Correcting below Schroeder, different target level

[radialMelt]

Hi all,

New here. Registered to pose a question to the forum re: approach when creating a correction filter for bass frequencies (below Schroeder.)

I'm working a very less-than-ideal space with some significant nulls in the low end. In order to achieve a "flat" response, I am lowering the target level in the REW EQ quite a bit, to the tune of 9-12dB, in order to attenuate everything else except for the null areas.

Anyway, because I am just applying filtering to below 350Hz (in my case) and because the target level of the filter is 9-12dB less than the level of my measurement, I end up with a great level disparity between the <350Hz range and the >350Hz range. See below. Blue as measured and averaged, purple is with filters below 350Hz applied, cyan is the target level of the filters.

I've been trying to use some additional filters to bring the >350Hz range down in overall level (to match the new target) but it is coming with unintended consequences.

How do you all approach correcting only a part of your room/speaker's frequency response and ensuring overall level continuity between filtered and unfiltered ranges? I.e., how is the boundary between corrected frequencies and uncorrected frequencies managed?

Thanks in advance!

 

[Sokel]

I would draw the same line at the 80dB mark and be done with it if I were you.

The 200Hz valley does not look like the room (as the physical building) , maybe positioning, fat couch/table/furniture?
Or, small mains with sub? This looks like a huge canceling like reversing an x-over or something.
Tried any other positioning?

 

[Daverz]

Looks like you want a cut off to be at a point where the response crosses 80 dB. I would suggest cutting off at 500 Hz rather than 400 Hz. You may even want to create several different filter sets with cut offs at 200 Hz, 500 Hz, on the right side of that 1 kHz peak, and even 2 kHz and then select the one that sounds most neutral.

 

[DWPress]

If you applied those filters you made in REW and remeasured your room it will not look like the prediction. The nulls are created by the room and objects in it and they will still be present and any +EQ you did to get it flat will be sucking amplifier power. As suggested, Try 80dB and only cut peaks, no boosting EQs and maybe a notch for that 1khz bump. Hopefully speaker placement can fix the big dip at 300hz.

Always best to try a filter set, remeasure, observe, repeat, repeat, repeat while noting any changes with your hearing correlated with the graphs. It's very helpful to have a way to switch between different filters for measuring and A/B listening, how are you implementing your filters (hardware/software)?

 

[kemmler3D]

I'll just chime in to recommend using MMM with noise to do bass EQ if you aren't already, single sweeps and positions end up with over/ under correction at certain frequencies IME.

 

[Keith_W]

You have a nasty dip between 200-400Hz. That is too high for you to fill with a sub. Before you even think about DSP, you need to diagnose the cause of this dip. It is almost certainly a room acoustics issue. More rarely it may be an issue of incompetent loudspeaker design (i.e. cancellation at the XO point between woofer and mid) and sometimes a measurement artefact. Assuming it's a properly taken measurement of a properly designed speaker, this kind of broad and deep dip at low-midrange frequencies is usually due to two or more cancellations at the same frequency band. It may be two SBIR frequencies, e.g. the distance between speaker to front wall being roughly equivalent to the speaker to side wall. Or it may be coincidence between an SBIR frequency and a room mode. Or two room modes.

Diagnosing the cause requires you to move the microphone and speakers. Please read the REW eBook, Ch 4.3 (p.40) on bass frequencies. There is a procedure in there to diagnose the cause of dips in the freq response.

Re: the blend between bass correction and upper frequencies, sometimes you have no choice but to apply full spectrum correction. One way to avoid this is if your speaker has dual binding posts:

This is a Sonus Faber Amati Futura that I helped to DSP for a friend. His speakers have "bi-wire" binding posts, and because he is rich he has a lot of amplifiers at home. I separated his crossover and bi-amped his speaker. This allowed me to increase the gain of the woofers (red) and subwoofers (not shown) to achieve a flat response whilst leaving the mid/tweeter (green) free of DSP. The final result is shown in black. This is a pretty expensive speaker - about USD$50k. He was pretty impressed with the result - the bass tightened up noticeably whilst leaving the tonality of his speaker unchanged.

