midrange peak, a problem in room acoustics

[Zathrez]

I have a room acoustics problem, but I don't understand it enough to even be able to describe it well. So I guess I can at least give you the REW mdat files (attached zip folder contains two files). One shows what the speaker (only the R) is actually doing, umik-1 was 1m away and at tweeter height, and pointed at the tweeter. The second is made with the umik at the MLP pointed directly up (with that calibration file of course). Both of these without EQ, and no subs. In Denon's words -- Pure Direct mode. Same chirp, just different mic positions.

The problem is that big fat peak at around 815 Hz. It is very "hot" and in a range that's frequently used in music. Makes anything that plays in that range sound sharp, glassy, metallic, and if it's acoustic piano, piercing. In short, highly annoying. And everyone who has listed to this room hears it.

My other problem is that I'm a REW newbie and have just enough experience to understand that I don't know what I'm doing. And I don't know how to use this big complex tool to help me find the problem(s) in the room so that I can fix them. Any help appreciated.

For context, Denon AVR3800h driving Ascend Acoustic Sierra Towers V2 (not the ELX towers, sadly). The Ascend Acoustic URL tab "measurements" has the tower's published spinorama, so you can compare that with my mdat file from the attached zip folder.

The room is basically a sealed rectangular shoebox, roughly 13.7' x 20.5' x 7.9'. It's "well treated" with IIRC 15 absorber panels, 12 diffuser / absorber panels, and floor to ceiling tri-trap "bass traps" in all four corners. All from GIK Acoustics. There's three doors (closed during use), and three windows. Two outside walls (long wall to left of MLP, and short wall to front (behind TV). One side wall window and one door are "conveniently located" at first reflection points -- both have acoustic panel coverage. Ceiling has lots of absorption. Oh, yes, like most basement rooms this one has an HVAC duct running down the two outside walls. It's behind soffit of 1x4s, with 5/8" wall board (lots more lumber than just 16" on-center just because that's how the carpenter did it).

Speakers positions are basically about 1m (39 inches) from side walls and from the front wall. Puts them mostly but not entirely in front of the TV to minimize reflections off TV (ribbon tweeter has huge horizontal dispersion). The L/R pair are angled in slightly. I've played with positioning of the speakers and this too makes minimal difference to the 815 Hz peak.

I have pretty much eliminated the idea that this peak is coming from a resonance in the room. I've dragged my head all over that room, floor to ceiling, and can hear the effects of this peak everywhere, but can't locate the source. I've touched every light fixture, switch plate, electrical outlet, door, door knob, all the individual panels, all the windows, etc. during chirps but nothing changes the REW results. I've covered the couch (MLP) with our two biggest thickest blankets, then the TV (front and back), and nothing I can do seems to make much of a change, and what little change there is is probably because I'm in the sound field blocking / absorbing some sound. But that 815 Hz peak doesn't waver.

So I gave up on resonances, and started looking at reflection points that weren't already covered. Yesterday I went all in on this idea, and moved panels from other areas in the room (because I don't have any panels that aren't already installed), to systematically cover possible reflection points one at a time. Six in all. All that happened is that each step made the peak either a little taller or a little wider. Often they changed other frequencies making some better and some worse. But not at 815 Hz -- that peak only got progressively (a little) worse each test.

So I'm a little lost now. I'm pretty sure that REW has tools that would help me track this down better, but I don't know what they are or how to use them. Basically, I just don't know how to troubleshoot this problem. Please shove me in the right direction; it will be much appreciated.

 

[Chazz6]

No experience with REW here – but a different attack on the problem. Do you have a simple parametric equalizer capability, whether in software or a device? I'd use it to pull down at 815 Hz. How much, how wide - trial and error by ear.

In my case, I have an RME ADI FS 2 DAC. It has a five-band parametric equalizer built in.

 

[kyuu]

The speaker does have a broad 2dB peak in its response centered right around 800Hz. How big of a peak are we looking at in your measurements? Posting an image of the graph would be helpful (many of us don't load up REW regularly).

815Hz seems pretty high for it to be a room-induced issue, though that broad peak does show up in the front-wall reflections. The directivity through that area is good enough that the simplest approach would likely be to simply use EQ.

 

[Sigberg Audio]

Have you tested what happens if you move the microphone, for instance to the two other positions in the couch?


For those not loading up rew:

1m from the speaker:

L+R at MLP:

L+R Average MLP:

 

[olieb]

My guess would be, that for several early reflections the path length difference to the direct distance is a multiple of 42cm (≈ wavelength for 820Hz).
reflection path - direct path ≈ n · 42cm
I would try to absorb some of the first reflection points and measure again.

