V5.20.14 early access build

[John Mulcahy]

Here it is with the adaptive timescale, I've done that for the next build.

 

[sm52]

Where are you placing the cursor to align them?

the cursor is at 2.87kHz in your post. In this case, the impulse graph is very strongly compressed. Can it be stretched somehow? I put the cursor in different places, but it doesn't change anything.

 

[sm52]

I've done that for the next build.

while I was writing the answer, you published a new post.

for the usual subwoofer range frequencies

now I understand. At subwoofer frequencies, the impulse graph was automatically stretched. And on HF it was compressed.

 

[thothsong]

I put the cursor in different places, but it doesn't change anything.

FWIW, here's your mdat with the current build, cursor arbitrarily at 82 Hz, and then clicking on Filter IRs at cursor:

 

[John Mulcahy]

Even with adapting the timescale it may be a bit hit and miss trying to align at high frequencies in what looks like a very reflective measurement environment.

 

[sm52]

This possibility of alignment by watching the movement of the impuls graph is a very good thing. Dare I ask John, is it possible to add the ability to see how the Step Response graphs behave when aligning? Maybe a button to switch from impulse to Step Response.

 

[John Mulcahy]

This possibility of alignment by watching the movement of the impuls graph is a very good thing. Dare I ask John, is it possible to add the ability to see how the Step Response graphs behave when aligning? Maybe a button to switch from impulse to Step Response.

No, step responses are too slow to generate. I don't think they are at all helpful for alignment, try looking at the step response for the aligned sum at a few different delays.

 

[sm52]

try looking at the step response for the aligned sum at a few different delays

Yes.

 

[serko70]

Here it is with the adaptive timescale, I've done that for the next build.

View attachment 60339

Is "Align IRs at cursor" button aligning for peaks, starts or with cross correlation?

 

[John Mulcahy]

From the help: The Align IRs at cursor button applies the filters at the cursor frequency and cross-correlates the responses to determine their best alignment.

 

[serko70]

From the help: The Align IRs at cursor button applies the filters at the cursor frequency and cross-correlates the responses to determine their best alignment.

Sorry, missed that.

Do you think cca method is optimal for "time aligning" a sub with a speaker crossed over with that sub?

 

[John Mulcahy]

"cca method" meaning what? The plot looks very misaligned, but hard to tell the behaviour at the crossover frequency from unfiltered responses.

 

[serko70]

"cca method" meaning what? The plot looks very misaligned, but hard to tell the behaviour at the crossover frequency from unfiltered responses.

Above plot is the result when left speaker and subwoofer are cross correlation aligned for the left sp. Both have fdw:9 applied. There's no crossover active during measurment. It looks not aligned but the sum of left sp and sub produces better step response and GD with cca than aligning for IR starts or IR peaks. I am a little confused.

Edit: this is for a home cinema setup. I am not interested in phase aligning the sub with the left speaker.

 

[John Mulcahy]

That is not what cross-correlation alignment is for. Correlation determines where responses are most similar, it is pretty much meaningless to look for similarities between responses with such different bandwidths, they have nothing in common. The impulse alignment mode first filters both responses so they have the same bandwidth, it then becomes meaningful to correlate them. Step response is no more useful for alignment than unfiltered impulse responses.

 

[John Mulcahy]

Builds updated today with these changes:

• Added: Distortion graphs support Fit Y to data
• Added: Graphical filter edit shows filter settings
• Changed: Adapt the time span shown in the Alignment tool in impulse mode to suit the alignment frequency
• Fixed: StormAudio equaliser used the wrong number of filters
• Fixed: ASIO input channel names could be shown incorrectly on the Measure dialog
• Fixed: 64k sweeps at 44.1 kHz starting at 20 Hz and ending above 11.025 kHz gave incorrect results

 

[sm52]

It looks not aligned but the sum of left sp and sub produces better step response and GD with cca than aligning for IR starts or IR peaks.

I don't know what method it is, but the result you are talking about should be correct. In my experience, don't align to peaks or impulse start, it won't give the best result.

 

[thothsong]

Added: Distortion graphs support Fit Y to data

It's a minor point, but in the main Distortion graph it seems like it takes all of the harmonics into account, even if they aren't selected for display?

 

[John Mulcahy]

The number of harmonics to use for THD is in the controls, it doesn't depend on which harmonics are shown.

 

[sm52]

Changed: Adapt the time span shown in the Alignment tool in impulse mode to suit the alignment frequency

It looks like the behavior of this instrument on LF and HF has been reversed. Now on the HF impulse is clearly visible, but it is compressed at the LF. The user still cannot stretch the impulse graphs along the horizontal axis. There are no plus and minus buttons that are on all charts.

 

[John Mulcahy]

That graph is not for viewing impulse responses, there is an overlays graph for that. It is to view the filtered responses which will be aligned, its width is a fixed number of cycles of the alignment frequency. If you drag the cursor along the frequency response graph you will see the lower graph width change correspondingly.

 

[sm52]

It is to view the filtered responses which will be aligned, its width is a fixed number of cycles of the alignment frequency. If you drag the cursor along the frequency response graph you will see the lower graph width change correspondingly.

Everything is as you explained. It's just that the expectations from the appearance of this tool were another. It was important for me to see how the impulse graphs are located during the alignment. Detailed. But it will be useful for someone as it is now implemented. Thank you.

 

[John Mulcahy]

If you don't wish to apply the filter you can just use the cursor frequency as a zoom control.

 

[thothsong]

The number of harmonics to use for THD is in the controls, it doesn't depend on which harmonics are shown.

I'm talking about Fit Y to data, not THD. I start with 9th as highest harmonic shown:

Note that some of the harmonics extend below the bottom of the graph view.
Now I change the control so that the 2nd is the highest harmonic shown:

Now I hit Fit Y to data:

The Y axis got zoomed out, rather than zoomed in.
Now I change the control so that the 9th is the highest harmonic shown:

All of the harmonics now fit in view, so the zoom out was apparently done to make all of the harmonics fit, even though they weren't displayed at the time Fit Y to data was performed.

 

[John Mulcahy]

Understood, I have fixed that for the next build. It was also looking at data across the whole span instead of the displayed range, that has been fixed as well.

 

[sm52]

you can just use the cursor frequency as a zoom control.

Yes. Thus it is possible. The impulse graph before the peak and after it, it seems to me, few people will be interested in this tool. You can increase the scaling to 2x or 4x the default to have 1-2ms to the left and 4-6ms to the right of the peak.