Hornresp

[tb46]

Hi David,

Thank you for your continuing improvements of Hornresp. The easy to use impulseresp tool is great. Could a "Compare Previous" button be added to this function? :-)

And, the additional number of data records in combination with an improved find function is quite helpful.

Thanks again,

Regards,

[tb46]

Hi Sabbelbacke,

Thanks for all the simulations.

Hi again David,

It seems that the impulse response curve is derived from the last SPL calculation, and not the particular driver plus enclosure combination in general. That means that it is really important to get the path length right for the combined response graph (e.g.: back-loaded horns).

I wonder if this is a feature or a bug? :-)

Regards,

[Eva]

The phase and group delay plots in Hornresp still represent the response at the driver (useless) rather than the response being shown in the SPL graph, regardless of the way in which it was last calculated.

Could this be fixed? Couldn't phase and group delay be derived from the SPL plot every time it's calculated?

[David McBean]

Originally posted by Sabbelbacke
I did some comparison to akabak and LSPCad regarding impulsresp.

Hi Sabbelbacke,

Thanks for the comparisons, they are very interesting indeed. Jean-Michel calculates impulse response using a method different to AkAbak and LspCAD, so it is not surprising that there may be some variances in the results.

Kind regards,

David

[David McBean]

Originally posted by tb46
Could a "Compare Previous" button be added to this function?

Hi Oliver,

See my post #471.

Kind regards,

David

[David McBean]

Originally posted by tb46
It seems that the impulse response curve is derived from the last SPL calculation.

Hi Oliver,

In effect, yes.

Kind regards,

David

[David McBean]

Originally posted by Eva
The phase and group delay plots in Hornresp still represent the response at the driver (useless) rather than the response being shown in the SPL graph, regardless of the way in which it was last calculated.

Could this be fixed? Couldn't phase and group delay be derived from the SPL plot every time it's calculated?

Hi Eva,

Sorry to hear that you consider some of the Hornresp results to be useless, and that they need fixing. If system phase and group delay charts were provided, how would you use this information? (Bearing in mind that the phase results would be shown in wrapped form).

Perhaps you could use AkAbak instead, if system phase and group delay are important to you?

Incidentally - the existing phase and group delay charts give an indication of the behaviour of a driver when horn loaded, and are also entirely relevant to a direct radiator loudspeaker system. The Hornresp results have been validated against data published by Marshall Leach.

Kind regards,

David

[Eva]

Knowing the actual phase and group delay of the SPL plot that is being shown is very important for sizing rear chambers for flatter group delay at lower cutoff and for crossover design, for designing horns allowing for easier phase matching (MF, HF). Propagation delay (either derived from horn length or just a value asked to the user) should be substracted for getting more meaningful phase plots without many wraps. Group delay should be meaningful without any adjustment.

The phase and group delay at the driver have little meaning because nobody is listening there, because the other drivers and horns the system is interacting with are not there (except in unity-like systems) and because the impedance ripple of the horn causes horrible phase and group delay ripple at the driver that does not actually appear at the horn mouth in most cases, it's just flattened through the horn.

In general the effective phase and group delay response in the bottom octave (after the horn) is quite different from shown in these plots.

[Jmmlc]

Hello,

As David wrote, the method used to derive the pulse response in Hornresp is different. Specially IFFT is not used.

A discrete inverse Fourier transform based on a log scale of frequency as the one in Hornresp (533 frequency points) is used.

As I told to David, the initial hypothesis is a continuous spectrum. When both the SPL and phase spectrum in Hornresp are continuous, then the calculated pulse is perfectly correct (and we can recover by FFT the same SPL and phase curves as initially caclulated in Hornresp).

Now there is 2 cases where in its actual version the pulse response is questionable.

1) when the phase spectrum is not continuous and shows 360° rotations through -180° and +180° limits.
As an example I took the TH example given by Oliver (tb46) in

http://www.diyaudio.com/forums/attac...amp=1239376483

You'll find in attached file the phase plot that shows several 360° rotations from a frequency point to the next.

This should be fixed if we unwrap the phase (easy to say not so easy to do...). I'll work on it...

2) when the combined response between a front wave and a rear wave is calculated the pulse response shows artifacts.

This is most probably due to the fact that Hornresp doesn't recalculate the phase response which remains unchanged. May be David can see if this "bug" can be fixed.

About Eva's remark, I second David, the hypothesis of a minimum phase behaviour allowing the use of the Hilbert transform to derive the phase from the SPL curve is questionable in the case of long resonant horns (having tubed pipe behaviour) , and in the case of a combined response.

Best regards from Paris, France


Jean-Michel Le Cléac'h

Originally posted by David McBean


Hi Sabbelbacke,

Thanks for the comparisons, they are very interesting indeed. Jean-Michel calculates impulse response using a method different to AkAbak and LspCAD, so it is not surprising that there may be some variances in the results.

Kind regards,

David

[David McBean]

Hi Jean-Michel,

Originally posted by Jmmlc
This should be fixed if we unwrap the phase (easy to say not so easy to do...). I'll work on it...

Many thanks :-).

Originally posted by Jmmlc
This is most probably due to the fact that Hornresp doesn't recalculate the phase response which remains unchanged.

Just to clarify, the phase shown on the phase response chart is not the same as the phase used in the impulse response calculations. The phase for the combined response impulse response is recalculated, and hopefully is the phase of the combined outputs.

The phase shown on the chart does not change because as Eva has pointed out, it simply shows the phase shift through the driver only. It is not updated at any time.

Kind regards,

David