Opinion? Recording interviews with MP3 320kbps at 48kHz

[Noiz2]

Paul, I agree with most of what you say - but not this bit.

PCM digital recording can perfectly recreate the original as long as the sampling freqency is more than double the highest frequency - IE: which is why 44.1kHz or 48kHz is used for an audio frequency response up to 20kHz.

There is *no* data thrown away at all - *nothing* is lost at all.

No "data" by definition because it's not "data" till it's digital, but information is most certainly lost. In theory IF you only listen to simple sine tones you can accurately guess at the original. The problem is if you don't have a sine tone then trowing away the information alters the waveform. A 20 khz square wave can not be accurately reproduced by a 44.1 sampling frequency. It will come out as something close to a sine wave. It won't be start getting really square till it's much lower since you need to hear the overtones.

But even if the reproduction is imperceptible from the original, you have still tossed away information. And yes I know technically people don't hear over 20k and for most of us it's a lot lower. But we can tell the diference between a square wave and a sine even at very high frequencies so we can hear the effect of frequencies that are higher then we can technically hear. Just as we can not see infrared but we can sense it.

 

[Noiz2]

Absolutely. Saying that common "PCM" is "throwing away" data is completely contrary to the commonly accepted definition of ordinary digital audio.

To be sure, we can't take so many samples that EVERY point on the waveform curve is covered. But then when it comes right down to that, we can't do that in analog (magnetic tape), either. The waveform is "sampled' by the individual microscopic magnetic dipoles in the tape oxide. The linear speed of the tape determines the "sampling rate". And the random orientation of the dipoles contributes to random noise (aka. "tape hiss").

I and I believe Paul are not comparing the quality of analog recording to digital recording but digital recording to the actual original sound.

I would agree that a decent digital recording is more "accurate" to the original sound than an analog recording, but that has nothing to do with the point I was making.

I do think you are probably correct, linguistically, and I should be talking about the loss of "information" not data. Information is lost in digitally encoding a sound.

We can show theoretically/mathematically that sampling at a sufficient rate results in reproduction waveforms that have sufficient accuracy ("fidelity") to be indistinguishable from the original waveform. This work was developed over 50 years ago by Nyquist and Shannon and others at Bell Labs in the earliest days of digital audio research. This is why saying that linear PCM "throws away" data because of sampling is not an accepted view by most people.

I don't mean to be rude but that is the same as saying that because most people today (a long time ago) think the world is flat, that it is.

In fact most people probably don't think there is any perceptible loss in MP3 at the kind of crappy data rates Apple uses.
There has been a lot of work done since the original work, partly because the original was not correct, enough. This is why A/D converters used to day do not sample at 44.1 or 48, they over sample by quite a bit then average down to the sample rate they "record" at.

Even with recording equipment and methods that record straight to compressed formats like MP3, the process BEGINS with linear PCM sampling and THEN the linear information is subsequently DATA REDUCED by the compression scheme: AAC, ATRAC, MP3, FLAC, Ogg Orbix, WMA, etc.

This again was NOT what I was talking about.

This whole tangent was started because I said more data would be better and IF you recorded straight to a compressed format (at the SAME data rate we were using for uncompressed recordings) we would be getting 5 to 10 times the data.

This was a HYPOTHESIZED recording method that does not exist as far as I know today.
Y
ou keep saying that is not how we do it today and I keep saying THAT IS NOT WHAT I AM TALKING ABOUT!

There are NON-lossy audio compression schemes as well as LOSSY compression schemes. Because of the way the LOSSY compression schemes work, renders them less than suitable for most audio production which involves further manipulation in post-production editing steps.

That is I believe almost verbatim what I said a few posts ago.

 

[Noiz2]

By that method then, there is NO method of recording sound (or picture, for that matter) that is perfect.
For that matter, even your own ears (and eyes) introduce errors from what the true audible/visible signal is while on the fly.
You can split hairs here, but you (and Scott) have rather distracted from the original topic. If you want to argue how good digital audio is, go for it. I'm not particularly interested in that pissing match.
But the original question was LOSSY DATA COMPRESSED vs. UNCOMPRESSED digital audio. And there can hardly be any controversy over that.

RE: Alex,

Your right posted the others before I got to yours.

Back to the original posters question.

The "controversy" as such is that there is one opinion that any lossy form of recording is "garbage" and you might as well not even record it.

The other opinion is that using a lossy format is not the best or first choice but if done carefully it can sound fine in many situations. And that there are scenarios where it is the only practical way to record something.

Both opinions have been expressed by folks who have a fair amount of experience and should know what they are talking about.

Why the difference? Probably experience. I have seen both sides (awful recordings and ones indistinguishable from DAT recordings). If I had not had some experience with good recordings I would probably be in the "don't use at any cost" camp.

I don't personally record compressed, generally. I don't store sound in a compressed format (a lot of video editors rip SFX CD's into MP3 files to "save" on storage. But I have recorded in compressed formats and I did a series of tests with MiniDisk recordings to see if we could get acceptable results.

I don't see right now many times going forward when I would choose to use a compressed format. Though I did loan the MiniDisk to a radio producer last year for a show that involved having the recorder in a very high risk environment.

I have also recorded to cassette. I know a film that got an special Oscar for sound that had some of the signature sounds recorded on a Sony Walkman Cassette recorder.

In the end it's not the tools it's how you use them. I personally don't eliminate any tool, even if I am very unlikely to ever use it.

 

[paulears]

A poorly made wav file can easily sound bad - People always worry over digital zero - and I end up doing plenty of repair jobs, either using plug-ins, or quite often, deleting a huge spike, and drawing in by hand a new one. A .wav file after this kind of savaging can sound as if no repair has been done. If somebody has pegged the meter to full scale, it can sound horrible. Carefully managed, I don't think anyone has said that any of the compressed/reduced/limited data sounds as good. What I suspect many people are finding, using their ears alone, is that some of the mp3s sound extremely good. I've come back to projects I have recorded and discovered afterwards that I have accidentally recorded in mp3 instead of .wav format - and even side by side with others, I did not notice. On the other hand I have mp3s that sound dull and 2nd rate.

The compression system used on MD, is something I'm very aware of. Some MD recordings sound great, while others don't. MD recorders seem to have the ability to record at almost flat out, with the result being what sounds like dynamic compression (and pretty severe compression at that) - looking at a waveform shows just a horizontal rectangle of white! It doesn't sound distorted, just squashed.

In a live situation, I'd be very happy to record something at say 320K as an MP3, then edit it, and then use the re-rendered file into a PA. If I'm recording a CD, then I'll capture at the highest sensible setting I can that makes sense. For me .wav at 24/96 would be ok, but I can't hear the difference if I dropped down to 48. As others possibly could, although I'm a sceptic, 96 isn't an issue.

Throwing data away - and I'm using data to mean everything that is captured by a microphone, isn't always bad from the sonic perspective. It's bad if the audio is compromised. I'm by no means one of the vinyl is best brigade - but if we compare what goes in vs what comes out, then none of the current systems are accurate. Data is lost, and the only thing that matters is if we can hear it.

Using an SM58 instead of something truly special as a microphone is (in my view) yet another example of lost data. The question is how much, and does it matter?