Topic: Effect of MD compression on .wav comp

[tms]

Hi,

There’s been some recent discussion about the use of MiniDisc for taping.  One of the downsides of MD is the fact that it uses compression and is therefore lossy.  I was curious the effect of the compression and so decided to do a comparison test. 

If anyone is interested in participating in this comp I’d love to have you weigh in.  It would especially be nice if a diverse range of listeners would participate. 

My gut feeling is that this should be an easy one to differentiate, and it should be interesting to hear the comments.

The .wav files are located here:  www.eng.buffalo.edu/~tmsnyder/MDTest  .  There are about 90mb total. 

Here are the instructions, also posted in the download directory:

This is a comparison test to study the effect of MiniDisc (MD) ATRAC compression and decompression on 16 bit, 44.1 wav files.  Only the compression is included, no A/D or D/A conversions are included in the comparison.  All transfers were done over optical connections (all digital transfers). 

Studio produced music cd’s were played on a Sony cd player.  The optical output was captured as a .wav with a JB3.  This is the control sample referred to below as the unmodified wav file. 

The same song was then played again and captured on a portable MD player, a MZ-S1 which uses ATRAC Type R compression and has an optical input.  The disc was then removed and played on a MD deck, a SONY MDS-JE480 which was modified with an optical output and sends data at 16 bit / 44.1 khz.  The resulting wav file is therefore the same size as the control.  The optical output was captured on the same JB3 as the control. 

So the only difference between the two sources is that one includes MD compression and decompression.

There were 3 CD’s used; classical, rock/blues, and bluegrass.  One song from each CD was captured using the two sources.  The songs were split into 30 second tracks so that first one source can be played, and then the other source can be played from approximately the same part of the song. 

For each song, designations of Source A and B were randomly chosen for which source was MD lineage and which was unmodified wav.  So the designations of A and B randomly vary from song to song.  A is either the MD source or the unmodified source, and B is the other remaining choice.  The designation of A or B doesn’t vary within the song.

Song 1 (Bach):

1-1A:  Song1Track1SourceA (30 seconds)
1-1B:  Song1Track1SourceB (30 seconds)

1-2A:  Song1Track2SourceA (30 seconds)
1-2B:  Song1Track2SourceB (30 seconds)

1-3A:  Song1Track3SourceA (30 seconds)
1-3B:  Song1Track3SourceB (30 seconds)

Song 1 choices (please choose one of the following):

*1* Source A is the unmodified wav file (Source B was MD compressed)
*2* Source B is the unmodified wav file (Source A is was MD compressed)

Certainty (1-5):  (1=No clue, total guess, 2=Slight suspicion, 3=Strong suspicion. 4=Pretty sure, 5=I'd bet my life on it)

Observations:

____________________________________________________________

Song 2 (Stevie Ray Vaughan):

2-1A:  Song2Track1SourceA (30 seconds)
2-1B:  Song2Track1SourceB (30 seconds)

2-2A:  Song2Track2SourceA (30 seconds)
2-2B:  Song2Track2SourceB (30 seconds)

2-3A:  Song2Track3SourceA (30 seconds)
2-3B:  Song2Track3SourceB (30 seconds)

Song 2 choices (please choose one of the following):

*1* Source A is the unmodified wav file (Source B was MD compressed)
*2* Source B is the unmodified wav file (Source A is was MD compressed)

Certainty (1-5):  (1=No clue, total guess, 2=Slight suspicion, 3=Strong suspicion. 4=Pretty sure, 5=I'd bet my life on it)

Observations:

____________________________________________________________


Song 3 (Del McCoury):

3-1A:  Song3Track1SourceA (30 seconds)
3-1B:  Song3Track1SourceB (30 seconds)

3-2A:  Song3Track2SourceA (30 seconds)
3-2B:  Song3Track2SourceB (30 seconds)

3-3A:  Song3Track3SourceA (30 seconds)
3-3B:  Song3Track3SourceB (30 seconds)

Song 3 choices (please choose one of the following):

*1* Source A is the unmodified wav file (Source B was MD compressed)
*2* Source B is the unmodified wav file (Source A is was MD compressed)

Certainty (1-5):  (1=No clue, total guess, 2=Slight suspicion, 3=Strong suspicion. 4=Pretty sure, 5=I'd bet my life on it)

Observations:

____________________________________________________________




Please describe your playback system:

____________________________________________________________

[twoheadedboy]

I think this is irrelevant now that A) you can record PCM 16-bit 44.1khz audio onto MD (1GB Hi-MD format) and B) with the Hi-MD format has arrived newer, higher quality compression settings.

No offense intended, just my two cents.

[madman]

I'd like to see a test with Hi-SP.  I bet with live, or even pre-recorded, that most people wouldn't be able to tell the difference by ear.

[itook2much]

I'd like to see a test with Hi-SP.  I bet with live, or even pre-recorded, that most people wouldn't be able to tell the difference by ear.

I use Hi-MD, & I can't.  Hi-SP recordings I've done sound as good to my ears, & the frequency range shows up just as broad when I edit.

[zowie]

Thanks for posting these.

It was interesting to compare.  I had no trouble telling which was which, however.

[jpschust]

it's irrelevant in the sense that no matter what it is a lossy compression.  you lose dynamic range- how much more do you need?

[tms]

Thanks for posting these.

It was interesting to compare.  I had no trouble telling which was which, however.

