new here - time alignment with REW and UMIK

[st0nehead_ss]

hey everyone
ive sent a few days reading most everything on here with regard to time alignment using the above.
I have a few questions, maybe im getting confused.

I am trying to test a 4 way reggae sound system (4 scoop subs, 4 es18bph kicks, 2x vertically stacked 2x12_hf paraflex line array style cabinets) they are set up in a mono stack.

I understand that the loopback isnt possible when using a USB mic, as the sources are different. its a shame because loopback is making more sense to me in determining IR delay than does the acoustic timing reference (mainly as i dont know how to use that in my case, except for maybe placing a monitor speaker off of the mixer's "booth" channel and placing it very close to the mic...)
where i am getting confused is that the "chirp" that would play through the Midtops for example, or the full range monitor speaker if i were to use it, also would be playing the full range sweep to be measured..thus interfering with the graph, as i wouldnt be able to see what is the timing speaker and what is the (for example) sub?
or is it likely to have a discernable IR gap between the two on the time domain/IR window when i look at it?

we spent a day fidding and didnt come up with much but i assume this is because of the setup errors...

below is a pic of the stack in case it helps...it is powered and controlled DSP wise via powersoft x8 with armonia+

 

[skid00]

I'm not sure how much help I can be... Can you send the sweep thru just the subs, just the kicks, just the HF? You would need to add a full-range (placed equi-distant with the subs or kicks (what freq range are those?)) for the chirp to be useful. Are you trying to time-align everything?

But you might be able to use pink noise, if the stack doesn't have much in the way of reflective surfaces (for EQ).

 

[st0nehead_ss]

I'm not sure how much help I can be... Can you send the sweep thru just the subs, just the kicks, just the HF? You would need to add a full-range (placed equi-distant with the subs or kicks (what freq range are those?)) for the chirp to be useful. Are you trying to time-align everything?

But you might be able to use pink noise, if the stack doesn't have much in the way of reflective surfaces (for EQ).

Hey, thanks for your reply.

The data we obtained I am fairly certain did not work out as expected. We erroneously measured with crossovers in place. The speaker for the acoustic time signal was set beside the mic to try and make it a 0 time red for example. We had found a fellow via speaker plans who made a document on this.
As the rig is fairly large and heavy I can't drag it out and take them again without help from 1-2 friends.
Pink noise did make sense to me if we were to leave crossovers on, in order to get a time domain IR alignment match but we didn't get to that...

 

[John Mulcahy]

The timing reference is played before the measurement sweep, it doesn't produce any change in the measurement itself. REW uses it to synchronise capture, it doesn't appear at all in the result. If you use a separate speaker for the timing reference that is placed much closer you will need to use a timing offset to remove the difference in distance between the speaker reproducing the timing signal and the speakers being measured. REW can estimate that for you using the "Estimate IR delay" feature on a measurement of the system to be measured, that timing offset should then remain the same for all the subsequent measurements.

 

[st0nehead_ss]

The timing reference is played before the measurement sweep, it doesn't produce any change in the measurement itself. REW uses it to synchronise capture, it doesn't appear at all in the result. If you use a separate speaker for the timing reference that is placed much closer you will need to use a timing offset to remove the difference in distance between the speaker reproducing the timing signal and the speakers being measured. REW can estimate that for you using the "Estimate IR delay" feature on a measurement of the system to be measured, that timing offset should then remain the same for all the subsequent measurements.

Thankyou very much for your response, John.
So the chirp plays, but is not recorded back via the mic?
It is sent to, for example, the full range speaker being used for reference, as well as the sub...but mainly probably only audible on full range reference and mid range/tweeters being tested?

Essentially I know I am going to have to delay kicks and midtops to the sub delay to time align, I was just getting lost.
I understand now this needs to be done without any crossover filters in place.
Once they are aligned with ms delay the filter choices can be determined around their effects in phase trace?


Sidenote, as you can see each "way" is playing a different frequency if the music. Whilst the stack is aligned in space, each cabinet has a different horn path and of course the way sound travels the lower freqs will be lagging so I need to align to the subs which play 30-80hz..

 

[John Mulcahy]

The timing reference signal is captured and processed along with the subsequent measurement sweep, but only the measurement part of the signal is used to generate the response. The timing signal and any timing offset specified on the Measure dialog are used to work out where t=0 should be placed. If you use a separate speaker for the timing reference (probably easiest for your setup) it would typically be connected to a different channel than the system being measured, so the timing reference would only ever go to the reference speaker. The "Ref output" line on the Measure dialog is used to select where the timing reference gets sent. You can then measure the individual parts of the system, which is best done with the crossovers in place unless you plan to use the measurements to decide on the crossovers. Once you have the measurements you can also use the Alignment tool in the All SPL graph controls to help work out the delays, for the sub that's usually the easiest method.

 

[st0nehead_ss]

Thankyou, that's helpful.
So I could send left output to the timing ref speaker (active), right out to the amp and main speaker stack (using it's DSP to mute channels accordingly for testing) and use the umik as the only input.
In this case the left output takes the place of what would otherwise be the loopback? But would the active full way speaker on the left channel still receive the entirety of the sweep?
Or is this what the selection you mentioned does, selects it only for the "chirp"?

Is this correct?

 

[John Mulcahy]

Yes, left output would only be sent the timing signal, nothing else.

 

[st0nehead_ss]

And somewhere inside the software I can determine that yeah?

 

[John Mulcahy]

The "Ref output" line on the Measure dialog is used to select where the timing reference gets sent.

 

[sm52]

st0nehead_ss
to do it right, you need two output channels of the sound card. One to which the reference signal will be applied, whistling (ref output), and the second, to which the sweep tone will be applied to capture the response.

