Help with REW/Room treatment (mdat and layout provided)

[Derboy]

So I did a few new REW measurements and the Bass decay at the MLP is better than it was. Treble decay is marginally improved but the room is still slightly over-damped, but I actually quite like what I've ended up with. I've been messing around with Dirac target curves and been doing a fair bit of tweaking. A lift in the the top end from around 2khz all the way to the top has resulted in a sound that I'm really happy with.

@Keith_W I really want to thank you for your time and patience in educating me on REW and the detailed explanations that not only offered up advice but actually helped me learn more about what's going on acoustically and understand the "why". You are credit to the site.

And also thanks to everyone else who commented and offered some inside as well. It's very much appreciated.

 

[Derboy]

The kit in the room has had an overhaul since I posted this up. Changed out sound of the treatments. Upgraded to Monitor Audio Gold 6G speakers and Marantz AV20+AMP20 combo.

Placement is still the same but IACC is definitely better than it was but it's never going to be perfect. I think it sounds as good as it's going to get unless I rip the room up and start again but that's not an option.

 

[Derboy]

Now here's a question. If I set a flat target curve in Dirac, then run a measurement with REW. Can I then reduce the peaks between the FL & FR speaker and then design a target curve that reduced the peaks to try and achieve symmetry?

Here's FL & FR compared with my current chosen (non-flat) target curve loaded.

Im wondering what would happen if I reduced the peaks in the orange trace to match the blue trace, from a symmetry prespective and if they're any downside?

 

[Keith_W]

Now here's a question. If I set a flat target curve in Dirac, then run a measurement with REW. Can I then record the peaks between the FL & FR speaker and then design a target curve that reduced the peaks to try and achieve symmetry?

Here's FL & FR compared with my current chosen (non-flat) target curve loaded.

View attachment 538427

Im wondering what would happen if I reduced the peaks in the orange trace to match the blue trace, from a symmetry prespective and if they're any downside?

Your graph is too zoomed out. Use a 50dB vertical scale please.

But the more important point is that it appears that you measured at the MLP and got Dirac to force the L/R high frequencies to the same target curve. Do not do that. Wavelengths are very short at high freqs, and you would be equalizing at a very specific point in space AND you would be including all sorts of reflections and measurement artefacts into your correction. Dr. Toole specifically warned about this type of DSP in his book, and he used pretty strong language, only falling short of using the word "scam".

What you should do instead: you need to measure the anechoic response of your loudspeakers above a certain frequency from 1m distance and correct that to flat (as in straight 0dB line). Then measure again at MLP, and if you want, you can apply "broad, low-Q, tone control like equalization" (Toole's words) to taste. Most of us use a -4dB slope if the speakers don't already do that. AFAIK, Dirac can't do that but I may be wrong.

You should absolutely not correct every little wiggle that you see at high frequencies from an MLP measurement.

 

[Derboy]

Your graph is too zoomed out. Use a 50dB vertical scale please.

But the more important point is that it appears that you measured at the MLP and got Dirac to force the L/R high frequencies to the same target curve. Do not do that. Wavelengths are very short at high freqs, and you would be equalizing at a very specific point in space AND you would be including all sorts of reflections and measurement artefacts into your correction. Dr. Toole specifically warned about this type of DSP in his book, and he used pretty strong language, only falling short of using the word "scam".

What you should do instead: you need to measure the anechoic response of your loudspeakers above a certain frequency from 1m distance and correct that to flat (as in straight 0dB line). Then measure again at MLP, and if you want, you can apply "broad, low-Q, tone control like equalization" (Toole's words) to taste. Most of us use a -4dB slope if the speakers don't already do that. AFAIK, Dirac can't do that but I may be wrong.

You should absolutely not correct every little wiggle that you see at high frequencies from an MLP measurement.

Ok. I hear you loud and clear on this. Don’t do what I’m proposing. BUT in graph above, I’ve not done it yet. This is just an MLP measurement with Dirac running using the same target curve for taste across all speakers. Which is the norm. Are you saying that I shouldn’t do that?

Dirac corrects the response of each speaker at the MLP to the chosen target whether it’s flat or not. It can’t do it in the way that you describe.

The only way I can think of would be to measure rack speaker with REW and design a target curve to correct each one on that basis and then feed those curves into Dirac for it to apply then recheck near-field again with REW?

 

[smygolf]

Very informative thread indeed (thank you @Keith_W )

A quick question @Derboy , how many measuring points do you have and how big is the measuring area ?

 

[Keith_W]

Ok. I hear you loud and clear on this. Don’t do what I’m proposing. BUT in graph above, I’ve not done it yet. This is just an MLP measurement with Dirac running using the same target curve for taste across all speakers. Which is the norm. Are you saying that I shouldn’t do that?

Dirac corrects the response of each speaker at the MLP to the chosen target whether it’s flat or not. It can’t do it in the way that you describe.

Are you applying the target curve across the entire frequency spectrum? Is there a way to smooth the upper frequencies more in Dirac so that less correction is applied? You don't want to chase every little peak in the upper frequencies. If something sounds off, take an anechoic measurement and check that it's not the speaker. Otherwise, it is best dealt with using room treatment.

I don't use Dirac myself, I have only seen other people use it. So I don't know if there is some kind of setting hiding somewhere to do what I describe. I know for sure you can do it with REW though.

The only way I can think of would be to measure rack speaker with REW and design a target curve to correct each one on that basis and then feed those curves into Dirac for it to apply then recheck near-field again with REW?

Or ... just don't use Dirac and only use REW. I don't see the point of Dirac.

 

[Derboy]

Are you applying the target curve across the entire frequency spectrum? Is there a way to smooth the upper frequencies more in Dirac so that less correction is applied? You don't want to chase every little peak in the upper frequencies. If something sounds off, take an anechoic measurement and check that it's not the speaker. Otherwise, it is best dealt with using room treatment.

I don't use Dirac myself, I have only seen other people use it. So I don't know if there is some kind of setting hiding somewhere to do what I describe. I know for sure you can do it with REW though.



Or ... just don't use Dirac and only use REW. I don't see the point of Dirac.

I use REW to measure Dirac's results but I'm not sure how I could use it to apply changes.

As for Dirac ART it is drastically improving the bass response in my room and the overall frequency response. You can see the difference that it's making from the graphs above. If you're not familiar with it, then we may be talking from opposing view-points on room correction entirely.

And yes the target curve is full range. In my "before graph" you can see that there is variation in response across different speakers right up until 6khz so Dirac is taming that.

As far as "something sounding off". I'll be honest, nothing does. I'm just having fun chasing perfection as much as possible within the limitations of my room, plus I like learning and messing about with this stuff! Maybe it's as good as it can be now but it's good to experiment.