Since downsizing 5 years ago and leaving behind my LPs and vinyl playback equipment, I have gradually built a new digital music library. At Dr. David's request, I shared this traumatic journey in several articles titled Yes, There Really is Music After Vinyl, the first of which you can find HERE. Today, most new additions to my library come from digital downloads. And as a fanatic about sound quality, I've had to suss out a process that tends to lead me to the best sounding version of any given album. Once I identify a likely good new issue or reissue, then it's a matter of deciding what resolution offered for download will likely give the best sound in my system. This article is to share my current process of selecting that format and why.
Hint: it is not always the "highest" resolution.
I see a fair number of comments in online audio forums about "best file format," particularly as related to choosing what file to download. And it seems that a lot of conversation revolves around DSD256 versus DXD (352.8kHz 24-bit PCM). Many times, people who are fans of DSD reflexively download the DSD version of the file. I'm sure for many this is based on their experience of hearing something they like about this format.
I'm going to be the contrarian—all for what I hope is the good purpose of helping you better discriminate what format is likely to sound better in your system.
First, let me stake myself out. I fully endorse the notion that Pure DSD256 is the finest sounding digital audio format we currently have available to us. It is just observationally closer to what the microphones hear than any other format.
But, my listening experience has convinced me that many of the differences we hear are related to how our equipment handles the digital signal stream, whether PCM or PDM (the format that includes DSD). The hardware and the algorithms that decode the digital stream vary from piece of gear to piece of gear. Some do better handling the PCM stream, others the PDM (DSD) stream. Some convert everything to PDM as part of their internal processing, as is the case with my Playback Designs MPD-8 DAC (circa 2019).
And, I've become convinced that sound differences are also impacted by the numbers of file conversions between the earliest capture of the signal and what we now have to hear. Every conversion in format leaves artifacts. The fewer conversions, the better the sound quality. If this sounds a lot like sound reproduction in analog, you're right.
Because there are so many variables, to keep this as simple as possible, I'm going to limit my comments just to Bob Witrak's High Definition Tape Transfers (HDTT) releases in what follows...
Bob releases some Pure DSD256 albums transferred from the source tape (often 2-track 15ips) directly to DSD256 using a Merging Technologies analog-to-digital converter (ADC). He releases the resulting DSD256 file without any PCM processing. It is Pure DSD256 from the analog tape. These are frequently the best sounding files HDTT releases. He will additionally convert these files to various flavors of PCM (96kHz, 192kHz, and 352.8kHz) to meet market demand. Most of these Pure DSD256 albums are jazz recordings because he has more success finding clean sounding jazz source tapes that can be released without further post-processing. For me, these are the "Gold Standard" for sonics among albums in the HDTT catalog.
But I am a classical music lover. And I love his catalog of classical recordings, with some of the greatest recordings ever made. So, what happens here?
Many of the classical source tapes need work. Not all—but the tapes that can be released without some post-processing are the exceptions, even the direct dubs from studio masters. So, Bob starts as he always does: a DSD256 transfer of the source tape. Then it gets converted in the Pyramix Digital Audio Workstation software to 352.8kHz/32-bit PCM ("DXD") for clean-up. The type of cleanup and the software tools used in doing this will vary depending on what work is needed to bring the final product up to Bob's standards. Usually the issue is background noise that needs to be filtered. But sometimes it is pitch, sometimes a bit of frequency balance adjustment here or there. This is all done by ear to Bob's tastes. If the project requires some particularly intensive post-processing help, the work may get outsourced to the very capable John Haley, of Harmony Restorations, LLC, a restoration magician if ever there was one. Again, the work is all by ear.
Once a final PCM edit master has been completed, the file is ready for publishing in any of the various formats released by HDTT: 352.8kHz/24-bit (DXD), 192kHz, 96kHz, DSD64, DSD128, DSD256. The web page for the release ALWAYS tells you whether the file has been post-processed in DXD. If it is Pure DSD, the page will clearly state that.
Now…, why am I going into all this long detail? So that this next statement may make some sense.
In my primary system, from among the various HDTT releases, I always hear what I consider to be the most accurate music reproduction when I listen to the edit master.
