dbode results in phase shift that is wrong (I believe)

I'm looking at bode plots of linear phase FIR filters and I believe I am getting the wrong phase plot with filter order > 5.

In particular, I'm looking at a moving average filter.

% FIR phase error ?

% LFIR.m

Ts= 1E-3;

n=6;

num= (1/n)*[ones(1,n)];

den= [1 zeros(1,n)];

sys=tf(num,den,Ts);

figure(1); pzmap(sys);

figure(2); dbode(num,den,Ts);

From the pzmap it can be seen that for small positive frequencies on the unit circle the phase contribution should be near zero. However the dbode plot shows the phase at w=0 r/s to be 360 degree. If I use n=5 or less, I do get 0 deg phase at w ~= 0 rad/sec. Is this a problem with dbode?

MATLAB Version: 9.10.0.1739362 (R2021a) Update 5

Here are the pzmap and dbode for n=6:

Answer 1

[Paul]

dbode() callse bode(), which in turn has an algorithm that tries to "unwrap" the phase by adding multiples of 360 deg across the frequency range. But the phase at each frequency is still correct. We can see this, for example, at low frequency

Ts= 1E-3;
n=6;
num= (1/n)*[ones(1,n)];
den= [1 zeros(1,n)];
sys=tf(num,den,Ts);
[m,p,w] = bode(sys);
p(1)

ans = 357.9946

pcheck = angle(polyval(num,exp(1j*w(1)*Ts))/polyval(den,exp(1j*w(1)*Ts)))*180/pi

pcheck = -2.0054

pcheck + 360

ans = 357.9946

If you prefer the phase to always be between +-180, one approach is to use bodeplot() with phase wrapping on

opts = bodeoptions;
opts.PhaseWrapping = 'on';
bodeplot(sys,opts)

[Brian Tremaine]

Thanks for the very clear and concise answer. Since most my work is on closed loop control systems I'm interested in the unwrapped phase shift below the Nyquist frequency.