Phil, as you know doubt know, Cookie Marenco at Blue Coast has been leading a small crusade to bring transparency and integrity to digital files. It may be a lost cause, but I hope not.
Thanks for your insightful posts, and welcome back to dCS, Phil.
I wasnāt aware of this but itās good to know Iām not alone in my desireā¦ 
chuckle thanksā¦
Yes, I know, just the the morning ramblings of a buffoon ā¦
Phil
This has been a very interesting exchange. The only comment I can add, since Somethinā Else (recorded in 1958) was mentioned, is that for both CDs and SACDs the quality of analogue source and of the digitisation process can make such a difference to the final result whether it is 44.1 kHz PCM or basic DSD (Iāve never heard the āfancyā kind). Sort of obvious, but it can be a mystery to the music purchaser what he/she is buying and whether higher resolution is genuinely better in any particular case. For example, I have both the Verve SACD of the Cole Porter Songbook, sung by Ella Fitzgerald, and the DCC CD reissue. The latter wins hands down in terms of the quality of sound. I think Iāve seen the engineer, Steve Hoffman, claim with regard to another recording that he used the original master tape to produce the DCC CD whereas the record company only used a copy for the SACD. Perhaps that is what happened in this case.
So, in conclusion, as a listener Iām more concerned about the quality of source, and of the A to D conversion, than I am about very high resolution formats. I expect similar quality considerations apply to digital recordings. I think it is not controversial to say that a good SACD will beat a good CD (or similar, for the streamed equivalents), all things being equal. But sometimes they are not: hence my experience with the Ella Fitzgerald recording.
3 Likes
Just adding some background to @Philās statement, because I think Paul is quite knowledgeable and have learned a lot from his videosā¦ Paul states, to be precise, that Octave Records records in native DSD and they can move the native DSD recording to PCM for editing without any losses (via Zephyr filter). However, he states that they incur minor losses on the return to DSD because there does not (yet?) exist a completely lossless SDM. He also states that during this process, aggregate losses are less than using PCM. I donāt have the technical knowledge to prove or disprove these statements:
Looking forward to it
Unfortunately, I have to disagree. Throughout his YouTube videos over the years, Paul often makes āfactualā statements that are not factual at all. 
In that video you linked for example, he states - āDSD is the best recording medium on the planet, it just isā. From an Engineering point of view, DSD vs. PCM has been disputed for as long as DSD has been around, for recording or otherwise. There are no facts of one being better than the other.
The dCS folks would know for sure, but I believe mathematically, thereās no way to convert from DSD to PCM completely losslessly (audibility of the losses aside).
So, Paulās discussion about the proprietary Zephir filter converting āperfectlyā is highly suspect.
If Iām not mistaken, the Zephir filter is based on algorithmās within BitPerfectās DSDMaster application, for which there is little technical details, but this is what they have to say about DSD to PCM conversion;
Unlike conversions within the PCM format ( WAV to FLAC, etc ), conversions from DSD to PCM - or the other way round - cannot be done losslessly. This is a mathematical fact, even though some otherwise well-informed individuals can be heard to assert otherwise.
1 Like
Thanks for your insights @Anupc, as always.
To be fair, even though he asserts it strongly, I interpreted Paulās DSS v PCM comment as his opinion.
I am definitely interested in the factual statement/question of whether DSD can be converted to PCM losslessly via Zephyr. Maybe @Phil or @Andrew can shed some light?
2 Likes
Iāve never met Paul, Iād love to and Iād love to get to just spend a couple of days chewing the fat with him because I think it would be interesting.
Thereās a fair chunk of what he comes out with that I personally disagree with but thereās nothing wrong with that - Iām happy for someone to have a different opinion and different viewpoint, some things are provable and others arenāt and this is a hobby ā¦ I used to disagree with a lot of what Max Townshend (God rest him) used to say/do but that didnāt stop us having some awesome conversations at various events (or in the bar afterwards) and me thinking he was totally a top bloke.
OK - so MY UNDERSTANDING is that it isnāt possible to losslessly convert between the two formats in either direction but if thereās any documentation to the contrary that anyone can point me to then Iām happy to take a look and re-evaluate with an open mind.
P
2 Likes
Thank you @Phil, yours was my understanding too.
This is what Paul wrote on his blog in January, 2022, and August, 2021, respectively. I donāt have any documentation supporting or negating:
January: Itās a look-forward filter that can convert DSD to PCM with zero loss. It wonāt work in real time which means we have to record in DSD and edit and do all that stuff in DSD to keep the quality of the recording where we want it. Then, once all the edits have been done (and overdubs etc.) we use the Zephiir filter to perfectly convert to PCM and we then can do the mixing in the digital domain without having to suffer the degradation of the analog domain mixing and conversions."
