average of the peaks

[teyhyrh4r]

I'd like to propose an alternative method of averaging, which may be particularly suitable for averaging measurements in multi-seat scenarios.
This approach is based on the assumption that peaks are prominent, while dips are obscured.
Let's consider an example with two "frequency responses" from two seats, presented here in a highly simplified manner to facilitate comprehension:

the current averaging methods differ slightly but essentially perform the same operation (in this case, (a+b)/2):

if we generate a correction for this and convolve (multiply) our two initial curves with it, the resulting curves are as follows:

Indeed, we have enhanced our responses, resulting in smaller peaks and dips... although the number of peaks has doubled.
Nevertheless, in accordance with our initial assumption, these reduced peaks still remain noticeable.

Now, if we compute the average of the peak values from both measurements, depicted by the red line "on top":

isolated:

and subsequently generate a correction for it, which we then convolve with the measurements, resulting in:

our corrected responses exhibit deeper dips compared to the standard averaging method, YET we no longer have any peaks.

In my opinion, it is definitely worth attempting this type of averaging.

 

[John Mulcahy]

That's a maximum rather than an average and would impose the worst case corrections on the output even if they only applied in one of the locations measured, degrading all other positions.

 

[teyhyrh4r]

That's a maximum rather than an average and would impose the worst case corrections on the output even if they only applied in one of the locations measured, degrading all other positions.

I'm not quite sure I grasp your meaning. In my opinion, I have demonstrated that this method would result in less degradation across multiple positions compared to averaging.

 

[John Mulcahy]

As an example, suppose one position has a big peak and all the others are flat. Would the final outcome be an improvement? One position fixed and all others with a big dip?

 

[teyhyrh4r]

As an example, suppose one position has a big peak and all the others are flat. Would the final outcome be an improvement? One position fixed and all others with a big dip?

Well, to begin with, we should acknowledge that this room has an issue that cannot be resolved solely through EQ, right?

Anyway, let's suppose that the average method is superior in this scenario... Does that imply it will always be superior? No, because I presented a case where it is clearly inferior.
However, will the outcome with the peak method truly be worse? With the average method, you will still have an annoying peak at that specific seat, and you will also have dips in the others. Both approaches are likely inadequate solutions in this situation.

 

[John Mulcahy]

EQ is a compromise because the response is different at different measurement positions but the EQ is the same for all. The approach of using the maximum from all positions means every position gets over-corrected. The underlying assumption that dips in the response don't matter is false.

 

[teyhyrh4r]

ok, we're going around in circles here. you simply don't want it. thanks