Topic: Sony PCM-M10 (Part 4)

[it-goes-to-eleven]

2) To my ears, 24 bit is of a great benefit when recording for the purposes of a lower noise floor, allowing for easier post-production work that may require the floor to be somewhat raised without hopefully raising too much noise with it as well.

24 bits does allow more resolution to be preserved when increasing gain in post, but it does not lower the noise floor.

[Mike Davis]

Ozpeter and Artstar,

Thanks for the quick response!

My gut tells me that y'all have this figured out and, believe me, I'm already concerned about the possibility of SQ damage done by converting FLAC to WAV without changing either the sampling rate or the bit depth.  At the end of the day, when I finish researching this, I hope that the right answer is to just leave files as close as possible to the way I received them.

All that said, I've been doing a lot of reading about the filtering that is performed by DACs and although the finer points completely escape me in some of the material I've studied, I'm certain there are some smart people out there who believe there is a lot to be gained in the accuracy of playback (and lack of distortion) by upsampling before playback.

One argument is that a processor fast enough to perform DAC functions with recordings that were ADC'd at 96-kHz will be handling a 44.1-kHz stream twice as fast as a 44.1-kHz DAC would - if and only if that 44.1-kHz data were disguised as 96-kHz data.  That's an oversimplification born of my lack of understanding all the stuff about anti-alias filtering and such, but it kind of makes sense to me.

One guy put it like this (paraphrasing): "If you are translating a foreign language to English, would't it be great if the person you are interpreting spoke at half the speed you're comfortable with?  You're translation would be far more accurate."

If you've heard arguments like this before and can point me to a good reference that disputes this position, I'm still educatable.   

My other question hasn't been answered (or maybe it has) - What do think about reducing the file size of 96/24 recordings (obtained from Linn or HDtracks) by 33% using dbPoweramp to reduce the bit depth from 24 to 16? As I said earlier, I don't believe the loss of dynamic range can be detected audibly, but I'm concerned about over-processing the file. Do you have any insights to offer on this?

Again, I am a lump of clay at this point - willing to be taught if the argument makes sense.

Thanks!

Mike

[rjp]

My gut tells me that y'all have this figured out and, believe me, I'm already concerned about the possibility of SQ damage done by converting FLAC to WAV without changing either the sampling rate or the bit depth.

 

Decoding a FLAC to a WAV results in a bit-for-bit identical copy of the original WAV file.

There is a potential issue with metadata inside the WAV. My LS-10 puts some Olympus-specific metadata in its WAV files. If I use the --keep-foreign-metadata option when encoding the original WAV to a FLAC, and then also use that option to decode the resulting FLAC back to a WAV, I will get an identical file back.


russ@blackbird:~/flactest$ ls -l
total 67420
-rwxr-xr-x 1 russ russ 68961280 2010-12-23 16:59 LS100274.WAV
russ@blackbird:~/flactest$ flac -8 --keep-foreign-metadata LS100274.WAV
NOTE: --keep-foreign-metadata is a new feature; make sure to test the output file before deleting the original.

flac 1.2.1, Copyright (C) 2000,2001,2002,2003,2004,2005,2006,2007  Josh Coalson
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

LS100274.WAV: WARNING: legacy WAVE file has format type 1 but bits-per-sample=24
LS100274.WAV: wrote 37096678 bytes, ratio=0.538
russ@blackbird:~/flactest$ mv LS100274.WAV LS100274.ORIG.WAV
russ@blackbird:~/flactest$ ls -l
total 103688
-rwxr-xr-x 1 russ russ 37096678 2010-12-23 16:59 LS100274.flac
-rwxr-xr-x 1 russ russ 68961280 2010-12-23 16:59 LS100274.ORIG.WAV
russ@blackbird:~/flactest$ flac -d --keep-foreign-metadata LS100274.flac
NOTE: --keep-foreign-metadata is a new feature; make sure to test the output file before deleting the original.

flac 1.2.1, Copyright (C) 2000,2001,2002,2003,2004,2005,2006,2007  Josh Coalson
flac comes with ABSOLUTELY NO WARRANTY.  This is free software, and you are
welcome to redistribute it under certain conditions.  Type `flac' for details.

