Cross Correlation and Vector Average Measurements

mmaresg6 · Oct 29, 2024

REW newbie here reaching out to advanced REW users. I reached out to OCA for Cross Correlation Time Alignment and Vector Average advice and recommendations to time align and EQ my 4 subs for 3 seats. I did three measurements for each of my seats, followed the steps to Cross Correlate and Vector average the measurements. EQed my subs to my preferred house curve, ran the predict function but when I measured to verify my predict house curve the verified measurement looked totally different that the predict. For my REW measurement, I used 0 to 24000Hz range, used "Used acoustic timing reference for my Timing. Does anyone have any idea as to what why the actual and the predict are different and what I might be doing wrong? The verified measurement has a null at 62.5 Hz and levels off up to the 100Hz. I'm very confused.

sm52 · Oct 30, 2024

The main reason for the discrepancy is the room with its unpredictable reflections. There is no program yet that could predict how reflections will change. The way is so. Made changes and checked them. Made other changes and checked them. At some point you will understand which option you can choose.

mmaresg6 · Oct 30, 2024

sm52 said:
The main reason for the discrepancy is the room with its unpredictable reflections. There is no program yet that could predict how reflections will change. The way is so. Made changes and checked them. Made other changes and checked them. At some point you will understand which option you can choose.

Okay, I understand better now why the discrepancy. I'll do some more testing. Thanks for the feedback!

moedra · Oct 30, 2024

If you make correction filters from vector average responses instead of dB or RMS responses, there will always be differences because the vector average is not a true representation of magnitude like a dB or RMS average is.

JStewart · Oct 30, 2024

mmaresg6 said:
REW newbie here reaching out to advanced REW users. I reached out to OCA for Cross Correlation Time Alignment and Vector Average advice and recommendations to time align and EQ my 4 subs for 3 seats. I did three measurements for each of my seats, followed the steps to Cross Correlate and Vector average the measurements. EQed my subs to my preferred house curve, ran the predict function but when I measured to verify my predict house curve the verified measurement looked totally different that the predict.

Was this what OCA recommended? I suspect not for multiple seats. Could be wrong, though. Can you link his video or whatever you got the method from?

It is possible to measure, use a target curve and calculate filters that will yield a matching measurement at each of multiple locations. To do that you need to either use the freeware MSO multi-sub optimizer at

Multiple Subwoofers: Optimize Them With Multi-Sub Optimizer Software

The benefits of multiple subwoofers are maximized using Multi-Sub Optimizer software for lowest seat-to-seat frequency response variation and highest SPL.

www.andyc.diy-audio-engineering.org

or use hardware that will run Dirac Live Bass Control which you will pay for.

Dirac Live Bass Control

Dirac Live Bass Control. Automate subwoofer optimization for a strong, smooth, and refined bass response in all listening positions.

www.dirac.com

JStewart · Oct 30, 2024

Here’s a post I did about MSO back in 2017 before I could spell optimize

Thread 'MSO - Multi-Sub Optomizer'
https://www.avnirvana.com/threads/mso-multi-sub-optomizer.1891/

mmaresg6 · Oct 30, 2024

What's got me confused is the discrepancy between the predict and actual only happens when I use the multiple measurement cross correlation and vector average. When I use the single mic measurement method there is no discrepancy. See attached.

mmaresg6 · Oct 30, 2024

JStewart said:
Was this what OCA recommended? I suspect not for multiple seats. Could be wrong, though. Can you link his video or whatever you got the method from?

It is possible to measure, use a target curve and calculate filters that will yield a matching measurement at each of multiple locations. To do that you need to either use the freeware MSO multi-sub optimizer at

Multiple Subwoofers: Optimize Them With Multi-Sub Optimizer Software

The benefits of multiple subwoofers are maximized using Multi-Sub Optimizer software for lowest seat-to-seat frequency response variation and highest SPL.

www.andyc.diy-audio-engineering.org

or use hardware that will run Dirac Live Bass Control which you will pay for.

Dirac Live Bass Control

Dirac Live Bass Control. Automate subwoofer optimization for a strong, smooth, and refined bass response in all listening positions.

www.dirac.com

This is the link to the video where I got the recommendation from OCA. BTW, towards the end of his response he mentions and make recommendations of MSO for 4 sub arrangements.

mmaresg6 · Oct 30, 2024

My next move is to dig deep into MSO, learn how to use it and try it and see what results I get for my open concept living room and 4 sub setup. Thank you for your feedback and suggestions. Appreciate it.

