Another OCD bitrate discussion

[MDane]

OK, I figured I would put some things out there...to see what you guys have found...as I am still in the process of simplifying my music collection as much as possible.

1. For starters, it seems to make sense on the surface that recordings made in real time (with hardware) would fare somewhat better than those done via software, assuming they are being encoded to the exact same bitrate. Can others second this?

2. Would real time recording be better if done via analog inputs or via digital input? Of course one would automatically think digitally would be superior, but myself as well as a friend that I discussed this with a while back both concluded that music recorded via analog seemed to sound a little smoother, particularly on higher frequencies. All I can figure is that there may be some minor differences in the way the sound was processed that would cause this. Or does Sony make an adjustment in the encoding process (when recording via analog) that colors the final result in a way that seems better (to the ear)? Or is this all in my head, lol.

3. When making an MD the other day with my JE440, I noticed that it actually says "Atrac DSP Type-R / Atrac 3" on the front (never paid attention to this all of these years). So am I to assume that the DSP Type-R only applies to SP recordings (hence the slash) or does it apply to the Atrac 3 modes as well. If not, I would then wonder if there is any (acoustical) advantage of using a home deck to record LP modes versus simply using SS, a even versus a portable unit, etc...other than any advantage that might exist recording in real time (as asked in question 1).

Although this is not exactly the same thing, I made a test recoding using the deck in LP4 mode, and compared that to one of the same material encoded to 64 kbps (Atrac 3+), and the 3+ blew the doors off of the LP4. Doing this comparison a while back recording talk material however didn't seem to leave such a gap between the two methods. Even though I realize that the +3 encoding is said to be superior, this still confusses things a little more...and makes me wonder if there really is an advantage to Type-R, real time recording, etc. Any insight?

4. I have "toyed" with the idea of simply converting all of my Atrac music to 320kbps MP3 for car use - as I currently have the dilema of some of my stuff being on MP3, the other half on ATRAC (MD and files) as some may know from my other thread. In other words, if I did this conversion, I could simply buy any deck with USB and load up a HD with everything and forget about it (or get an SD card deck) and accomplish the same thing.

Anyway, I have to wonder how well a high MP3 bitrate would fare. From what I have read, MP3 is pretty decent in high bitrate...but "decent" is a relative term. I just have to wonder how audio would fare if converted back from ATRAC to MP3 (rather than the other way around). Anyone else tried conversions back to Mp3? Chances are this would sound like mud, but I wanted to put it out there to see what others have found.

[Avrin]

1. Software-encoded recordings are better than hardware-encoded. The reason is simple - hardware encoders are simplified, since processors inside units are not powerful enough to encode the full frequency range in good quality on the fly, Hence, SP and Hi-SP are cut off at 19 kHz, and what is left is encoded. Software (SonicStage) encoding, on the other hand, keeps the entire frequency range when encoding to ATRAC3plus @ 256 kbit/s and above.

2. In an ideal situation a recording made though a digital input will be better. But in reality, especially when using modern CDs (and this article http://www.cdmasteringservices.com/dynamicdeath.htm is quite old and optimistic), a good CD deck may actually improve the sound of a CD, and output this improved sound via analog, after which it gets digitized again by a minidisc deck/portable, and recorded in somewhat better quality than the original. But of course you can rip the CD on your computer, clean/adjust/restore the sound of it, and put it onto a minidisc through SonicStage, achieving even better results.

3. The ATRAC Type-R codec only applies to SP recordings. But ATRAC Type-S is a chip containing the ATRAC Type-R codec for SP, plus improved LP codecs.

4. If you don't mind losing gapless playback - go ahead and convert your music to MP3. I would advise using the console LAME.EXE encoder with some front-end (like RazorLame) to control it from Windows. LAME.EXE is known to support gapless encoding (and decoding of files encoded by it), but the only hardware player known to support gapless playback of these files is Rio Karma, if you can find it.

Edited December 2, 2008 by Avrin

[MDane]

On the first point, I figured that recording in real time would have the advantage of having more time to encode, versus, say a conversion in SS that will encode, say a 4 minute song in 30 seconds. Also, I totally agree that Atrac 3+ has a newer and more thorough encoding method that previous ones...but what about deck LP2 versus SS LP2 for example?

Also, if one has ever messed with equalization, they would find there is probably no musical sound above 19khz. For that matter, very little data even exists above 15Khz for anything but maybe test tones, sound effects, etc. So if one is not worried about how the newly encoded file looks on a graph or whatever, I cannot see how that (19khz chop) would effect anything to the human ear. Or does this somehow color everything else in the final product?

I am not sure I follow on the forth point, meaning I don't know what you mean by gapless. In other words, I figured would use my M200 and upload all MD based music to the computer which would produce tracks exactly as they were on the MD (but in WAV). Then I figured I could use some pretty standard MP3 software to convert these files over from WAV to MP3. Then, if playing these files back on a car deck, why would it not be seamless from track to track (assuming no intentional sound breaks). Please elaborate as to how this would not work exactly as I had expected.

[Avrin]

1. Signals above 15 kHz are usually not "heard" by a human ear directly, but they make the sound more natural. And the ability to record higher frequencies gives the capability to better preserve phase information for lower frequencies. That's why vinyl sounds much more natural than CD.

Encoding time cannot be compared directly between a unit and a computer. The power of the unit's processor is in all probability less than that of a first-generation Pentium. What the unit's processor does in realtime, may be done is split seconds by a modern computer processor. So, there's no need to simplify any algorithms.