 

[DWPress]

I'll just chime in to recommend using MMM with noise to do bass EQ if you aren't already, single sweeps and positions end up with over/ under correction at certain frequencies IME.

MMM method link. Very helpful to help position your speakers as well, faster than sweeps.

 

[radialMelt]

Thanks all. I appreciate the advice. This setup is in a far-from-ideal room and I sadly don't have a lot of options for placement/treatment - it's small and asymmetrical. Here's a floor plan. Ceiling is 2.47m high. Units = meter

If I understand what you're all saying, there is no cut and dry way of managing the transition between low end correction and the rest of the frequency range if the target level (of said low-end correction) is drastically different from the measurement level. I have been experimenting with leaving the target level as 80dB RMS (since that is what it was measured at) and the results are definitely more natural sounding. I like the idea of tailoring the filter range to coincide with an 80dB point. I will try that.

As for the measurements, I'll do some reading and more experimentation. I agree there's funky stuff going on. Will check out the MMM method. Thanks for the link.

 

[Daverz]

Thanks all. I appreciate the advice. This setup is in a far-from-ideal room and I sadly don't have a lot of options for placement/treatment - it's small and asymmetrical. Here's a floor plan. Ceiling is 2.47m high. Units = meter

View attachment 497377

If I understand what you're all saying, there is no cut and dry way of managing the transition between low end correction and the rest of the frequency range if the target level (of said low-end correction) is drastically different from the measurement level. I have been experimenting with leaving the target level as 80dB RMS (since that is what it was measured at) and the results are definitely more natural sounding. I like the idea of tailoring the filter range to coincide with an 80dB point. I will try that.

As for the measurements, I'll do some reading and more experimentation. I agree there's funky stuff going on. Will check out the MMM method. Thanks for the link.

I don't know if rotating the system 90 degrees counter-clockwise is practical. That would also put the listening position further away from a wall.

 

[Philbo King]

Hi all,

New here. Registered to pose a question to the forum re: approach when creating a correction filter for bass frequencies (below Schroeder.)

I'm working a very less-than-ideal space with some significant nulls in the low end. In order to achieve a "flat" response, I am lowering the target level in the REW EQ quite a bit, to the tune of 9-12dB, in order to attenuate everything else except for the null areas.

Anyway, because I am just applying filtering to below 350Hz (in my case) and because the target level of the filter is 9-12dB less than the level of my measurement, I end up with a great level disparity between the <350Hz range and the >350Hz range. See below. Blue as measured and averaged, purple is with filters below 350Hz applied, cyan is the target level of the filters.

I've been trying to use some additional filters to bring the >350Hz range down in overall level (to match the new target) but it is coming with unintended consequences.

How do you all approach correcting only a part of your room/speaker's frequency response and ensuring overall level continuity between filtered and unfiltered ranges? I.e., how is the boundary between corrected frequencies and uncorrected frequencies managed?

Thanks in advance!

View attachment 497316

The logical thing would be to lower gain and/or EQ on the stuff above Schroeder too. Of course, the bass nulls will magically reappear at the new lower level. The comments suggesting to use 80 dB as the baseline are the way to go.
Summary: You can use EQ to reduce peaks, but nulls due to acoustical problems stubbornly remain.

 

[radialMelt]

The logical thing would be to lower gain and/or EQ on the stuff above Schroeder too. Of course, the bass nulls will magically reappear at the new lower level. The comments suggesting to use 80 dB as the baseline are the way to go.
Summary: You can use EQ to reduce peaks, but nulls due to acoustical problems stubbornly remain.

Yeah, logical indeed (re: lower gain on the stuff above 350Hz) but I can't see a way to simply do that. The way REW is handling the transition from the low portion to the high (unaffected) portion is not clear to me. I was trying to add a shelving EQ on the region above 350Hz but I couldn't get it to line up properly-- either it leaves a bump in and around the transition, or it rolls off above 12kHz too much. In other words, I can't get a simple filter do cleanly do the job of bringing the target level of the high end (>350Hz) down to around the target level of the low end (<350Hz) after it's been corrected.

 

[radialMelt]

I don't know if rotating the system 90 degrees counter-clockwise is practical. That would also put the listening position further away from a wall.

Unfortunately, no. I am totally aware and willing to admit the physical layout and room will be the limiting factor here, and I don't expect "perfection" no matter how much work I do on the DRC side. Even with the odd looking curves, the bit of correction I am doing seems to be improving the imaging and punch. I'm motivated to try out a few more things, and learn something along the way.

 

[radialMelt]

Just gave the MMM approach a try. Quite efficient. They do corroborate the awful measurements I shared earlier though, hah.

L = blue, R = red. Var smoothing. 80dB measurement level. I suspect what we're seeing (the "offset" of room mode in the right channel) is a consequence of the fact that it's basically in a "deeper" part of the room. The back wall in that location is quite a bit further back than at the left position.

The downside to this is that it doesn't seem to work with the method I've been following that uses a vector average which I then bring into rePhase. I'll still give this a go with miniDSP PEQ

 

[DWPress]

Yup, welcome to the rabbit chase. Pay attention to your X/Y scaling and use it consistently - the pic above has 1dB Y scale making things look worse than the typical 5dB scale, keep the bottom and top SPL consistent as well to help yourself and others more easily interpret the data.

Use MMM to optimize new speaker positions for best response then take a new sweep and come back with results. No use trying to generate filters to fix this if things can be better optimized pointing speakers further in/out from LP, playing with height or distance from sidewall/back wall. If you have flexibility it could make EQ work easier.

 

[radialMelt]

Going to play with what limited placement options I have (toe-in, distance between speakers) to see if I can come up with better uncorrected measurements.

I will say that correcting up to 1200Hz ish (just right of that big 1k peak) and leaving the target level at 80dB has resulted in a much, much more balanced sound. I no longer hear the transition between the corrected and uncorrected ranges.

 

[radialMelt]

I think this is about as good as I can get it with speaker positioning alone. Var smoothing. The 125-300Hz region is a disaster but I think it must be the asymmetry of the room, considering the right side is much worse than the left. The side wall next to the left speaker is where I have my record shelves, which is likely helping with the first order reflections there, whereas the right side has nothing really in the way of absorptive or diffusing material. Will do some sweeps and see where I get with it.

 

[DWPress]

It's unfortunate the dip can't be improved more but that's really not bad especially if you added a sub or two. I would still use 80Hz as a basis for the EQ curve, it already looks like a +4 Harmon curve in the bass region if you plug that into REW EQ parameters. Take out the peaks at 600 & 900Hz and maybe try a very low Q filter at 4k to take out that slight rise. Just noted you are using rePhase too, as you discovered no time info in those MMM plots but still very useful as you discovered.

 

[Good Vibes]

Hi,

I was experiencing dips in the 150-500hz region until I followed some advice to move the speakers very close to the wall. I moved them so that the rear of the cabinet is approx 10cm away from the wall and it boosted the 150-500hz region. It left me with a dip in the 65-45hz region which I can remedy with a single subwoofer (in an impractical position). I had been led to believe that having speakers that close to the wall would have a negative impact on the overall sound but I tame the resulting peaks with EQ and it sounds better than ever.

 

[kyuu]

Agree with above, those dips above 100Hz could be SBIR. When you were playing around with speaker positioning did you try moving the speakers very close to the wall?

 

[radialMelt]

Thanks for the tips. I have not tried a position with the speakers very close to the rear wall as that's generally been taboo in my experience. They have rear facing ports and are already pretty bass-heavy so I didn't bother. That said, now I'm curious, so I'll give that a try and report back.

The only caveat being that the "rear wall" is not particularly well defined - behind the left speaker is a very deep closet, and behind the right speaker is essentially a corridor that leads to a stairwell, with the "rear wall" being in between these two entryways. There is a rear wall but the speakers are not entirely in front of it. Will do some more experiments though!