August: "It is true most modern ADC/DAC use a multi-bit version thatās sort of like PDM but is in actuality PCM. They definitely use SDMs (Sigma Delta Modulators) and therein can lie many problems. Look in the last or second to last issue of Copper Magazine where Richard Murrison gives a much better explanation than I.
The A/Ds we are building for Octave studios are running at 128fs and then rely upon the TI ADCās internal SDM to produce the 2X DSD PDM stream. Once we have that we will use Richard Murisonās Zero phase decimatorās low pass function to then produce a perfect 352kHz PCM version of the original 128fs DSD capture. This then allows us to edit and mix in a souped up DAW without any sonic loss. The final 2-channel master then runs through the same A/D and we get the final 2-channel master in 2X DSD which we can then do what we want with it.
This will be an absolute first and cannot wait to share the results with folks."
Thereās some interesting stuff being said and also not being said here and I think both need to be looked atā¦
(I donāt have any issue with DSD as an actual format or with Octave just to be totally clear - Iām just looking through whatās been listed as their processā¦)
So, as a āfirst skimā extract we see that Paul says that they can convert from DSD to PCM with zero loss although it isnāt a real time process but hey, thatās cool if they can.
He then says as it cant be done in real time they have to āedit and do all that stuff in DSD to keep the quality of the recording where we want itā ā¦ well, the only editing you can do in DSD without losing quality is, IIRC, straight cuts, copies and pastes. I donāt think even level adjustment can be done without loss as DSD so I donāt see an actual advantage here over doing that as part of the PCM phase (which apparently he says you can get to losslessly anyway) ā¦ I maybe smell a bit of gilding the Lilly here myself, perhaps trying to suggest that less is being done in the PCM domain than in the half-a-veil-better DSD domain? 
We then skip any actual detail of all the mixing and editing work being done as PCM but he does note that doing editing as analogue would involve degradation (so Iām going to take from that that mixing it as PCM DOESNāT involve degradation while looking a little quizzical at the inference that perhaps actually EDITING in PCM does involve degradation?)
OK so what I read theyāre doing DSD capture of the analogue source but then going straight to āperfect 352kHz PCMā which can be edited and mixed in their āsouped up DAW without any sonic lossā so here weāre saying that editing and mixing in PCM is sonically losslessā¦
I presume their DAW tops out at handling 384kHz/24bit or maybe 768kHz/24bit ā¦ we were using 192kHz/24bit internally for editing some 25 years ago in the vocal studio I managed, yes itās moved on but not exactly earth shatteringly so given back then we were using Pentium P133ās with maybe 16megs of RAM (megs, not gigs) and now we have some monstrous hardware, RAM and storage to fall back onā¦
They then output that as analogue audio from the DAW and run it back through their DSD capture device again (rather than doing a convert on the data) to create the final DSD format file ā¦
Iām left wondering whether their work ultimately is as applicable to both PCM and DSD - especially as their output from their editing station seems to be being converted back to analogue and then resampled ā¦ wouldnāt that make the PCM source data āmore accurateā than (or at least āas accurate asā) the DSD resampled version?
As far as I see it the main thing here would be the application of gentle, sympathetic and intelligent editing and mixing rather than clumsy and heavy handed gain riding and audio compression that is normally applied to general mass market stuff - not the actual file format - but isnāt that the point? Isnāt that EXACTLY what we want, the best quality audio, or am I missing something?
Phil
Glad that you are having fun with it still
1 Like
Hey Jon, are you still back in the stone age rockinā your CDS 
btwā¦itās Colin from KL
Thatās the point.
1 Like
Thanks for that. Looks like Paulās reference to Richard Murisonās article is this one;
https://www.psaudio.com/copper/article/dsd-is-it-pcm-or-isnt-it/
Which is a back-to-basics on what DSD is. Nothing there about lossless DSD-to-PCM conversion.
So I did a quick search. Looks like Richard Murison first covered DSD on PS Audioās Copper Magazine much earlier into the Magazineās inception, in a three part piece back in 2015 (Part 1, Part 2, Part 3).
Part 3 is especially relevant with this paragraph;
In effect, by passing the DSD bitstream through a low-pass filter, we end up converting it to PCM. This is how DSD-to-PCM conversion is done. You simply pass it through a low-pass filter. The resultant PCM representation can be very close to a perfect copy of the original signal, limited only by the accuracy of the low-pass filter used.
In order to get PCM from low-pass filtering DSD, one has to use a decimation filter - which explains the āZero phase decimatorā that Paul alludes to.
