Is 16BIT enough vs 24BIT

killfilm said:
yes 24bit is better for more headroom in processing, but seriously, how much processing are you gonna do on a vocal track, IF it was recorded properly, perhaps noise reduction, compression, equalization, thats about it.
Click to expand...

Not headroom in processing - headroom in recording!

Noone's arguing that 16-bit can be suficient. It is, and has been, and will in many cases continue to be.

That said, 24-bit is at least as good, and for some things, better. These days, there's no reason not to use 24-bit if it's just as easy.

Finally, though our theoretically perfect 16-bit system may be sufficient, that doesn't mean that audio recorded by the camera at 16 bits will be good. Acceptable for some things, but not great. Cameras are electrically noisy environments, and don't have great-quality analog components - the combination of these 2 leads to significant noise. The camera will never match the quality of halfway-decent dedicated audio recording gear.

Early films used phonograph records with strange mechanical devices that synced them to film. Noise and dynamic range characteristics have kept improving since then, and will continue to improve. There will always be examples of older films with great sound made on "outdated" equipment, and there will always be the possibility of cleaner sound with new technologies. So what? Do the best with what you've got... when you can really know when and where you can get away with one technology, or when you really need a newer one, you'll be sure about it.
 
killfilm said:
the info was taken from tomlinson holman's book "sound for digital video", he states that through the mic inputs and to the analog outs there's 67 usable db on the dvx100
Click to expand...

Why rely on (missundestood) hearsay, when you can easily analyze the camera's audio performance yourself like I explaned in the HDV200 thread. All you need is a decent mixer and audio editor (like Audition). Record a sample with (near) clipping and silence (line in, pots down), analyze. With XH-A1 I got almost 94 dB DR which is more than enough for everything. I see no reason why DVX100 would be much worse.
 
Petrus - I think your technique tests something like the dynamic range of the converters, but not of the entire analog system. A fairer test would be to record near-clipping sound and silence WITHOUT changing the gain on the pots - that's why decent specs will specify dynamic range at a given gain setting.

If this is what you meant, forgive this message - I think your post can be read either way, but on my first reading I assumed "(line in, pots down)" applied to the silence, but not the clipping level....
 
Pots down on the mixer, of course, same level on camera all the time. How else could this measure camera's DR????

When using mixer all level setting are done with the mixer (after the initial level matching). As we are testing the camera only, not mixer/camera combo, all ways of getting a minimal signal to camera's line-ins are allowed, short of pulling the cables off, of course.

The way I did it with SD302 was to record FS test tone (+20 dBVU or digital clipping level), some dialog at normal levels and then turn the mixer pots down to kill the signal.
 
Last edited:
Petrus,

Good points and to expand a bit. Different equipment is going to sound different depending on maintenance handling etc. So doing the test Petrus talks about gets you two important things. An indication of how well that model camera does AND more importantly how well that specific camera performs. If it sounds good you don't have to spend a lot of time trying to figure out how the technical papers support (or don't) your conclusion. Even if all the other cameras sound like crap your is fine. And that's all that really matters. If it doesn't sound fine it really doesn't mater what the specs say or why it "should" sound great, it just doesn't. And as I think I mentioned many of the post benefits of 24 bit are from working at 24, whether your source was 24 or 16.
 
The biggest advantage 24-bit offers is in post (assuming you got your levels right) because every time you process sound there is a round-off error that effectively loses 1/2 bit. That can add up quickly. Also, when converting sound from 16 to 24 bits during post the extra space you use is temproary, so it doesn't really matter much. A great way to visualize this is to use levels against 24-bit and 48-bit images in Photoshop - if you apply them to 24-bit images the histogram comes out all chopped up. If you first convert them (24-bit files, that is) to 48-bit and then apply the levels command the result comes out smoother (even if you convert back to 24-bit) because round-off errors are comparitively insignificant. The same principle applies to sound. By the way, even if you always go back to lower resolution when saving, converting back to higher resolution when processing generally yields better results - although it's better to stay in 24-bit until the project is finished. Some programs already do this by converting audio data temporarily to a higher resolution when processing - but it only really helps if you are doing multiple audio processing steps in one pass because you still have one round-off error loss every time you convert back to 16-bit. In other words, the advantage of 24-bit is best realized if all audio processing - each step of which produces round-off errors - is done in the 24-bit domain. The extra 8 bits allows you bunches of extra audio processing steps before you will hear the same sound degradation introduced by just one processing step in 16-bit.
 
Last edited:
All decent audio editors do all processing using 32 bits, no matter what bit depth the original file is. The result is only rounded out when the file is saved to disk. Thus 16->24 conversion is not deeded, editor will do it for you, even better.
 
Hey Petrus... sorry to be a pain, that's not 100% true. Also, we should be clear that many software DAWs use 32-bit FLOATING POINT, which has a whole different set of issues and benefits. But the arguably most popular DAW in pro circles, ProTools, uses a 48-bit integer internal representation (ProTools LE, though, uses 32-bit floating point).

But I'm not arguing your main point, that most DAWs convert to a better resolution than either the recording format or output format for internal processing.

The I/O sample rates do raise their head again if you use any outboard processors in your editing workflow...
 
I am as a big an audio-wanker as anyone, but I think everyone should agree that the answer to the original question is just a simple: Yes.

This issue is so far down the chain for most people that it just does not matter. Almost everything else (technique, gear, environment) matters much more than your bit depth when capturing audio.
 
You must log in or register to reply here.
Back
Top