JStewart · Oct 31, 2024

mmaresg6 said:
What's got me confused is the discrepancy between the predict and actual only happens when I use the multiple measurement cross correlation and vector average. When I use the single mic measurement method there is no discrepancy. See attached.

This is expected behavior.
If you move the mic just 6” the response will likely be different.
An average of multiple locations will often make the MLP worse and will never make it as good as it could be.
In the 1st 15 min of the linked video, only measurements at and very close to the MLP are being averaged. This makes sense as all the measurements are where one’s ears might be so you are reducing the effect of response anomalies within this area (although some well credentialed/respected individuals will say not helpful or necessary). A single response measurement will still not equal the average response measurement, however, and EQ doesn’t change that.

MSO and DLBC are entirely different methods that use computing power to zero in on a set of filters that yield closest to desired result at each location.

mmaresg6 · Oct 31, 2024

JStewart said:
This is expected behavior.
If you move the mic just 6” the response will likely be different.
An average of multiple locations will often make the MLP worse and will never make it as good as it could be.
In the 1st 15 min of the linked video, only measurements at and very close to the MLP are being averaged. This makes sense as all the measurements are where one’s ears might be so you are reducing the effect of response anomalies within this area (although some well credentialed/respected individuals will say not helpful or necessary). A single response measurement will still not equal the average response measurement, however, and EQ doesn’t change that.

MSO and DLBC are entirely different methods that use computing power to zero in on a set of filters that yield closest to desired result at each location.

Thank you for pointing out the 1st 15 min of the video and for helping me understand the single and multiple response measurements and its compromises. I think for now and util I learn how to use MSO effectively; I am going to do multiple measurements, 6"-8" apart and optimize for the MLP. Once I learn how to use MSO comfortably, I will optimize for 3 seats. Thank you again for your feedback and explanation.

mmaresg6 · Oct 31, 2024

Just to be sure when I do the multiple measurements at the center seat to time align my 4 subs I need to do Cross Correlation Align and Vector Average correct?

JStewart · Nov 1, 2024

mmaresg6 said:
Just to be sure when I do the multiple measurements at the center seat to time align my 4 subs I need to do Cross Correlation Align and Vector Average correct?

Yes, cross-correlation and subsequent vector averaging can be used to get a spatially averaged measurement to apply EQ to.

Aligning subwoofers to each other is a separate process from this and would be done first using REW’s alignment tool. With 4 subwoofers start with the furthest two and then save the aligned copy to which to align the third farthest. Same process for #4.

I would emphasize one thing OCA said and that is you are aligning for maximum SPL (=most in-phase or phase aligned) and that may or may not also be the smoothest response.

You’re always welcome to post a REW .mdat file with questions and if you do have questions it may also be helpful to know what hardware is handling the DSP and what model(s) of subwoofers.

Gerry Iaria · Nov 2, 2024

I'm yet to try cross-correlation and vector average, it sounds difficult, probably it isn't, I've a Mac, and so, don't use MSO, it's my wife's laptop, and don't want to mess it up for her, if I know what's good for me :laugh:

mmaresg6 · Nov 3, 2024

I spent some more time Friday, trying Cross Corr Align and Vector Average to get a spatially averaged measurement and applied EQ to my preferred house curve. Like JStewart suggested, I timed aligned my 4 subs using REWs alignment tool which it always blows me away on how precise this tool is. I did a total of 6 measurements around the MLP(Center seat), 8 inches to left and right, and 6 inches to front of MLP. I must say that this procedure resulted in an improvement for less seat-to-seat variation for my three seats but for the center seat in particular as compared to the single measurement at the MLP. I'm still going to try MSO to see if I get better results for my three seats. And as always, any support from this Forum is greatly appreciated. BTW. my 4 subs are RLS Speedwoofer 10s, connected to a miniDSP 2x4HD.

mmaresg6 · Nov 3, 2024

JStewart said:
Yes, cross-correlation and subsequent vector averaging can be used to get a spatially averaged measurement to apply EQ to.

Aligning subwoofers to each other is a separate process from this and would be done first using REW’s alignment tool. With 4 subwoofers start with the furthest two and then save the aligned copy to which to align the third farthest. Same process for #4.

I would emphasize one thing OCA said and that is you are aligning for maximum SPL (=most in-phase or phase aligned) and that may or may not also be the smoothest response.

You’re always welcome to post a REW .mdat file with questions and if you do have questions it may also be helpful to know what hardware is handling the DSP and what model(s) of subwoofers.

