Popular Post soekris Posted April 24, 2019 Popular Post Share Posted April 24, 2019 I just want to throw in a couple of comments.... PeterSt, you don't seems to understand the Sign Magnitude R-2R principle, please do a little research before posting those claims.... It's correct that a normal non sign magnitude DAC need very precise resistors as the MSB bit toggle all the time when the signal crosses zero, so THD is always relative to full level, while in a sign magnitude DAC ony the bits used by the signal changes, so THD is relative to the actual signal level. The PCM1704 was sign magnitude and so is the soekris line of DACs and DAC modules. You can just visit SBAF, they have done measurements on a number of my dac1541 showing signal level to be correct down to -140 dB, with harmonics following the level down, the dam1021 will be the same, while a non sign magnitude R-2R DACs signal would be distorted.... On Noise level, the advantage of discrete DACs is that you can use lower resistance part, and the noise level of a R-2R DAC is directly related to the resistors noise, while chips use larger value resistors. The dam1021 noise is that of a 625 ohms resistor. And just for the record, the dam1021 module is manufactured in Denmark, and the important parts, the resistors, are manufactured in the EU. Teresa and Danny Kaey 1 1 Søren Link to comment
Popular Post soekris Posted April 25, 2019 Popular Post Share Posted April 25, 2019 17 hours ago, PeterSt said: OK, so the debate about what's required for precision is still going on, eh ? To be clear, I was never involved in the initial DIYAudio thread - I only followed it. Today's point is: you seem to suggest that I relate my "math" to MSB toggling. No idea where you got that from, unless you saw "MSB" in my Excel line. ... For monotonicity the MSB('s) need to be more precise than the LSB('s). That's all. People discussed it back then and maybe I am not attempting that today (why would I - and I am also not saying it is necessary), but all I talked about was that with the least possible words (mentioning IMD which refers to linearity which comes down to monotonicity, etc. and now try that with 24 bits) ... that is, if you talked about that. Maybe I misunderstood. I could also refer you to some larger thread about the glitching of some R2R DAC, but a few people won't like me for doing that. So I won't. Haha. But anyway, you must be talking about that (glitching). I was not ... Maybe you referred to me implying (!) that the PCM1704 can measure better than your board showed at various places. So yes I sure implied that. But you know, I was trying to stay far away of being "competition" and just brought it in neutral fashion. That could have stayed like that until a. people start to bring forward claims which can not be justified and b. someone explicitly tells me to better stay out because being competition. The two don't go along well. The main problem seems to be that we talk about the very same thing with the very same objectives. Spreadsheet time again: ... just to show that I am serious, including low resistance of the ladder (I put your 625R in there and it will be linear to the 24th bit but mind the required accuracy = "Tolerance"). I am not trying to be competition at all and merely appreciate what some people are capable of; When I finished my R2R discrete project and had all the 2700 parts in the house for a 2 channel trial, I just gave up because I knew it couldn't be better than the 1704 based proven design. It's up to you or anyone what to think of this, but one thing is relatively important: I don't have to make up any kind of justification for using a discrete ladder because the best chip ever made for the job went EOL. This is how you see me talking in the other direction: "the 1704 is better because ...". (and the inside information is that I bought all the 1704's at some stage) So you see, we talk about the same. My discrete design included all what was Sign Magnitude and more. Much more. But also too much because of that. Unmanageable (IMO). And obviously when I would not have had the idea that the discrete ladder would be for the better, I would not have started the whole project (kept me from the streets for 2 years). I just don't see how the currently running design can be improved upon. And the stupid thing is: all I made up for the discrete design for improvement, can also be applied to (added to) the silicon design. I will always lose from myself. (THD+N is just too good) Figures are not everything although for me it is a base. I always envisioned the sound of discrete as "analogue". And maybe I am jealous that some got there. You sure did. Thanks. Let me explain in details the sign magnitude magic, not that I'm particular good at that, I have had many people not getting it, I'm a little tired of people claiming that with 0.01% resistor it can't be more than a 13-14 bit DAC, while the measurements of my dam1021 module show a dynamic range of at least 130 dB.... We can agree that a digital audio signal crosses zero all the time, and that mean for the normal R-2R DAC, the MSB bit toggle each time the signal crosses zero, no matter the signal level. That error between the value of the MSB bit and all the other bits causes distortion. A sign magnitude R-2R DAC only have the bits actual carry signal changing, so t.ex. with a -60dB signal, the 10 MSB bits all stays at zero, resulting in errors following the signal level down, resulting in low distortions at lower levels, and that matters as music is very dynamic. And that mean for a sign magnitude DAC, the resistor precison is not as critical as a regular non sign magnitude DAC. I dont' really understand your spreadsheet, but it doesn't matter anyway, my DACs don't have 2R as 1250 ohms, I said my DAC have an output impedance of 625 ohms. We can agree that the pcm1704 was the best audio DAC chip ever made, unfortunately it was expensive and TI discontinued it as everybody, except the high end, wanted cheaper chips, the pcm1704 was expensive to manufacture due to its bipolar process and need for laser trimming.... I don't know about you, I can't base a business on parts that's not available anymore, and I don't count china and ebay as useable sources.... Unless you of course managed to purchase a lot of chips before they got EOL'ed.... So I couldn't make a DAC based on the pcm1704, and then started to think, it can't be that complicated to make a discrete sign magnitude DAC. Which I then did, and offered the dam1021 modules at pretty low cost. And a finished DAC is more that the parts used, there are plenty of high end DAC selling for thousands using $5 chips.... I'm sorry for you that you didn't manage to complete your own discrete DAC, but why 2700 parts ?? My dam1021 module uses 510 parts.... When you run out of pcm1704 stock, take a look at my OEM modules, perfect replacements.... Pure Vinyl Club, Superdad and Axial 3 Søren Link to comment
Popular Post soekris Posted April 25, 2019 Popular Post Share Posted April 25, 2019 27 minutes ago, mansr said: Dynamic range is not the same thing as linearity. No, we cannot. If it was, they'd still be making it. Complicated, no. Accurate? Also no. You need the linearity to get the dynamic range. And the key here is level linearity, which is where the sign magnitude DAC really shines.... Just because the pcm1704 was the best, does not mean that TI was making money on it.... And semiconductor manufacturer like to keep their expensive fabs used for money making parts.... So what DAC chip do you consider the best ? And the dam1021 is accurate enough to that the customers like it, just read the dam1021 thread at diyaudio.com.... Seems like the audio community is divided in two parts, those that trust their ears and those that believe in numbers.... Pure Vinyl Club and Teresa 1 1 Søren Link to comment
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now