limiter/dithering

[Peel]

In digital, any signal processing you do degrades the signal to some degree (well, this is true in analogue too, but in a different way -- I'm talking about loss of precision due to rounding). In 16-bit, this is significant enough that you really don't want to do any processing if you can avoid it -- although the degradation of simple processing like normalization might not be noticable if the resolution of the original was high enough. 24-bit gives you quite a bit of "spare" resolution for signal processing, and in 32-bit (the bit depth used internally in most sequencers), I believe you don't really need to worry about this at all in any real-world applications.

The problem is that people know that the more a signal that utilizes the full 16/24/32 bits (ie the peak is at 0dB), the higher the resolution and audio quality. This is true, and this is why you should record as hot as possible without clipping.
But many people go on to assume that normalizing a quiet audio file to 0dB will also increase the quality in the same way. This is untrue; it will simply be a louder version of the same signal, but with some added distortion due to the signal processing.
Basically, any destructive changes in volume level before the mastering stage are, as platinumears said, pointless at best.

[Paresh]

I need to reread these posts, but I think someone said that you don't have to worry about distortion if I record w compression (I still have vocals in mind). I'm very careful not to record into the red zone but I've never heard the sort of harsh digital distortion that i read about when setting levels & they're still too hot. As a result I have a safe but too quiet vocal track. I'm wondering if this means I can push more into the red zone if I record w compression, like I cd w analog tape recording? Thanks

[Lunch Money]

[A bunch of useful (and as far as I know, true!) stuff]

Agreed; however, normalization is such a linear and basic math operation that I challenge anyone to tell the difference between a normalized and non-normalized file.

Ie. if your original peaks at -5 dB, give it 5 dB of gain so that it's equal in level to the normalized file. I don't know that there even IS a difference, and you certainly won't hear it.

If it's easier for somebody to normalize (it doesn't have to be to 0db.. maybe your 'mixed' signal will hover around -5 dB, in which case you might as well only normalize to -5 dB) before final mixing, I say go for it. If I've accidentally recorded a signal that's not hot enough but it's a brilliant take that I can't do without, I'd rather run a noise-reduction algo (destructively, and I may or may not bother keeping a backup depending on how much I know and trust the algo) and normalize to the ballpark before mixing. Beats farting around with a stack of plug-ins that will ultimately do the same job.

Of course, I'm also not producing a pro record. The pros will have already set their levels properly and if their artist doesn't use the mic properly, they'll have the time and know-how to coach them on it. I totally agree that it's WAY better to just record at optimal levels; however, I also feel there's room for destructive editing for certain functions. Heck, you can always keep a spare "destroyed audio files" folder somewhere for backups just in case you ever need to go back to them.

Greg

[audiobot202]

Of course, I'm also not producing a pro record. The pros will have already set their levels properly and if their artist doesn't use the mic properly, they'll have the time and know-how to coach them on it. I totally agree that it's WAY better to just record at optimal levels; however, I also feel there's room for destructive editing for certain functions. Heck, you can always keep a spare "destroyed audio files" folder somewhere for backups just in case you ever need to go back to them.

Agree with Lunchmoney on this. At 16 bit all the way I can sort of understand the logic behind it. Having said that, I can't imagine any possible DSP degradation being particularly audible. I record everything at 24 bit anyway, so it may not affect me as much. Like I've said before, Compression is far more deadly to audio than any normalisation process will be. It squashes dynamics by nature and there are plenty of commercial examples of overcompressed and flat sounding recordings. I'd be far more worried about using the right amount of compression than simple normalisation of a waveform.
Still....I guess it all depends on what you produce. If you want pro, you're far more likely to follow current mastering trends and use a fair amount of digital squashing to achieve the desired result. Which is unfortunate....

[slasi]

Dither basically makes the lowest bit of the reduced wordlength signal toggle on and off in a way that statistically re-creates some of the information that would have been lost otherwise. In other words it can actually increase the apparent bit-depth of CD audio, some say to 18-bit or better.

I'm not sure this part is correct. My understanding of dither is that it decouples the distortion introduced from truncation of word length by introducing a very low level noise signal. It doesn't "re-create some of the information that would have been lost otherwise" as implied above.