LS100274.flac: done
russ@blackbird:~/flactest$ ls -l
total 171108
-rwxr-xr-x 1 russ russ 37096678 2010-12-23 16:59 LS100274.flac
-rwxr-xr-x 1 russ russ 68961280 2010-12-23 16:59 LS100274.ORIG.WAV
-rwxr-xr-x 1 russ russ 68961280 2010-12-23 16:59 LS100274.wav
russ@blackbird:~/flactest$ cmp LS100274.ORIG.WAV LS100274.wav
russ@blackbird:~/flactest$ sha1sum LS100274.ORIG.WAV LS100274.wav
cc8f9f361af5b1a605e865829597c8d0f8acbff8  LS100274.ORIG.WAV
cc8f9f361af5b1a605e865829597c8d0f8acbff8  LS100274.wav
russ@blackbird:~/flactest$


After encoding the original WAV, I renamed it and then decoded the FLAC to a second WAV file. The cmp command generated no output, which means the files are identical. I also ran SHA-1 checksums against both the original WAV and the recreated one, and those are identical as they should be.

[Mike Davis]

Hi RJP,

Decoding a FLAC to a WAV results in a bit-for-bit identical copy of the original WAV file.

Hence, the software titled "Exact Audio Copy."   I was just trying to drive home the point that I am, at heart, a guy who prefers not to over-process files, in the belief that every operation can take its toll (nearly every operation, anyway).  But thanks for putting in the effort to illustrate that we can trust the software you use for converting between FLAC and WAV.

With respect for your command of the subject, I'd really appreciate your jumping in on the questions I was asking, above.  Any comment there?

Thanks,

Mike

[rjp]

Exact Audio Copy is a CD ripper. It will try very hard to make a bit-perfect copy of the audio on the CD, and put it into WAV files. Converting the WAV files to FLAC will make them take less space, without compromising audio quality (and it will also allow you to add tags to the files, very handy for organizing them in a player like foobar2000). Decoding the FLAC will produce precisely the same audio stream that was originally encoded. Playback quality will depend on the playback equipment.

As for upsampling a file (WAV or FLAC), I don't see any point in it. The upsampler might try to interpolate the samples it adds, but "interpolate" is just a fancy word for "guess." It certainly won't increase the high frequency resolution, since higher frequencies (beyond what the original sampling rate can resolve) were never there to begin with.

My $0.02 is that sampling rates beyond 48 kHz aren't worth the extra space needed, unless you are: a) actually recording at the higher sampling rate, b) have equipment that accurately captures, passes, and plays ultrasonic audio, and c) are recording for scientific or engineering reasons that require ultrasonic resolution, or for an audience that can actually hear ultrasonic frequencies (such as dogs or bats).

[Artstar]

24 bits does allow more resolution to be preserved when increasing gain in post, but it does not lower the noise floor.

The noise floor is pertinent to the highest common denominator. Ergo, if the analogue stage has a noise floor lower than the AD can handle in terms of resolution, then the lower bit resolution will raise that noise floor as it is the highest common denominator. A higher resolution doesn't help to raise the upper limit, as that's 0dBFS. What the higher resolution offers is the ability to quantise levels lower than that of it's lower-resolution counterpart and consequently, have a reduced noise floor provided the analogue stage is up to the task as well.

So, if the OP has a 44.1/16 file, he obviously won't benefit from upconverting it to a 48/24 file (for example) as there is nothing in the source that will take advantage of the higher resolutions. Clearly, that won't raise the noise floor. I think we both agree on that.

[Artstar]

All that said, I've been doing a lot of reading about the filtering that is performed by DACs and although the finer points completely escape me in some of the material I've studied, I'm certain there are some smart people out there who believe there is a lot to be gained in the accuracy of playback (and lack of distortion) by upsampling before playback.

But you're still forgetting that it won't be any more accurate than the source material. Therefore, all you're going to do is chew up even more battery power for the processing required to deal with an upconverted format.

One argument is that a processor fast enough to perform DAC functions with recordings that were ADC'd at 96-kHz will be handling a 44.1-kHz stream twice as fast as a 44.1-kHz DAC would - if and only if that 44.1-kHz data were disguised as 96-kHz data.

Not at all, as the processor will simply work at a slower speed with lower sampling rates which will inherently consume less power. Better for your battery life. Otherwise, the processor will still only produce results that are as accurate as the source.

