Can Filling Nulls with DSP Actually Work?

[levimax]

While there is a lot of discussion / disagreement about the pro's and con's of DSP for room correction and the best techniques to use, I have found that the recommendation to not try to fill "nulls" with DSP gain to be widely accepted and backed by scientific theory i.e. pumping energy into "nulls" will not work because they are caused by destructive interference so more energy will just cancel out and not fill the dip. In addition any "boost" to a frequency range will reduce headroom and possibly overload the speakers and cause increased distortion. This makes practical and theoretical sense to me so for room corrections I just knock down the "peaks" and leave the dips alone.

Recently I was playing around with a room correction and I decided to try boosting a null and then took a measurement and much to my surprise the correction "worked" and visually filled in the null and listening it sounded fine. This then led me to trying REW's room correction filter generator and rather than use "no boost" I allowed it to create filters with individual boost up to 12 dB. I then tried these filters and again much to my surprise the filters all worked (filled in dips and cut peaks) and the measured response was almost perfectly matched to the target curve. I had to reduce the digital gain by 9 db (maximum boost of the generated REW filters) but listening to these filters they sounded fine on a wide range of different music.

So now I don't know what to think. My system is 2.2 with 2 KH310's and 2 very large subs co-located with the mains which allows me to cross over relatively high (200 Hz). The room is not good with alcoves, half walls, inconsistent width and multiple ceiling heights with only rugs and furniture for "room treatment". The speaker location also is not good. There is a large 30 Hz mode and a large suck out around 100 Hz and these show up pretty much no matter where I put the speakers or the LP but using the REW generated filters all these issues appear to be "corrected" (on axis at least).

See below MMM measurements (Var smoothing) of the KH310's by themselves (aqua), with 2 subs (purple), and with 2 subs and DSP(red).

So what is going on that the dips are all "filled" with DSP? Do large subs located close to the LP (6.5 feet away) have the ability to create "direct sound" that can over power room modes? Is the "can't fill dips with DSP" axiom actually not correct? Or is something else going on?

 

[Keith_W]

The rule that nulls can not be filled with DSP needs to be seen with a bit of nuance.

First, we should distinguish between nulls that are produced when a wave interacts with its own reflection (SBIR / Allison effect or room modes) and when a wave interacts with another speaker. For example, subwoofer and woofer, or between two subwoofers. In the latter case, the null can be ameliorated by adjusting the phase of the band producing the cancellation. You can do this with the usual DSP tools of delays, polarity inversion, and all-pass filters.

In the former case, where a wave interacts with its own reflection, increasing the amplitude of the null frequency with PEQ also increases the amplitude of the reflection. A perfect null is caused by two waves that are 180deg out-of-phase. But not all nulls are perfectly 180deg out of phase. In fact, if it's between 180deg +/- 90deg (i.e. 90 to 270deg) out-of-phase, it will produce a null that is gradually less severe, with the most severe null being at 180deg. The null at 180deg is called a zero in the Z-plane unit circle, and zeros require infinite energy to fill. As you approach the zero, the energy requirements exponentially go up. How well you can successfully do this depends on how close the null approaches 180deg, and how much headroom you have in your speakers/amps.

 

[digitalfrost]

I can't answer your question, but you are absolutely crazy and I love it. WTF are those subs? Making KH310 look small like that is a feat.

 

[DVDdoug]

+12dB requires 16 times as much power! If you aren't pushing your amplifier (or speakers) you might get-away with it.

If you reduce your overall level by 9dB, you are boosting by only +3dB (double the power) but you still have to listen a LOT lower than your maximum.

 

[levimax]

+12dB requires 16 times as much power! If you aren't pushing your amplifier (or speakers) you might get-away with it.

If you reduce your overall level by 9dB, you are boosting by only +3dB (double the power) but you still have to listen a LOT lower than your maximum.

Even though I set max boost to 12 dB REW only used 9 dB which is why I lowered the level by 9 dB. With 200 Hz crossover to the subs (60dB per Octave Linear phase) all the extra gain is handled by the subs and since the KH 310 can "only" put out ~100 dB, the subs, even with 9 dB attenuation, can still put out more than that so no no loss of maximum volume.

The reason I am curious about this is that I have struggled with LF response in this room and tried 1, 2, and 4 subs (sub crawling) with various MSO programs, REW no boost, manual corrections, and DIRAC DLBC. None worked particularly well. Using REW with boost and 2 co-located subs seems to be working MUCH better and outside of needing 2 powerful subs, this REW with boost method is fast and easy and retains stereo bass.

 

[levimax]

I can't answer your question, but you are absolutely crazy and I love it. WTF are those subs? Making KH310 look small like that is a feat.

These are GSG "Full Monty" subs assembled from a flat pack with 18" LaVoce pro drivers and a crown 1300 watt per channel amp. I had some small sealed SVS subs but was disappointed with their performance so I decided to take Hoffman's iron law to the extreme and see how large subs perform. While huge they are not expensive if you assemble them and their performance is on a different level proportional to their size

 

[amirm]

Nulls are never perfect so you can indeed fill them until you run out of amplifier power and or speaker dynamic range. In my tests of Dirac, it too attempts to do that. It will pull all peaks down and then boost the whole range up. This actually caused my powerful amplifier run out of juice. So it can be done, you just have to be judicious about it.

