Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Why do DSF files not sound as good as the SACDs they were made from? (Read 1260 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Why do DSF files not sound as good as the SACDs they were made from?


I recently started extracting my SACD collection to DSF files so I could have the awesome SACD quality sound available to me wherever I go. 

I wanted to check the resulting DSF file quality against the original SACD to see how they sound so I did a test:

I first played the original SACD in an Oppo UDP-205 and ran it into a high-quality recorder.
I then played the extracted DSF file off a USB drive in the Oppo UDP-205 and ran it into the high-quality recorder.

I then compared back and forth between the SACD and the extracted DSF file at the same exact levels.
I did this test with several SACDs (only stereo SACD layers as I am not going to remotely be listening to surround SACD).

What I hear is that DSF files sound good but not as good as the original SACD.
The DSF file sounds a little lower-res, not as "open" as the SACD, as though it went through some process.
Of course, the DSF file did go through a process to get extracted from the SACD, but it seems like the sound quality change should be less noticeable.

I am using SACD Extract on Windows 10 and a Sony BX510 to extract the DSF file.

I am new to extracting SACDs. I don't really see any options to change how the DSF files are made. I have padding selected, but I don't know that it would change the general sound of the files. I also see that the DST De-Compression is permanently selected in SACD Extract and I don't know if that would change the sound quality but there's also no option to change that selection and an SACD with DST would need de-compression somewhere along the line before playing back.  (These last points are just comments as I don't know that they have any bearing on what I am hearing.)

Any advice on this would be appreciated. I feel that what I have with DSF is nice, but it should be even closer in sound to the original SACD.  If there is anything I can do to improve my DSF files to make them closer sounding to the SACD I'd be interested in trying it.


Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #1
It's possible that the extracted DSD is not low-pass filtered (as it probably is with the player) and the ultrasonic noise (inherent to DSD) is mucking-up the recorder.

Quote
I then compared back and forth between the SACD and the extracted DSF file at the same exact levels.
That's good but you skipped the blind part and the statistical repeatability.   ;)

Quote
The DSF file sounds a little lower-res, not as "open" as the SACD, as though it went through some process.
I don't know what you are hearing but "loss of resolution" is heard as quantization noise.  If you make an 8-bit WAV file with Audacity you'll hear it.    Like regular analog noise, quantization noise is worse with quiet sounds, but unlike analog noise it goes-away completely with "digital silence".

...Oh, if you make an 8-bit file with Audacity, turn-off dither.   Dither is added noise that's supposed to sound better than quantization noise. 

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #2
What DVDDoug said.
First, unless you have blinded your tests, your eyes will fool you. We are wired that way. And this forum requires you to actually check before making claims whenever that is possible, terms of service item #8.
Since you have recorded files, you can easily do that test. Use e.g. the foobar2000 player with the ABX comparator. https://www.foobar2000.org/components/view/foo_abx

So try that, and if you still can distinguish - then we are off to possible explanations like the noise from the DSD format. (It is not a good format at all - but done right it is good enough, and sometimes, good masterings are only available in one single fomat, so SACDs might sound stellar for that reason.)


As for DST, it is a lossless compression, so it doesn't matter. DST-encoded material must be uncompressed somewhere in the playback chain anyway.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #3
When DSD goes wrong on hardware player...
https://goldensound.audio/2022/04/06/benchmark-dac3-b-measurements/

One thing to consider is that ABX test is used to verify audible difference, but not preference. A marginally distorted version could be identified as "warmer", more "analog like", more "inner details" and so on.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #4
Thank you all for the bright input and the opportunity to put this to some testing.

I used the Foobar2000 ABX Comparator to do some tests of the Michael Jackson Thriller SACD song "Billie Jean".   I compared the SACD disk vs. the DSF file I extracted.  I had three buddies get tested who are kind of into sound.  I did 16 "trials" on each of them.  The results:

Listener 1: 11/16
Listener 2: 10/16
Listener 3: 10/16

I am not a statistician and those results don't seem very conclusive, but each person tested said they could tell a difference, even if it was subtle.  

