FIR filter sampling frequency

I have some questions regarding the sampling frequency of a digital filter?
This is the rate at which the internal stages are clocked, why does this
have to be twice the incoming data rate. I don't understand where Nyquist
comes in the filter design.Our incmoing data was generated by sampling a
analog at twice frequency, and now my filter must run at twice the data
rate.Why?

Zeph80 wrote:
> I have some questions regarding the sampling frequency of a digital filter?
> This is the rate at which the internal stages are clocked, why does this
> have to be twice the incoming data rate. I don't understand where Nyquist
> comes in the filter design.Our incmoing data was generated by sampling a
> analog at twice frequency, and now my filter must run at twice the data
> rate.Why?

In a nutshell, the reason for the sampling requirements are to prevent
aliasing, which is a condition where higher frequency components mask
as lower frequency components.  The mathematical basis of this involves
the periodic nature of signals in discrete time and if you are really
interested in it, I would suggest getting an introductory signal
processing book as there are several key concepts that you need to be
familiar with to understand the derivation.

The operation, or rather clock rate, of your digital filter depends on
other factors.  If you wish for your filter to operate in real time,
your system must be operating or processing at a clock rate faster than
you are sampling the data so that it can keep up.  On the other hand,
if you are only post processing aquired data there is no such
constraint on the processing rate of your filter.

>Zeph80 wrote:
>> I have some questions regarding the sampling frequency of a digital
filter?
>> This is the rate at which the internal stages are clocked, why does
this
>> have to be twice the incoming data rate. I don't understand where
Nyquist
>> comes in the filter design.Our incmoing data was generated by sampling
a
>> analog at twice frequency, and now my filter must run at twice the
data
>> rate.Why?
>
>In a nutshell, the reason for the sampling requirements are to prevent
>aliasing, which is a condition where higher frequency components mask
>as lower frequency components.  The mathematical basis of this involves
>the periodic nature of signals in discrete time and if you are really
>interested in it, I would suggest getting an introductory signal
>processing book as there are several key concepts that you need to be
>familiar with to understand the derivation.
>
>The operation, or rather clock rate, of your digital filter depends on
>other factors.  If you wish for your filter to operate in real time,
>your system must be operating or processing at a clock rate faster than
>you are sampling the data so that it can keep up.  On the other hand,
>if you are only post processing aquired data there is no such
>constraint on the processing rate of your filter.
>
>
Yes, I understand aliasig and sampling.But I havent done digital filter
design. My question was related to the latter part of your answer.Most
places I read they refer to this clock rate as oversampling and in a real
time system state that it should run at least twice . I was just wondering
why at least twice

Zeph80 wrote:
> I have some questions regarding the sampling frequency of a digital filter?
> This is the rate at which the internal stages are clocked, why does this
> have to be twice the incoming data rate. I don't understand where Nyquist
> comes in the filter design.Our incmoing data was generated by sampling a
> analog at twice frequency, and now my filter must run at twice the data
> rate.Why?

There seems to be a bit of confusion. The sample rate has to exceed 
twice the highest frequency that the analog signal might contain. Once 
sampling is accomplished. the sample rate _is_the data rate.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

Jerry Avins wrote:

> Zeph80 wrote:
> 
>> I have some questions regarding the sampling frequency of a digital 
>> filter?
>> This is the rate at which the internal stages are clocked, why does this
>> have to be twice the incoming data rate. I don't understand where Nyquist
>> comes in the filter design.Our incmoing data was generated by sampling a
>> analog at twice frequency, and now my filter must run at twice the data
>> rate.Why?
> 
> 
> There seems to be a bit of confusion. The sample rate has to exceed 
> twice the highest frequency that the analog signal might contain. Once 
> sampling is accomplished. the sample rate _is_the data rate.
> 
> Jerry

Jerry having cleared that up, here's another issue, though:

You say that you're sampling the analog at 'twice frequency'.  Nyquist 
is a theoretical limit, and only applies if you have perfect filters and 
perfect patience -- because a perfect brick-wall filter will take an 
infinite amount if time to settle.

