Jump to content
The Klipsch Audio Community
Chris A

Using REW to Determine Time Delays Between Drivers

Recommended Posts

I recently received a request to lay out how to set the time delays on multi--amped (bi-amp, tri-amp, quadruple-amp, etc.) driver channels in a single loudspeaker using a DSP crossover--such as a Xilica XP or XP series (or any other DSP crossover for that matter).  Below you'll find a step-by-step discussion of how I approach that task. 

 

Note that setting the delays is only one part of a group of actions and that to set the delays properly, you must also check the polarities of the drivers versus the type (Linkwitz-Riley, Butterworth, Bessel, etc.) and order (first...6 dB/octave, second...12 dB/octave, third...18 dB/octave, etc.) of the DSP crossover.  The higher the order (steeper the slopes), the more you'll be dealing with polarity issues of the drivers as you go about dialing them in.  So to start this process, I'll be using a little graphic called a QFD House of Quality (HOQ) to show you how each step of the process traces back to what the goals (WHATs) are:

 

  1. determine which crossover filter types and order that you will be using
  2. dial-in the initial delays as determined by REW measurements and plots,
  3. fix any polarity issues that crop up, and
  4. flatten SPL (amplitude) response of the setup to achieve smooth, flat response that's also time-aligned
     

491825628_SettingdelayusingREWandDSPcrossover(Xilica).thumb.JPG.9493743f68378109bb58420250706b9e.JPG

 

The "Hows" across the top of the QFD HOQ matrix show how to achieve each of the "Whats" that we want.  The "Hows" also just happen to show the initial sequence that I use them, namely:

 

1. Flatten individual driver/horn SPLs and set relative channel gains:

 

This step is relatively simple: run REW's EQ facility to flatten the frequency response of each driver in the loudspeaker (one driver/horn at a time).  After you have accomplished this, each driver/horn in the loudspeaker will be ready to integrate together and to set their relative delays.  After you've flattened the response of each driver (using mostly attenuating PEQs), then run one sweep with all channels on to see their relative channel gains.  You can set the approximate crossover frequencies and crossover filter types/slopes to see the relative SPL of each driver without having excessive overlaps between channels.  Having the driver channels at the same relative gain will facilitate reading the spectrograms to see the delays more clearly.

 

2. Set crossover filter type & slope:

 

Now that you have the individual drivers flattened in SPL and relatively close in terms of overall channel gains, now you can choose the preferred crossover slopes.  Look at the driver overlaps in the frequency bands where you intend to cross them over by taking individual sweeps of each driver channel and then looking at the combined frequency response plots within REW using the second button, called "All SPL" on the button bar just above the plot.  The width of the these natural bands of overlap will help to determine how steep or shallow a set of crossover filters that you should use.  In the case of drivers/horns that can just barely make it to the crossover frequency on each side of the crossover frequency, you might be forced to use a steeper filter to bridge the crossover frequency region--the "crossover interference band".  In the case of two drivers ( and horns as the case may be), they can have significant natural response overlaps, such as the following plot of a Cornwall bass bin with an ESS Air Motion Transformer (AMT-1) on top, both showing their two-octave overlap in frequency response:
 

2004631049_CornwallBassBin--ESSAMT-1crossoveroverlapplot(SPLvs.frequency).thumb.jpg.0832e5c361706198ca366d2114d3aa96.jpg

 

For the above case, it is possible to use a very low order crossover filter set, (high pass for the higher frequency driver, low pass on the lower frequency driver) because of the width of the crossover interference band--approximately two octaves from ~600 Hz to 2400 Hz.  It's usually wise to choose somewhere in the center of the interference band as the center crossover frequency. If you wish to use higher order filters for other reasons than just driver interference band overlap width, that's certainly at the discretion of the person dialing in the loudspeakers.  I personally recommend also trying out first-order filters (where possible) if using higher order filters to hear the difference, and set up two presets on your DSP crossover to be able to switch back and forth between them to hear the differences (if any). 

 

I've already matched the relative gains of each driver channel so that the output frequency response is relatively flat across the full frequency spectrum of the drivers that you're crossing (~40-20,000 Hz in this case). 

