DSD: Worst Practices
It shouldn’t be necessary to talk about the worst ways in which to produce and release DSD on disc or as files. But there are a few “high-end” record labels that see big dollars in releasing in the new “hot” DSD format at premium prices. I don’t begrudge others for trying to maximize their upside on older recordings that they may have in their vault. But it does irk me when they misrepresent these old standard definition recordings as something that they’re not.
Sony intended DSD to be an archive format. Our discussion over the past few days pointed out that digital transfers of older analog tapes works fine when using 1-bit DSD so long as you don’t have to do anything with the archival copy. I also presented the best-case scenarios for new DSD productions. The first uses live analog mixing to DSD and the second captures performances using high-rate PCM and then downconverts to DSD during the final output stage. Neither of these methods compromises the actual recording quality. But recording to a standard resolution format does.
Let’s start with analog tape. I mentioned T. Bone Burnett and his engineer Mike Piersante are avid analog tape fans. They choose analog 24-track technology because they prefer the “sound” of the format. No argument from me. They record to tape, the mix through an analog console and capture the output to either analog tape or Pro Tools at 96 khZ/24-bits. The records that T. Bone Burnett has produced are very successful and wonderful. And he believes they are high-resolution…but they’re not.
Other engineers and labels capture to multichannel analog tape, mix through an analog console (with PCM digital reverb), and capture to DSD 64 or DSD 128. The files are then made available for downloading at hefty prices. This “worst” case production technique guarantees that the final delivered file is as good as we had 30-40 years ago. Is getting the best possible archived version of an aging master worth $40 per album download? I certainly don’t think so.
The DSD format at 64, 128, 256 or even 512 is still highly flawed because it requires numerous conversions during the production process. And we know converting from one encoding format to another results in a loss of fidelity. The promise of DSD purity, the magic of “analog like” 1-bit encoding, and the avoidance of “brickwall” low pass filters aren’t achievable in the real world. The diagrams that continue to accompany websites and magazine articles show only the ideal case of DSD…which was abandoned over 10 years ago.
I’ve attended numerous DSD panels that hype the format and have promised that production tools are just around the corner for more than 5 years. It’s not going to happen. No one is working on 1-bit production tools because there’s no market for them. The best Merging Technologies and now Sound Magic can offer is a hybrid approach. They use DXD (PCM) to do all of the post production steps and then handle conversion back to DSD or any lower resolution PCM spec.
Why bother do all of this conversion? Just record in the best available high-resolution PCM format you have available and leave it in that format.
Mark,
Please post an article discussing 96/24 PCM and the digital filters needed to make it work well.
In particular, the filters companies such as Arye Acoustic design and use that don’t have much
ringing in the time domain and don’t have pre-echo, which is not a sound that occurs in nature.
Thanks,
Scott
Scott…thanks for the suggestion. I’ll try to get to that soon.
Don’t forget to take the frequency of pre-echo into account, because nothing matters if it is inaudible.