sample rate change of a narrow band signal

I want to change sampling rate of a narrow band signal by a fractional
factor of 15/13. The original signal is -60dB or lower after f=0.08 (1
corresponds to half the sampling rate). I use matlab upfirdn function and
design the anti-imaging/anti-aliasing filter using remez. The problem is
that i get small droop in the passband no matter how sharp a filter i have
got. I know this sounds a bit vague but i think may be someone can help
this has to do something with the signal being narrow band.

fahim skrev:
> I want to change sampling rate of a narrow band signal by a fractional
> factor of 15/13.

OK. How many samples are present in the signal?
What is the relativ bandwidth of the signal? What is
the center frequency of the signal?

>The original signal is -60dB or lower after f=0.08 (1
> corresponds to half the sampling rate).

60 dB lower than what? How did you measure this number?

> I use matlab upfirdn function and
> design the anti-imaging/anti-aliasing filter using remez.

How have you implemented the re-sampling? What are
the specs of the anti-alias filter?

> The problem is
> that i get small droop in the passband no matter how sharp a filter i have
> got.

What exactly do you mean by "droop"? How do you see this?
How sharp filters do you design?

> I know this sounds a bit vague but i think may be someone can help

You are right in that your description is vague. A little
more information might enable people here to help.

> this has to do something with the signal being narrow band.

Maybe, and matbe not.

Rune

Hey Rune,

Thanks alot for your reply.I will try to make it more clear in response to
your questions.


>fahim skrev:
>> I want to change sampling rate of a narrow band signal by a fractional
>> factor of 15/13.
>
>OK. How many samples are present in the signal?
>What is the relativ bandwidth of the signal? What is
>the center frequency of the signal?

Well the exact number of samples are 32768=128(samples) x 256 (upsampling
factor). The signal has not been upconverted, it is at baseband (center
frequency=zero). relative bandwith is 0.08 after which signal is down by
60 db below zero. 


>>The original signal is -60dB or lower after f=0.08 (1
>> corresponds to half the sampling rate).
>
>60 dB lower than what? How did you measure this number?
>

I measure this by plotting magnitude response of signal using freqz
function.

>> I use matlab upfirdn function and
>> design the anti-imaging/anti-aliasing filter using remez.
>
>How have you implemented the re-sampling? What are
>the specs of the anti-alias filter?
>

I believe you are refering to "resample" function of matlab. Well i
thought its filter may not meet the narrow band signal requirements. Any
way with 200 taps provided to "resample" performance is bad as compared to
custom made filter via remez. As far as the specs of remez filter goes,
since i upsample signal by 15, the passband cutoff is at 0.7/15 and
stopband cutoff 0.9/15. 

>> The problem is
>> that i get small droop in the passband no matter how sharp a filter i
have
>> got.
>
>What exactly do you mean by "droop"? How do you see this?
>How sharp filters do you design?
>

Once again i observe the droop after plotting magnitude response of
resampled signal using freqz function. Instead of monotonically decreasing
curve just after passband there is an 'S' like curve its fairly small but
should'nt we get the same frequency response if nyquist theorem is
satisfied, signal being fairly narrowband.

>> I know this sounds a bit vague but i think may be someone can help
>
>You are right in that your description is vague. A little
>more information might enable people here to help.
>
I understand. hopefully now its more clearly stated.The problem has been
bothering me for a while now.

>> this has to do something with the signal being narrow band.
>
>Maybe, and matbe not.
>
>Rune
>
>

Once again thanks alot for your time and reply.
fahim