-Shehryar

[mauseoleum]

hmm - i have also seen/read some *bit-enhnacement* theories regarding dither - while technically it's still the same 16bit info. so in all my mighty smartness galore i couldn't figure out *exactly how* this could be achieved in plain 16bit word without some serious woodoo - which apparently isn't there.

so - i settled my gray cells with notion it must have been some sort of psychoacoustic comparison - stating that *dithered* material sounds more equivalent to 18bit audio than plain undithered 16bit audio.

OR NOT ????!!!!

[IIRs]

The best explanation of dither I've yet come across is in Bob Katz's book "Mastering Audio", but as he credits it to someone else anyway I hope he won't mind me reproducing it here:

Here's a simple thought experiment that explains why dither is necessary and how it works. Lets create a basic A/D converter. We'll make it sensitive to DC and bipolar, so it responds to both positive and negative analogue inputs, and we'll give it a very big LSB threshold of 1 volt to make the numbers easy. We'll construct our ADC so that an analogue source over the range between +.5 volts and 1,5 volts produces an output of 1, and so on. If, without applying any dither, we present a 0.25 volt DC (continuous) signal to the input of the ADC, the output of the ADC will be a string of zeros. In fact any signal between -0.5 and 0.5 volts will result in an ADC output of zero. Any information below the LSB threshold is completely lost.

Remove the 0.25 volt signal and apply dither to the input of the ADC in the form of a completely random signal (i.e.,noise) centred around 0 volts. Its peak amplitude randomly toggles the LSB of the ADC. The output of the ADC will be a stream of very small random values. However the average of all these values will be zero.

Now lets apply our 0.25 volt signal again (with the dither on). The two analogue voltages sum together, the dither and our signal. At each sample point (in time), the 0.25 value of our analogue source is added to the random dither value. The output stream wil again look like a stream of very small random numbers, but guess what? The AVERAGE of all those numbers will now be...you guessed it, 0.25. We have thus retained the information that was previously lost (even though its buried in "noise"). In other words, our resolution has improved. The conversion is still essentially random, but the presence of the 0.25 volt signal biases the randomness. Put another way, the characterization of the system with dither on is transformed from completely deterministic to one of statistical probability. The periodic alternation of the LSB between the states of 0 and 1 results in encoding a source value that is smaller than the LSB. In other words, on the average, the LSB puts out a few more ones than zeros because of our +0.25 volt signal. We say that dither exercises or toggles or modulates the LSB.

With the dither on, we can now change the input signal over a continous range and the average of the ADC will track it perfectly. An input signal of 0.371476 volts will have an average ADC output of (the binary equivalent of) 0.371476. The same will hold true of inputs going over the LSB threshold: an input of 3.22278 will have an average ADC output of 3.22278. So not only has the dither enhanced the resolution of the system to many decimal places, but it has also eliminated "stepping" quantisation effects!

Now go and buy Bob Katz's book.

[donkey tugger]

Anyone for doing any music perchance? Or is this the f**king open university, you beardy weirdo brown cords and sandals soldering iron twats?

[kilroy]

[IIRs]

The other thing Bob has to say about dither:

"..if the mix isn't good, or the music isn't swinging, then dither probably doesn't matter very much.

You have a point Mr Tugger, but stop to think: if no-body worried about this stuff you wouldn't have an affordable good sounding DAW with which to do your music..

[kylen]

Anyone for doing any music perchance? Or is this the f**king open university, you beardy weirdo brown cords and sandals soldering iron twats?

Hehe - you just made rosin squirt out my nose...

[donkey tugger]

The other thing Bob has to say about dither:

"..if the mix isn't good, or the music isn't swinging, then dither probably doesn't matter very much.

You have a point Mr Tugger, but stop to think: if no-body worried about this stuff you wouldn't have an affordable good sounding DAW with which to do your music..

Fair play! Err...can someone mend me guitar for me...?

[CypherOne]

I think it has something to do with the flugelbinder not cross referencing the dylithium crystals, but hey, what do I know?