Average of measurements

[Marzolino]

As one only measurement is not enough to get a valid EQ I use to make 4 measurements, around an half circle centred on the loudspeaker. So I get four measurement files that I name Meas-1, Meas-2, Meas-3 and Meas-4 and then I proceed as follows:

• open Meas-1 and Meas-2 in the All SPL window , generate the average (A+B)/2 and save it as: Meas AVG 1+2
  
• open Meas-3 and Meas-4 as before, generate the average (A+B)/2 and save it as: AVG 3+4
  
  
• finally I open the Meas AVG 1+2 and Meas AVG 3+4 and generate the average saved as:AVG 4Meas.

The last file AVG 4Meas could be the average of the 4 original measurements, but is it true? I doubt that the procedure is correct, I'm afraid that the last “AVG 4Meas” doesn't save all the information previously held in the original single measurements, including SPL, distortion, phase, IR and so one. Wath do you think about?

Another way to get an averaged measurement could be, I think, to merge the single filters derived from each measure, and calculate the average filter. Is there a way to get from 4 filters one single averaged? Some ideas about it ?



Thank you to everybody

 

[JStewart]

The last file AVG 4Meas could be the average of the 4 original measurements, but is it true? I doubt that the procedure is correct, I'm afraid that the last “AVG 4Meas” doesn't save all the information previously held in the original single measurements, including SPL, distortion, phase, IR and so one. Wath do you think about?

Not an expert in this area but with vector averaging phase data is definitely correct for the average. Since frequency and phase are mathematically derived from the impulse using FFT I believe the impulse must also be accurate.

For EQ purposes the Moving Mic Method might be a better approach as you will be applying the EQ to the steady state response within the measurement area.

You can use your favorite search engine or you can start here:

MMM mouving measurements method, which one is best with REW?

Hi, I saw this video: And I was surprised how fast they seem to move the mic. I am only an amateur here, but I followed the following method instead: https://mehlau.net/audio/room-correction-peq/index.html Measure speaker response Set RTA > Averages > forever Hold the mic at the main...

www.avnirvana.com

 

[jtalden]

As one only measurement is not enough to get a valid EQ I use to make 4 measurements, around an half circle centred on the loudspeaker. So I get four measurement files that I name Meas-1, Meas-2, Meas-3 and Meas-4 and then I proceed as follows:

• open Meas-1 and Meas-2 in the All SPL window , generate the average (A+B)/2 and save it as: Meas AVG 1+2
  
• open Meas-3 and Meas-4 as before, generate the average (A+B)/2 and save it as: AVG 3+4
  
  
• finally I open the Meas AVG 1+2 and Meas AVG 3+4 and generate the average saved as:AVG 4Meas.

The last file AVG 4Meas could be the average of the 4 original measurements, but is it true?

That results in a technically correct average for the 4 measurements. I have see may recommend averaging this way, but a simple average rather than the vector average seems to work better for EQ purposes in my opinion. The MMM method also agrees very closely with the simple average of many mic positions around the same area.

I'm afraid that the last “AVG 4Meas” doesn't save all the information previously held in the original single measurements, including SPL, distortion, phase, IR and so one. Wath do you think about?

REW will correctly calculate that vector average and the charts are technically correct, but I think they are misleading visually as there is more irregularity in the traces that may raise a concern. It is hard to say that is the way we would hear the sound as our ears are not vector averaging 4 locations at once. The SPL difference may be trivial for EQ purposes in many cases however, so it may work fine for you.

Another way to get an averaged measurement could be, I think, to merge the single filters derived from each measure, and calculate the average filter. Is there a way to get from 4 filters one single averaged? Some ideas about it ?

REW doesn't support this automatically. This sounds to me like another another possible approach, but I don't foresee any particular advantage compared to the the current recommended methods for EQ. In general, there is no value in looking at LP area phase response for the purpose of PEQ determination.

Some hobbyist do remove the direct sound phase rotation using an FIR filter and report a sound improvement. I do this, but in my listening experiments I have found no significant benefit to the sound quality. Possibly my hearing limitations or my room setup prevents this.

 

[Marzolino]

Not an expert in this area but with vector averaging phase data is definitely correct for the average. Since frequency and phase are mathematically derived from the impulse using FFT I believe the impulse must also be accurate.

For EQ purposes the Moving Mic Method might be a better approach as you will be applying the EQ to the steady state response within the measurement area.

I'm not a mathematic analyst so I must stay at low height, I do'nt know the difference between aritmetic average and vector average but I see that after an average is generated some caracteristics of a .dat file, available on the original measurement file, are lost, i.e. Distorsion, Waterfall, Impulse Response,

 

[JStewart]

I'm not a mathematic analyst so I must stay at low height, I do'nt know the difference between aritmetic average and vector average but I see that after an average is generated some caracteristics of a .dat file, available on the original measurement file, are lost, i.e. Distorsion, Waterfall, Impulse Response,

No math required, thankfully.
This section from REWs help files explains:

All SPL Graph

 

[Marzolino]

I'm not a mathematics analyst so I must stay at low height, I don't know the difference between arithmetic average and vector average and so I must believe that the resulting averages from single measurements are good!

I appreciate and use the Moving Mic Measurement, I usually make first the low frequencies EQ using the REW method (windowed sweep) that makes a selection of minimum phase range,where the EQ can be applied. So I get a set of EQ filters for low frequencies and apply it with my equalizer, then, with low frequencies equalized, make a Moving Mic Measurement, with low motion method widely in the room that gives me the high frequencies measurements. With these results I try to have a good, balanced sound analyzing graphs and listening to the music. I'm enough happy of the results.

Thanks




It seems to me an odd method that I've seen in the picture you have posted because of high mic speed and narrow space examined

 

[Marzolino]

No math required, thankfully.
This section from REWs help files explains:
All SPL Graph

Thank you JStewart, now it's all clear !