Hornresp

[Eva]

Hi David

I have tried the 22 version too. At last I can see group delay and phase of the acoustical output, even when using combined response The jumps in grup delay in tapped horns (and in the bottom octave of midrange horns) that cause crossover headaches are now evident. The changes in phase and group delay depending on spatial loading are now evident too.

I suggest adding delay compensation to the phase response, and a dialog allowing the user to enter the desired amount of delay, and a checkbox to optionally keep this value for further calculations (as long as the record is not changed). By default that delay could be the one associated to total horn length.

I also suggest adding a range select dialog to the group delay plot, like in the electrical impedance plot but allowing to type the values and with a check box to optionally don't reset the values in the next calculations (again, as long as the record is not changed). This is to overcome the tall and narrow spikes that sometimes appear. Allowing to enter a value that would be just substracted from all the values in the plot would be useful too. This value could be the same used in the phase plot. Remember that many of us are using DSP delay to compensate horns and tweaking the delay sometimes allows minor phase correction.

Also, it seems that phase and group delay don't get updated when calculating frequency response with the directivity option.

Thank you very much for the update

[David McBean]

Originally posted by Jmmlc
The equivalent within Hornresp could be a phase correction by a linear phase equivalent to the mean group delay inside the bandwith.

Hi Jean-Michel,

While your suggestion could certainly be implemented in Hornresp (I have already tested it in the development environment - see attached screenprint) I think I nevertheless would still prefer to display and export 'raw' -pi / +pi wrapped phase instead.

The way things are at the moment, users can process exported Hornresp phase data any way that they wish. As far as I can see, this would not be possible if 'derived' corrected phase was exported instead.

However - if enough users were to indicate that they preferred your corrected phase, then I might be persuaded to change my mind :-). Before anyone asks though, displaying and/or exporting both phase presentations, is not an option that I am willing to consider.

Kind regards,

David

[David McBean]

Originally posted by Eva
I suggest adding delay compensation to the phase response, and a dialog allowing the user to enter the desired amount of delay, and a checkbox to optionally keep this value for further calculations (as long as the record is not changed).

I also suggest adding a range select dialog to the group delay plot, like in the electrical impedance plot but allowing to type the values and with a check box to optionally don't reset the values in the next calculations (again, as long as the record is not changed).

Also, it seems that phase and group delay don't get updated when calculating frequency response with the directivity option.

Hi Eva,

Your suggestions would require significant changes to Hornresp and a considerable amount of work to implement. I will keep them in mind for the future, but I am not promising anything :-).

You are correct - phase and group delay don't get updated when calculating frequency response with the directivity option. I have no plans to change this at the moment. The next release will however allow impulse response to be calculated off-axis and exported as a wave sound file, as requested by Jean-Michel.

Kind regards,

David

[Jmmlc]

Hello David,

"The way things are at the moment, users can process exported Hornresp phase data any way that they wish. As far as I can see, this would not be possible if 'derived' corrected phase was exported instead."

I am sorry we disagree on that and prefer others to comment on the usefulness of the phase as presented in the actual version of Hornresp...

Best regards from Paris, France

Jean-Michel Le Cléac'h

[soongsc]

View phase without the wrap back caused by distance/delay is a much more useful format.

[David McBean]

Originally posted by Jmmlc
"The way things are at the moment, users can process exported Hornresp phase data any way that they wish. As far as I can see, this would not be possible if 'derived' corrected phase was exported instead."

I am sorry we disagree on that and prefer others to comment on the usefulness of the phase as presented in the actual version of Hornresp...

Hi Jean-Michel,

The problem I see in exporting corrected phase is that it has already been processed using what could reasonably be described as an "arbitrary rule". How can the 'raw' phase information then be retrieved from the exported corrected phase data, if required, without having to also export information regarding the correction / normalising rule used?

Kind regards,

David

[David McBean]

Originally posted by soongsc
View phase without the wrap back caused by distance/delay is a much more useful format.

Hi soongsc,

Does this mean that you would be happy for the Hornresp phase chart to use Jean-Michel's method to display phase?

Kind regards,

David

[Eva]

How about a checkbox to tell Hornresp if phase has to be unwrapped or not?

[Jmmlc]

Hello David,

If we let any travel time effect in the phase curve then it is obvious that at high frequency it will correspond to many hundreds or thousands periods and thus rotate the phase many time 2pi.

Then, with rising travel time warping will also affect low frequency and this is something we don't want because the effect of the travel time will mask the phase variation due to the acoustic loading we want to analyse.

IMHO, when it comes to the phase of propagating waves, the only reference that we should consider is the phase as reflected at the origin of the propagation (travel time = 0) though I know this becomes a bit difficult when multipath waves exist (e.g.: combined load...).

You know perfectly all of that, but this leads to the additional question of "phase curve what for?". Having quite a long experience on that subject, I don't follow Eva on the usefulness of the knowledge of the phase curve to program a digital filter:
1) simulation is not measurement
2) phase values taken in an absolute way are mostly useless ...

For practical purpose as crossover set up (delays, phase correction..) Hornresp's group delay curve is much more important to consider IMHO (specially when the group delay is expressed in equivalent distance travelled at the speed of sound).

The main importance of the phase curve I can see in Hornresp is when performing optimisation, e.g.: optimisation of rear load when trying to compensate reactance... as details on the phase curve are more easily visible than on the group delay curve.

But I'll be interested to know how the way other people are using the phase curve given by Hornresp.

Best regards from Paris. France

Jean-Michel Le Cléac'h

Originally posted by David McBean


Hi Jean-Michel,

The problem I see in exporting corrected phase is that it has already been processed using what could reasonably be described as an "arbitrary rule". How can the 'raw' phase information then be retrieved from the exported corrected phase data, if required, without having to also export information regarding the correction / normalising rule used?

Kind regards,

David

[David McBean]

Hi Jean-Michel,

Originally posted by Jmmlc
The main importance of the phase curve I can see in Hornresp is when performing optimisation, e.g.: optimisation of rear load when trying to compensate reactance... as details on the phase curve are more easily visible than on the group delay curve.

Thanks for your insights into phase and group delay. There is one thing though that I still don't quite understand:

Your corrected phase is calculated as follows -
for I = 1 : Nfreq;
PUNW(I) = PUNW(I) - Dphase * (FREQ(I) - FREQ(1));
end

Why is it not calculated as -
for I = 1 : Nfreq;
PUNW(I) = PUNW(I) - (PUNW(Imag1) + Dphase * (FREQ(I) - FREQ(Imag1)));
end

Assuming that in both cases -
Dphase = (PUNW(Imag2) - PUNW(Imag1)) / (FREQ(Imag2) - FREQ(Imag1));

Any further comments would be appreciated.

Originally posted by Jmmlc
But I'll be interested to know how the way other people are using the phase curve given by Hornresp.

Me too :-).

Kind regards,

David