I don't understand one thing that has long been on my mind.

[bobolix]

When I open the "Offset t = 0" window in the measurement, the non-zero value "Cumulative shift" is displayed, even if the time shift has not been manipulated in any way. These are usually values in the order of some microsecond, which are different for each measurement and have nothing to do with the pulse time delay. What does this mean and what to do if, for example, I want to derive the exact phase response of a speaker driver by subtracting the calculated delay given by the microphone distance from the speaker cabinet? Should I enter the calculated delay as an offset or as a cumulative shift?

 

[John Mulcahy]

That comes from adjustments that were made to align with the timing reference or with the peak if a timing reference was not used. Enter any other delays you want as an offset.

 

[bobolix]

I still don't understand. I enclose as an example the measurement of a tweeter at a distance of 500 mm. The time delay that needs to be subtracted to obtain the correct phase response is 1.4534 ms at an air temperature of 20.5 ° C. The initial cumulative shift that is shown is 6.3 microseconds without any manual adjustment. The measurement was performed with a loopback time reference. What is this initial cumulative shift of the measurement itself? I assume that the time delay to be subtracted should be entered as Offset, right?

 

[John Mulcahy]

The measurement is shifted only to align with the loopback peak. The delay relative to the loopback is shown in the measurement panel and as System delay in the Info panel. You can remove that delay by entering that figure, or some other figure if you think that is wrong. You could also enter it as a timing offset for further measurements so it gets applied when you make the measurement. You do not need to make use of the cumulative delay figure, the system delay shows timing relative to the reference.

 

[bobolix]

Somehow we don't understand each other. I need to derive the absolute phase response of the driver to include the transport delay of the driver, which is caused, for example, by the diaphragm being shifted to a depth relative to the flange or front wall of the speaker cabinet. From this point of view, the overall impulse response does not start with the peak of the response, but somewhat earlier. The system delay shown by REW is unusable in this respect. Moreover, in the specific case of the tweeter, the measurement of which I have enclosed, the situation is complicated by the fact that the impulse response has two peaks, first positive and then slightly higher negative. REW determines the system delay according to the delayed negative peak because it is higher, so the resulting "system delay" is completely out. The delay to be used for time correction needs therefore to be determined by calculation according to the distance of the microphone from the front wall of the baffle, of course for the speed of sound at the current temperature. I would need to introduce this calculated delay into the measurement using the panel called as Offset t = 0, but I don't know if to enter it as Cumulative shift or as Offset, because when opening the panel the non-zero Cumulative shift is always immediately displayed, which is different for each measurement, while the displayed Offset is zero (what exactly is the Cumulative shift in this case, if the loopback is still the same?). Maybe I'm completely stupid, but this just doesn't sit me.

 

[John Mulcahy]

As I said above, the cumulative shift figure comes from fractional sample adjustments that were made to align with the timing reference, which would be the peak of the loopback IR for your measurements. It has no relevance for what you want to do so please ignore it. The system delay for loopback measurements shows the relative times of the loopback and measurement peaks. That may not be the best choice, I could change it to operate in the same way as the acoustic timing reference, which uses the estimated delay of the measured IR if its peak is close to the IR start, or the estimated IR start if the peak is later (typically the case for measurements of low bandwidth systems or drivers). You can do this manually by using the Estimate IR delay button, and then choosing the option to either shift the IR or to both shift the IR and use the delay estimate as the timing offset for subsequent measurements. For the measurement you posted above the estimated delay is 1.4954 ms, 513 mm. Here is how the IR looks when shifted by that amount, shown overlaid on the minimum phase IR since REW uses cross-correlation with the min phase IR to estimate the delay.

Again, if you think that figure is wrong, enter whatever figure you want to use instead, as an offset for t=0 and/or in the timing offset field on the Measure dialog for subsequent measurements.

 

[Breeman]

As I said above, the cumulative shift figure comes from fractional sample adjustments that were made to align with the timing reference, which would be the peak of the loopback IR for your measurements. It has no relevance for what you want to do so please ignore it. The system delay for loopback measurements shows the relative times of the loopback and measurement peaks. That may not be the best choice, I could change it to operate in the same way as the acoustic timing reference, which uses the estimated delay of the measured IR if its peak is close to the IR start, or the estimated IR start if the peak is later (typically the case for measurements of low bandwidth systems or drivers). You can do this manually by using the Estimate IR delay button, and then choosing the option to either shift the IR or to both shift the IR and use the delay estimate as the timing offset for subsequent measurements. For the measurement you posted above the estimated delay is 1.4954 ms, 513 mm. Here is how the IR looks when shifted by that amount, shown overlaid on the minimum phase IR since REW uses cross-correlation with the min phase IR to estimate the delay.

View attachment 44410

Again, if you think that figure is wrong, enter whatever figure you want to use instead, as an offset for t=0 and/or in the timing offset field on the Measure dialog for subsequent measurements.

Hi John,

Apologies for hijacking the thread but this caught my interest. Instead of changing the functionality of the cumulative delay, won't it be a better option to add the other feature you mention and let the user choose the one they would like to use depending on the use case?

 

[John Mulcahy]

I am not making any change to the cumulative delay, but I have added a new Analysis preference for the next build to choose whether the system delay when using a loopback timing reference is based on the peak times or the estimated delay.

 

[Breeman]

Thanks for considering this John