min. sampling frequency of DTMF decoder

Given the following frequencies in DTMF :


lfg = [697 770 852 941]; % Low frequency group
hfg = [1209 1336 1477];  % High frequency group


what is the min. sampling frequency and the number of samples required
(using FFT) to detect the keys?


I tried Nyquist sampling frequency (2*max freq = 2*1477 ~3000 )
and for the min. freq resolution requirement of 73Hz (from 697-770Hz)
the number of samples I took is 45 (given by 3000/73 approx)


of course, I am not able to detect all keys, since there will be
leakages.


Is there any other considerations I must take into account when
determining the min sampling freq. of the DTMF decoder?


thanks in advance 


lionel lewis

Lionel Lewis wrote:
> Given the following frequencies in DTMF :
> 
> 
> lfg = [697 770 852 941]; % Low frequency group
> hfg = [1209 1336 1477];  % High frequency group
> 
> 
> what is the min. sampling frequency and the number of samples required
> (using FFT) to detect the keys?
> 
> 
> I tried Nyquist sampling frequency (2*max freq = 2*1477 ~3000 )
> and for the min. freq resolution requirement of 73Hz (from 697-770Hz)
> the number of samples I took is 45 (given by 3000/73 approx)
> 
> 
> of course, I am not able to detect all keys, since there will be
> leakages.
> 
> 
> Is there any other considerations I must take into account when
> determining the min sampling freq. of the DTMF decoder?
> 
> 
> thanks in advance 
> 
> 
> lionel lewis
> 


It seems like time-domain techniques might be a better match for DTMF 
detection since you are only interested in the power values at a handful 
of frequencies, compared to the power contained within a given time 
window.


Disclaimer: I've never made a DTMF detecter myself.

Mark Borgerding wrote:
> Lionel Lewis wrote:
> > Given the following frequencies in DTMF :
> >
> >
> > lfg = [697 770 852 941]; % Low frequency group
> > hfg = [1209 1336 1477];  % High frequency group
> >
> >
> > what is the min. sampling frequency and the number of samples
required
> > (using FFT) to detect the keys?
> >
> >
> > I tried Nyquist sampling frequency (2*max freq = 2*1477 ~3000 )
> > and for the min. freq resolution requirement of 73Hz (from
697-770Hz)
> > the number of samples I took is 45 (given by 3000/73 approx)
> >
> >
> > of course, I am not able to detect all keys, since there will be
> > leakages.
> >
> >
> > Is there any other considerations I must take into account when
> > determining the min sampling freq. of the DTMF decoder?
> >
> >
> > thanks in advance
> >
> >
> > lionel lewis
> >
>
>
> It seems like time-domain techniques might be a better match for DTMF

> detection since you are only interested in the power values at a
handful
> of frequencies, compared to the power contained within a given time
> window.
>
>
> Disclaimer: I've never made a DTMF detecter myself.

DTMF detector algorithms often use Goertzel filters designed for 8 ksps
data. There are detailed specifications that a DTMF detector should
meet if it is to be connected to the PSTN, and there are test tapes
with DTMF on top of speech, unbalanced tones, short tones, noisy tones,
etc. Getting the detector to meet spec is not trivial. 

John

john wrote:
> DTMF detector algorithms often use Goertzel filters designed for 8 ksps
> data. There are detailed specifications that a DTMF detector should
> meet if it is to be connected to the PSTN, and there are test tapes
> with DTMF on top of speech, unbalanced tones, short tones, noisy tones,
> etc. Getting the detector to meet spec is not trivial. 

Unfortunately you can no longer obtain either the Mitel or Bellcore test 
tapes.

Regards,
Steve