Some advice for general room correction - I've read quite a few tutorials, but I still have some que

[Pygmy]

I've built a pair of speakers for in my kitchen.
The kit I've built is the "Samuel HQ" by Heissmann Acoustics (original kit measurements are on this page):
https://heissmann-acoustics.de/en/samuel-hq-sph175hq-xt300-xt25/

My kitchen is 5.5x5 meters, and sadly I have no choice in speaker positioning - they're completely in the corners, nothing I can do about that.
My usual listening position means they're at approximately 45 degrees instead of 30, and the right speaker can't be completely toed in to point at me.
As can be expected this results in less than optimal sound.

Maybe an important note:
I am not looking for audiophile perfection.
I just want to make an agreeable best of a bad situation, averaged for a wide listening position (give or take a meter :-)).

A friend borrowed me his Umik1 for as long as I like, so I started to read up on room correction and REW.
I've read quite a few tutorials on different forums and I think I have a fairly decent understanding now - for a beginner ;-)

I've tried a few different approaches so far - one sweep measurement at the listening position, 9 sweep measurements on/around the listening position, and averaged RTA measurements left and right of my head at the listening position.

I'm doing my measurements and tests using a laptop and Equalizer APO, but in the end I am using a Raspberry Pi with a Hifiberry DAC+ DSP for the permanent eq setup.

My questions:

I read on multiple occasions that REW EQ correction is best used on the low end - say below 400hz.
But my measurements show a significant boost in the high end that I would like to tame. (I'd prefer to end up with something like Olive/Toole house curve)
I've been able to do that perfectly fine using target matching to 20khz, but as mentioned a lot of people say it's bad to do correction above the bass frequencies.
Am I going the wrong way applying EQ all the way up and if so - why, and how could I get my sound okay without doing that then?

Next, I have some dips in bass response. I can fix them by boosting, but generally I read that boosting is ill-advised because it puts more strain on the amplifier.
Okay, but , I *do* want to get the correct response! :-)

Another thing - I read that having lots of filters is a bad thing, indicative of a wrong approach.
Yet if I tell REW I want to get close to 1dB matching I get a really nice prediction - but lots of filters.

I realize this is all much more complex than my understanding of it as of now.

Can you give me more understanding on how to go about all of this?

Edit:
Here's the graph of my listening-position measurement using RTA of the right speaker (which is positioned most problematic), using var smoothing.
I'd think the 46 hz dip would be a room mode response, but this was an RTA measurement of a 30x30x30cm area to the right of my head - shouldn't the area averaging take care of room modes?
What would be your approach to correct this speaker's response?

 

[Wayne A. Pflughaupt]

You’ve obviously read a lot of opinions on this subject. Here’s one more.

IMO any fixed house curve, like the one Olive/Toole recommends, is a potential problem because no curve will fit every room. Smaller rooms need a steeper low-freq slope than large ones, and a close-up seating position needs a steeper high-end roll out than one 20 ft. away.

Certainly, the Olive/Toole curve is fine for a starting point, but if it doesn’t sound good to you, then feel free make adjustments. Don’t assume it’s “right” just because it’s from a highly esteemed source, if your ears tell you it’s “wrong.” The article on house curves I wrote (a few years before the Olive/Toole paper) gives some tips on determining the house curve your room needs, although it mainly concentrates on the lower frequencies.

You’ll probably find that “1-dB matching” with lots of filters will sound worse than the resulting graph looks. That’s one of the things I learned early on experimenting with REW, that a graph that looks good doesn’t necessarily sound good. My rule of thumb is to avoid filters with a Q higher than ~5.8 and gain values 3 dB or less, as they will not be audible with program material. You might also want to wade through my other dissertation on mimimal EQ.

As far as EQing above 400 Hz, it can be tricky, but if done properly and it gets an audible improvement, then I don’t see a reason not to do it.

Typically you’re focusing on the worst problems, as they are the most audible, and not every little ripple in the response curve. Again, this typically means problems that can be fixed with lower-Q filters and significant gain values, probably more than 4 dB. Just make sure that filters above ~2-300 Hz for the front L/R pair match for both channels, even if their separate measurement are different. EQing the L/R channels independently will result in imaging problems.

Again, improved sound quality is the goal. If a graph appears to need a 7 dB cut at a certain place, but that doesn’t sound as good as a 4 dB cut, then go with the latter. The high-end “hill” in your graph is a no-brainer. A filter ~1.4Q at 4 kHz cut ~ 5-6 dB should do the trick. There’s no way that won’t sound better, assuming things sound bright to you now.

Feel free to ignore any advice about not using any boosted EQ filters. It’s a myth that just won’t die.