One guy put it like this (paraphrasing): "If you are translating a foreign language to English, would't it be great if the person you are interpreting spoke at half the speed you're comfortable with?  You're translation would be far more accurate."

Therefore, both the foreigner and the interpreter are working at the same speed. In which case, that analogy proves that upconverted files are more strenuous since the foreigner is talking faster but provided the interpreter can keep up (and in turn consume more energy - Powerade?) the translation remains the same because the same words are still being spoken. Right?

What do think about reducing the file size of 96/24 recordings (obtained from Linn or HDtracks) by 33% using dbPoweramp to reduce the bit depth from 24 to 16? As I said earlier, I don't believe the loss of dynamic range can be detected audibly, but I'm concerned about over-processing the file. Do you have any insights to offer on this?

I think you answered your own question there. If you don't believe there is a noticeable difference going down to 16-bit, then you're not over-processing the file. Over-processing the file (ie. messing with the quality) suggests that you can perceive the difference in quality. If you can't hear the difference, then the conversion is transparent and you're a happy man.

[Mike Davis]

Thank you Artstar and RJP for explaining your convictions.   

My search for truth is settled on this subject - nothing can be gained by upsampling. 

I did my best to present a minority argument that I've read here and there on mutliple forums, but that school of thought just doesn't make sense.  I'll stand with the majority. 

And I like your position regarding my question about reducing the bit depth 24 to 16.  As space is not an issue (yet), I think I will just allow my 96/24 files to stay as they are (for now).

Thanks again!

Mike

[morst]

My $0.02 is that sampling rates beyond 48 kHz aren't worth the extra space needed, unless you are: a) actually recording at the higher sampling rate, b) have equipment that accurately captures, passes, and plays ultrasonic audio, and c) are recording for scientific or engineering reasons that require ultrasonic resolution, or for an audience that can actually hear ultrasonic frequencies (such as dogs or bats).

two more for ya:
D) planning to slow the recording down several octaves in order to shift hypersonic information down to audible frequency range
E) desiring future-compatibility with systems like the Plangent Process wherein 192kHz-sampled analog tapes can be time-aligned to the tape head bias frequency. http://www.plangentprocesses.com/

I agree that upsampling is not advised, but I would strongly caution against bit reduction from 24 to 16. Perhaps your playback gear is not allowing you to hear the difference. It's rather dramatic, actually. Also, as an M10 owner, I've noticed that it's built-in D>A converters are not very good-sounding compared to my desktop computer's built-in audio. Very handy for portable use, of course, but doesn't sound like top-notch audiophile gear. YMMV of course.

[Mike Davis]

YMMV of course.

I guess so! 

I love the way my PCM-M10 sounds with Line Out to a Meier Audio Corda Stepdance and I don't find it's headphone out to be shabby, either, but a far cry from the Stepdance.  Perhaps there's something wrong with my ears.   



Mike

[morst]

Oh, I never noticed the Sony's line out being sub-par until I did an A/B. Let me clarify - it's fine, it's just not as good as my other option. It's easy to be happy with gear if you don't a/b, I do that all the time. I still play the sony in the car, and at other people's houses after shows to blow their minds!! 

[Sloan Simpson]

Got my M10 and just taped its first show, but now have a problem: when connecting to my PC via USB, it cycles between connected and disconnected rapidly, never staying connected long enough to even see the directory structure, much less transfer a file.  Haven't had the problem with this PC and a PCM-D50.

Anyone seen this behavior or did I just get a dud unit?

EDIT: Problem solved.  I was using the D50's USB cable, when I switched to the one provided with the M10 it worked.

[spyder9]

Just swapped batteries on both of mine last night at a show.  Clocks did not reset.

[12milluz]

Got mine about a week ago and I love it. Haven't taped any shows but made some recordings around the house and it's easier to use than my old DR07.

[dphirschler]

I finally caved in and bought a PCM-M10.  It should be here any day.  In the mean time, I've been reading these forums.  Thanks all for the wealth of info being shared here.  I am surprised about the clock issue.  I hope mine does not exhibit that problem. 

In my search for info, I discovered some photos on an M10 taken apart.  Here's a link:

http://4ltrs.in/esem/?pcmm10


Darryl