Great, thanks for listening to them!  Would you mind PM'ing me with your answers and comments?



Todd in Buffalo

[tms]

it's irrelevant in the sense that no matter what it is a lossy compression.  you lose dynamic range- how much more do you need?

It might not be irrelevent to everyone though, depending on their application.  For instance, I carry an MD as a backup to my JB3.  And the MD was my first recorder, as it is for a lot of people.  It's small, inexpensive, waterproof, and records for 7 hours on a single AA battery. And I find that I actually listen to my MD recordings a lot more than my JB3 recordings because it's so easy to pop one in my deck.  95% of my JB3 recordings are still on my PC harddrive waiting to be tracked and burned to CD.

Since it turns out I'm still using the MD even though it is obsolete, I was wondering how much is lost by the compression, that was the reason for the comp.  I figured that others might want to hear the change caused by ATRAC as well.

[tms]

it's irrelevant in the sense that no matter what it is a lossy compression.  you lose dynamic range- how much more do you need?

You're right, for achiving purposes you shouldn't use an MD.

[tms]

I'd like to see a test with Hi-SP.  I bet with live, or even pre-recorded, that most people wouldn't be able to tell the difference by ear.

I use Hi-MD, & I can't.  Hi-SP recordings I've done sound as good to my ears, & the frequency range shows up just as broad when I edit.

A Hi-MD test would be kind of pointless, the data are very nearly bit perfect when recorded optical-in and recovered with HiMDRenderer. 

What are the specs on Hi-SP?  Is it better quality than SP?

[zowie]

Those are my comments.  I can hear the difference easily.  The ATRAC doesn't suck, but it's clearly worse.  I don't have time to fill out the whole shebang.

Hi-SP is not bit perfect, just the oppositem it is another lossy compression algo.  The question presented by Madman is whether (or not) Hi-SP sounds better than old SP, which would be interesting to know IF one is going to use any lossy compression in the first place, which of course is undesirable for live recording except under exceptional circumstances (i.e., its still better than nothing)>

[tms]

Those are my comments.  I can hear the difference easily.  The ATRAC doesn't suck, but it's clearly worse.  I don't have time to fill out the whole shebang.

Hi-SP is not bit perfect, just the oppositem it is another lossy compression algo.  The question presented by Madman is whether (or not) Hi-SP sounds better than old SP, which would be interesting to know IF one is going to use any lossy compression in the first place, which of course is undesirable for live recording except under exceptional circumstances (i.e., its still better than nothing)>

Right, Hi-SP is not bit perfect, but Hi-MD recorded as PCM and recovered by HiMDRenderer is.

Right, Hi-SP is still compressed, which is definitely not as good as uncompressed.  I just wasn't sure if it was more compressed or less compressed than SP.  I think it's even more compressed (lower bitrate) than SP.

Please PM me with your answers, I'm interested to see how you did. 

Song 1:  Were the A tracks MD or .wav?
Song 2:  Were the A tracks MD or .wav?
Song 3:  Were the A tracks MD or .wav?

[tms]

Those are my comments.  I can hear the difference easily.  The ATRAC doesn't suck, but it's clearly worse.  I don't have time to fill out the whole shebang.

Hi-SP is not bit perfect, just the oppositem it is another lossy compression algo.  The question presented by Madman is whether (or not) Hi-SP sounds better than old SP, which would be interesting to know IF one is going to use any lossy compression in the first place, which of course is undesirable for live recording except under exceptional circumstances (i.e., its still better than nothing)>

Right, Hi-SP is not bit perfect, but Hi-MD recorded as PCM and recovered by HiMDRenderer is.

Right, Hi-SP is still compressed, which is definitely not as good as uncompressed.  I just wasn't sure if it was more compressed or less compressed than SP.  I think it's even more compressed (lower bitrate) than SP.

Please PM me with your answers, I'm interested to see how you did. 

Song 1:  Were the A tracks MD or .wav?
Song 2:  Were the A tracks MD or .wav?
Song 3:  Were the A tracks MD or .wav?

[Brian Skalinder]

Hi-SP is not bit perfect, but Hi-MD recorded as PCM and recovered by HiMDRenderer is.

Have you done any controlled testing with HiMDRenderer to verify its bit-accuracy?  I'm sure the author has improved it from when I last checked into it, but...last I looked into it HiMDRenderer did not prove bit-accurate.

[poorlyconditioned]

Hi-SP is not bit perfect, but Hi-MD recorded as PCM and recovered by HiMDRenderer is.

Have you done any controlled testing with HiMDRenderer to verify its bit-accuracy?  I'm sure the author has improved it from when I last checked into it, but...last I looked into it HiMDRenderer did not prove bit-accurate.

The problem is not whether HiMDRender is bit accurate.  I think it is, and *a small number of* my tests confirm this.  The problem is that the MD *resamples* the digital input.  So, for the purist, you need a NJB3 for digital recording.

Back to whether to us MD.  I would say yes.  I've seen *much* bigger differences from changing microphones, microphone powering (phantom vs. battery box), and microphone placement.  If you're using anything like CSB omnis or AT853, I think it is perfectly reasonable to use MD (compressed or not).  Oh yeah, I would rather have analog in (line/mic) to a MD than to a NJB3 any day.  MD has proper gain controls and no HD noise on analog in.  So, as noted in my "sig" below, I would either use analog in to MD or digital in (eg., UA5) to NJB3.

  Richard