Yep, which is why the script scans each track individually without making assumptions, as I mentioned a few posts back.
-
WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!
The sound quality of DSD?
- Thread starter Saidera
- Start date
Bow_Wazoo
Addicted to Fun and Learning
- Joined
- Jan 3, 2021
- Messages
- 783
- Likes
- 684
What do I do ideally with DSF / DSD files that I can not get in any other format?
Convert to Flac 24bit / 96khz?
Is there a program that is particularly recommended for conversion?
I'm not comfortable with all this high-frequency garbage....
Convert to Flac 24bit / 96khz?
Is there a program that is particularly recommended for conversion?
I'm not comfortable with all this high-frequency garbage....
What do you mean? What high-frequency garbage?
What SW are you using for playback?
If you have software like Foobar (with plugin) or Roon, or many others, it will playback DSD for you. Most DACs will accept it and either convert it to PCM internally, or play it directly and convert it analog. As part of the conversion to analog, the "high-frequency garbage" gets filtered out.
Bow_Wazoo
Addicted to Fun and Learning
- Joined
- Jan 3, 2021
- Messages
- 783
- Likes
- 684
This is what I mean.
Probably insignificant.
But as far as I've heard, the experts disagree about whether it has any effect.
My Neutron player has no problem with playback and conversion.
I do signal processing because of EQ.
Neutron converts there in real time in PCM, but the high frequencies are not touched.
Probably insignificant.
But as far as I've heard, the experts disagree about whether it has any effect.
My Neutron player has no problem with playback and conversion.
I do signal processing because of EQ.
Neutron converts there in real time in PCM, but the high frequencies are not touched.
mkt
Senior Member
- Joined
- Jan 6, 2020
- Messages
- 338
- Likes
- 465
If you want to convert to PCM, I don't think there is a 100% correction answer because the conversion is lossy and there is the choice. of filtering. There are probably a bunch of methods that are fine. I have used sox and Saracon. (To me the most tedious point is choosing the gain. In theory 6db should be right. In practice, not always.)
Some discussion here (probably already cited in this thread)
ANALYSIS: DSD-to-PCM Conversion 2015 - Windows & Mac OS X
A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.
archimago.blogspot.com
MEASUREMENTS: A look at DSD and using SoX-DSD as a standard for test signal conversion. The 1/10-decade Multitone 32 test. And retro-pop remixes... ;-)
A blog for audiophiles about more objective topics. Measurements of audio gear. Reasonable, realistic, no snakeoil assessment of sound, and equipment.
archimago.blogspot.com
If you don't want high frequency garbage, then your best bet is to convert everything in your library to 48khz/24bit PCM at time of playback/in your music software.
That should get rid of pretty much everything above 20khz - aka garbage - leaving a bit of headroom.
Last edited:
"Ideally?" People have different ideas about that.
But keep in mind that DSD was originally designed to be converted to PCM at integral multiples of 44.1kHz SR. (44.1, 88.2, etc).
(This is an "ideal";but if you choose to convert to the other common PCM SR series (48, 96, etc) it's very unlikely you'll hear any deficit)
Not necessarily. It depends on whether the hardware (or software) has a final lowpass filtering step.
DSD has tons of ultrasonic hash. Obviously, proper converting of DSD to a chosen PCM sample rate will involve filtering out what's above its half rate. That means if you convert to any SR value above 44kHz, any ultrasonic noise above 22 kHz remains, up to the half rate of the value.
It helps to know the distribution of noise in the DSD file. (If it concerns you.)
Hardware SACD players have (had) a lowpass filter as the last output step to filter out some of this. 50 kHz cutoff was common.
Last edited:
julian_hughes
Addicted to Fun and Learning
- Joined
- Aug 23, 2020
- Messages
- 657
- Likes
- 903
You can actually do it quite easily with ffmpeg. See my older post at https://audiosciencereview.com/forum/index.php?threads/dsd-is-it-of-any-value.10682/post-756813
Currently I use a tiny shellscript (am on linux, but this would likely work fine on Mac and could very easily be adapted into a windows batch file):
#!/bin/bash
for i in "$@" ; do
ffmpeg -i "$i" -af volume=4dB,aresample=resampler=soxr:precision=28:dither_method=lipshitz:dither_scale=0.5:cutoff=1 -ar 44100 -sample_fmt s16 "${i%.*}".flac
done
If you prefer a 24-bit output or different sample rate then you just change the argument i.e. "-ar 44100" might become "-ar 48000" and "-sample_fmt s16" might become "-sample_fmt s24"
If you read that older linked post you'll see something about why I settled on these particular options. I find the conversions indistinguishable from the originals to my ears, and at least as good as the SACD's CD layer. I usually convert from my own SACDs, ripped to iso, then the 2 channel audio extracted to a DSDIFF Edit Master file with cue sheet, then I convert that large DSDIFF file with the above script and only then split into tracks with the cue file. The result retains true gapless playback, but if you convert collections of single tracks then this cannot be guaranteed as each of reducing bit-depth and/or reducing sample rate is inevitably not lossless.
julian_hughes
Addicted to Fun and Learning
- Joined
- Aug 23, 2020
- Messages
- 657
- Likes
- 903
DSD has a much lower level than red book. When converted to/via PCM the peak level is about 6dB below 0dBFS, so using a scale of 4dB leaves a full 2dB of headroom. I've never achieved clipping using that scale. If you use a 6dB scale you probably will. 5dB? Maybe. 4dB is fine. You won't get any clipping using the above script. And it's hard to see how going from 24-bit to 16-bit, dithered and sample rate converted, might be "wasting bits".
edit:typo
I've started to upsample everything in roon to DSD512.. my DAC (Smsl D300) supports DSD512.. so I hope (still reading about all this stuff) that upsampling to DSD512 in roon might offer better results than the DAC doing some pcm upsampling.
The DAC has a cutoff filter for DSD anyways.. so it cuts off everything above 52khz in the highest setting anyway.
Last edited:
My Smsl D300 supports DSD512. The DAC has 3 cutoff filters. 13khz, 26khz and 52khz. That should get rid of the high frequency garbage anyway shouldn't it?
Archimago has done lots of tests on DAC PCM/DSD upsampling.
Absolutely no audible difference either way, no matter how you slice it - not with modern DACs.
It's rare that raw DSD-->PCM conversion will be clipped (the only cases I've seen are when the SACD was improperly authored -- Michael Jackson's Thriller. Its instantaneous clipping on a couple fo tracks and it's inaudible ). More commonly you're 'wasting bits'. I do it conversion foobar2k's SACD plugin (which is also a DSD-->PCM converter). It lets you lower PCM output level by as much as +6 dB in 1dB increments. I start at +6 and check for clipping in the Console as its converting; if a flag goes up, I set it to the next. lowest level. Repeat as necessary (usually just once if at all)
let's review
benchmarkmedia.com
Audio Myth -"DSD Provides a Direct Stream from A/D to D/A"
This myth goes something like this: "DSD provides a simple and direct digital path between the A/D and D/A." DSD does not deliver a simple and "direct" path.
I have a question. I understand Nyquist and how given two sample points there is only one solution to "draw" a sine wave through them, but how does the DAC actually fit the sine wave to these two points? How do they correctly interpolate?
Similar threads
