help of 3-bit jayant quantizer.

i am studying jayant quantizer. but i couldnt find appropriate stuff
on how to choose the multipliers and delta value. for eg., say i have
a speech signal, if i have to perform adpcm, using 3-bit jayant
quantizer, how do i choose the values of multipliers and delta value?
should i get the probabilities of values by examining all the sample
values of speech? i have read sayood. i have also reffered jayant and
noll, but found it very confusing. any help please?
thakyou.

sameerkumarkaushik@gmail.com (blue) wrote in message news:<18361d26.0504100608.2c3401@p
osting.google.com>...
> i am studying jayant quantizer. but i couldnt find appropriate stuff
> on how to choose the multipliers and delta value. for eg., say i have
> a speech signal, if i have to perform adpcm, using 3-bit jayant
> quantizer, how do i choose the values of multipliers and delta value?
> should i get the probabilities of values by examining all the sample
> values of speech? i have read sayood. i have also reffered jayant and
> noll, but found it very confusing. any help please?
> thakyou.

i assume this is a delta sigma modulator thing. if adpcm steps to
small then slew rate limiting leads to low pass filtering of signal,
where as steps too big then lots of high frequency quantization noise.
whats your bit depth of signal in? do you have a post integration
filter in the feedback loop? why would you not use a 3 times
oversampled 1 bit converter with a butterworth low pass in the
feedback loop (high order).

Simon Jackson, BEng. wrote:
> sameerkumarkaushik@gmail.com (blue) wrote in message
news:<18361d26.0504100608.2c3401@posting.google.com>... 
stuff 
have 
value? 
sample 
and 
>
> i assume this is a delta sigma modulator thing. if adpcm steps to
> small then slew rate limiting leads to low pass filtering of signal,
> where as steps too big then lots of high frequency quantization
noise.
> whats your bit depth of signal in? do you have a post integration
> filter in the feedback loop? why would you not use a 3 times
> oversampled 1 bit converter with a butterworth low pass in the
> feedback loop (high order).

i dont know what is delta-sigma modulator. but as far as i know. this
system has no such component. i have a simple .wav file which has some
5000 samples. i have to contruct a adpcm with 3-bit jayanth quantizer.
so i'm struct here. i dont know how to choose the step size and
multiplier values. any help?

"blue" <sameerkumarkaushik@gmail.com> wrote in message news:<1113321947.127833.188930@z
14g2000cwz.googlegroups.com>...
> Simon Jackson, BEng. wrote: 
> news:<18361d26.0504100608.2c3401@posting.google.com>... 
> stuff 
> have 
> value? 
> sample 
> and 
> noise. 
>
> i dont know what is delta-sigma modulator. but as far as i know. this
> system has no such component. i have a simple .wav file which has some
> 5000 samples. i have to contruct a adpcm with 3-bit jayanth quantizer.
> so i'm struct here. i dont know how to choose the step size and
> multiplier values. any help?

delta sigma: input to comparator ( > or < ), output most significant
bit, filter output and apply to feedback of comparator,
input-feedback>0 out 1 else out 0.

set end ranges for both values, and use half way values for lowest rms
error over 5000 samples. converge to the optimum root solution.