I have another question regarding Cross Corr Align and Vector Average. Is it recommended to do multiple measurements at different positions around the MLP for each sub, cross correlate and vector average to time align subs and use the vector average result to apply EQ? All the reading I've done about cross corr and vector average suggest you do the spatial averaged measurements for EQ but after you get your subs phase and time aligned? I want to make sure I'm doing this right. Thank you in advance for all your help!

Gerry Iaria · Nov 3, 2024

mmaresg6 said:
I spent some more time Friday, trying Cross Corr Align and Vector Average to get a spatially averaged measurement and applied EQ to my preferred house curve. Like JStewart suggested, I timed aligned my 4 subs using REWs alignment tool which it always blows me away on how precise this tool is. I did a total of 6 measurements around the MLP(Center seat), 8 inches to left and right, and 6 inches to front of MLP. I must say that this procedure resulted in an improvement for less seat-to-seat variation for my three seats but for the center seat in particular as compared to the single measurement at the MLP. I'm still going to try MSO to see if I get better results for my three seats. And as always, any support from this Forum is greatly appreciated. BTW. my 4 subs are RLS Speedwoofer 10s, connected to a miniDSP 2x4HD.

Nice job

JStewart · Nov 3, 2024

mmaresg6 said:
I have another question regarding Cross Corr Align and Vector Average. Is it recommended to do multiple measurements at different positions around the MLP for each sub, cross correlate and vector average to time align subs and use the vector average result to apply EQ? All the reading I've done about cross corr and vector average suggest you do the spatial averaged measurements for EQ but after you get your subs phase and time aligned? I want to make sure I'm doing this right. Thank you in advance for all your help!

Your reading is correct. With the subs aligned to each other they act as a single sub that can be EQ’d.
Different filters for different subs will lead to unexpected results when the subs are combined. This is because filters change both magnitude and phase and therefore the way the subs will sum together.
With enough computing power you can use different filters on different subs to effect a desired result. This is what MSO does. Many thousands of calculations to arrive at a solution that sums properly.

Your chart looks good.

How’s it sound?

mmaresg6 · Nov 3, 2024

JStewart said:
Your reading is correct. With the subs aligned to each other they act as a single sub that can be EQ’d.
Different filters for different subs will lead to unexpected results when the subs are combined. This is because filters change both magnitude and phase and therefore the way the subs will sum together.
With enough computing power you can use different filters on different subs to effect a desired result. This is what MSO does. Many thousands of calculations to arrive at a solution that sums properly.

Your chart looks good. How’s it sound?

Ok thanks! There is so much to learn about REW and MSO and figuring stuff out is always fun. I ended up with a 10dB boost EQ house curve. My system sounds awesome. The bass sounds clean and tight. Thanks again!

mmaresg6 · Nov 3, 2024

JStewart said:
Yes, cross-correlation and subsequent vector averaging can be used to get a spatially averaged measurement to apply EQ to.

Aligning subwoofers to each other is a separate process from this and would be done first using REW’s alignment tool. With 4 subwoofers start with the furthest two and then save the aligned copy to which to align the third farthest. Same process for #4.

I would emphasize one thing OCA said and that is you are aligning for maximum SPL (=most in-phase or phase aligned) and that may or may not also be the smoothest response.

You’re always welcome to post a REW .mdat file with questions and if you do have questions it may also be helpful to know what hardware is handling the DSP and what model(s) of subwoofers.

I think I have a good idea what might've caused the discrepancy for the verified house curve response and that prompted to initiate this thread. I just realized that I did multiple measurements, one for each seat(3), each sub, did a Cross Corr Align and Vector averaged the three measurements for each sub. I then Time Aligned the subs using the Vector Average response. As I have stated that all reading I've done about this suggested that you do the spatial average measurements and apply to EQ after you have the subs time aligned to each other and you have corroborated that on paragraph 2 of your reply. So what I have learned is that you don't do multiple measurements, you don't do Cross Corr Align and Vector Average to time align subs to each other. I tried this method after watching this video from OCA (link attached). It is also possible I may have misunderstood the video tutorial too.

Cross Correlation and Vector Average Measurements

Member

Attachments

Member

Member

Member

Senior AV Addict

Senior AV Addict

Member

Attachments

Member

Attachments

Member

Senior AV Addict

Member

Member

Senior AV Addict

Moderator

Member

Attachments

Member

Moderator

Senior AV Addict

Member

Member

Popular tags