It makes sense to me that if one took a file from a SACD and put it into another format there would be some very small change to the quality.   After all, the file is going from a spinning disk being read by a laser, through an ethernet cable, through a network switch, another ethernet cable, and then into a PC, and getting processed in some way to output it as a DSF file.  

Are there any other methods or variables in extracting the DSF file that I might try to improve my results?  For example, I am using SACD Extract with a Sony BX510.  Is there any reason one would get a different result using an Oppo to extract?  (I wouldn't think so, but I don't know all the technical details and maybe someone has experience or more technical information on this,) 

I'm not ready to try magical amulets, but there might be something more down-to-earth that could improve the process.  For example, when working with SACD Extract I am converting directly from SACD to DSF.  Is there any error correction in this process, such as the error correction that exists when using "Exact Audio Copy" to extract CDs to FLAC?  Would it be better to convert from SACD to ISO and then from ISO to DSF to reduce possible data errors?

There was mention of using a low-pass filter, but it's not clear to me how I would implement that.  I don't have software for processing DSF audio.  It seems that the SACD disk and the DSF file would both be processed in the same way with the same filtering upon playback from my Oppo.

Cheers.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #5
So try that, and if you still can distinguish - then we are off to possible explanations like the noise from the DSD format. (It is not a good format at all - but done right it is good enough, and sometimes, good masterings are only available in one single fomat, so SACDs might sound stellar for that reason.)

Thank you for the ideas on this.  Can you elaborate on why you feel DSD is not a good format at all?  I'm interested. 

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #6
Quote
Listener 1: 11/16
Listener 2: 10/16
Listener 3: 10/16

I am not a statistician and those results don't seem very conclusive, but each person tested said they could tell a difference, even if it was subtle. 
Here is a probability chart.

Not "terrible".   You'd get 11/16 about 10% of the time by guessing and 10/16 about 20% of the time by guessing.

But usually statisticians like to be around 95% confident (or better).

Were the listeners conferring or was it a "secret ballet".   It's interesting that all 3 results were so close.   I don't know what that means statistically.   But intuitively, I'd expect more variation unless everybody was near 50% (everybody "guessing") or everybody was nearly 100% right.

But if you can't get it right almost 100% of the time, is it worth worrying about?

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #7
It makes sense to me that if one took a file from a SACD and put it into another format there would be some very small change to the quality.   After all, the file is going from a spinning disk being read by a laser, through an ethernet cable, through a network switch, another ethernet cable, and then into a PC, and getting processed in some way to output it as a DSF file.
It doesn't make sense to me. It's digital, the DAC receives exactly the same data no matter how it gets there. The only "processing" is removing the SACD TOC and inserting the DSF header, and neither of those will be sent to the DAC.

It seems that the SACD disk and the DSF file would both be processed in the same way with the same filtering upon playback from my Oppo.
I certainly hope so, but I don't know if I would count on it...

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #8
Here is a probability chart.
A successful ABX test, according to the chart above, should be below 0.05, for example:
Why 13 trials?
A p-value of <5% is reached with 10/13
So, if different listeners, multiple trials often show a trend of < 0.05, then the chance of audibility is high.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #9
If there are three listeners doing independent trials (rather than switching back and forth while they discuss), then this is 31/48, and the chances that cointosses would produce 31 or more is < 3 percent. If this independence between listeners holds up, we should conclude that the signals are audibly different, even if we cannot conclude that any single of the three listeners is able to distringuish.

But since one was sent to a hardware player, they were sitting in the same room and listening, while someone would change source (so effectively it isn't double-blind)? Did they have the opportunity to discuss?


If there was indeed a difference, then first guess is to put it down on the noise in the DSD signal, and the Oppo being able to filter it away when it received it from the SACD but not when played from file. If it got the file delivered as PCM (no "pass-through") then it isn't the Oppo's fault for not knowing.

As for noise in DSD, look at the link @bennetng posted, or the charts by the same user in this thread: https://hydrogenaud.io/index.php/topic,123597.0.html

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #10
Thank you. I need to consider these notes further.