 

[st0nehead_ss]

yes, thankyou.

i was not certain where to place the full range speaker used for time alignment, i had seen it placed next to the mic (assuming to simulate loopback, which would be instant, vs distance from mic if aligned with the speaker stack)

 

[st0nehead_ss]

The timing reference is played before the measurement sweep, it doesn't produce any change in the measurement itself. REW uses it to synchronise capture, it doesn't appear at all in the result. If you use a separate speaker for the timing reference that is placed much closer you will need to use a timing offset to remove the difference in distance between the speaker reproducing the timing signal and the speakers being measured. REW can estimate that for you using the "Estimate IR delay" feature on a measurement of the system to be measured, that timing offset should then remain the same for all the subsequent measurements.

john, im just going back over this, and through some other peoples threeads...on the impulse graph, where the line with REF is set, is this little spike indicating that this is the position of the timing reference chirp? i didnt thing from what you wrote here that it should show up in the graph?

 

[John Mulcahy]

It shouldn't, but crosstalk in the interface can produce an artefact ahead of the measured response.

 

[st0nehead_ss]

It shouldn't, but crosstalk in the interface can produce an artefact ahead of the measured response.

thanks, i just re-looked at the thread i was referring to and the fellow taking the measurements was taking a measurement of two boxes at once, and has set t=o at cursor of the later arriving box, which had a spike.
i assume usually the "REF" line is where the acoustic timing reference is signalling t=o?

 

[John Mulcahy]

The Ref is at the peak. The latest builds have an indicator at the bottom of the graph showing the location of the timing reference, that is usually where zero is set for the time axis unless a timing offset is used.

 

[st0nehead_ss]

The timing reference signal is captured and processed along with the subsequent measurement sweep, but only the measurement part of the signal is used to generate the response. The timing signal and any timing offset specified on the Measure dialog are used to work out where t=0 should be placed. If you use a separate speaker for the timing reference (probably easiest for your setup) it would typically be connected to a different channel than the system being measured, so the timing reference would only ever go to the reference speaker. The "Ref output" line on the Measure dialog is used to select where the timing reference gets sent. You can then measure the individual parts of the system, which is best done with the crossovers in place unless you plan to use the measurements to decide on the crossovers. Once you have the measurements you can also use the Alignment tool in the All SPL graph controls to help work out the delays, for the sub that's usually the easiest method.

john, if i am running the sweeps with the current crossovers in place, and use the feature of the alignment tool labelled "align phase at cursor" or whatever it is (at crossover frequency basically) will this give the necessary delays to align the phase response not only at the cursor position/crossover frequency, but an octave either side?

 

[John Mulcahy]

REW uses the phase data within one octave either side of the chosen frequency to find a delay figure that gives the best overall alignment. The range over which that remains valid depends on the phase responses themselves, but you can see from the aligned sum how well the delay works.

 

[st0nehead_ss]

Great, thanks

I assume this can be "simulated" ie show you the sum without needing to apply the suggested delay and run another sweep?
I'm sorry for so many queries I just need to have things in order before I drag out the (rather large) stack again

 

[John Mulcahy]

Yes, that is what the Alignment tool is for.

 

[st0nehead_ss]

Yes, that is what the Alignment tool is for.

Thankyou very much John for all your help

 

[st0nehead_ss]

One more thing is I was planning to run tests per this video for time alignment of IR

But this seems not to consider the phase between mids and tops, and rather just align IR of all cabs to the sub...although is seemingly straightforward for my application

So would it be sensible to run the following sweeps

Sub solo
Kicks solo
Midtops solo

Then use the alignment tool kik to sub, midtops to kicks (plus the delay of kik to sub) to ensure phase alignment at crossovers....but does this also align the IR so that timing of impulse arrivals is done?

Then IR delay per the video (sweep sub and kik together, then delayed kick and midtop etc)
And perhaps then compare this to having used the alignnt tools suggestions?

What to do if these are not like one another, or there is a vast discrepancy?

the fellow in the presentation screenshot below says to run each passband with no filters (i assume this means one should use the filter sweep options to sit the start and stop just outside each passband to avoid any damage)
he seems to be saying, as far as i can tell, that once things are IR aligned here, this wont change with the application of crossover filters, and thus you can choose which ones give best phase response based on their graphs

 

[John Mulcahy]

Aligning impulse responses of drivers with different bandwidths/frequency ranges is difficult to do correctly since the shapes and widths of their impulse responses will be different. One way to do it is to use a bandpass filter at the crossover frequency for each pair, REW (5.20.14 builds) offers that if the Impulse mode is selected in the alignment tool. The results should generally be the same as when using phase alignment, but filtered impulse alignment can make it easier to see when a driver should have its polarity inverted to get good alignment.

 

[st0nehead_ss]

Aligning impulse responses of drivers with different bandwidths/frequency ranges is difficult to do correctly since the shapes and widths of their impulse responses will be different. One way to do it is to use a bandpass filter at the crossover frequency for each pair, REW (5.20.14 builds) offers that if the Impulse mode is selected in the alignment tool. The results should generally be the same as when using phase alignment, but filtered impulse alignment can make it easier to see when a driver should have its polarity inverted to get good alignment.

Right...

So given all of the above, and the specifications of my system from OP, what do you think the best way to measure for IR alignment whilst maintaining phase matching at crossover +/- 1 octave would be?

As you can see there is confusion around the best method and the most direct way to align for best sum

 

[st0nehead_ss]

would it also be wise to (rather than using peak arrival time) use initial impulse arrival time (ie, beginning of the upward wave?)
also, on another forum, someone mentioned that REW starts sweeps at _half_ the specified measurement frequency....is this so?