2. Gapless = no silence where it shouldn't be between tracks. Very few MP3 devices support true gapless playback. Most devices that claim this are actually masking gaps by short crossfades. That's the inherent problem of the format - it was never meant to be gapless. Only additional extensions (such as binary tags from LAME.EXE) allow to encode/decode music to/from MP3 so that information required for gapless playback is maintaned, and can be used.

Edited December 2, 2008 by Avrin

[culp4684]

Where do you get your info that SP recordings made through hardware (decks) are cut off at 19 khz? There are also more factors than just frequency response when it comes to the sound quality of recordings.

[MDane]

1. Signals above 15 kHz are usually not "heard" by a human ear directly, but they make the sound more natural. And the ability to record higher frequencies gives the capability to better preserve phase information for lower frequencies. That's why vinyl sounds much more natural than CD.

2. Encoding time cannot be compared directly between a unit and a computer. The power of the unit's processor is in all probability less than that of a first-generation Pentium. What the unit's processor does in realtime, may be done is split seconds by a modern computer processor. So, there's no need to simplify any algorithms.

3. Gapless = no silence where it shouldn't be between tracks. Very few MP3 devices support true gapless playback. Most devices that claim this are actually masking gaps by short crossfades. That's the inherent problem of the format - it was never meant to be gapless. Only additional extensions (such as binary tags from LAME.EXE) allow to encode/decode music to/from MP3 so that information required for gapless playback is maintaned, and can be used.

1. I still can't see how chopping off what little may exist above 19khz makes a (noticeable) difference however - or be the reason one compression scheme would excel over another. However, if the chop point were, say 12.5khz, I could see it. I am looking at it like this - I can take an equalizer with 10+ bands, and go to either extreme (+ or - 12db) on the highest band, and not really tell a difference most of the time (assuming it's centering frequency is quite high, for example 18khz).

2. That makes a little more sense now. However, I could probably still make a case for my scenario however...as my SS devoted computer is only a PII, and it can still make, say a 5 minute LP2/LP4 file in maybe 30 seconds. So going with your explanation, that would put the "relative processing speed" of my home MD deck along the lines of maybe 25mhz by comparison to the computer...assuming only double the processing ability of the computer per second when Atrac file creating...and that doesn't even factor in the other drains on the computer's resources.

My point is that it's a tough theory to swallow...but you may be right regardless. I just feel like my circa 2000 MD home deck should be a little faster than that. When time permits, I will make some back to back comparisons to see what I can hear acoustically speaking. Chances are I will be better off to use LP4 to see any differences.

3. I believe that due to my relatively limited experience with MP3 I may not be following this completely. Even in the good ole' free-for-all Napster days, I typically just converted such files to MD, and never once bought a devoted Mp3 player - so I am not sure as to any difference I would encounter versus using ATRAC equipemtn. However, I do use my E105 to play back MP3 podcasts...but never notice gaps, lack of gaps, etc. However, once I get a handle on this, I can experiement with it...as it pretty much plays everything.

For now however, I am assuming this is what you mean: Say I had a recording of a live concert that runs continuously, but I put track marks in to seperate songs. Then if conveted and played back in MP3 there would be an (unnatural) sound break between tracks. Is that what you mean?

[Avrin]

1. Not that little. There is quite a lot of data (including noise). And this data is the hardest to encode. LAME.EXE also cuts off high frequencies by default, but this can be controlled and/or disabled.

3. Yes.

Where do you get your info that SP recordings made through hardware (decks) are cut off at 19 khz? There are also more factors than just frequency response when it comes to the sound quality of recordings.

I got this info by simply looking at the Frequency Analysis window in Adobe Audition with these recordings opened, and comparing what I see to what was there for the original CD.

And yes, there are more factors. And a complicated computer algorithm on a modern processor is able to take care of these factors much better.

[samueldilworth]

Hi MDane. Just to be clear, the speed of your computer enables it to encode faster than a MD recorder, but the exact speed of your computer has no effect on the sound quality of the files it encodes. The encoder does everything it is programmed to do, taking as long as it needs to, regardless of the speed of your computer's processor. Therefore the fact that it takes only 30 seconds (or whatever) to encode a track doesn't compromise the sound quality.

A desktop computer's processor is massively more powerful than a MiniDisc recorder's processor, but it's a general-purpose processor, whereas the MD has a digital signal processor designed specifically for encoding ATRAC (or at least for processing digital audio). So the quality differences, if any, may be less than you might expect, though probably in favour of the software and computer combination where present.

I too have my doubts that removing 19+ kHz information is problematic. I wonder if the average person could ABX this, or indeed, if anyone could with a typical music sample? With a 15 kHz low-pass filter it would be another matter, though still not as easy to ABX as many people imagine.

Regarding transcoding ATRAC to MP3: this is a lossy to lossy conversion and should be avoided if possible, because sound quality degradation due to the compression builds up from one generation to the next, and may become easily audible. That said, MP3 at high bitrates is indeed very good, but it does depend on the specific codec you use. Versions 3.97 or 3.98 of the LAME codec mentioned by Avrin are generally considered the best choice, and are certainly much better than many MP3 codecs.

[THIS SUCKS]

didnt sony get the maximum frequency encoded to minidisc sp up to 20khz with type-r encoding for sp?

[Avrin]

No.

Pre-recorded discs, on the other hand, may carry the full frequency range, since they are pre-encoded before mastering, and the equipment used may be powerful enough to handle the entire range.