The very early true 1-bit converters, e.g, Philips TDA1547 DAC7 used in some of the early CD/SACD players uses a single SDM (Sigma Delta Modulator) to process 1-bit data. All modern DAC chip uses Multi-SDM architecture; 4 to 8-bit to further improve SNR inside the chip design.
You are right R2R is designed for PCM, but Holo Spring DAC has a separate hardware DSD section specially designed to process 1-bit data. The way R2R process data depends how it is wired to process PCM or DSD. This process is called Discrete DSD conversion, Chord, dCS and MSB all use this form to convert DSD to analogue. There’s no material at the moment that does DSD512 but many use HQP to up sample to DSD512. So having a DAC that does DSD512 is a plus.
Don’t get confuse the word ‘native DSD’, this term is usually used to describe how data is being transported from the host to a DAC. Most DACs can accept DoP and native modes. In both cases they are bit perfect. Jeff Zhu, the designer of Holo Spring DAC has a in depth discussion at Kitsunehifi:
The way R2R process data depends how it is wired to process PCM or DSD. This process is called Discrete DSD conversion, Chord, dCS and MSB all use this form to convert DSD to analogue. There’s no material at the moment that does DSD512 but many use HQP to up sample to DSD512. So having a DAC that does DSD512 is a plus.
Thanks a lot for the detailed reply and useful links. Much appreciated.
However, Jussi does mention that there is no point for an R2R dac to support his HQplayer DSD512 only if the bits actually go as-is to the resistors.
I assume what Jussi’s described also indicates the final SQ result difference.
For instance, some users of Rockna wavedream I chat with feel that there is no big difference among DSD64, DSD128, DSD256 and DSD512, or even between DSD and PCM when using HQplayer upsampling and filters.
In true 1 bit converter DAC like Jussi’s own DCS1 or T+A DAC8 DSD or even some multi-bit sabre DSD dac like exasound, users like me (I’m using T+A DAC8 DSD and exasound) can feel the big SQ improvement from DSD64 all the way up to DSD512. In my opinion, DSD512’s SQ improvement on T+A is like a magic with HQplayer.
In other words, Can I understand that R2R’s DSD512 is more like a marketing feature just for its compatailibty instead of improving sound quality by HQplayer’s own purpose?
I’m not sure Jussi’s own DCS1 is a true 1-bit converter (Single SDM). As I mentioned above, almost all modern DACs are designed around Multi-SDM to improve the SNR, including Burr-Brown, Cirrus Logic, ESS Sabres and AKM.
The ways to convert digital to analog, R2R and Multi-SDM as being the most common. R2R is a fixed resistors architecture that convert digital to analog The precision of conversion depends how well the resistors are matched. SDM works on a principle of feedback to continuously correct itself from errors. They tends to be more noisier than R2R design, to reduce the noise issue in 1-bit converter, they turn to Multi-SDM and higher order of noise shaping >5th order. R2R on the other hand is much quieter and some designs requires just an analog LPF in the final output stage.
I’ve tried listening to DSD64, DSD128 and DSD256 in its native recordings they all sound different in Holo Spring DAC. I managed to up sampled using Roon 1.3 to DSD128 (My PC just max out to DSD128 processing power) from a 44.1k, and it sound gorgeous.
I see what you mean, I believe your experience since Jussi also mentioned Holo spring DAC might possibly have a specific resistor array for DSD (but he is not sure). So I guess it’s a little bit hybrid R2R dac.
Haven’t tried Holo ring dac yet, but you made me curious.
From your point of view, if Jussi’s DCS1 is not true 1-bit, maybe T+A is also a little bit multi-SDM. (then what’s the point of calling it true one bit converter as their promoting feature?)
I guess Jussi might pop up for further argument?
I’ve tried exasound e12 (ESS multi-bit DSD) vs T+A DAC8 DSD (true 1-bit converter), big difference especially in DSD512, T+A with HQplayer shines big time.
Suppose Holo spring dac should also sound amazing in DSD512? However, from the review of Ted_b on CA, gives me a little bit mixed feeling. “Best Of Both Worlds? Best of Neither in terms of PCM and DSD?”
At the moment, he’s trying to get a T+A loan for another useful review. It could be a useful reference for us.
Not trying to de-value any commerical brand, but I think the combo of T+A DAC8 DSD + HQplayer outperforms quite a few dac which is more expensive.
There’s another unique feature that Holo Spring DAC stands out the rest…the ability to bypass the hardware over sampling digital filters inside the DAC. This means you can use HQP to do all the software up samplings as well different type of digital filters, then pass all these directly to the digital to analog conversion stage of Holo Spring.
I agree. Any NOS dac will benefit from HQplayer. Either R2R or SDM dac (good example: exasound, since exasound doesn’t put any hardware for upsampling)
But the difference is HQplayer’s upsampling setting has two sections, one for PCM, the other for SDM.
For R2R, users should pick the PCM upsampling section while for SDM DAC, pick the SDM setting part.
I have no idea how your Holo spring will perform under those two, since it’s a little bit hybrid DAC.
For R2R dac like ROCKNA dreamwave, using SDM upsampling to its DSD512 is meaningless, this can be also proved by some users’ experience. HQplayers’ SDM upsampling doesn’t give any bonus to Rockna compared to HQplayer PCM’s upsampling.
If I’m correct, this “no SQ effect” DSD512 is because Rockna uses FPGA for the processing. So instead of saying DSD512 feature on this R2R dac is a plus, I would prefer to consider it’s just a marketing trick let consumers think it can support all the formats on the market so far.
No idea of your holo spring since it might have a dedicated DSD resistor array for DSD procesing.
Anyway, looking forward to the big educational moment from Jussi.
Yes, if the DAC already built-in FPGA processing, such like PS Audio DirectStream DAC. It actually convert all inputs whether PCM or DSD and up sampled to ‘high speed DSD’ then this is fed to their customed designed ‘1-bit converter’. Now if you doing HQP in the source, your effectiveness is now limited by the FPGA processing of DAC. That may explain there’s little SQ effect here.
What we want is to do HQP at the source (up sampling, digital filters) and fed directly to a digital to analog conversion stage, whether it can be R2R or SDM. Simply put it this way, we want to do all processing in software level, conversion from digital to analog is done by the hardware.
I’m upsampling Redbook to DSD 512 into the Holo Audio Spring using HQP with native DSD. It sounds great. Jussi uses SDM as a term, not a prescription. I’ve never seen him say that R2R can’t or shouldn’t be used with SDM settings.
As I said, some Rockna wavedream users I know feel no difference in terms of DSD512 compare to DSD64, 128,256, or even think PCM sound is better than DSD512.
So for pure R2R dac, using SDM setting will be also pointless. Because for a R2R dac, in order to support DSD512, most of them using FPGA.
However, Holo spring dac is an exception since its resistor array might be dedicated for DSD (I’m just guessing), but is this hybrid architecture performs equivalent in DSD512 to some dedicated pure-1 bit converter DAC? That’s the question to be answered. That’s why I’m waiting for Ted_b’s review on CA.
Some serious R2R dac won’t support DSD to be honest, do you find Metrum support DSD? No.
As I said, for R2R dac, either we can understand DSD support is a plus feature or meaningless marketing trick.
The Holo Audio Spring had 4 resistor networks. Two handle PCM and are R2R. Two handle DSD in a similar fashion to Jussi’s SC-1. Within each group one network performs the processing and the second is used to correct linearity.
The surface mount PCB with the resistor networks is double sided, repeating on the underside: