Need Help: KH120A + Dual KH750DSP Nulls in untreated room

[ENEN]

Thanks for reading my thread! I'm seeking help about eliminating the nulls in a non acoustic treated room.

I'm running the Neumann KH120A + Dual KH750DSP setup (with MA-1 Calibration), I've been working on them for several days now, and the REW Result still doesn't seem promising, with several nulls. Since my house is rented, I don't have the permission to drill holes in the walls (yet), so before requesting for any invasive work, I'd like to ask the ASR community for some advice.

My Room (Height: 2.4m (94.4882 inch)):
(All Doors / Windows Always Closed)

Below are the best 2 REW Results I got so far (All FR Not Smoothed unless stated):
Case 1.
Two subwoofers placed on the front wall (MA-1 Uniform, non-Stereo mode), KH120A's placed on IsoAcoustics ISO-130 stands on a tabletop, tilted upwards to face the ears.

REW Results:

MA-1 Result:

In this case:
~10-15dB Null @ 82Hz / 20-30dB (?) Null @ 344Hz

Case 2.
Two subwoofers placed on the front wall (MA-1 Uniform, non-Stereo mode), KH120A's placed on IsoAcoustics ISO-130 stands on a tabletop, using long bars to make tweeters height at ear level, no tilt.

REW Result:

MA-1 Result:

In this case:
~10dB Null @ 82Hz (Better than Case 1) / ~15dB Null @ ~310-350Hz (Better than Case 1) / ~20dB Null @ 499Hz (Was not in Case 1)

I'm currently at a loss, constantly creating new problems (nulls) while trying to fix the existing ones (which is understandable given the lack of acoustic treatment). Furthermore, I haven't attempted to move the table significantly for measurements because it's acting as a cable organizer underneath, moving the table back and forth would be extremely time-consuming. I'm out of ideas. Any suggestions are welcome! : D

--------------------
FYI, here are REW .MDAT files if you'd like to take a look : D
Case 1 File
Case 2 File

* Edit: Replaced Images

 

[kyuu]

I'm not one of the measurement experts here, but until one comes along I'm curious if, for the first setup you confirmed that second "null" at 344Hz is actually there or if it's a measurement artifact. Easy way to check would be to use a tone generator to sweep up and down in that area to see if you can detect the amplitude changing. If it's not audible, then it's not a problem. The dip at 82Hz is almost certainly a room mode.

 

[staticV3]

My advice:

-Use REW's Room Sim for figuring out the ideal placement. Much less tedious

-Use the Moving Microphone Method for measuring in-room frequency response

-set the response graph to 50dB vertical range (max SPL displayed minus min SPL displayed = 50dB)

-lastly, accept that in the real world, things will never be quite perfect. There's an excerpt in Olive's Harman target paper where various professional mastering studios were measured and the differences especially in bass response were horrible.
Don't obsess over a few dB.

 

[ENEN]

I'm not one of the measurement experts here, but until one comes along I'm curious if, for the first setup you confirmed that second "null" at 344Hz is actually there or if it's a measurement artifact. Easy way to check would be to use a tone generator to sweep up and down in that area to see if you can detect the amplitude changing. If it's not audible, then it's not a problem. The dip at 82Hz is almost certainly a room mode.

Thanks for your reply! In the Case 1, I checked 344Hz using a online tone generator, there's a significant audible "volume drop" near that range.

* Note: The 344Hz dip only occurs in L+R Combined Measurement, in Left Channel Measurement it's 250/395Hz (smaller dips, not too bad), in Right Channel Measurement it's 328Hz (Deep).
View attachment 487730

 

[ENEN]

My advice:

-Use REW's Room Sim for figuring out the ideal placement. Much less tedious

-Use the Moving Microphone Method for measuring in-room frequency response

-set the response graph to 50dB vertical range (max SPL displayed minus min SPL displayed = 50dB)

-lastly, accept that in the real world, things will never be quite perfect. There's an excerpt in Olive's Harman target paper where various professional mastering studios were measured and the differences especially in bass response were horrible.
Don't obsess over a few dB.

I appreciate your advice! I’ll give the new method a try during the daytime : D
Here is the updated FR (no smoothing) according to your suggestion (bass boosted due to the MA-1 preference curve):

Indeed, I suspect I've become overly obsessed with achieving the perfect sound, and I'm willing to try if it can get as close as possible.

* Edit: Replaced the wrong screenshot

 

[Joachim Herbert]

Two subwoofers placed on the front wall (MA-1 Uniform, non-Stereo mode)...

My desktop setup is KH120/KH750 (1x). Never felt the need for room calibration in my nearfield (80cm) setup, though I own a MA1 kit. Having said this, I would use stereo mode when using MA1, as both channels behave differently. But maybe I just misunderstand "uniform, not stereo".

I am in no way questioning the usefulness of room eq. I use GLM with Genelecs in the living room, which is a pleasure to use (in contrast to MA1).

 

[amirm]

I would ignore those sharp nulls. Many rooms have them. You have to have just the right note to hit them which is going to be very rare.

Your room is also not a rectangle so no standard/simple simulation is going to predict the response.

The MA-1 results look fine to me. Does it not sound good?

 

[ENEN]

My desktop setup is KH120/KH750 (1x). Never felt the need for room calibration in my nearfield (80cm) setup, though I own a MA1 kit. Having said this, I would use stereo mode when using MA1, as both channels behave differently. But maybe I just misunderstand "uniform, not stereo".

I am in no way questioning the usefulness of room eq. I use GLM with Genelecs in the living room, which is a pleasure to use (in contrast to MA1).

Thanks for your reply : D In my room, without MA-1 or any other calibration, there's a severe peak in the 100-300Hz range and more nulls caused by room modal / SBIR (I assume), which is very uncomfortable for my ears. After calibration, the abnormality in this range was almost corrected, and the effect was quite significant : D Happy to hear that your room sounds good even without calibration.

Regarding the Uniform mode I mentioned, I think it's only meaningful for multi-subwoofer setups. In my dual subwoofer layout, Uniform mode allows the two subwoofers to sound in a "Plane Wave Bass Array" configuration. According to Neumann, this has a chance to reduce the abnormalities caused by the room mode.

* Screenshot captured from Neumann KH750DSP Manual - Page 8

 

[ENEN]

I would ignore those sharp nulls. Many rooms have them. You have to have just the right note to hit them which is going to be very rare.

Your room is also not a rectangle so no standard/simple simulation is going to predict the response.

The MA-1 results look fine to me. Does it not sound good?

I really appreciate the info! : D Actually the sound is quite good with the current settings, but when checking with the Tone Generator, the frequencies near nulls seem to be muted (noticeably lower volume). Considering that the nulls fall within frequency ranges that most musical instruments cover to some extent (especially the 500Hz one), I'm quite worried they might severely impact sound reproduction (or maybe my perfectionist side is overthinking it; after browsing many posts, I feel that nulls / dips of 10dB or more are soooo bad).

The good thing is, after hearing your suggestion, my anxiety about the severity of these nulls seems to have lessened somewhat.

 

[AnalogSteph]

With your settings you are aiming straight for maximally ragged graphs:
L+R
Single mic position
No windowing
No smoothing

Things start looking a lot less dramatic once I apply a FDW of 9 cycles, but I would really recommend following the MMM tutorial and seeing what that gets you.

Given all the stuff on your desk, I would think that the higher stand configuration is the better one. The window behind the desk has me slightly worried... some absorbers behind the monitor area may prove useful (should be thick enough to remain useful to ~300 Hz, so I'd say 4" min).

 

[custom]

Place the subwoofers at 25% and 75% of the width of the room. Measure the room width with the bathroom door closed. Then rerun MA1 with the Mono Sub configuration. This subwoofer configuration should produce the best results. You can rotate the sub 90 degrees if it makes placement easier.

You can try to lift the subwoofer to 25% of room height. I think, this would help most about the null between 80-100hz.

However, don't expect perfect results without room treatment.

 

[HiMu]

but when checking with the Tone Generator, the frequencies near nulls seem to be muted (noticeably lower volume).

When you try that with tone generator, move a bit left-right/up-down to see how much that changes what you hear. In my room there are spots in that frequency region where left ear hears the signal and right doesn't.

Vector average from your Case2 measurements. Ignore the room mode bars at the bottom of the chart, those are from my room's dimensions.

Without any acoustic treatments that is pretty good starting point.

 

[levimax]

Sharp LF dips look bad on a graph but usually are not very audible. Try psychacoustic smoothing to see how you will perceive the sound.

 

[kyuu]

Sorry, I may have caused you more anxiety with my tone generator advice. I should have specified that you should be sweeping through fairly quickly so that only lower Q (broad) issues will stand out. Obviously if you are moving slowly and lingering on the problematic frequency you will hear it as the tone generator is extremely narrowband. As Amir said, very little actual content is like that and very high Q (narrow) nulls like that are generally not audible. At least we know it's really there and not an artifact of the measurement setup.

If you're inclined to mess with it more, it's almost certain that the null at 344Hz is SBIR. If you love the speaker closer or farther away to the boundary the frequency and amplitude should change. You could also do something like temporarily place something directly behind the speakers to check if that ameliorates the issue.

 

[somebodyelse]

Your room is also not a rectangle so no standard/simple simulation is going to predict the response.

If you can make a 3d model of your room there's a tool to calculate room modes:
https://www.audiosciencereview.com/forum/index.php?threads/room-eigenmodes-calculator.38734/
https://github.com/Ylio7/room_eigenmodes_simulator

 

[ENEN]

A minor update on November 14: Today I tried moving the main desk to the left and adjusting the speaker placement, I decided to replace the long IsoAcoustics stand with the shorter one. I tested and found that the 500Hz dip was caused by reflections from the hydraulic lift platform on the desktop (How I find it: I covered this area with my thick pillow, suddenly the 500Hz became flat!).

Place the subwoofers at 25% and 75% of the width of the room. Measure the room width with the bathroom door closed. Then rerun MA1 with the Mono Sub configuration. This subwoofer configuration should produce the best results. You can rotate the sub 90 degrees if it makes placement easier.

However, don't expect perfect results without room treatment.

Thanks! Today I adjusted the subwoofer's position as much as possible, following your method (which is also Neumann's official recommendation). Unfortunately, the low-frequency characteristics are similar to before and have not been fixed. I suspect it's due to the protruding walk-in closet in the room. Low Freq always hard to fix

With your settings you are aiming straight for maximally ragged graphs:
L+R
Single mic position
No windowing
No smoothing

Things start looking a lot less dramatic once I apply a FDW of 9 cycles, but I would really recommend following the MMM tutorial and seeing what that gets you.

I apologize for the way I presented my REW data, and thank you for pointing out these issues! Today I tried the MMM method (along with frequency sweep) for measurements. If I kept the microphone stationary, the nulls still looked very deep in the RTA Graph, but when I moved the microphone slightly following the MMM method, apart from these low-frequency nulls, the overall FR curve seemed less unpleasant. If possible any feedback on this latest result would be greatly appreciated : D

Pic 1: MMM - L + R

Pic 2: MMM - L

Pic 3: MMM - R
Link to: .mdat and screenshots

When you try that with tone generator, move a bit left-right/up-down to see how much that changes what you hear. In my room there are spots in that frequency region where left ear hears the signal and right doesn't.

Yes, for today's latest measurement, volume of the 300Hz Dip will change if I move my head : D

Without any acoustic treatments that is pretty good starting point.

That puts my mind at ease, thank you! My perfectionist side thinks that for further optimization, some invasive acoustics adjustments to the room might be the next step.

Sharp LF dips look bad on a graph but usually are not very audible. Try psychacoustic smoothing to see how you will perceive the sound.

