Parametric Sum – Thoughts

[jtalden]

JohnM,
I tried out the new parametric sum feature using my SW and midrange measurements. The new feature works very well.
Here is a suggestion and some additional thoughts FYI:

Suggestion:
Selecting the best timing using unsmoothed traces is very difficult in many cases. There are lots of room interactions in this frequency range and it is difficult deciding the maximum SPL support. Smoothing the traces to 1/3 octave is very helpful in finding the best SPL support. We can smooth the SW and midrange traces prior to opening the parametric sum panel, but the parametric sum trace is still unsmoothed. It would be helpful to provide some method of the smoothing it also. Possibly automatically pick up the smoothing selected for the 2 input traces and use that for the parametric sum trace.

Additional thoughts:

1. The all SPL panel can be very busy if there are many measurements selected. It may be helpful to automatically/temporarily deselect all except the 2 measurements that are chosen when the panel is activated.
2. It may be better to close the parametric sum panel with a cancel button, or maybe only for other actions taken within REW, so we can redirect our attention to other activities like; taking a screen snip of the panel, checking notifications, changing focus to our browser, etc.
3. Adding an option to both create the parametric sum and to also invert and shift the impulse per the panel settings would be useful.
4. I like to apply a 5 octave FDW and open the phase overlay window and keep it on top (using 'always on top'). That way I can see the phase tracking relationship as the SW impulse is shifted. Since this new feature is not updating the actual impulse position the phase is not being updated.
5. A couple of these thoughts make me think it may be better to actually move and invert the actual impulse as selected. A cancel or undo button may be enough to recover or exit without changes. This probably isn't practical though.

I have no clear suggestions for changes based on these thoughts. These are just first impressions in case they are helpful.

 

[John Mulcahy]

Selecting the best timing using unsmoothed traces is very difficult in many cases. There are lots of room interactions in this frequency range and it is difficult deciding the maximum SPL support. Smoothing the traces to 1/3 octave is very helpful in finding the best SPL support. We can smooth the SW and midrange traces prior to opening the parametric sum panel, but the parametric sum trace is still unsmoothed. It would be helpful to provide some method of the smoothing it also. Possibly automatically pick up the smoothing selected for the 2 input traces and use that for the parametric sum trace.

The lack of smoothing on the preview is a bug, it should pick up the smoothing of trace A (as the generated result does), I've fixed that for the next build.

The all SPL panel can be very busy if there are many measurements selected. It may be helpful to automatically/temporarily deselect all except the 2 measurements that are chosen when the panel is activated.

That wouldn't work when a third trace is being used as a reference or target, such as comparing the sum of the measurements to a summed measurement as on the co-located subs thread. Better left for the user to choose which traces to select.

It may be better to close the parametric sum panel with a cancel button, or maybe only for other actions taken within REW, so we can redirect our attention to other activities like; taking a screen snip of the panel, checking notifications, changing focus to our browser, etc.

In the next build I have made the parametric sum controls a normal dialog that can be re-positioned and will not disappear when clicking elsewhere.


Points 3 - 5 would produce side effects on source measurements when generating the result, which is outside the scope of what a trace operation should do. A possibility might be to create copies of the affected source measurements with the selected changes applied to them. Happy to hear opinions on that or how such a feature could work.

 

[jtalden]

Thanks John. This is a great feature to ease the process of finding favorable delay timing for a crossover.

A possibility might be to create copies of the affected source measurements with the selected changes applied to them. Happy to hear opinions on that or how such a feature could work.

My normal process is to first create a copy of the SW and move/invert that copy to find improved timing. That way I always have the original measurement for comparison or other analysis. In this case I like your thought that it would be helpful to copy both main and SW as we do not know which the user will move or invert. The copied measurements can always be deleted if not needed. The copies will prevent mistakes of; moving the wrong measurement, moving the wrong way, or inverting the wrong measurement when manually creating the new timing to investigate phase tracking, group delay, or other characteristics of the new alignment.