I remain confused with how a DSD signal is played by the R-2R DAC.
@dvorak appears to be stating one uses two of the R-2R’s inputs, each with an equivalent value but one negative and one positive. (A two-input R-2R).
@tedsmith seems to be stating one connects to only one R-2R input and switch this single input between positive and negative.
Are both of these methods viable? (both logically/theoretically and in physical reality)
Another potential source of confusion is my understanding as to how an R-2R ladder DAC works. My understanding may be incorrect, making my confusion worse.
Mu understanding is each input of an R-2R DAC is is mapped to a specific bit-level. That is, for example, the top level bit goes to input 1; the second bit, input 2; etc. This way each bit contributes its information, from big to small, and the final output represents all of the information captured by the PCM’s data stream.
Where am I wrong? It could easily be in more than one way.
With an R-2R DAC you play DSD merely by picking two PCM values you like to set the level (and offset if that’s needed) and then you pick the more positive when there’s a one and the more negative when there’s a zero. E.g. if PCM 1 maps to full scale, then you might use 0.5 for DSD 1 and -0.5 for DSD 0. You could equally well use .978 for one and -.978 for zero to get a different output level. And for the weird you could use 3/4 and 1/4 if you want the output to be centered around 1/2 full scale.
All I’m saying is that if you can come up with a scheme to get your R-2R circuit to output +X volts when DSD bit is 1 and -X volts when DSD bit is 0, and you can do that fast enough to keep up with the bits as they come in, then you have a working DSD-to-analog converter which doesn’t “convert DSD to PCM”. There are at least several different ways you could implement that, some which would be optimised for that specific purpose and others which would just be a brute force marvel of modern materials science and engineering.
PCM differs from DSD in that its digital form does not directly represent its analog form in the way that DSD’s does. Every 1 in DSD represents the same analog value, every 0 represents the same alternative analog value. But with PCM you need knowledge of the specific encoding scheme, the number of bits per sample and the location of the sample boundaries. If you treated PCM like DSD and just sent it through an analog low-pass filter you would get nothing but noise. No signal at all.
If you have a multi-bit sample stream at the same sampling frequency as a DSD bitstream, the only difference in the encoded signal is the quantisation noise. So it does not matter if you “convert” each DSD bit to one of two different multi-bit sample values without changing the sampling rate, and then use a PCM decoder to translate them back to two alternating voltages. That’s roughly what we’re suggesting these DACs are doing.
There are other possible tricks, where they might use a dedicated switched line to directly energise the MSB – or all the bits – on the positive and negative halves of the ladder without running through the normal PCM decoder path. I don’t know if any DAC does this, but that’s really an implementation optimisation that doesn’t change the fact the DSD is eventually coming out in its true form.
Cool. Maybe the missing piece is to say that you can’t use just any R2R DAC to do this. Apart from the sampling frequency involved (at least 2.8MHz) there’s a logic component that has to recognise a DSD stream, and then perform whatever action that particular DAC designer has implemented to translate each bit into the chosen output voltage.
Most PCM DACs get their registers loaded serially from a data line then as the final bit is loaded there’s a separate input toggled to cause the output lines to change their state and match whatever their corresponding input bit is now set to. Bear in mind that there’s a positive and negative half to a decoder like this, so when the sample value changes polarity, while one side is getting told to set its value the other half gets hit with a reset line that pulls it all back to zero. Hypothetically, the same (unsigned) 15, 19 or 23 bits could get loaded into each half and then the polarity of the sample determines which half gets the activation signal and which gets the reset signal.
For DSD, the bit represents polarity only. So the MSB, or indeed all bits, for each half would be permanently set to 1, and the DSD sample bit would just toggle which half of the decoder is active. Instead of toggling the active/reset lines every 16/20/24 clock cycles you toggle them every cycle.
That’s just one example of a physical scheme that could work. Hope that helps.
It just occurred to me that one of the factors hampering my understanding was that I was assuming an R-2R ladder DAC had to be used as a ladder DAC. As DSD does not have bit depth, as does PCM, my conclusion was a ladder DAC cannot be used to convert DSD to analog.
Both of you, of course, knew one is not required to use the ladder but can instead use just a small part of the ladder.
I do not have a “DSD DAC”, but it converts DSD64 to 40bit/384 kHz pcm. Why? I have no idea. I’m with @tarheelneil and @jazznut on this one, there are too many acronyms involved, I sometimes think I’m reading a Scrabble forum rather than an audio one, and who knows what matters, whether hardware or software. I’m happiest with my head buried in the sand on the technology and enjoying the music. I chose D/A conversion by listening to machines as there seems too much information for Mr Average Punter and there are almost as many different theories how best to do D/A as there are audio companies.
Of course you’re confused. All of audiophiles get confused. In fact, I’m confusing myself right now! (And jazznut is scratching his head saying “huh??” !)
Think your issue was whether it’s the hardware or software, or both, that we need to understand.
@tarheelneil I had to look up DoP and I think I’ve already forgotten. I don’t know what PCM means, but I’ve been listening to it since 1983.
As someone who has bought the vast majority of his audio from bricks-and-mortar stores, and trusted “audio consultants”, as a rule they NEVER get technical on this stuff. I have never heard an audio consultant talk about DSD or talk acronyms, sampling rates or the like. I was in a store on Saturday and they just talked about high quality audio. One dealer explained the reason is because 99% of his customers are not audiophiles, just want an audio system.
I do wonder if there are any audiophiles who go to a store, listen to a few systems or components, make a decision based on what they hear and pay up on the spot, without wondering if there is DSD to PCM conversion going on?
Yes, finally the consumer doesn’t want to know about this, at least the typical audiophile just wants best sound (however achieved).
But we know behind good quality there’s mostly a lot of effort, technical knowledge and such differentiation that leads to acronyms (even if they’re not used for marketing).
Those who care (and only who cares gets best quality) at least have to get a rough overview and make up an opinion or decide whom to trust. I personally trust little in people who sell and want to know the basics.
Yes, I understand we’re talking about nearly all, rather than all . Is the DirectStream not purely DSD because your upsampling scheme sounds better than straight DSD?
I have appreciation of the complexity of design and manufacturing processes and how it is harder to make things appear simple. I’m not sure a lack of acronymic know-how ever stopped anyone not getting the best consumer product for them, whether audio, dishwashers or TVs (all of which I bought this weekend past).
I switch between wanting to understand a product to just wanting to listen and enjoy music. There are loads of good value phono’s about and I have arranged a loan to compare two solid state, the Stellar Phono vs the Primare R35, described by the dealer as detailed vs musical, but I so enjoy the EAR Phonobox, I wonder why bother. Maybe I should save up for the Zanden my dealer recently acquired, its only €30,000.
The upsampling is required for PCM. I could rip out all upsampling code and be a DSD only DAC (it’s only software.) The high rate is simply so that 192k and DSD can be treated the same. The simplest thing to do if you have the resources is to find the least common multiple of 192kHz and 2.8224MHz which is 28.224MHz and convert everything to that. For slightly unrelated reasons I now go to double that (56.448MHz).
)
. Is the DirectStream not purely DSD because your upsampling scheme sounds better than straight DSD?