Question about REW L+R vs L vs R speaker measurements

[janbth]

For in-room frequency response measurements, I'd highly recommend you use the Moving Microphone Method with no additional smoothing, and if you want to measure combined L+R response, use the beta version of REW and in the Generator->random PN-> turn Uncorrelated On.

Uncorrelated turned On, instead of Off? Doesn’t that defeat the purpose of measuring L+R combined, as opposed to L and R separately and then summing?

 

[Keith_W]

I would actually flip this. I think we would finally deserve a solution that works above the room transient frequency. For god's sake we have been to the moon and back so many times and still struggling with this?

The limitation is not DSP, but measurement capability. You need to be able to obtain an anechoic measurement of your speaker if you want to correct it full range. Such capability already exists. I know of two guys who own a machine able to take such measurements, and they do it at home: Amir and Erin.

There are ways around this, but you will always have a low frequency limit dictated by the size of your speaker, proximity of drivers to the floor, and your measurement setup. Remember that a 20Hz sound has a wavelength of 17.2m - there's no room big enough for you to obtain a clean 20Hz - 20kHz measurement. What I suggest is to try to obtain a quasi-anechoic measurement of the loudspeaker alone down to the Schroder frequency and correct that with DSP. Anything below Schroder can be corrected together with the room. Even that can be very difficult.

 

[Keith_W]

Now what else can/should I improve based on these measurements of my response with DSP:

You need to pay attention to what @OCA said about that all-pass filter in your left speaker. Remove it.

This is the unwrapped phase of your right speaker, with and without DSP.

And this is for your left speaker (yellow = no DSP).

I don't know what your software thinks it is correcting, but I can't see any good reason for that AP Filter to be there. All it would do is create weird left-right phase effects and ruin centre imaging.

You may also want to consider some room treatment. I don't know how big your room is, but 600-700ms is pushing it. The appearance of your ETC suggests you have a rather large room, I can see a large reflection at 5.5ms and 25.3ms. You will need to calculate the appropriate RT60 for your room size. You will find instructions on how to interpret the RT60 and do this calculation in that eBook I linked earlier.

 

[nadrealista]

You need to pay attention to what @OCA said about that all-pass filter in your left speaker. Remove it.

View attachment 489822

This is the unwrapped phase of your right speaker, with and without DSP.

View attachment 489823

And this is for your left speaker (yellow = no DSP).

I don't know what your software thinks it is correcting, but I can't see any good reason for that AP Filter to be there. All it would do is create weird left-right phase effects and ruin centre imaging.

View attachment 489824

You may also want to consider some room treatment. I don't know how big your room is, but 600-700ms is pushing it. The appearance of your ETC suggests you have a rather large room, I can see a large reflection at 5.5ms and 25.3ms. You will need to calculate the appropriate RT60 for your room size. You will find instructions on how to interpret the RT60 and do this calculation in that eBook I linked earlier.

I will check on L vs R phase with DSP and get back to you - the measurements you are looking at were generated with 1/2" delta between L and R speaker distance in AVR calibration. Cancelation/FR dip was gone after I set the same distance to L and R.

Room is 24 x 17.5 x 8.75ft in the basement used as HT and playroom with concrete foundations so floor (carpet over pad on concrete), right and back wall are 2x4 drywall on concrete, front and side wall are regular drywall. Left wall is open to long hallway just before my MLP where some of the front speakers first reflections would hit. There is also overhead soffit for AC duct right above my MLP. All of those make some interesting room modes distributions.

Room has zero treatment, I am thinking about building DIY panels for the first reflection points plus front of the soffit, any thoughts on that?

