What is the correct method to do room measurement to create Convolution Filter?

[Ducnguyen2k10]

Dear whom knows best about room acoustic measurement,

I’m getting into creating convolution filter by REW to use with Roon. I’ve read few online instruction and I did 2 methods of measurement:
1. To use pink noise to perform 9 measurements/chanel while moving the RTA Mics around listening point. Then do RMS aveage and apply EQ on that FR result.
2. To use full range sine sweep with 3 measurements/per point/chanel with 5 points at and around the listening point. Then do Vector Average and do EQ on that FR result.

I don't know which method is correct. Yet the first method produces the sound I like, and the 2nd method sounds weird to me.

My measurement setup:
- RTA Mic: Behringer EMC8000 with generic calibration applied
- Soundcard: Roland UA-55
- Software: REW

Is any one here knows the best and correct method?

 

[ernestcarl]

Dear whom knows best about room acoustic measurement,

I’m getting into creating convolution filter by REW to use with Roon. I’ve read few online instruction and I did 2 methods of measurement:
1. To use pink noise to perform 9 measurements/chanel while moving the RTA Mics around listening point. Then do RMS aveage and apply EQ on that FR result.
2. To use full range sine sweep with 3 measurements/per point/chanel with 5 points at and around the listening point. Then do Vector Average and do EQ on that FR result.

I don't know which method is correct. Yet the first method produces the sound I like, and the 2nd method sounds weird to me.

My measurement setup:
- RTA Mic: Behringer EMC8000 with generic calibration applied
- Soundcard: Roland UA-55
- Software: REW

Is any one here knows the best and correct method?

If it’s a moving average, why do you even need 9 sets per channel? Seems excessive. A single or two averaged sets would be enough. And with the normal sweeps, you are better off using the rms avg. I use the vector average as well with frequency dependent windowing, but I overlay the rms average on top to see the EQ effect on both.

 

[DWPress]

For a quick and easy measurement to get things dialed in close try the MMM method from listening position on your left and right channels inidividually and EQ LR separately. When you create your WAV filter use the EQ result for a stereo filter. See how that compares.

Did you note much difference in your 9 point/channel MMM measurements? Usually mine is consistent between measurements when no other changes are made. MMM will give you amplitude corrections but lacks other information like phase and decay. Creating filters and EQ in REW is a dance between flatness and not losing the dynamics of your listening space. I usually go at making filters from the measured file a few days in a row and apply the results (Roon in your case) randomly afterwards - our subjective impressions change with mood among other things.

Find the sound you like but then bust out the mic again and verify your results.

 

[Ducnguyen2k10]

If it’s a moving average, why do you even need 9 sets per channel? Seems excessive. A single or two averaged sets would be enough. And with the normal sweeps, you are better off using the rms avg. I use the vector average as well with frequency dependent windowing, but I overlay the rms average on top to see the EQ effect on both.

thanks for ur respose. So which method should I use? Using pink noise or using the sine sweep?

 

[ernestcarl]

thanks for ur respose. So which method should I use? Using pink noise or using the sine sweep?

Either method works fine. Log sine sweeps matter if one wants to look at the time component. MMM is faster, but it will not have the time information — which is fine if you don’t need it/don’t know what to do with it.

 

[Ducnguyen2k10]

For a quick and easy measurement to get things dialed in close try the MMM method from listening position on your left and right channels inidividually and EQ LR separately. When you create your WAV filter use the EQ result for a stereo filter. See how that compares.

Did you note much difference in your 9 point/channel MMM measurements? Usually mine is consistent between measurements when no other changes are made. MMM will give you amplitude corrections but lacks other information like phase and decay. Creating filters and EQ in REW is a dance between flatness and not losing the dynamics of your listening space. I usually go at making filters from the measured file a few days in a row and apply the results (Roon in your case) randomly afterwards - our subjective impressions change with mood among other things.

Find the sound you like but then bust out the mic again and verify your results.

I did the method 1 extensively. After put the Convolution filter into Roon, I did measure again using pink noise for each chanel and found the actual result is very close to the Predicted FR in REW EQ.
There is not much difference in 9 point/channel MMM measurements. So I think 3 points at and around the ears are sufficient.
Today I did the sine sweep method without changing anything in the room.
I do compare subjective listening btw 2 methods and found sine sweep sound really weird to me.
I also compare both methods with Anthem ARC (9 points measured) and the Anthem ARC sounds also very different.

 

[Ducnguyen2k10]

Either method works fine. Log sine sweeps matter if one wants to look at the time component. MMM is faster, but it will not have the time information — which is fine if you don’t need it/don’t know what to do with it.

I know with sine sweep, I can look at the waterfall graph to see the decay time.
The one step I skipped when doing the sine sweep method is not doing rephase. Maybe that makes my system sounds weird.

 

[ernestcarl]

I know with sine sweep, I can look at the waterfall graph to see the decay time.
The one step I skipped when doing the sine sweep method is not doing rephase. Maybe that makes my system sounds weird.

You can use rePhase with or without any kind of phase “correction”. You can also drag and drop exported measurements from REW with/out the phase information of either MMM or single sweeps to create your convolution file.

Why it sounds weird is probably because of overcorrection above the room’s transition zone ~150-250Hz

 

[Ducnguyen2k10]

You can use rePhase with or without any kind of phase “correction”. You can also drag and drop exported measurements from REW with/out the phase information of either MMM or single sweeps to create your convolution file.

Why it sounds weird is probably because of overcorrection above the room’s transition zone ~150-250Hz

You’re right. Something is really wrong with the sine sweep method. I should do it again!

 

[DWPress]

It might help if you showed us your measurements and correction filters so we can see what you're hearing. Like I said, going back and redoing or tweaking the EQ filters is often beneficial to hear the subtlety different approaches.

 

[Ducnguyen2k10]

It might help if you showed us your measurements and correction filters so we can see what you're hearing. Like I said, going back and redoing or tweaking the EQ filters is often beneficial to hear the subtlety different approaches.

Thank you! I’ll do the measurement again and post the results for you guys to check!