Why does the magnitude response of freqz change with the value of the input argument n?

I am using the freqz command, and the second argument, n, is documented as N-point frequency response. Therefore, I take it to mean that as n gets larger, the resolution of the frequency response should get finer and more smooth. But when I run this command, this is not what happens. Instead, I see the magnitude values changing. This makes no sense, and the documentation is quite poor.

Answer 1

[Cyrus]

OH OH OHHHH, I think I found the problem. it should be the third input argument. When I had it as the second, it was changing the actual FIR filter itself!

freqz(B,1,n)

[Star Strider]

Please note: you did not ever post the code you used in your freqz call.

Our ability to read minds is in general severely lacking.

[Cyrus]

my apologies Star!

Answer 2

[Cyrus]

To be more clear, when I run freqz(B) the amplitude is nearly 1 until the cutoff, and then falls off to zero - as it should. It's a low pass filter of order 80 using fir1.

When I do freqz(B,2000), the amplitude from zero to the cutoff is now no where near 1, its down to 0.00003. This makes no sense.

Answer 3

[Cyrus]

Run this code and change freqz(B) to instead say freqz(B,2000) and see what happens:

rps2Hz = 1/(2*pi);
fs     = 1000;
M      = 15;
fpass  = 0:0.1:fs/(2*M);
frej   = fs/(2*M):0.1:fs/2;
order = 80;
wn    = 1/M;
B     = fir1(order,wn,hamming(order+1));
[h,w] = freqz(B);

figure(2);
plot(fs*w*rps2Hz,abs(h),'linewidth',2.0);
xlabel('freq (Hz)');
ylabel('Amplitude');