artifact of DSD to PCM?

Topic: artifact of DSD to PCM? (Read 455 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

artifact of DSD to PCM?

2019-08-06 21:02:34

I recently had cause and opportunity to examine the same 5.1 recording released in DVD-Audio (96/24 PCM) and SACD (DSD) . The recording is Elton John's Goodbye Yellow Brick Road. The DVD was released before the SACD (and a BluRay version has since appeared as well). I was able to directly rip both respective discs (which I own) straight to digital. In the case of the SACD, the iso2dsd DSD rip was converted to 88/24 WAV using foobar2000's foo_sacd; the DVDA was ripped directly to the native 96/24 using DVDA Audio Extractor.

To compare them, I loaded the track 'Grey Seal' into Audition 3.0 and ran frequency analysis of the front left channel, as well as visualized the waveforms (for frequency analysis the 88kHz file was upsampled to 96kHz, to match the DVDA). Unsurprisingly, the frequency scan curves were virtually exact parallels out to 22kHz, suggesting no significant mastering difference (first attachment)

But the waveforms showed a difference I'd like to understand. They are essentially the same dynamically compressed audio, with the limiting (not clipping) to -0.5dBFS obvious in the perfectly flat top and bottom of the low-magnification DVDA waveform (2nd attachment). But the SACD (converted to PCM) waveform is 'fuzzy' there (3rd attachment). I'm wondering what could cause this. Would it be a normal artifact of my DSD-->88/24 PCM conversion, or something intrinsic to the mastering itself (e.g., was the audio taken through an analog stage as part of the creation of the DSD version)?

Re: artifact of DSD to PCM?

Reply #1 – 2019-08-07 01:23:00

I saw his happening to a degree when hard clipped material is resampled to no clipping levels.
My guess is the DSD version is based on this 24/96 or higher master used for the DVD.
The hard and flat peaks change the shape simply because of the PCM->DSD->PCM penetration and the added HF noise.
Maybe some other tweaks are applied that are hard to see in simple graphics to create the SACD sound.
I lately did stumble across DeltaWave from https://deltaw.org/ but didn't play much with it. There seem to be many options and one may be able to isolate the dirt in the peaks.
I wouldn't be surprised it is almost only the rising DSD noise but i have to admit i never digged into this special case.

Re: artifact of DSD to PCM?

Reply #2 – 2019-08-07 14:52:49

Just checked. I did hefty clip a 24/96 recording, encoded it to dsf and back with SoX but couldn't create such fuzz. I am not familar with foobars way of decoding it.

Re: artifact of DSD to PCM?

Reply #3 – 2019-08-07 18:30:12

I would guess it's just phase shift/intersample peaks.

Re: artifact of DSD to PCM?

Reply #4 – 2019-08-07 20:50:48

Quote from: Wombat on 2019-08-07 01:23:00

I saw his happening to a degree when hard clipped material is resampled to no clipping levels.
My guess is the DSD version is based on this 24/96 or higher master used for the DVD.

Yes, it seems virtually certain to me that the SACD was sourced from the DVD-A (or they shared the same digital master), given their virtual identity (in the audible region) in average frequency data, and similar compression.