I was playing in python and wrote some code to do speaker placement optimization and visualize the room 1st reflection point. Can share rest of the code with you guys if you are interested and have some coding skills

here is post with speaker placement calculator code: https://www.audiosciencereview.com/...lator-many-get-this-wrong.66373/#post-2416324

and here is my room:

top view with first reflection points:

side view:

soffit reflections:

 

[nadrealista]

@Keith_W
retook all measurements, attached.

here is the L vs R phase with dsp:

and without DSP:

In both cases there is additional phase shift on the R speaker between 800 and 900 hz?
Should I worry about this?


however running L and R together phase looks fine:

 

[Keith_W]

Please post the MDAT.

 

[nadrealista]

Please post the MDAT.

now attached in previous post

 

[Keith_W]

There are no non-DSP measurements in your MDAT to compare. There's only L and R DSP, and L and R "DSP Large".

 

[nadrealista]

ugh, correct file attached now. let me know if you see anything else I should be concerned about/address.

 

[Keith_W]

This is what you need to compare:

Yellow = No DSP
Red/Blue = DSP and "DSP Large" whatever that is.

You can see that the only difference is that both DSP curves have delay applied. In fact, "L no DSP Large" is exactly the same as "L DSP Large" with the exception of delay.

It's the same with the right, with the exception of a tiny bump at 18kHz in "R No DSP Large".

 

[nadrealista]

So DSP is not messing things up anymore now that distances for L and R are same?

Large = L and R speakers set to Large so they run full range without 80Hz XO to subs

I am assuming DSP is using delay for integration with two Subs?

Additional phase change in the R around 800Hz is not a problem?

 

[Keith_W]

So DSP is not messing things up anymore now that distances for L and R are same?

When you look at a phase graph, you have to realize that half of what you see is the speaker, and the other half is the room. Attempting to correct phase at high frequencies with an all-pass filter is a mistake, unless you take very specific measures: you need to obtain the anechoic response of the loudspeaker, and whatever you do to the left, has to be done to the right. Otherwise you will get weird L-R phase issues and it will sound off. What you hear is unbalanced imaging, or "phantom centre image drift" - where a vocalist appears to be centred when singing lower notes, and off to the side when singing higher notes.

When I look at a phase graph for high frequencies, I ignore all room effects. You tackle those by better speaker positioning, not by attempting to DSP them out. All I want to see is that no DSP has been inappropriately applied. And in this case, before/after are the same - so it's all good.

All this is different for low frequencies, especially modal frequencies. Because long wavelengths produce predictable patterns in the room, you can try to DSP the heck out of it. But once again, you need to know what you are doing and how to look for problems. I said in my eBook that for beginners, it is sufficient for you to look at the frequency response alone and ignore phase. You only need to look at the phase response if you are going to manipulate it. In your case, your DSP has inappropriately thrown in an all-pass filter at 7kHz which is why we have to look at your phase graph. Otherwise I would tell you to ignore it.

I am assuming DSP is using delay for integration with two Subs?

Who knows what your DSP is doing, man I'm sorry if that's not helpful, but automated DSP systems are a bit of a black box. I prefer to do it manually so that I know what has been done.

Additional phase change in the R around 800Hz is not a problem?

It's either speaker XO or room asymmetry. But since the phase discontinuity is different between L (280Hz) and R (800Hz), my money is on the room.

 

[nadrealista]

thank you for the insights and for looking into this.

I took a look at the impulse responses, and it is good to see that most peaks align with my first reflection's top view chart I generated. Looks like ceiling reflection followed by opposite side wall and back wall are the highest ones all about -16 to -20 dB down. Also interesting is that ceiling and back wall reflection levels are different between the speakers? There are few reflections that happen after back wall (25ms) so not sure where those would be coming from?

Hallway on the left side wall does eliminate R opposite side wall reflection

here is L impulse:

here is R impulse:

and reflection points along with calculated delays:

It is also interesting to see predicted vs measured room modes and how 2 subs address most of it:

 

[Keith_W]

That's good, you are learning I find it satisfying when my prediction from calculation matches the measurement. It gives me more confidence that I have identified the problem. Anyway i'll leave you for now - take more measurements, read that eBook, and check back here if you have more questions.