For instance, let’s say your subwoofer has a peak at 45 Hz that needs to be tamed. The situation is that your subwoofer calibration (i.e. its level relevant to the main speakers) is based on that peak. Once you eliminate the peak with equalization, you will find that the sub’s level is now too low, and you have to increase it to compensate.

Well – say goodbye to any headroom you thought you “saved” by using only cutting filters. Gain is gain as far as the signal that passes to the amplifier is concerned. It doesn’t matter if it comes from EQ filters or the sub’s level control.

What’s apparently escaped some folks is the electrical response of the equalizer as filters are added. Let’s take the case of that 45 Hz peak. If you use a bunch of filters to boost everything above and below 45 Hz up to the level of the peak, the electrical response of the equalizer, passed to the sub’s amp, would look something like this:

However, if you had merely used a single filter to cut the peak, the electrical response seen by the amplifier would look like this:

What’s the difference between the two? For all practical purposes, nothing: In both cases, the signal passing to the amp has a big hole where 45 Hz “used” to be, and is relatively flat on either side of it (save for the natural roll-out of the filters in the first picture).

And - it should be a no-brainer that multiple cutting filters leave you with peaks between the filters!

So at the end of the day, cut-only filters accomplish nothing as far as “headroom saving” is concerned. The simple truth is, virtually any equalization taxes both amplifier and driver headroom, so you have to have enough to spare going in. This is why it’s critical for the main speakers to limit EQ to as much as possible, as most systems have far less amplifier power and more delicate drivers than the typical high-performance subwoofer. Not to mention, distortion from clipping is more readily audible.

That said, EQ boosting of nulls is a big no-no. You can tell if you have a null if there is no change in the measurement after EQ. It looks like you have one at ~48 Hz. However, the depression at 100 Hz can probably be lifted.

Regards,
Wayne

 

[Pygmy]

Cheers for the thorough explanation!
I've got some re-measuring and eq'ing to do now using this new info :-)

 

[Pygmy]

Okay, so I think I'll go forward applying individual filters per side for the low frequencies (as my speakers have different corner positioning / toe-in), and the same filters for middle- to high frequencies.

Edit:
After reading quite some more forum posts I think the best idea for me is to measure using RTA with infinite averaging, measuring a decent area around my listening position,
Using pink periodic noise, FFT size of 65536, rectangular window.

This should give me a pretty good idea of the response without getting stuck with room modes that a single (or low count multiple) position measurement would probably show me.

Do this individually for both speakers then use the individual speaker measurement to create the bass eq per speaker. (Say up to 300 - 400 hz)

Then combine the individual speaker measurements, and use that to create the same EQ setting for both speakers for everything above that frequency - trying to use minimal EQ to get somewhat close to my preferred response.

In the end I'll go over the suggested EQ settings, and remove filters with a Q higher than ~5.8 and gain values 3 dB or less, and test whether that makes any audible difference.
If removing doesn't degrade the sound quality in my opinion, they're right out :-)

doesn this sound like a good strategy?
If so, and I like the results, I think I'll probably write up a "Room correction for dummies with a stereo setup" post on my blog with this procedure explaining how I pieced together this info and why I made each decision.. :-)

 

[Wayne A. Pflughaupt]

Also, if your run a test signal with both speakers running, you may find that the highs in the measurement droop a bit compared to their individual measurements. If so, it’s an anomaly, so don’t try to correct that.

Regards,
Wayne

 

[NBPK402]

I will say that it is very important to get the Mic in exactly the same spot every time you run a test with Audyssey, YPAO, Dirac etc. If you are using multiple positions...make a template of some sort too so the Mic is in exactly the same spot. I say this because we would run Audyssey or Dirac, and look at the measurements, and listen it. Then we would run it again, and the measurements would be different, and we could hear the difference. Once I made a template for positioning the Mic exactly the same every time the sound stayed the same, and the measurements too.

It never hurts to have the Mic in exactly the same position even with REW when you are setting up crossovers either IMO.

 

[Pygmy]

I use RTA over a 1m2 area around my usual listening position exactly to *avoid* the measurements being so specific for the mic position.
I mean, my ears won't be at the exact mic position anyway (or even in the exact position every time I sit down), so tailoring the sound exactly for one position but not for an inch away doesn't make much sense to me.

 

[NBPK402]

The reason I said to put the Mic in exactly the same position was so that you have a measurement that is repeatable. You can do as wide if a spread as you like just make sure the Mic placement is identical to your previous one. By doing it this way you know that whatever changes you made to the system or placement of the speakers are what you are seeing, and not due to a Mic placement difference.

 

[Pygmy]

Using the RTA method I get results that are very repeatable.