Vivaldi upsampler's future

That is not correct ( unfortunately). There are indeed options. However they only concern the mode of upsampling. They do not enable non-upsampling replay. The options are to select between DSD , DSDx2 and DXD upsampling. There is no Clone Mode which is required for upsampling to be avoided. NB; I have just looked through the latest User Manual for Rossini DAC; v.03 . See pages 23 and 39 in the manual and also the menu function map p.32.

1 Like

The manual seems to differentiate DSD upsampling vs DXD oversampling.

Yes, they are different. One is a form of encoding using delta sigma modulation with the trade name DSD ( Direct Stream Digital) and the other, DXD, uses pulse code modulation ( PCM).

The manual says DSD upsampling is optional. Selecting DXD causes the dac to not do upsampling just oversampling. All dCS dacs do oversampling of pcm data. The manual never says the Rossini upsamples before oversampling. Based on the manual I’d say the statement “Can’t shut off upsampling on Rossini” is false. Not sure if there is a functional difference between upsampling + oversampling and oversampling.

I am sorry Katzsky but you are just playing with semantics. To put it very simply; if you send a redbook 16/44.1 file to Bartok or Rossini the sample rate will be changed (upsampled) to DSD64 or DSD128. Those two options are what the manual refers to as optional 1-bit upsampling ( my emphasis and this is for PCM sources, ). If those options are not selected then there is default PCM upsampling implemented; to 24/352800 ( DXD). In short , speaking figuratively, you cannot put 16/44.1 in and get 16/44.1 out. With Vivaldi you can do this as it can be used without an upsampler or be set to Clone Mode.

Upsampling is a form of data conversion and differs in intent from the upsampling used as part of the Ring DAC process;

All dCS DACs do use an upsampling process as part of the way that Ring DAC works. All incoming data is upsampled to 706.8kHz or 768kHz as this is how it is fed to the 5 bit modulator as the mapping process ( how the data is distributed to the current sources) is 5 bit. However this differs from an upsampler as the resulting output from the DAC is not converted to another sample rate. 16/44.1 goes in, there is some processing which includes upsampling and, again figuratively, 16/44.1 comes out. I say figuratively as what actually comes out is an analogue signal. I have been careful to differentiate the two when writing about my particular misgivings about the sonic results, as I hear them, when selecting upsampling in contrast to no upsampling :

Unfortunately I’m not capable of playing with semantics on this topic. DCS refers to DXD as oversampling. If it’s just semantics then it’s dCS.

What you’re saying is “DXD upsamples to 352khz using oversampling sequence then oversamples pcm data”? Maybe I expect too much from technical writers since this isn’t communicated via the plain text. Either way… how is it functionally different to upsample to DXD when the process is exactly the same code used to oversample?

The only thing I know for sure is that the writer is very careful using the word “oversampling”. I don’t know the difference but the writer is telling me there is a difference.

Yes indeed, the difference between oversampling and upsampling has always been made by dCS and as I mentioned elsewhere when they introduced the first home use upsampler they both invented the term and published a white paper on the difference between upsampling and oversampling which, frankly, I found pretty incomprehensible back then. Sadly I don’t think it is still available. If anyone has a link I would appreciate it.

However, whatever the precise wording, the bottom line is that DXD is always 24/352800 and not 16/44.1 or 24/44.1 or 24/96 or any other PCM rate. There is no option to deselect this except by selecting one of the the two 1-bit upsample rates in which case the PCM is then sample rate converted to DSD64 or DSD128. NB: native DSD files are not upsampled in this way.

I guess dCS uses the word ‘oversampling’ when applied processing is ‘only’ changing the frequency and adding zero + dithering at the end of the 16 bit data sample. In short, PCM In gives PCM Out after processing. DXD is PCM data. So 16/44.1 over sampled to 24/352.8 is oversampling according to dCS.

However dCS seems to prefer referring to upsampling when PCM data is converted to a DSD audio stream.(quite different, in nature, from the PCM data).

They probably have a good reason to do so.

Over… versus Up… (sampling); well, this is always manipulation/processing of the original data. The ones Peter is not Fan of … (same for me, although I never tested with a genuine dCS upsampler itself).

I have spent a little time this afternoon looking on the web at answers to the question “what is the difference between upsampling and oversampling ?” Frankly I think that Hans Beekhuyzen comes closest. After explaining that technically they are similar at one point he concludes that the words are selected in accordance with how the manufacturer chooses to use them :smiley:.

John Atkinson in Stereophile touches on the subject but I don’t think moves far away from explaining why he thinks that upsampling improves the sound.

I am posting links to both but , from my viewpoint or from my hobby horse if you like, I am taken by a point made by the well known recording engineer Bob Katz in a letter following Atkinson’s piece. He remarks " some people may prefer the sound of the asynchronous converter on some recordings, because the vagueness of imaging is aesthetically pleasing (shades of vinyl?)." According to Katz the dCS upsamplers of the time ( 2001) used a Crystal Semiconductor CS 8420 chip which was asynchronous. I have no idea what ingredients are in the current upsampler recipe but, irrespective of cause, I find it heartening that the artefact of odd imaging which I am very aware of was recognised 20 years ago, at least in professional circles .I no longer feel so isolated. I owned the dCS Purcell mentioned by Katz and, as I have written in other threads, that was around the time that I started to have my concerns on this issue

1 Like

Pete, slight mistake there.

The actual quote from Bob Katz in his letter responding to that Stereophile article was:

I note that you have been exploring upsamplers. Of the commercial units you have reviewed, to the best of my knowledge, all of these samplers and upsampling converters, except for the dCS 972 and Purcell models, are based on the Crystal Semiconductor CS 8420.

