EQing in stereo vs Eqing in mono

[Rednaxela]

Sorry for the necro post, but how about this approach.

Assuming MMM measurement technique and a 10 band PEQ.

We have three measurements, left, right, and left + right playing together.

Then in addition to the measured sum, we calculate the L + R sum as well.

We compare the two sums. Exactly around the 60dB line a difference trace is displayed. Of course the actual difference trace fluctuates around 0dB.

The difference trace represents what we expect to happen minus what actually happens when both speakers are playing together. This expectation I believe is the underlying assumption to the separate channel EQ method, and the difference trace might represent additional corrections needed to make this assumption reality.

We export the difference trace.


We take one channel and set out to EQ it to a simple -0.6dB target curve, just as an example. In this image the HF section was EQ-ed already. The focus is on the sub-250Hz portion.

We load the exported difference trace as a house curve. This changes our target considerably.

We EQ to the adjusted target.

We repeat this for the left channel.

The end result should now match the result of EQ-ing the measured L+R response to the original target with one correction applied to both channels. The difference being that the individual speakers now have the same in-room response as well. Not really sure how desirable that is to be honest though. With the other method you make your speakers do the same thing and let the individual room responses fall where they fall.

Of course all this can and should be verified with post-EQ measurements. Working on that. In the mean time I'd be most interested in your thoughts.

 

[Rednaxela]

What kind of strikes me about his is that some of the contours of this difference trace are also visible as peaks and dips in the individual L and R in-room responses. Not all, but some are so striking that it's hard to believe it's coincidence.

So where one would think an additional correction is needed somewhere because the difference trace shows a little bump there, the bump is already present in the individual response. Almost like you get the required extra correction for free. See for instance how you hardly have to do anything right around the 90Hz peak, say at 70Hz and 110Hz, most notably in the right channel.

Am I onto something? I feel like someone who just worked out that 4/2=2, after having been taught that 2+2=4. Like a very big duh moment is about to occur.

Where is the catch?

 

[Geert]

Sorry for the necro post, but how about this approach.

Intuitively I have 2 remarks:

1. The difference graph is the delta between L+R summed sound levels, which you later on apply to each channel. Wouldn't you need to apply the half of the difference to each?
2. You apply the summed difference to both speakers, which means you don't take into consideration to what extent each speaker contributes to the difference. If I'm not mistaken you can have scenarios where as a result you will apply the most eq correction to the speaker with the flattest response (as that speakers response deviates the most from the difference graph).

What kind of strikes me about his is that some of the contours of this difference trace are also visible as peaks and dips in the individual L and R in-room responses

I think this relates to my second remark. Where a speakers response deviates the most, it will puts its stamp on the difference graph. If the other speaker has a neutral response at that frequency band, the difference response will equal the deviating speakers response. And the essence of my remark; the result is that you're going to correct the neutral speaker as the deviating one already looks fine (compared to your difference graph).

 

[Rednaxela]

Thank you so much for your thoughts @Geert, highly appreciated!

Please allow me to first clarify the objectives of the exercise a bit further.

---

When we wonder what's better - dual mono or linked stereo EQ, I think we are going back and forth between the benefits of

1) having each channel meet a certain in-room target response such as Harman, Toole, etc. with the appeal of 2) having an equal in-room response for each channel

versus

3) having a combined in-room response that meets the target

respectively.

With my exercise I try to reach 2) and 3), at the expense of 1).

(Ignoring for a moment if this makes sense to pursue in the first place!)

---

With regards to your remarks.

1. The difference graph is the delta between L+R summed sound levels, which you later on apply to each channel. Wouldn't you need to apply the half of the difference to each?

Thank you for pointing this out. This was my intuition as well. However I couldn't work out what this 'half of the difference' would translate to exactly i.e. how to correctly derive it from the original difference trace. Then I started to think that the application of the eventual EQ can alternatively be done in two stages, which should theoretically lead to the same result.

Stage 1: A dual mono EQ, getting the individual channels' in-room responses to meet a certain target. Basically what is in the fourth picture of post 21.

Stage 2: A linked stereo EQ to correct the actual combined in-room response for deviations from a(n imagined) combined target.

Now my theory is that the difference trace of the third picture of post 21 is exactly the EQ one would apply in Stage 2. If you agree, I'd argue that this EQ is applied as is to the individual channels, i.o.w. no halving is taking place. If you agree with that I'd further argue that this is equivalent to adjusting the targets of Stage 1 with this difference trace. Which is what happens from picture 5 on.

I guess if there is a catch, it's somewhere in that theory.

2. You apply the summed difference to both speakers, which means you don't take into consideration to what extent each speaker contributes to the difference. If I'm not mistaken you can have scenarios where as a result you will apply the most eq correction to the speaker with the flattest response (as that speakers response deviates the most from the difference graph).

I will have to think about this a bit more. But in general you're right. In this exercise I assume the anechoic responses of the speakers to be the same, I only EQ towards in-room targets, and I do not worry about which speaker contributed to what extent to a deviation from an intended sum.