Usually one samples at much higher than Nyquist, with a tradeoff being 
made between sampling rate, settling time, anti-aliasing filter 
complexity and the purity of the sampled signal.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google?  See http://cfaj.freeshell.org/google/

"Zeph80" <surabhi_talwar@hotmail.com> wrote in message
news:G_SdnaGLfeb3AcrZRVn-pA@giganews.com...
> >Zeph80 wrote:
> >> I have some questions regarding the sampling frequency of a digital
> filter?
> >> This is the rate at which the internal stages are clocked, why does
> this
> >> have to be twice the incoming data rate. I don't understand where
> Nyquist
> >> comes in the filter design.Our incmoing data was generated by sampling
> a
> >> analog at twice frequency, and now my filter must run at twice the
> data
> >> rate.Why?
> >
> >In a nutshell, the reason for the sampling requirements are to prevent
> >aliasing, which is a condition where higher frequency components mask
> >as lower frequency components.  The mathematical basis of this involves
> >the periodic nature of signals in discrete time and if you are really
> >interested in it, I would suggest getting an introductory signal
> >processing book as there are several key concepts that you need to be
> >familiar with to understand the derivation.
> >
> >The operation, or rather clock rate, of your digital filter depends on
> >other factors.  If you wish for your filter to operate in real time,
> >your system must be operating or processing at a clock rate faster than
> >you are sampling the data so that it can keep up.  On the other hand,
> >if you are only post processing aquired data there is no such
> >constraint on the processing rate of your filter.
> >
> >
> Yes, I understand aliasig and sampling.But I havent done digital filter
> design. My question was related to the latter part of your answer.Most
> places I read they refer to this clock rate as oversampling and in a real
> time system state that it should run at least twice . I was just wondering
> why at least twice

You are getting confused with terminology.
You can 'clock' your digital filter at pretty much any rate you want. For
real time operation this rate will typically be higher than the sample rate
of the incoming digital data. Depending on the hardware implementation, this
can be *much* higher than the incoming data rate. This is related to the
*hardware* design of digital filters.
There is another aspect - the actual design of the filter properties - this
deals with design of the right set of coeffs for the system. This design
refers to the sample rate and has nothing to do with the clock rate which
your hardware uses. I suspect you've seen literature that refers to this
sample rate and that it must be twice the highest frequency of the analog
signal. Your clock rate could even be 10x slower than your sample rate if
you are post-processing captured data - there isn't a relationship for clock
rate and the input signal frequency content.

Hope that clarifies some confusion.

Cheers
Bhaskar

>Jerry Avins wrote:
>
>> Zeph80 wrote:
>> 
>>> I have some questions regarding the sampling frequency of a digital 
>>> filter?
>>> This is the rate at which the internal stages are clocked, why does
this
>>> have to be twice the incoming data rate. I don't understand where
Nyquist
>>> comes in the filter design.Our incmoing data was generated by sampling
a
>>> analog at twice frequency, and now my filter must run at twice the
data
>>> rate.Why?
>> 
>> 
>> There seems to be a bit of confusion. The sample rate has to exceed 
>> twice the highest frequency that the analog signal might contain. Once

>> sampling is accomplished. the sample rate _is_the data rate.
>> 
>> Jerry
>
>Jerry having cleared that up, here's another issue, though:
>
>You say that you're sampling the analog at 'twice frequency'.  Nyquist 
>is a theoretical limit, and only applies if you have perfect filters and

>perfect patience -- because a perfect brick-wall filter will take an 
>infinite amount if time to settle.
>
>Usually one samples at much higher than Nyquist, with a tradeoff being 
>made between sampling rate, settling time, anti-aliasing filter 
>complexity and the purity of the sampled signal.
>
>-- 
>
>Tim Wescott
>Wescott Design Services
>http://www.wescottdesign.com
>
>Posting from Google?  See http://cfaj.freeshell.org/google/
>
Thank you, my confusion's cleared. I was under the impression, they meant
sample rate of filter is twice data rate.So it has to be at least more
than twice highest freq in analog signal. AAnd since you wont have a ideal
brick wall filter, you choose more than 2fs.This is the lower bound for
your D/A and digital filter.

