Obviously, I'd like to hear about any impressions in comparing the Audial S; what, if any, are worthy competitors? Also, it looks like you've tried some original 90's dacs, perhaps with some modifications, such as a NOS conversion; what did you try and can you give some of your impressions?
I'm not sure if purrin has the TDA1543 mixed up. Maybe, maybe not. I don't believe he has them mixed up based on a small conversation I had with him a couple weeks back. I know he had some doubts about the TDA1541As datasheet at least in terms of reported INL/DNL.
The TDA1541A is a whole different ball game compared to the TDA1543, as I'm sure you know. Based on what little experience I have with DACs that use either chip and what research I've done, just in terms of tone and such, the 1543 is warmer, more forward and aggressive, and otherwise just generally less capable on all technical levels. But it can make for an enjoyable listen from a budget DAC. When I see a TDA1543 DAC costing over $500 or, worse, $1000, I just have to shake my head. I don't care how many chips you parallel together. It's still a TDA1543.
I believe TDA1541A, standard grade, is rated for worst case scenario +/- 1 LSB, per the datasheet (still trying to dig for more info and clarification on 1541 lsb error). The various "grades" of the chip have different performance tolerances, some which are rated at worst case of 0.5 to 1LSB depending on bit, I believe. That's not to say you can't find a regular 1541A chip that performs on par with the S1 or S2 grade chips, just that the S1/S2 chips are guaranteed for a higher level of performance.
FWIW, Pedja of Audial hand picks the chips from a very specific batch of TDA1541As, from a '98 run in Taiwan. They're not S1/S2 grade, but subjectively and measurably he says they should perform as such. I have to trust him, because, as far as I know, he's well known in the DIY world for his experience working with and knowledge on the TDA1541A, and I've seen him post comparative measurements of various graded chips. (Can't say how much he knows/doesn't know compared to anyone else, so not saying he's necessarily the bestest.)
From what I've read according to those that design DACs around the TDA1541A, you can get somewhere around 15-15.5 ENOB (effective number of bits) on a good design and assuming you get a good chip from a good batch. I think chips like the PCM63 or AD1862 can reach a higher ENOB, but I also believe they are rated higher than 16-bits. TDA1541A is rated at a max of 16-bits. Compare this to the rumored chips that Metrum uses, which could be anywhere from 12-14 ENOB, which wouldn't be too far off from the TDA1543 (not 1541). TotalDAC with 0.01% resistors may only be able to get 14 ENOB, maybe. I don't have enough knowledge to explain how/why the TotalDAC does so well with static measurements but is rumored to really have only 14-bit performance max. 6-moons wrote a bit about this a while back, and there are other rumors floating around. I get why it might be limited to 14 ENOB, but not sure about the static measurements. Same thing for that new soekris DIY discrete ladder DAC.
Would I have gone with a PCM63-based non-oversampling DAC if available? I definitely would have considered it. That or the AD1862. I just didn't see any available or appealing options that at least looked like they were well designed and came close to maxing out said chip's performance. I know MHDT labs is/was going to put out a non-oversampled, AD1862-based DAC, but A) I have no idea when it will be available, B) I don't know what it will cost, and C) based on reviews I read of other MHDT DACs, I decided the design might not be to my tastes or as good as I'd like. The Metrum Pavane might have better accuracy than the Hex or similar, but it's too pricy, and the listed THD still concerns me about its true performance.
Another thing I've read from DIY DAC designers is that getting the TDA1541A to reach its specified, max potential is rather difficult. And when you're going for a non-oversampled design, it becomes that much more difficult to get it to look good with objective measurements. It seems the only thing holding the Audial Model S back is variance between the left/right channel performance on the chip, which I've read is common or to be expected on the TDA1541A or TDA1543. But, as I said, the main different between the left channels is almost exclusively in the 2nd order distortion, per my (unpublished) measurements. It still looks a whole hell of a lot better in measurements than the Metrum Hex did, and sounds it too. I'm not going to split hairs if the left channel reports 0.007% THD at 0dB 1khz tone and right reports 0.002%, and only due to 2nd order differences. For the most part, looks good on paper, and sounds great to my ears.
Even outside of good performance on paper/datasheets, the TDA1541A has a following because a lot of people simply like its tone. It's just very natural sounding and fatigue free without being boring, slow, too colored, glossing over details, and so on. As I understand it, though, you really need a good, beefy design to really bring out its true dynamic performance. Then you have fans that prefer the PCM63 or AD1862 for different reasons, general tone and different technical qualities. They all look good on paper, so assuming you put them in a good implementation, it really comes down to personal tastes after that point.
ANYWAY...let me know if you want further details beyond the Model S impressions I already provided. I know I already wrote a lot, more than intended, but I will be publishing a full review in the near future. I only have the Theta Gen Va on hand to directly compare it against, though I will do what I can to compare it against the Metrum Hex, Audio-GD NOS1704, and Classe DAC-1 based on memory.
Never tried any mods on these DACs, though it would have been cool to try to get the Classe DAC-1 to run in non-oversample mode. I'm not an electrical engineer of any sorts, or what have you, so that's way outside my expertise. (And, of course because of this, call me out if I misunderstood or got something wrong, please.)
Don't get me wrong...something like the Yggy definitely will beat the Audial Model S in terms of sheer techincal proficiency and accuracy. Had I not required a non-oversampled DAC for my primary setup, I easily would have gone that route. It's such a no-brainer vs. the hunt I had to go through to find the Model S. And I'm sure I'll get a DAC with trickle down Yggy tech in the future as a secondary/complementary DAC. So, anyway, that's just a lot of unnecessary details about the TDA1541A and how I came to my decision to get the Model S, but the vast majority of people are going to be best served by something like the Yggy.
urrin