Edit: After a bit of thought, you might check, whether this idea of constructive interference has anything to do with your problem, by moving the mic 20cm back/forward, left/right or up/down. If that does not (significantly change the FR then it is very improbable that early reflection superposition has something to do with it.

 

[Sernyl]

"reflection path - direct path ≈ n · 42cm"
I suppose a problem of this sort with speakers separation, but can't affirm as it is in the modal zone...

 

[Mishaiger]

I had a look at the graphs. I don't think the 815 Hz peak is a resonance in the usual sense.

As others have already pointed out, your speakers do have a small broad rise around this same frequency range. However, that alone does not explain the much larger peak seen at the MLP from both speakers, so this still looks more like a room/listening-position issue than a fault with either speaker.

I also don't see anything particularly unusual around 800 Hz in the decay plots. It does not seem to ring longer than the neighbouring frequencies. The filtered IR peaks are different for left and right, even though the tonal problem is similar, so I would not immediately blame one rattling object or one specific reflection point.

What does catch my eye is that the 100–600 Hz region seems to decay a bit faster than the range above it. I do not mean the frequency response itself here, but the reverberation balance. That can make a modest rise around 800 Hz sound much more obvious and unpleasant than it looks in the speaker's own response.

I don't know which panels you have or where they are placed, but diffusion does need some distance. Around 815 Hz the wavelength is about 42 cm, so as a rough rule a conventional diffuser would want at least about 1.25 m from the listener to be useful in that region, and preferably more. If any diffuser-absorber panels are close behind the MLP, or very close at early-reflection locations, I would try temporarily replacing those with plain broadband absorption and measuring again. If they are on the front wall several metres away, I would be much less suspicious of them.

That said, I am very much in favour of using EQ when it solves a real problem. I am currently listening in a temporary room with poor, asymmetrical acoustics, in nearfield at about 80 cm from the speakers. Without correction it sounds terrible. After measuring and applying separate EQ to each speaker, it sounds remarkably good for such a bad room. So I would not avoid EQ on principle, especially for a clearly audible problem that is easy to verify in measurements.

Before applying it, I would measure a small area around the MLP rather than just one exact point — centre, left/right, forward/back, perhaps slightly above/below — and average those measurements. If the 800–850 Hz rise remains, a moderate PEQ cut there is entirely reasonable. I would start with less than the full correction suggested by one measurement and adjust from there.

Since you said you are new to REW, there is also a lot of useful free material in the REW documentation and here on ASR about taking and interpreting measurements. It is worth learning enough to make one small change, measure it, and understand whether it helped — for example after shifting the MLP slightly or swapping a pair of panels — rather than having to post another set of measurements for every experiment.

Good luck tracking it down.

 

[Zathrez]

The speaker does have a broad 2dB peak in its response centered right around 800Hz. How big of a peak are we looking at in your measurements? Posting an image of the graph would be helpful (many of us don't load up REW regularly).

815Hz seems pretty high for it to be a room-induced issue, though that broad peak does show up in the front-wall reflections. The directivity through that area is good enough that the simplest approach would likely be to simply use EQ.

Seems to be in the 6-10 dB higher than 75 dB (REW wants it's output calibrated to 75 dB).

I have tried to EQ it with Audyssey. And Audyssey can not handle it. It can take it down 3-4 dB. What's left can be clearly heard. It's an improvement but not enough. What I think I need to do is take the peak down to the point that Audyssey can handle it. Or find a parametric EQ I can insert between the amp and the speaker. I'd need two of those, or a stereo pair that could be independently adjustable. I know the L/R are not exact duplicates.

 

[Zathrez]

My guess would be, that for several early reflections the path length difference to the direct distance is a multiple of 42cm (≈ wavelength for 820Hz).
reflection path - direct path ≈ n · 42cm
I would try to absorb some of the first reflection points and measure again.

Interesting thought. I didn't know it could work like that. I'll try and measure some tomorrow.

OTOH, the first reflection points are reasonably well covered. L/R, ceiling, and even the L speaker's right side first reflection zone, and the R speakers left zone. But there a fair amount of wall not cover in panels. More coverage to the front of the room (at least in front of the MLP) and less in the rear. I couldn't afford to cover everything, and who would want to listen in there if I did? I've spent a couple of bizarrely bad weeks working in an anechoic chamber; I for sure do not want that.

 

[Zathrez]

I had a look at the graphs. I don't think the 815 Hz peak is a resonance in the usual sense.

I've spent hours trying to find a resonance to no avail. I'm glad I'm not the only one to doubt a resonance.

As others have already pointed out, your speakers do have a small broad rise around this same frequency range. However, that alone does not explain the much larger peak seen at the MLP from both speakers, so this still looks more like a room/listening-position issue than a fault with either speaker.

Can a positioning issue, speaker or MLP, cause a peak this big at a frequency this far above the Schroeder freq.? The room mode apps I've used have calculated Schroeder freq. to be around 220 Hz.

I also don't see anything particularly unusual around 800 Hz in the decay plots. It does not seem to ring longer than the neighbouring frequencies. The filtered IR peaks are different for left and right, even though the tonal problem is similar, so I would not immediately blame one rattling object or one specific reflection point.

What does catch my eye is that the 100–600 Hz region seems to decay a bit faster than the range above it. I do not mean the frequency response itself here, but the reverberation balance. That can make a modest rise around 800 Hz sound much more obvious and unpleasant than it looks in the speaker's own response.

I did not know that. Maybe more treatment of the rear wall is in order.

I don't know which panels you have or where they are placed, but diffusion does need some distance. Around 815 Hz the wavelength is about 42 cm, so as a rough rule a conventional diffuser would want at least about 1.25 m from the listener to be useful in that region, and preferably more. If any diffuser-absorber panels are close behind the MLP, or very close at early-reflection locations, I would try temporarily replacing those with plain broadband absorption and measuring again. If they are on the front wall several metres away, I would be much less suspicious of them.

All the panels are GIK Acoustics. Mostly four inch (10 cm) panels, just less than half of those with diffuser plates on. There are three six inch (15 cm) panels on the rear wall to absorb a touch lower on the frequency scale. All the diffuser / absorber panels are at least a couple of meters away from the MLP. Luck, mostly. I didn't know to space them like that.

That said, I am very much in favour of using EQ when it solves a real problem. I am currently listening in a temporary room with poor, asymmetrical acoustics, in nearfield at about 80 cm from the speakers. Without correction it sounds terrible. After measuring and applying separate EQ to each speaker, it sounds remarkably good for such a bad room. So I would not avoid EQ on principle, especially for a clearly audible problem that is easy to verify in measurements.

Before applying it, I would measure a small area around the MLP rather than just one exact point — centre, left/right, forward/back, perhaps slightly above/below — and average those measurements. If the 800–850 Hz rise remains, a moderate PEQ cut there is entirely reasonable. I would start with less than the full correction suggested by one measurement and adjust from there.

Since you said you are new to REW, there is also a lot of useful free material in the REW documentation and here on ASR about taking and interpreting measurements. It is worth learning enough to make one small change, measure it, and understand whether it helped — for example after shifting the MLP slightly or swapping a pair of panels — rather than having to post another set of measurements for every experiment.

Good luck tracking it down.

I need all the luck I can get. I'm thinking this will also take a lot of work, and probably in the end, an untold number of measurements. Sigh... Oh well.

Or, you know, a decent PEQ on each channel. I'm not against doing that, but at this point I'm more curious about what weird thing(s) might be happening in the room. Curiosity is a great time waster sometimes. I hope this isn't one of those times. I'm hoping it will be more of a "learning experience".

 

[Ype]

I've spent hours trying to find a resonance to no avail. I'm glad I'm not the only one to doubt a resonance.

Can a positioning issue, speaker or MLP, cause a peak this big at a frequency this far above the Schroeder freq.? The room mode apps I've used have calculated Schroeder freq. to be around 220 Hz.

I did not know that. Maybe more treatment of the rear wall is in order.

All the panels are GIK Acoustics. Mostly four inch (10 cm) panels, just less than half of those with diffuser plates on. There are three six inch (15 cm) panels on the rear wall to absorb a touch lower on the frequency scale. All the diffuser / absorber panels are at least a couple of meters away from the MLP. Luck, mostly. I didn't know to space them like that.

I need all the luck I can get. I'm thinking this will also take a lot of work, and probably in the end, an untold number of measurements. Sigh... Oh well.

Or, you know, a decent PEQ on each channel. I'm not against doing that, but at this point I'm more curious about what weird thing(s) might be happening in the room. Curiosity is a great time waster sometimes. I hope this isn't one of those times. I'm hoping it will be more of a "learning experience".

What you hear is in line with the spinorama, just EQ from there first and ABX it yourself.
What a microphone picks up is not what you hear above the transition frequency.
Do not chase your tail before you EQ'ed the peak from the Spinorama measurements.

If it still does not sound good then take another look at your room, Floyd Toole has a lot of great suggestions in his book on where and how to treat a room.
To much treatment never sounded right to me, I simply use a nearfield seating location if I want to have a more direct sound field.

All I can tell you is what I learned from the research and ABXed myself, as hearing is believing in this case.
Just make sure to ABX and use the spinorama as your guide to EQ.

 

[Truth Ness]

Quick question:
As you were taking the 1m measurement, it seems the speaker was (~20 cm) farther away from the front wall compared to the other measurements?
Is that correct?
According to the mdat it was taken almost two weeks later so you may have changed their position while testing several things during that time?

