REW measurements. Is it possible to fix 100-1000 Hz range with treatment?

[Keith_W]

Try moving front to back

Anyway if you REALLY want to find out where it's from, measure the time lapse between zero and the peak of the reflection. You may be able to see it on the "RT60 Decay" view. Or you could take time slices in the Spectral Decay plot. Use whatever seems easiest. In this spectro view, you need to zoom in quite a bit. Then using the speed of sound, convert the time-of-flight delay to distance. Then look around your room and see what might be causing it. Or you could slap an absorber on that wall and see if it goes away.

 

[janbth]

… the wavelet showed something going on around 500-600 Hz in all of my measurements.

What is going on in there?

So you have a desktop setup, as shown in your first post? The 500-600 Hz dip is possibly a desktop reflection. Try to cover the desktop with something and re-measure to see what happens.

 

[Tupisac]

You may be able to see it on the "RT60 Decay" view. Or you could take time slices in the Spectral Decay plot. Use whatever seems easiest.

RT60 decay and spectral slices are beyond my comprehension.

Is this method valid?

 

[Tupisac]

So you have a desktop setup, as shown in your first post? The 500-600 Hz dip is possibly a desktop reflection. Try to cover the desktop with something and re-measure to see what happens.

I'll try that tomorrow and report back. Thanks.

 

[Tupisac]

So you have a desktop setup, as shown in your first post? The 500-600 Hz dip is possibly a desktop reflection. Try to cover the desktop with something and re-measure to see what happens.

Ok, it is a desk.

Red is standard listening position, green with panel on the desk and blue with panel on the desk and matress on back wall:

Spectrograms:

 

[kyuu]

The "treatments" seem to have helped a bit with the issue between 500Hz and 700Hz, but have done very little with the big null at 250Hz. Normally I might suspect SBIR issues, but you already have the speakers basically against the wall, right? That's about the best you can do to ameliorate SBIR. Could still be a reflection issue with the right speaker due to it being in a corner, although I don't see evidence of that looking back at the individual measurements of the speakers.

Due to it being such a small room it is possible that you're suffering from room modes at higher frequencies than we'd normally see. If that's the case, it's going to be a rough problem to resolve. You could try using PEQ to fill in the dip (and knock down peaks of course); I'd try doing a pre-gain of -10dB and then see if putting 10dB into that dip helps any. Alternatively, if you don't want to spend time messing with manual PEQ (or learn to use REW's functions for generating PEQ filters), you could try a trial of Dirac to see if it helps you any.

I'll now go and hide before Keith throws something at me for suggesting an automated DSP solution.

 

[janbth]

Ok, it is a desk.

Red is standard listening position, green with panel on the desk and blue with panel on the desk and matress on back wall:

View attachment 551604


Spectrograms:

View attachment 551605
View attachment 551606View attachment 551607

OK - that’s a convincing demonstration of a desk reflection! Is it very audible? If so, I hope you will be able to make some useful modifications like moving the speakers, or something like that.

 

[janbth]

Regarding the null at 250 Hz (or slightly less?), I second @kyuu on a possible room mode issue. However, I fear it’s due to the listening position being in the null, rather than the speakers. This means that fixing it with PEQ will be difficult. You could always try it, but listener-null will probable need you to move the listener position to somewhere else.

I had my friend Claude generate a figure for the four lowest lying modes between walls with 3m distance:

The relevant mode is n=4. As you can see, it has a null at about 1.1m from the wall. This applies to both width and depth directions of your room. Is your MLP close to 1.1m distance from the front or side wall? If so, it will explain the dip in the frequency response.

 

[Tupisac]

The relevant mode is n=4. As you can see, it has a null at about 1.1m from the wall. This applies to both width and depth directions of your room. Is your MLP close to 1.1m distance from the front or side wall? If so, it will explain the dip in the frequency response.

Thanks, that's quite interesting.

MLP is 1 m from the side wall and 1 m from the front wall.
Between side walls it's 3,3 m and between front and back around 2,85 m.

I screwed the previous measurement drawings. Here is correct version:

 

[janbth]

MLP is 1 m from the side wall and 1 m from the front wall.
Between side walls it's 3,3 m and between front and back around 2,85 m.

Quite close, I would say. Real world measurements are never exactly what the theory predicts. I think the case can be made that you suffer from these nulls.

 

[Tupisac]

Quite close, I would say. Real world measurements are never exactly what the theory predicts. I think the case can be made that you suffer from these nulls.

Well... I guess it's time to accept the facts and dust off my pair of trusty headphones.

Thanks everybody for all the help. I learned quite a lot from your answers.

 

[Tupisac]

Lol, turns out stupid ham-fisted approach yields best results...

 

[kyuu]

Huh, seems almost too effective. How does it sound?

 

[Tupisac]

I'm just going through my playlists and can't hear anything wrong. Sounds nice and full.

Maybe it's the magic of near-field? There is just this one peak directly screaming at the ears and overpowering all the reflections coming back? I don't know...

 

[Keith_W]

Are those "after" graphs actual acoustic measurements, or are they a simulation?

If we look at these PEQ's ... that -11dB cut is OK. But boosts at 392, 483, 589Hz of up to 9.7dB ... be careful. It is likely that you are fixing the measurement instead of fixing a real acoustic issue.

 

[Tupisac]

Are those "after" graphs actual acoustic measurements, or are they a simulation?
If we look at these PEQ's ... that -11dB cut is OK. But boosts at 392, 483, 589Hz of up to 9.7dB ... be careful. It is likely that you are fixing the measurement instead of fixing a real acoustic issue.

Those are quick and dirty but actual measurements. No simulations. I didn't bother with measuring tape or SPL calibration, but I have a mark on the floor left from previous sessions and roughly set the knobs at similar positions. I've attached the mdat file.

I mostly care about 250 Hz hole as it sits right on the open B string of a guitar. It does coincide with the ringing you've pointed out on the waterfalls earlier on, so maybe it's not just a null. Now when I test it with slow sine sweeps whole area got audibly more consistent, maybe with a bit of a narrow peak. But so far I'd call that an improvement.

 

[Keith_W]

All I am saying is to be careful with DSP that high up the frequency range, particularly in a small room. You have to be sure that it's real. A single measurement from a fixed position won't show it, you have to move the mic around a bit and average the responses. Or you could do an MMM.

 

[kyuu]

They did perform an MMM. Or at least they labeled one of their graphs saying they did:

 

[Tupisac]

Yep, I did the MMMs. All is in the attached file. Pink periodic with crest factor minimized, 75% Hann window and all that jazz. 45 averages to fit between passing cars. I moved the mic around 20-30 cm in each direction.

What might partially explain the overall improvement is that I didn't compensate for volume drop. Graphs were aligned later.

Here are the normalized waterfalls before and after: