naturerecordists
[Top] [All Lists]

Re: dither

Subject: Re: dither
From: Walter Knapp <>
Date: Wed, 11 Dec 2002 03:20:26 -0500
Rob Danielson wrote:
> Thank for the comments, Walt.  Please see inserts
> Rob D.

>>I'm not inclined to buy in so big to the low end being somehow unique in
>>number of bits actually used. A high intensity sound that has little
>>dynamic range will be just the same. And in most cases we are really
>>talking sounds that are only about half way down in the dynamic range,
>>not stuff way down at the bottom.
> 
> 
> I was thinking of lower levels, say around 2-4% saturation in the 
> orig field recording-- about the amount I'd quess Rich would be 
> working with if he's hearing dithering artifacts.

I'm not sure that's all that likely. Anyway, If I found myself in that 
area I'd be thinking more about getting closer or something, it would be 
hard to even monitor what was going on. As I noted, what I've been able 
to tell we don't normally get down in the lower end of the range of our 
recorders much. Even if it's a 94 dB DAT, that's a lot of dynamic range 
to play with.

Rich can tell us, but my impression was that he was worrying over how to 
avoid dithering artifacts but did not have any clear cut ones. Or more 
precisely what way to set up his software in light of using better mics.


>>The dynamic filter I use most of the time works on all frequencies at
>>once. It would be nice to have a filter that combined a multi-band eq
>>with dynamics. As it is most eq just moves the gain moving the whole
>>range at once. In my case being able to remove, say, loud insects at a
>>certain frequency range while preserving the lower level ambiance at
>>that range is what I'm talking about. Or maybe being able to set the
>>upper and lower limits of the dynamic range for that band that will be
>>passed would be a good way to have it.
> 
> 
> 
> Perhaps something like Waves C4?  I'm trying to learn how to use C4 
> effectively for this but not with much success.

I can do it in Spark XL's filter setup by creating multiple filtering 
paths. But it would be pretty complex to set up and quickly very 
processor intensive. Right now the nicest dynamic range filter I've 
found in stuff I have is in Soundhack, but it only provides for one cut 
point and does all frequencies. It's nice for cleaning up small messes 
has to be used very carefully or your main sound will suffer. I'm 
learning a bit about the structure of sound playing with it, and have 
made a little practical use of it.

When you figure out Waves C4 I'd be interested in a report. I can't 
afford to buy every piece of software.

Spark XL has just been upgraded. And there are some new filters to find 
out about. Once my copy arrives. Sounds like they have upgraded their 
active noise filter a fair bit.

Part of the problem with detailed dynamics filtering is that I don't 
think it is possible except with a computer. The sound world is still 
mostly running on imitations of what used to be hardware and will be 
slow to move into new areas. Got to think outside the comfortable old 
hardware box.

>>I would certainly agree that much of the noise we deal with is in the
>>lower frequencies, but this is a different issue. Bit depth is sound
>>intensity. A frequency filter will change intensity, but is not
>>specifically filtering along the intensity line.
> 
> 
> 
> My guess is a good part of the choppiness that stems from the low bit 
> rate becomes most audible in the low range,.. have no proof of this, 
> just that selective reduction works best in that range.

I would guess the opposite as higher frequencies have much less in the 
way of samples to define the waveform. At 100 hz you have hundreds of 
samples per cycle. Any error would tend to average out there. Up at the 
high end you may only have a sample or two to define the waveform, 
that's really going to show any error. It's also what should show the 
greatest degradation from dither too. Selective reduction probably works 
best in the low range because you have so many more samples to work with.

I've always maintained that virtually the upper half of the available 
frequency range in digital is very crude. When that's important the best 
thing you can do is go to higher sampling rates. That's probably more 
important than bit depth in that case. Much of what's thought to be bit 
depth problems would go away with higher sampling rates.

For me that's not a big problem, 8khz is about the top for the frogs I 
deal with. Though there is at least one frog that goes up into the 
ultrasonic. I'm also old enough that I don't hear high frequencies all 
that well so am unlikely to take up recording high frequency calls.

Walt




________________________________________________________________________
________________________________________________________________________



Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/ 


<Prev in Thread] Current Thread [Next in Thread>
Admin

The University of NSW School of Computer and Engineering takes no responsibility for the contents of this archive. It is purely a compilation of material sent by many people to the naturerecordists mailing list. It has not been checked for accuracy nor its content verified in any way. If you wish to get material removed from the archive or have other queries about the archive e-mail Andrew Taylor at this address: andrewt@cse.unsw.EDU.AU