If the edit master is Pure DSD256, then that's the one. But if the original DSD256 transfer from the source tape has been post-processed in DXD to do some cleanup, then it is this DXD file that is the edit master. That DXD edit master file is then the one I want to hear. If Bob were to provided the 32-bit version of that DXD file (as NativeDSD does on some releases), then it would be the 32-bit file. Heard through my Playback Designs MPD-8 DAC, the edit master DXD file will sound better than the DSD256 file that Bob has created from that DXD edit master file.
Why? Two reasons: (1) one fewer file conversion, there are always artifacts; and (2) my DAC is going to convert the DXD file into DSD256 automatically and it does a better job converting from DXD to DSD than does the Pyramix software Bob uses.
So, for me, in my system, the DXD edit master file will sound better that the DSD256 file derived from that DXD edit master via Pyramix Album Publishing.
On the other hand, in my wife's office system, using a Teac 508 DAC, the HDTT provided DSD256 file sounds better. Why? Different DAC, different internal hardware, different algorithms. The Teac DAC is optimized for playing DSD and just sounds better playing a DSD file.
All of this is just to say: everything we hear is system dependent. I believe DSD256 is objectively superior in sound quality to PCM. But what we actually hear in our systems is system dependent, not absolutes.
After much comparative listening, for my primary system, this is my selection tree:
- First, the Pure DSD256 file, if available
- Second, the edit master file, in whatever format it exists (96kHz wav or DXD, I want the source—and, yes, there are Pure DSD edit master files that have been mixed in Pure DSD, read HERE.
- If both a DXD and a DSD256 file are offered and if I'm pretty sure the file has been post-processed (as most will have been), then the DXD is likely superior in my system because my DAC does a better job converting to DSD256 than the ubiquitous Pyramix Workstation. But for my wife's office system, then the DSD256 will likely sound better in that system.
Now, in your system, this may not be the best selection tree. If your DAC is based on a chip of some sort (Burr-Brown, ESS or AKM, for example) and particularly if it touts playing DSD512, you may get consistently better sound by always choosing the DSD256 or DSD512 file. I've been told by folks who I believe know what they're talking about that DSD512 capable chips, particularly, are optimized for DSD playback. And they often will sound a little cleaner playing DSD512 files than playing DSD256 files because of the way the chip is filtering the data stream. (Note: this does not mean that DSD512 sounds better than DSD256 in any absolute sense. It only means that when played using these DAC chips, the DSD512 may sound cleaner that the DSD256 in that chip.)
But, if your system includes a DAC that internally converts all signals to DSD256 as part of its internal processing chain, you might be like me and find that your DAC does a better conversion to DSD256 than getting the Pyramix processed DSD256 file that comes from a DXD edit master file.
Whew! Got all that? What a mess. But this is no different than everything we navigate in the vinyl playback world.
The only way to have a clue about what will sound best in your system is to do your own listening tests. I've said this over and over in other articles. Today, due to Bob Witrak's very kind support, I am able to take this one step further and offer you two "free to download" comparison tracks. Download them and listen critically. See what sounds best to you in your system.
Comparison One
A Pure DSD256 edit master track compared to DXD of the same track ("Marche du soldat" from the album Stravinsky L'histoire du Soldat, Ars Nova, 1956). Download free sample tracks here.
The original album can be found HERE.
Comparison Two
A DSD256 track derived from a DXD edit master track ("Angel Eyes" from the album Frank Sinatra Sings For Only the Lonely, 1958). Download free sample tracks here.
The original album can be found HERE.
These tracks are provided by the generous support of High Definition Tape Transfers.
These tracks have not been cherry picked to prove any point. The selection is rather random based on something that would have a range of instruments, have some general musical interest, and be under 4 minutes long.
My suggested next challenge for you is to listen to a 96kHz edit master file and compare it to the upsampled DXD iteration or DSD modulation. None available at HDTT, however. For this you will need to explore the NativeDSD catalog. Look for recordings from Linn Records or Alpha Classics or Fuga Libera. If there is enough interest, we might be able to get comparison samples for this as well. Let us know in the comments section.
I hope your explorations prove fruitful for your future listening enjoyment.