 

The next step is to set the crossover filter types and slopes within your DSP crossover (using the XConsole application in the case of Xilica crossovers).  Note that you don't have to worry about calculating relative delays at this point if you wish to just take a REW upsweep measurement and look at the results of the measurement to directly measure the needed delays.  Many people will select something like 24 dB/octave Linkwitz-Riley filters (i.e., fourth order).  The Xilica can also go up to 8th order (48 dB/octave) filters--but at the expense of degrading the resulting impulse response of the loudspeaker.

 

3. Take REW sweep with all drivers and chosen crossover filters:

 

At this point, you've got all drivers relatively matched in gain, their frequency responses flattened, and the chosen crossover filters set, so it's time to run a full sweep using REW to see the relative delays of each driver/horn within the single loudspeaker.  What you will see is uneven frequency response and phase curves within REW.  This is normal and indicates that time alignment of channels is now needed.

 

4. Read spectrogram relative channel delays:

 

From the full sweep measurement with all drivers and crossover filters, select "spectrogram" view from the plot menu bar just above the plot (the next-to-last button on the plot menu bar).  After setting the plot preferences  (frequency vertical, dark background, and scaling zoomed in to 0.5 ms horizontal divisions), you'll see something like this:

 

Cornwa II Impulse Spectrogram.jpg

 

The above spectrogram is from a 1979 Cornwall measurement, showing the time delay between the K-77 tweeter impulse energy (at the top of the plot), the K-55 midrange horn/driver at about 750 µs behind the tweeter, then the direct radiating woofer at about 250 µs behind the tweeter, and 500 µs in front of the midrange (to find this, look at 400 Hz at the peak energy plot line for this particular measurement).  So the channel delays must be set relative to the latest arriving channel--which is the midrange in the Cornwall (and is usually the bass bin in other loudspeakers).  So the tweeter channel will need to be delayed 750 µs and the woofer by 250 µs in this example. 

 

5. Set  Xover delays and PEQs, rerun REW sweep:

 

When those values of delay are inserted into the output channels:

 

475754197_XConsoleoutputchanneldelaycolumn.JPG.a06b280a1f6f09fa6a3ed8c6f81346ce.JPG

 

After those delays are set and another REW measurement upsweep is made, something like the following will be seen in the spectrogram plot:

 

Cornwall bass bin + AMT-1 6dB per oct spectrogram on-axis.jpg

 

Note the gradually increasing lag of the peak energy time as the frequency decreases.  This is the characteristic that is desired in order to get the best impulse response.  The smoother the peak energy time curve, the better the impulse response. 

 

6. Look at driver-to-driver crossover band FR & phase:

 

After the delays have been set, take another upsweep using REW, then look at the SPL & Phase plot (the first selection button on the menu bar just above the plot). 

 

833409471_JubCrossoverPolarityNullandPhaseResponse.thumb.jpg.5e46b764cd2c5823c93d84c69c799149.jpg

 

7. Check step/phase/group delay plots & update delays/filters:

 

Note the big nulls at ~125 Hz and ~215 Hz.  These are caused from drivers not being time aligned.  This may indicate that you need to flop the polarity of the lower-frequency driver channel using XConsole.   This will depend on the crossover type and steepness that is chosen.  If you change the crossover type of slope, this will need to be revisited and adjusted again. 

 

The phase curve also goes through a tight little zig-zag at these nulls.  This also shows up on the group delay plots (the first derivative of phase with respect to frequency):

 

1810474443_JubCrossoverPolarity-InducedGroupDelayNulls.thumb.jpg.02f2012b3622372427cdab0eb879295d.jpg

 

These are indicators that something is still not dialed-in properly.  Go back to step 5 and re-set the time delay on the leading drivers until the frequency response, phase, and group delay plots are smooth.