>"Zeph80" <surabhi_talwar@hotmail.com> wrote in message
>news:G_SdnaGLfeb3AcrZRVn-pA@giganews.com...
>> >Zeph80 wrote:
>> >> I have some questions regarding the sampling frequency of a digital
>> filter?
>> >> This is the rate at which the internal stages are clocked, why does
>> this
>> >> have to be twice the incoming data rate. I don't understand where
>> Nyquist
>> >> comes in the filter design.Our incmoing data was generated by
sampling
>> a
>> >> analog at twice frequency, and now my filter must run at twice the
>> data
>> >> rate.Why?
>> >
>> >In a nutshell, the reason for the sampling requirements are to
prevent
>> >aliasing, which is a condition where higher frequency components mask
>> >as lower frequency components.  The mathematical basis of this
involves
>> >the periodic nature of signals in discrete time and if you are really
>> >interested in it, I would suggest getting an introductory signal
>> >processing book as there are several key concepts that you need to be
>> >familiar with to understand the derivation.
>> >
>> >The operation, or rather clock rate, of your digital filter depends
on
>> >other factors.  If you wish for your filter to operate in real time,
>> >your system must be operating or processing at a clock rate faster
than
>> >you are sampling the data so that it can keep up.  On the other hand,
>> >if you are only post processing aquired data there is no such
>> >constraint on the processing rate of your filter.
>> >
>> >
>> Yes, I understand aliasig and sampling.But I havent done digital
filter
>> design. My question was related to the latter part of your answer.Most
>> places I read they refer to this clock rate as oversampling and in a
real
>> time system state that it should run at least twice . I was just
wondering
>> why at least twice
>
>You are getting confused with terminology.
>You can 'clock' your digital filter at pretty much any rate you want.
For
>real time operation this rate will typically be higher than the sample
rate
>of the incoming digital data. Depending on the hardware implementation,
this
>can be *much* higher than the incoming data rate. This is related to the
>*hardware* design of digital filters.
>There is another aspect - the actual design of the filter properties -
this
>deals with design of the right set of coeffs for the system. This design
>refers to the sample rate and has nothing to do with the clock rate
which
>your hardware uses. I suspect you've seen literature that refers to this
>sample rate and that it must be twice the highest frequency of the
analog
>signal. Your clock rate could even be 10x slower than your sample rate
if
>you are post-processing captured data - there isn't a relationship for
clock
>rate and the input signal frequency content.
>
>Hope that clarifies some confusion.
>
>Cheers
>Bhaskar

Yes in the filter design,  I've seen sampling frequency being used is 84
khz for a 10.5 khz symbol rate.They refer to this as 8 samples per
symbol.What is this frequency and how does it affect the filter.
>
>
>
>
>
>
>
>

>"Zeph80" <surabhi_talwar@hotmail.com> wrote in message
>news:G_SdnaGLfeb3AcrZRVn-pA@giganews.com...
>> >Zeph80 wrote:
>> >> I have some questions regarding the sampling frequency of a digital
>> filter?
>> >> This is the rate at which the internal stages are clocked, why does
>> this
>> >> have to be twice the incoming data rate. I don't understand where
>> Nyquist
>> >> comes in the filter design.Our incmoing data was generated by
sampling
>> a
>> >> analog at twice frequency, and now my filter must run at twice the
>> data
>> >> rate.Why?
>> >
>> >In a nutshell, the reason for the sampling requirements are to
prevent
>> >aliasing, which is a condition where higher frequency components mask
>> >as lower frequency components.  The mathematical basis of this
involves
>> >the periodic nature of signals in discrete time and if you are really
>> >interested in it, I would suggest getting an introductory signal
>> >processing book as there are several key concepts that you need to be
>> >familiar with to understand the derivation.
>> >
>> >The operation, or rather clock rate, of your digital filter depends
on
>> >other factors.  If you wish for your filter to operate in real time,
>> >your system must be operating or processing at a clock rate faster
than
>> >you are sampling the data so that it can keep up.  On the other hand,
>> >if you are only post processing aquired data there is no such
>> >constraint on the processing rate of your filter.
>> >
>> >
>> Yes, I understand aliasig and sampling.But I havent done digital
filter
>> design. My question was related to the latter part of your answer.Most
>> places I read they refer to this clock rate as oversampling and in a
real
>> time system state that it should run at least twice . I was just
wondering
>> why at least twice
>
>You are getting confused with terminology.
>You can 'clock' your digital filter at pretty much any rate you want.
For
>real time operation this rate will typically be higher than the sample
rate
>of the incoming digital data. Depending on the hardware implementation,
this
>can be *much* higher than the incoming data rate. This is related to the
>*hardware* design of digital filters.
>There is another aspect - the actual design of the filter properties -
this
>deals with design of the right set of coeffs for the system. This design
>refers to the sample rate and has nothing to do with the clock rate
which
>your hardware uses. I suspect you've seen literature that refers to this
>sample rate and that it must be twice the highest frequency of the
analog
>signal. Your clock rate could even be 10x slower than your sample rate
if
>you are post-processing captured data - there isn't a relationship for
clock
>rate and the input signal frequency content.
>
>Hope that clarifies some confusion.
>
>Cheers
>Bhaskar