I haven’t peeked inside my dCS 974, but I seriously doubt dCS was using any commercial off-the-shelf ASRC, even at the time, it was always their own algorithms on FPGAs :stuck_out_tongue_winking_eye:

Yes, I too was surprised by the mention of the CS8420. Honestly my reading ability is getting increasingly compromised by my sight loss. Anyway my main point was that, irrespective of cause, the possibility of odd imaging related to upsampling was already recognised 20 years ago.

Actually Pete, you may have missed a little nuance to Bob Katz’ comments.

I have done the listening tests, and, under identical jitter conditions, recordings that have been upsampled via a good-quality synchronous converter sound about “two points” better, with a wider, more stable soundstage and imaging.

Of course, some people may prefer the sound of the asynchronous converter on some recordings, because the vagueness of imaging is aesthetically pleasing (shades of vinyl?). Professionals who upsample will generally avoid asynchronous converters because we want to get those last two points of sonic performance.

So, he was actually comparing the sonic results of a synchronous converter to an asynchronous one, not to no upsampling. He wasn’t against upsampling per se, just ASRCs.

Which begs the question, exactly how does dCS’ Upsampling interpolation work? I don’t think I’ve ever seen any technical explanation :thinking:

In a quick scan of all the Upsampler manuals, only the Purcell Manual has this to say;

This is achieved using extremely powerful and accurate digital interpolation filters, which introduce negligible levels of distortion.

No other manual even mentions the word interpolation :man_shrugging:t2:

Talking of nuances Katz does not say that synchronous sample rate conversion offers a stable soundstage and imaging, just more stable. So that leaves room for my experiences and also that such issues have the potential to be even more pronounced than I have heard if using a converter that is inferior to dCS’. Further I recall reading the preamble to Leedh Processing’s patent and that too mentions imaging aberrations related to upsampling. NB: Although Leedh are most closely associated with lossless digital volume controls their process(es) is/are also related to upsampling, filtering and file compression.

Well, Philips were 4x oversampling redbook from the earliest days of their CD player. In nearly 40 years, no one (else) seems to have documented any imaging aberrations or measured it? :stuck_out_tongue_winking_eye:

As the Philips oversampling could not be defeated no one could have been aware that what they were hearing may have been impaired especially as CD reproduction of the time was, as I recall ( and I owned three Philips players of succeeding generations) not all that great anyway - for the mass market players in any case. The sound of them was no doubt also compromised by some fairly basic analogue output stages. I only bought into CD because the classical labels were the first market segment to drop LP.

As I have also mentioned previously, if your listening diet is pop/rock you are not likely to be aware of this aspect as there is no imaging reference to prompt the observation.

Classical music on Philips: Brahms, third racket :laughing:

4 Likes

A minor twist to the subject but if it needs to go elsewhere pls advise. I have a paganini stack with vivaldi transport. I use up sampler only while streaming from a dedicated music server (infrequently). I am primarily cd/sacd and analog. I just upgraded to vivaldi dac and clock. Thinking to maintain the paganini upsampler to maintain the better quality coming into the dac from the server when used. Server is on dcs clock with balance of dcs equipment. Any input welcomed. Thx

If I read your post correctly you are now actually running a three box Vivaldi stack ( Transport, DAC, clock) . You have a Paganini upsampler but are asking if you should get a Vivaldi upsampler or continue with your existing Paganini upsampler. Is that right?

In the circumstances I would not want to use Paganini in that combination. Aside from the observation that it is not fully compatible as it cannot take advantage of the dual wordclock into provided by Vivaldi clock you then have no hi-res network access and, I would say more essentially, no dCS Mosaic control which as well as providing control of the replay of the files n your server adds access to streaming services ( subject to subscription) and internet radio . Vivaldi Upsampler is not just an upsampler. It is also uses more advanced overall technology than Paganini.

How do you currently navigate the files held on your server? I note you say that the server is " on dCS clock" which I assume means that you are not talking about an standard IT server as such devices do not have wordclock inputs. I would be grateful for your clarification.

Hi Pete. I was trying to be brief but happy to clarify. Today i have vivaldi transport and paganini 3 box stack. I use a “home built” dedicated server that is attached directly to my pag upsampler via usb i access the files via plex. This “server” is synched to my pag clock currently I use this network function almost exclusively for background music for parties (not through my main amps and speakers but via cat 5 and firestick or smart tv to localized areas of my home) and use the plex and streaming function primarily for bringing music, tv and video to my ipad or phone when on road trips and traveling. It is almost never used through the main system but it is a nice addition. My processor has a strong dac built in but i prefer to send the data through the pag stack when i do listen to the server in the listening room and it performs nicely. My listening is really 99% in a dedicated listening room and when digital (75%) is through the pag dac, clock and current vivaldi transport playing cds and sacds - (25%) analog. I have acquired a vivaldi dac and clock and will replace the pag stack in the next few days. My decision thinking at this point is whether i leave the pag upsampler in the setup to improve the server fed music to the main setup until such time that i purchase an upsampler planned for sometime over the next year, bypass getting a new upsampler altogether and run the server through my processor dac only and what dcs functionality and improvements am i giving up through this method. I never plan to pay for online streaming and have a massive digital collection on cd,sacd and to a lesser extent in lossless server content Thats long. I hope that it is somewhat clear. Thx

Thanks for the response. Unfortunately it leaves me just as confused albeit in a slightly different way :smiley:.

Anyway you are not really asking how best to use your network option as it is unimportant to you and is only used for background music at parties. So you may as well keep the Paganini.

As for your future plans if network audio remains a minor and less important pursuit and if you will never be interested in streaming services etc, then I would find the addition of Vivaldi Upsampler more difficult to justify financially. If course Vivaldi transport will upsample your CDs should such a thing appeal to you.

1 Like