I realized if I use the Psychoacoustic Smoothing, it's almost a perfect system (but not really xD)

Sorry, I may have caused you more anxiety with my tone generator advice. I should have specified that you should be sweeping through fairly quickly so that only lower Q (broad) issues will stand out. Obviously if you are moving slowly and lingering on the problematic frequency you will hear it as the tone generator is extremely narrowband. As Amir said, very little actual content is like that and very high Q (narrow) nulls like that are generally not audible. At least we know it's really there and not an artifact of the measurement setup.

If you're inclined to mess with it more, it's almost certain that the null at 344Hz is SBIR. If you love the speaker closer or farther away to the boundary the frequency and amplitude should change. You could also do something like temporarily place something directly behind the speakers to check if that ameliorates the issue.

Thank you for your detailed explanation! : D I also agree with your opinion on the 300ish Hz issue, yet there's no way for us to verify it at the moment but: I might buy some absorbers to try it out soon (today I tried walking around the room with my super heavy quilt to see if I could smooth out the 300Hz issue somewhere, but I oversimplified it and ended up sweating a lot).

 

[staticV3]

If possible any feedback on this latest result would be greatly appreciated : D

Pic 1: MMM - L + R

Pic 2: MMM - L

Pic 3: MMM - R

50dB vertical range please. And no need to show the Peak graph. Just RTA 1/48 Oct is fine.

 

[ENEN]

50dB vertical range please. And no need to show the Peak graph. Just RTA 1/48 Oct is fine.

Here are new screenshots based on your suggestion using 1/48 Smoothing.
* Using 50dB Vertical Range makes the dip looks even more terrifying

Also, I noticed a "Fit to Data" option in "Set Graph Axis Limits." Is it generally recommended to use this?

Thanks again!

Spoiler: MMM RTA 1/48 Smoothing - Left + Right

Spoiler: MMM RTA 1/48 Smoothing - Left

Spoiler: MMM RTA 1/48 Smoothing - Right

 

[custom]

The ripple in the high frequencies is caused by the comb filter effect from the reflections on the desk. You can try placing the monitor at ear level on the desk. Increasing the distance from the desk will reduce the effect somewhat. You can put a thick blanket on the desk and take a REW measurement. You should see a difference above 1khz. You can check for early reflections on the REW impulse graph.

The DIP at the 106hz on the right monitor can be caused by the right wall. You can try to increase the distance to the right and then the null should shift and decrease. An absorber at the reflection point on the right wall should decrease it as well.

I have similar system with 2xKH80 and 2xKH750. I posted some measurements in the past (2xKH80 and 1xKH750) here: https://www.audiosciencereview.com/...ll-us-how-your-bass-sounds.22567/post-1806762

 

[staticV3]

Here are new screenshots based on your suggestion using 1/48 Smoothing.

The measurements are 1/48 Oct straight out of the RTA already (since that's set to 1/48 Oct resolution).

There's no need to apply 1/48 Oct smoothing to them. That just loses resolution.

Also, I noticed a "Fit to Data" option in "Set Graph Axis Limits." Is it generally recommended to use this?

Fit to data is good for quick and dirty homing, but it can result in unnecessarily zoomed in y-axis if the measurement is flat enough, as well as unnecessarily zoomed out x-axis if the measurement sample rate is too high.

It's usually best to manually choose upper and lower y-axis limits that are 50dB apart and frame your measurement well, and then click the "20 ... 20k" button that appears when your cursor hovers over the bottom left corner of the graph.

Regarding your measurements, one thing that stands out is how treble rolls off much more in the L+R measurement.

As a fix, try enabling the "Uncorrelated" check box in the Generator window (requires REW Beta version here).
That's designed specifically for measuring stereo systems and may circumvent weird phasing issues that may happen otherwise.

Edit:
Oh also, in the RTA Window->Cogwheel->Appearance, turn off "Use Bars on RTA". That makes the measurements easier on the eyes.