Yes in the filter design example I've seen sampling frequency being used
is 84khz for a 10.5 khz symbol rate.They refer to this as 8 samples per
symbol.What is this "sampling frequency" and why is it so high?

"Zeph80" <surabhi_talwar@hotmail.com> wrote in message
news:PtGdnfLHkNleJcrZnZ2dnUVZ_sadnZ2d@giganews.com...
> >"Zeph80" <surabhi_talwar@hotmail.com> wrote in message
> >news:G_SdnaGLfeb3AcrZRVn-pA@giganews.com...
> >> >Zeph80 wrote:
> >> >> I have some questions regarding the sampling frequency of a digital
> >> filter?
> >> >> This is the rate at which the internal stages are clocked, why does
> >> this
> >> >> have to be twice the incoming data rate. I don't understand where
> >> Nyquist
> >> >> comes in the filter design.Our incmoing data was generated by
> sampling
> >> a
> >> >> analog at twice frequency, and now my filter must run at twice the
> >> data
> >> >> rate.Why?
> >> >
> >> >In a nutshell, the reason for the sampling requirements are to
> prevent
> >> >aliasing, which is a condition where higher frequency components mask
> >> >as lower frequency components.  The mathematical basis of this
> involves
> >> >the periodic nature of signals in discrete time and if you are really
> >> >interested in it, I would suggest getting an introductory signal
> >> >processing book as there are several key concepts that you need to be
> >> >familiar with to understand the derivation.
> >> >
> >> >The operation, or rather clock rate, of your digital filter depends
> on
> >> >other factors.  If you wish for your filter to operate in real time,
> >> >your system must be operating or processing at a clock rate faster
> than
> >> >you are sampling the data so that it can keep up.  On the other hand,
> >> >if you are only post processing aquired data there is no such
> >> >constraint on the processing rate of your filter.
> >> >
> >> >
> >> Yes, I understand aliasig and sampling.But I havent done digital
> filter
> >> design. My question was related to the latter part of your answer.Most
> >> places I read they refer to this clock rate as oversampling and in a
> real
> >> time system state that it should run at least twice . I was just
> wondering
> >> why at least twice
> >
> >You are getting confused with terminology.
> >You can 'clock' your digital filter at pretty much any rate you want.
> For
> >real time operation this rate will typically be higher than the sample
> rate
> >of the incoming digital data. Depending on the hardware implementation,
> this
> >can be *much* higher than the incoming data rate. This is related to the
> >*hardware* design of digital filters.
> >There is another aspect - the actual design of the filter properties -
> this
> >deals with design of the right set of coeffs for the system. This design
> >refers to the sample rate and has nothing to do with the clock rate
> which
> >your hardware uses. I suspect you've seen literature that refers to this
> >sample rate and that it must be twice the highest frequency of the
> analog
> >signal. Your clock rate could even be 10x slower than your sample rate
> if
> >you are post-processing captured data - there isn't a relationship for
> clock
> >rate and the input signal frequency content.
> >
> >Hope that clarifies some confusion.
> >
> >Cheers
> >Bhaskar
>
> Yes in the filter design example I've seen sampling frequency being used
> is 84khz for a 10.5 khz symbol rate.They refer to this as 8 samples per
> symbol.What is this "sampling frequency" and why is it so high?

This sampling frequency is the rate at which the incoming data gets sampled.
The symbol rate is another way of stating the BW of the incoming signal.
Nyquist theorem states that the sampling freq needs to be greater than twice
the BW of the input signal (some books use 2x the highest freq for
simplicity but it's really BW).

In digital comms, it's common to sample at 4x or 8x the symbol rate to help
with synchronization.

Hope this helps.

Cheers
Bhaskar