 

The step response (the plot is found under the Impulse Response plot window) should look something like this:

 

306159432_JubileeStepResponse(GoodResponse).thumb.jpg.f7ecc760bb32a7e1877193ccadaa1cf9.jpg

 

and not like this:

 

595895680_JubileeStepResponse(Not-So-GoodResponse).thumb.jpg.efd282e8adb8bc4a76fba9fdc248f448.jpg

 

_________________________________________________________________________________________________________________________

 

Once you've completed the adjustments in frequency response and delays for the first loudspeaker, now it's time to move to the other loudspeaker and complete steps 1-7.  XConsole (the Xilica PC/MAC control application) provides a "Copy" function under the "Setup" menu to copy settings from one channel to another (user selectable).  This saves a lot of time, and allows you to clone your settings to the other loudspeaker channels without keying mistakes.

 

8. Listen:

 

This is self explanatory, but when everything is dialed in for all channels, you should have both conscious and subconscious improvements in the way the loudspeakers now sound.  If you don't get a subconscious improvement in sound, try using first order crossover filters instead of the higher order filters that you might have been using and compare the listening results.  When dialed-in properly and playing most reasonably hi-fi recordings (fully acoustic instrumentation is the best choice to hear this, avoiding any electronic or electrical amplification which scrambles the phase information in the recordings), this subjective sound improvement will be most apparent.

 

Chris

  • Like 3

Share this post


Link to post
Share on other sites

So the above steps represent the basic mechanics of time aligning drivers within a loudspeaker using REW and a DSP crossover.  The devil (of course) is in the details.  This is where audience participation is required to highlight areas that may be obscure. 

 

All of the above was developed, whole cloth, from my experiences without an external guide to refer to, so some of the steps in the sequence need to be amplified a bit.  I intend to edit the above as time passes in order to make the process clearer and more intelligible.  What I need from those that are actually using these steps is specific questions for amplifying obscure areas (i.e., not overall assessments) to make it more usable.

 

Happy trails.

 

Chris

  • Like 1
  • Thanks 1

Share this post


Link to post
Share on other sites

Chris,

This is a very good explanation as to how to get it right, thanks for posting.

Share this post


Link to post
Share on other sites

Chris,

 

Thanks for this............I had asked Chris to do this.  I have been using True RTA for years.  I got a new PC and downloaded REW and anyone who has used both will tell you REW is far more comprehensive.  I have not dug in yet.

 

Anyhow, I got some new subs for my 5 way MCM setup and it radically changed the time alignment.  I thought it would be a good time to learn how to use REW to reset the system.

 

Question #1:  Chris, where did you place the mic for these measurements?

 

Thanks again.  I know there will be much more to come on this.

Share this post


Link to post
Share on other sites

Cornwa II Impulse Spectrogram.jpg

 

This roughly 550 milliseconds is approximately .5 of a second misalignment? That's pretty significant. 

Share this post


Link to post
Share on other sites

Both physical distance and crossover phase shifts to contend with, but it was micro, not milli.

Share this post


Link to post
Share on other sites
3 hours ago, Schu said:

This roughly 550 milliseconds is approximately .5 of a second misalignment? That's pretty significant. 

Microseconds...not milliseconds.  Think about it: the distance associated with 500 milliseconds of sound travel is 566.7 feet. 

 

The symbol for microseconds is "µs", for milliseconds is "ms'.

 

Share this post


Link to post
Share on other sites
1 hour ago, mark1101 said:

Question #1:  Chris, where did you place the mic for these measurements?

I set the microphone at listening height--just below the top of  the bass bin, one metre away from the front baffle of the Jub bass bin--in order to minimize in-room reflections which show up in the group delay plot. For loudspeakers that are over 6 feet tall, you will have to move back to the listening position and use as much absorption and time gating as you can to suppress first reflections obliterating the phase/group delay data.

 

Chris

Share this post


Link to post
Share on other sites
31 minutes ago, Chris A said:

Microseconds...not milliseconds.  Think about it: the distance associated with 500 milliseconds of sound travel is 566.7 feet. 

 

The symbol for microseconds is "µs", for milliseconds is "ms'.

So is .75 milliseconds even audible?