 

[DavidMcRoy]

i have a typical 65" TV between the front speakers, and every speaker I measure here seems to have funny peaks around 800ish to 1000ish Hertz that wander with the exact listening/measuring position. All the speakers are placed about a half meter or less ahead of the screen. I never did the math to see if screen reflections might be the culprit.

 

[Farenheit]

I had the EXACT same problem. My living room is treated, even the ceiling. It was solved by removing the thick carpet that covered the floor.

And don't ask me why. I still don't understand it.

 

[Zathrez]

Quick question:
As you were taking the 1m measurement, it seems the speaker was (~20 cm) farther away from the front wall compared to the other measurements?
Is that correct?
According to the mdat it was taken almost two weeks later so you may have changed their position while testing several things during that time?

That "1m from the speaker" measurement is a one-off and not related to anything else. And therefore I don't remember the speaker's exact location. All that was for was to narrow the search, and it does that. It shows that the peak isn't coming from the speaker, or anything upstream of the speaker. I posted it simply to save people time and avoid the "blame the speaker" thing that sometimes happens.

 

[raindance]

What type of seat are you using?

 

[Zathrez]

What type of seat are you using?

It's a low-backed loveseat. Just my wife and I, and she's very attached to it. It's very comfortable.

We're recovering it with a slip cover. There's a highly regarded family run business here that makes excellent slip covers. What's coming is a covering that looks like a recycled stage curtain. A thick velvet-textured dark gray. It will be about as much of a broadband absorber as I can make it be. Right now it's covered with our two thickest blankets because it's summer and we aren't using them for anything else. Said blankets did make a small but noticeable improvement in dialog clarity -- the center speaker in HT mode points right at the MLP, and therefore right at this couch. But it really doesn't do much for the stereo problem with this 815 Hz peak.

 

[Zathrez]

I had the EXACT same problem. My living room is treated, even the ceiling. It was solved by removing the thick carpet that covered the floor.

And don't ask me why. I still don't understand it.

Yeah, I've found fabrics of all kinds to be sorta tricky. Some that I would have thought would be good acoustically aren't, and some that I was going to pass on turned out to work pretty well.

Leather for example -- I had no idea it would be so reflective in the mid range. A leather wing back club chair is no place to listen to music. Don't ask.

 

[Truth Ness]

That "1m from the speaker" measurement is a one-off and not related to anything else. And therefore I don't remember the speaker's exact location. All that was for was to narrow the search, and it does that. It shows that the peak isn't coming from the speaker, or anything upstream of the speaker. I posted it simply to save people time and avoid the "blame the speaker" thing that sometimes happens.

Okay, i understand.

I would experiment with speaker position, because that is one of the most important things and has much influence on the frequency response etc. Try and measure and you will see how the response will significantly change at various speaker positions.

1 meter distance from baffle (front of the speaker) to the front wall is generally a bad thing. It will result in a cancellation at ~86 Hz.
The smaller the distance the higher this dip increases in frequency, the further away it decreases. You may want to look up "SBIR" and how it affects the frequency response. Sbir not only causes cancellations, but also peaks. To be more specific, a series of cancellations and peaks, their frequencies depend on the distance to the corresponding surface (wall), and that does not only happen for the front, but for every wall, floor and ceiling.

Here is a very helpful Sbir tool where you can enter your room dimensions, distances etc and it will tell you where dips and peaks are to be expected. You can lessen their effects with absorption on the corresponding surfaces, btw.

SBIR calculator

SBIR calculator

tripp.com.au

It is best to either place the speakers as close to the wall as possible to shift the first cancellation (null) to a high enough frequency where it does not cause issue in the important bass range (for example 35 cm => 245 Hz) and is generally not that severe, or so far away that it shifts down very low (for example 2 m => 43 Hz) But many don't have a large enough room for that.

Since you have a somewhat long room you may have the space to place the speakers 1,65 m away from the front wall? Which will result in a dip at ~52 Hz, and it may well be that this will even out your room mode at ~52 Hz, killing two birds with one stone. Definitely worth to try if you have the space.

However, as already said, try different speaker positions, toeing in etc, you may find one (or several) where that ~800 Hz bump evens out. It most likely is caused by a reflection (or Sbir which is also caused by reflection) or several ones that interact. Sometimes 5 or 10 centimeters can make all the difference. If that doesn't work then try to raise the speakers.

Cheers

 

[Zathrez]

Okay, i understand.

I would experiment with speaker position, because that is one of the most important things and has much influence on the frequency response etc. Try and measure and you will see how the response will significantly change at various speaker positions.

I did that. See what you think; REW mdat files attached.

The rest is off topic. Currently all I've got time for is that 815 Hz peak.