Thatās pretty much exactly how practically everybody* else converts DSD-to-PCM! 
Doing it non-real-time gives an opportunity to increase the FIR filter tap length substantially, and/or multistage (both achievable in-real-time with an appropriate FPGA or ASIC).
No, not by a long shot. PS Audio is not doing anything new or special, and no, itās not a lossless conversion. As usual, Paulās just playing with words to make it seem special
ps * : By āeverybodyā I mean Audiopraise, Aul Converter, Signalyst, SoX-DSD, etc., etc.
Thanks @Anupc and agreed. Iām curious about this equation:
(A) Losses from (Analogue to DSD recording -->DSD to PCM for editing ā PCM to DSD āfinalā)
= Greater or Less than
(B) Losses from (Analogue to DXD recording ā Editing ā DXD āfinalā)
Basically if one believes that DSD or 2xDSD has a higher initial capture rate than DXD, would the losses in round-tripping DSD to PCM and back, still lead to higher fidelity than keeping everything in PCM?
I guess you are going to just tell me āit dependsā ; )
The thing is that whatās being concentrated on here is the file format - DSD vs PCM (vs FLAC etc.) - and what the differences between the formats means to to final product, but thatās kind of like comparing the colours of the bucket carrying the content and debating what the colour of the bucket means to the final product (the audio file) rather than the content (the audio file) itself.
Surely whatās far more far more relevant is the actual content and what is done with the content?
The biggest influence on the quality of what you 're listening to is the post processing / editing / mixing of the audio and not the colour of the bucket thatās holding it and thatās where these small studios come in, they will generally have a process that is more targeted at making the best quality product rather than churning out cookie cutter blandness, think āMamaās home made lasagneā compared to a supermarket frozen microwave lasagneā¦
Thatās where I see the real value of these small studios is - hopefully making a great quality product which is what we audiophiles are really afterā¦
Phil
3 Likes
Iād have to agree wth Phil, I donāt think the capture/recording format (or ārateā per se) matters as much as the quality of the recording and the mixing/mastering.
That said, given the two paths you posed - one involving DSD for recording/delivery, vs. staying purely in the PCM domain - technically, all else being equal, I think the pure PCM path has a better chance of being least harmful to the source (assuming some mixing/mastering is returned).
Problem is, all else is seldom equal, so one ends up being ābetterā than the other.
For example, with all the commercial interest PS Audio/Octave Records has in DSD - including recoding equipment, their flagship DAC architecture etc. - I think they would struggle to differentiate if they were to swing fully to PCM for all workflow. So, they have to find some hybrid mode retaining DSD somewhere in the chain.
By the way, I donāt have a personal preference of one format versus another (which is why I picked dCS 20 years ago 
)
1 Like
Thanks - although a good disagreement can spark off some valuable discussion and debate if you keep an open mind.
IMO the underlying format of the file (within reason - weāre not talking 64kbit/sec MP3 here) has a disappearingly small influence on the quality of what you actually hear and I strongly suspect that few people would be able to reliably identify the same source material just presented in a few different formats ā¦
Ultimately the file format itself isnāt any indication of the quality of the actual audio that it contains - Iāve heard some shockingly good 44.1/16 material over the years and some absolutely dire high-bitrate stuff so, as always, listen with your ears not with your eyes.
As a manufacturer then they will constantly have the same problem that all manufacturers have and that is the availability of really good source material so - just as with some other manufacturers out there - actually having your own label / studio can be a good thing (itās rarely usefully profitable though).
But if the result is that there is high-quality material out there for the rest of us to enjoy then thatās awesome!
Personally there have been so many albums that have simply disappointed ā¦ U2s āJoshua Treeā I always felt simply didnāt āget goingā ā¦ it kind of suggested at first that it would, I always felt it should have been far cleaner and much more dynamic than it was, instead it ended up a squishy mush, same with Madonnaās āMusicā - that was appallingly produced.
Conversely Massive Attackās āProtectionā I bought many years ago and couldnāt listen to on the system that I had at the time because it sounded awful yet when I got myself a much better setup it really came to life and was most enjoyable.
Of course the obvious answer is to not compress and not gain ride music when itās being post processed but then you have the problem of the old (was it Telarc) CD of the 1812 Overture - the one that was issued without any dynamic range compression and had a warning on it that it contained music of high dynamic range. The label actually printed a disclaimer on the front saying that they werenāt liable for any damage caused ā¦ yeah great it had huge (ish) dynamic range but it was TOTALLY unlistenable because you ended up constantly turning the volume up and down so thereās a sweet spot that needs to be hit and it isnāt easyā¦
P
3 Likes