Share this post


Link to post
Share on other sites

A question was asked about how to read the relative delays from the spectrogram.  The annotated figure below shows the example above and where I got those delays:

 

1655244322_Cornwallspectrogramannotateddelays.thumb.jpg.703e33c220099a120193556caa04211d.jpg

 

Chris

Share this post


Link to post
Share on other sites
52 minutes ago, Schu said:

So is .75 milliseconds even audible? 

So since this subject obviously has come up again...how many wavelengths at the nominal 5000 Hz crossover frequency is represented by 750 µs of time alignment mismatch? 

 

(Hint: greater than two).

 

(Another hint: you need to be within 90 degrees of alignment to make two acoustic sources couple in order to behave like one source.  How many microseconds is represented by 90 degrees of phase at 5000 Hz?)

 

Chris

Share this post


Link to post
Share on other sites
1 hour ago, Chris A said:

A question was asked about how to read the relative delays from the spectrogram.  The annotated figure below shows the example above and where I got those delays:

 

1655244322_Cornwallspectrogramannotateddelays.thumb.jpg.703e33c220099a120193556caa04211d.jpg

 

Chris

 

So what is significant about the exact spots you chose here?  Are these the acoustical crossover points that were verified prior?  The X scale says Ms.  Where to get Us out of this chart?

 

Nevermind about the uSec.  I got it now.  I just need to know how those particular spots with the arrows are the magic points.

Share this post


Link to post
Share on other sites

The Cornwall's crossover frequencies are ~600 and 6000 Hz.  You need to look above and below those frequencies, just far enough away from the crossover frequencies to see what their relative time delay happens to be.  That's what is shown on the chart above.  If you are too close to the crossover frequencies, you see a blending of the two drivers in their interference bands.  If you're too far away from the crossover frequencies, the natural phase growth of the drivers themselves will lead to erroneous time delay readings.  So you need to read the time delays as close to the actual crossover points as you can without actually being in the interference band itself.

 

Chris

Share this post


Link to post
Share on other sites

One other way to find the channel time delays is to take the measurements of the natural response of the drivers/horns themselves and look at their phase (i.e., time delay) growth near the preferred crossover frequencies to make sure that their natural phase curves are not rotating rapidly at the planned crossover frequency.  Then you can combine the two compared drivers by reading their absolute time offsets at their crossover frequencies, and subtract the difference in arrival time at the microphone. 

 

The problem with that approach is that you'll need to calibrate your sound card/preamp delays (loopback) beforehand to eliminate the inherent the channel time delays in the electronics/digital chain.  That's a much more difficult way to get results that can be trusted. You'll usually wind up doing it the way that I discussed above because of errors in reading the absolute delays and phase offsets are not very accurate using the absolute delay reading method.

 

Chris

Share this post


Link to post
Share on other sites
8 hours ago, Chris A said:

 

 

... 😛

Share this post


Link to post
Share on other sites
15 hours ago, Schu said:

So is .75 milliseconds even audible?

Do any Klipsch factory speakers correct for time delay between drivers at all? 

Share this post


Link to post
Share on other sites
33 minutes ago, JL Sargent said:

Do any Klipsch factory speakers correct for time delay between drivers at all? 

 

Yes...the Klipsch Professional (cinema) models generally are.  This includes the Jubilee and MWM-based loudspeakers.

 

Generally speaking, the home theater series appear to be so, as do The One, The Three, and The Sixes.

 

Chris

  • Like 1

Share this post


Link to post
Share on other sites

OK, if I could just create that spectrograph now.......I'd be on my way.  :)  Not having such good luck with REW.

 

I think for my system which is 4 way and subs (so 5 way) I may be better off doing 2 or 3 drivers at a time rather than the whole thing.

Share this post


Link to post
Share on other sites

Perfectly times post Chris.  Thank you.  I will need to intigrate my newly installed bass bins once I finish the build.  This should come in very handy.

Share this post


Link to post
Share on other sites

So what do I have here? Who has been studying? Took me about 1/2 a hour just to make it. Wondering how I read it?

 

WhatDoIHaveHere.jpg.9dca89e1875dffcdefee11195e469a08.jpg

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...