The tests were done independently as double-blind tests using Foobar2000 (using the ABX comparator).  The three people tested did not get tested at the same time and were not in the room where the test was done at the same time.

As regards how the Oppo plays back the media, I am not familiar with any option that would filter the SACD playback but not the DSF file playback.  (But this certainly would cause a perceived sonic difference if different filters are applied.  If there is any suggestion on how to determine a difference in Oppo playback filtering I'll try it out. )

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #11
The tests were done independently as double-blind tests using Foobar2000 (using the ABX comparator).  The three people tested did not get tested at the same time and were not in the room where the test was done at the same time.
Sounds like test is passed.

As regards how the Oppo plays back the media, I am not familiar with any option that would filter the SACD playback but not the DSF file playback.
How is your playback chains concerning DSF files? Do you have DSD passthrough to the DAC? (Does fb2k even support that?!)
If whatever player makes PCM out of it, there is no way the Oppo can know that it was DSD.

(But this certainly would cause a perceived sonic difference if different filters are applied.
That should rather be "could" than "would". The filtering is not supposed to kill audible tones - but will filter off noise that could cause audible distortion. Intermodulation distortion shows up at a different frequency than those sounds that "intermodulate".
I don't know if the "volume of the noise" is enough to strain your playback chain, but if it were completely safe, then I doubt the SACD industry would have implemented that filtering.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #12
In order to create files to do an ABX test I ran the analog output of the Oppo 205 SACD and DSF file into a digital recorder. 
I used Pro Tools running at 96 kHz, 24-bit to capture the analog output of the Oppo 205.  This way I had two WAV files I could put into Foobar2000 to do the ABX tests.   Any change to the sound introduced by capturing the audio in this way was consistent for both the SACD and the DSF file.

It is possible that the engineering of the Oppo 205 might cause an SACD to sound slightly different from the DSF file.  I don't know enough about the circuitry.

As noted, any differences are very subtle.

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #13
In order to create files to do an ABX test I ran the analog output of the Oppo 205 SACD and DSF file into a digital recorder. 
But, how did you feed the Oppo the DSF file? By a USB memory stick ... ?

I used Pro Tools running at 96 kHz, 24-bit to capture the analog output of the Oppo 205. 
OK, so the difference isn't due to 60 to 80 kHz noise then.

This is getting weird. Were the WAVE files equally loud?

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #14
I put the original SACD disk into the Oppo 205.  I recorded the analog output into Pro Tools to make a WAV file.

Without changing any settings or cabling or any levels, I put a USB drive into the Oppo 205 and played the DSF file (which I had extracted from the SACD disk).  I recorded the analog output into Pro Tools to make a WAV file.

The levels are the same.  I loaded the two resulting WAV files into Foobar2000 to do the ABX tests. 

(Unless one had two identical Oppo players to play back both sources and could switch between them, I don't have another practical way to conduct this test,)

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #15
I think you can contact Archimago, he has the Oppo UDP-205.
https://archimago.blogspot.com/2019/01/measurements-oppo-udp-205-dsd-playback.html

FWIW his DSD results are not bad enough to cause audible difference.

Since it is not feasible to burn your own test signal SACD, when recording the analog output of the player, you can only capture a snapshot of the two playback methods. For example, if for some reasons an unstable firmware caused difference at the moment you captured the files, then you are always listening to the same snapshot. So perhaps you can try to make more snapshots, or switch playback order of disc vs file, power cycle and recapture etc, so you can have more snapshots to experiment with.

You can share the captured files with Archimago too, he should have more tools to analyze the files. As an owner, he is also more familiar with the features of the hardware than the rest of us too.

 

Re: Why do DSF files not sound as good as the SACDs they were made from?

Reply #16
Without changing any settings or cabling or any levels, I put a USB drive into the Oppo 205 and played the DSF file (which I had extracted from the SACD disk).
Now I am out of ideas. But yeah, give Archimago an e-mail.