Hilbert transform, minimum phase, Speaker Workshop, questions...

[mr_push_pull]

Short background: I'm trying to do acoustical measurements with Speaker Workshop, but I have a soundcard with unstable (varying) latency, so I can't use the measured phase in Speaker Workshop to design my XO. Instead I use minimum phase derived from amplitude response to calculate minimum phase and account for driver offset (I haven't figured out exactly how yet).

Now the questions. My mic preamp has a low-pass filter with a corner frequency of about 100Hz at the input, and I don't want to disassemble it to change the caps in the filter to a bigger value (it's in a case and it's kinda hard to get the PCB out because of some wires attached to it, don't ask ).
So this means that my nearfield measurements aren't usable.

Now the question. I don't know exactly how the phase is derived with the Hilbert transform (this notion of minimum phase system it's all fuzzy to me). Can I use only the gated farfield (down to approx 400Hz minimum because of my small room with lots of furniture) to derive usable phase data? Or the response in the lower frequency band influences the phase in the whole band and it will render my derived minimum phase useless?

[SY]

Rather than slog through a lot of work trying to get inappropriate hardware to work, why not just replace the hardware? A decent test mike can be fabricated for under $50, and a better-than-usable soundcard is not much more.

The best amateur-level explanations of minimum phase and Hilbert transforms can be found in Joe d'Appolito's book on measuring loudspeakers. Basically, a minimum phase device's phase can be calculated to be proportional to the derivative of amplitude with respect to frequency.

[phase_accurate]

Basically, a minimum phase device's phase can be calculated to be proportional to the derivative of amplitude with respect to frequency.

Unfortunately multiway speakers are usually ( > 99 % of them !!!!) not minimum phase and therefore making any attempt of calculating phase-response from amplitude response useless.

Regards

Charles

[SY]

Charles: Exactly!

[mr_push_pull]

Originally posted by SY
Rather than slog through a lot of work trying to get inappropriate hardware to work, why not just replace the hardware? A decent test mike can be fabricated for under $50, and a better-than-usable soundcard is not much more.

The best amateur-level explanations of minimum phase and Hilbert transforms can be found in Joe d'Appolito's book on measuring loudspeakers.

1. I have an ECM8000, and the preamp I built myself. It uses just two opamps and it's battery powered. I'd call it decent, except the fact that I haven't gone throught the trouble to calculate or simulate the cut-off frequency of that filter before putting it into the case. Basically it's the Paia preamp ( link to schematic for whoever it may interest ) without the voltage multiplier as phantom source, but batteries instead.
What is there inappropriate about it?

2. I already spent $50 on a soundcard bought only for measurement purposes (Terratec DMX Xfire 1024). I don't know if the latency problem is caused by the card itself or my PC.

3. I only work late hours because I have a job, and quite demanding one, and spending more time searching for, installing, configuring etc a SC that may prove to be as "inappropriate" as my current one is out of the question.

Thanks for the tip on Measuring Loudspeakers.

[claudio]

Hi,
since mr_push_pull is looking at the minimum phase for crossover simulation, I don't understand why should "any attempt of calculating phase-response from amplitude response useless": we are talking of each driver minimum phase calculation. In my HP I show that calculating minimum phase from the amplitude response works pretty well for crossover simulation.

About the unusable near field: are you going to cross below 400Hz ?
if not, try using the Normalizer of the FRD Combiner (I imagine you are using this program to calculate Minimum Phase from the frequency response) to extend the response: it is important to have a well extended response, to get a reliable minimum phase (see David Ralph articles).

Regards,
Claudio

[ucla88]

Originally posted by mr_push_pull
Can I use only the gated farfield (down to approx 400Hz minimum because of my small room with lots of furniture) to derive usable phase data? Or the response in the lower frequency band influences the phase in the whole band and it will render my derived minimum phase useless?

Yes,

As Claudio pointed out, and to clarify a bit more, loudspeaker driver units are virtually exclusively minimum phase devices and the Hilbert transform can be used. A loudspeaker with drivers and crossover usually is not minimum phase and the HB transform will not accurately give phase.

But your comment about the response in the whole band influencing HB phase is true. It won't be useless, but there will be some error in it. I'm not exactly sure how SW handles the HB transform. You can do it, but you'll have to fiddle with your results and probably do a couple of measurements and models to get a good result.

[phase_accurate]

The most necessary feature of HB transform when measuring a multiway speaker is the calcualtion of the excess phase.
It does so by comparing the actual measured phase against HB transform calculated phase.

Regards

Charles

[claudio]

ucla88 wrote:
I'm not exactly sure how SW handles the HB transform.

Actually SW measures the phase, doesn't calculate it from HT; if I remember well MLSSA does the same. However it is useless if a varying latency exists, as happens to mr_push_pull.

Regards,
Claudio

[ucla88]

Originally posted by phase_accurate
The most necessary feature of HB transform when measuring a multiway speaker is the calcualtion of the excess phase.
It does so by comparing the actual measured phase against HB transform calculated phase.

Regards

Charles

indeed.

http://my.starstream.net/mk/groupdelay/groupdelay.htm


But that doesn't exactly pertain to the orginal poster's question. Which is, how inaccurate will his HB transform be and how good will the modeling process be if he uses the HB plus a time delay and a correction for acoustic center offset. And again, the answer is very difficult to know for sure. It can work, but errors will accumulate along the way.