[sdiy] Plotting the phase of a filter with FFT or a Phase Comparator?

[cheater00 .]

Hi Neil!
Thank you very much for a very informed answer.

On Mon, May 6, 2013 at 4:16 PM, Neil Johnson <neil.johnson71 at gmail.com> wrote:
> cheater00 . wrote:
>>
>> Hi guys,
>> I have been looking for a good way to measure the phase shift of a
>> filter. One way is to use an FFT spectrograph, but it isn't always the
>> best idea:
>> - they're a bit cumbersome to use. tiny screens and even tinier
>> resolutions don't help.
>
>
> Which specific "FFT spectrograph"s (I guess you mean a spectrum analyser?)

Yeah - thanks! Living in Germany is really killing me.

> with tiny screens and even tinier resolutions are you referring to?

Well, the HP 35660A :-) for one thing. Those things just seem so
cumbersome, especially if I compare to a full-screen spectrogram  on a
TV!

> More importantly, what resolution are you looking for?

I think it's less about what resolution I'm looking for, and more
about what resolution can be had..

>> - it's difficult to find an FFT which does audio frequencies, and if
>> you do they're usually a bit expensive
>
>
> For a rough estimate you could use a PC sound card.

Sure, but that's not really going to give me any certainties. Agreed
though, a PC sound card is an OK first approximation.

>> - there's little resolution in the low end (not great if you want to
>> e.g. design a filter network for a subwoofer, or see how your vcf
>> behaves in the low end of the spectrum)
>
>
> Oh I don't know ... my HP 35660A can resolve down to microHz if you give it
> long enough.

Yeah, I have looked at the 35660A and 3561A. Which one would you
recommend and why? Is there any "trick" to finding spectrum analyzers
that go this low in frequency?

>> An alternative I've found is to use a phase comparator. First off, I
>> stumbled upon something called a "phase writer", which is one of those
>> wax roll printers. What it does is it takes two sine waves as input -
>> one is the reference and one is the tested wave, and plot the phase
>> shift between them (either in the range -tau/2..tau/2 or 0..tau).
>
>
> Sounds like a mechanical version of a network analyser.

I think it's not mechanical - fully electronic. I think the Tracor
905a might be analog while the Dranetz 305 is digital, but not sure?
I'll look into it.. there's some manual browsing in store for me I
guess.

>> The less cumbersome version is the Phase Comparator (not really sure
>> if there's much of a connection with the circuit used in PLL's). It
>> takes two sine waves, again, but this time it outputs a DC voltage or
>> current which is proportional to the phase shift. Two that seem to be
>> useful for low-frequency work are the Dranetz 305 (apparently very
>> precise, and somewhat inexpensive), as well as the Tracor 905a (google
>> thinks it's a tractor if you don't include the "a" so beware). Manuals
>> or brochures are available for both.
>
>
> HP did a phase-magnitude meter that seems quite popular with the audio
> crowd.  HP 3575A Gain-Phase meter, although it only goes down to around
> 40Hz.  Below that you're really looking at mechanical or FFT.

Thanks, I'll research the HP3575A and Gain-Phase meters in general.

>> One thing that's bothering me about phase comparators is that they are
>> made to compare the phase of static frequencies, and they have a
>> settling time, which means you have to turn off the Z axis or the
>> storage in the scope while the comparator is settling. Ideally you'd
>> want to have a sweep, and no gating at all.
>
>
> *ANY* frequency or phase mesaurement takes some time, and the lower the
> frequency the longer the time it takes to get, more so if you want more
> significant digits.  That's life.  Or start with an HP 141T and the
> 30Hz-300kHz plugin, and then you really appreciate what "slow" means....

Would you suggest the HP141T as a good spectrum analyzer?

>> Is there any other equipment useful for this purpose that you could
>> recommend? I know there's the impulse method, but again that's
>> deconvolution/fft. I'm not sure there's a box which does that
>> specifically, it's always integrated into fft spectrographs.
>
>
> Audio (specifically 20Hz to 20kHz) can be done on the cheap with a PC sound
> card and software.  Below 20Hz or above 20kHz and you start heading towards
> proper test equipment.
>
> The classic HP spectrum analyser for audio work is the 3580A, covering 1Hz
> to 50kHz.

Can you compare the 3580A to the 3561A? The 3561A does 125uHz-100kHz,
but not sure about much more.

>> BTW, can someone recommend good spectrographs which go down to 1Hz?
>
>
> Check out the FFT-based HP Dynamic Signal Analysers.  To properly measure
> phase you need a dual channel unit.  The 3562A still commands good prices,
> but if you hunt around you can find something like the 35660A or 35665A for
> reasonable prices.

Again.. would you be able to tell the differences that 3562A brings
in? And the 35665A?

Cheers,
D.