 

[Mort]

Genelec GLM 5.x does the same

 

[Leeken]

Nulls are never perfect so you can indeed fill them until you run out of amplifier power and or speaker dynamic range. In my tests of Dirac, it too attempts to do that. It will pull all peaks down and then boost the whole range up. This actually caused my powerful amplifier run out of juice. So it can be done, you just have to be judicious about it.

Have you ever tested on your analyser a dac and amp running a heavily eq’d signal? Always wondered if it would change the distortion characteristics etc?

 

[levimax]

The rule that nulls can not be filled with DSP needs to be seen with a bit of nuance.

First, we should distinguish between nulls that are produced when a wave interacts with its own reflection (SBIR / Allison effect or room modes) and when a wave interacts with another speaker. For example, subwoofer and woofer, or between two subwoofers. In the latter case, the null can be ameliorated by adjusting the phase of the band producing the cancellation. You can do this with the usual DSP tools of delays, polarity inversion, and all-pass filters.

In the former case, where a wave interacts with its own reflection, increasing the amplitude of the null frequency with PEQ also increases the amplitude of the reflection. A perfect null is caused by two waves that are 180deg out-of-phase. But not all nulls are perfectly 180deg out of phase. In fact, if it's between 180deg +/- 90deg (i.e. 90 to 270deg) out-of-phase, it will produce a null that is gradually less severe, with the most severe null being at 180deg. The null at 180deg is called a zero in the Z-plane unit circle, and zeros require infinite energy to fill. As you approach the zero, the energy requirements exponentially go up. How well you can successfully do this depends on how close the null approaches 180deg, and how much headroom you have in your speakers/amps.

I have played around with manual filters trying to boost dips in the past and have run into some that respond well and some that don't at all and some that are in between. I am sure that is due to how far out of phase they are as you mention. What I have never seen before are "dips" that are filled almost perfectly with auto generated filters from REW. In this case when REW sees a 6 dB dip in the measurement if creates a 6 dB boost in the filter and the 6 dB boost actually results in the dip being filled in based on an actual measurement (Not just in a simulation). REW didn't just fill in one "dip" but around 4 dips all filled almost perfectly. That is why I was wondering if something else was at play with the large subs or more direct sound or something else or maybe I am just lucky in this case.

 

[Elgrosso]

Impressive subs, and measurements!
I assume the measurements shows L+R, subs in mono but still EQed independently?
If so I would not be so surprised, they fill each other dips.
independent L&R measurement might not look so clean while still working well once combined.
Do you have bass panning around at some frequencies?
But they’re so closed it might not be an issue.

 

[Arindal]

In my tests of Dirac, it too attempts to do that. It will pull all peaks down and then boost the whole range up. This actually caused my powerful amplifier run out of juice.

It would be interesting to have a measurement and listening test at the exact position where the reflections or resonances at this particular frequency are not causing a null (like at the listening position), but a perfect addition after you have tried to filled the null.

Me thinks, this would be interesting, as in this case you can expect a resonance peak of roughly +20dB. Place a cupboard there with precious china, and you have fun.

So what is going on that the dips are all "filled" with DSP?

Depends on the root cause of the null. If it is a pressure minimum of a standing wave, and you try to correct it, you can bet on the fact that you will get crazy resonance boost at a different position, I would try to take additional measurements like 1/4 Lambda away after correction, in this case 1.44m or .77m away

 

[Jiraya369]

These are GSG "Full Monty" subs assembled from a flat pack with 18" LaVoce pro drivers and a crown 1300 watt per channel amp. I had some small sealed SVS subs but was disappointed with their performance so I decided to take Hoffman's iron law to the extreme and see how large subs perform. While huge they are not expensive if you assemble them and their performance is on a different level proportional to their size

im curious, how much did it cost per sub?

 

[Defi]

Nulls are never perfect so you can indeed fill them until you run out of amplifier power and or speaker dynamic range. In my tests of Dirac, it too attempts to do that. It will pull all peaks down and then boost the whole range up. This actually caused my powerful amplifier run out of juice. So it can be done, you just have to be judicious about it.

My first attempts with Dirac (Polk L800 and Anthem MCA225) brought my amp to its knees when trying to correct for a null somwhere in the sub bass region. I realize now that I was asking way too much of it. Now with subs they have vastly more reach down there and it's just another day at the office to punch through it.

I find my subs to be so ludicrously overpowered I can't imagine what 18" drivers in a 4' tall cabinet are needed for but I would love to hear them haha. MSO had to lop off like 15-20dB to match up to my mains and that's without touching the gain knob on the back of the subs.

 

[levimax]

im curious, how much did it cost per sub?

Flat pack is $314, driver $400, supplies (paint, glue, etc.) $60 so $774 each. I picked up flat packs at factory, if you had them shipped it would add $100 each to cost. I picked up amp on ebay for $500. So ~$1,000 each with amplification, freight would be extra. The kits go together well with just glue but you do need some clamps. https://shop.gsgad.com/collections/...ies/products/roundover-full-marty-single-unit I am not affilated with GSG in any way.

 

[levimax]

Impressive subs, and measurements!
I assume the measurements shows L+R, subs in mono but still EQed independently?
If so I would not be so surprised, they fill each other dips.
independent L&R measurement might not look so clean while still working well once combined.
Do you have bass panning around at some frequencies?
But they’re so closed it might not be an issue.

Subs are full range stereo. Measurements are MMM at LP both L+R combined. L and R speakers have similar (dip and peaks at same frequency) but not identical response if measured separately. EQ filter is "mono" in this case. I tried it individual but it was not a lot different and mono MMM measurements and corrections are quick and easy.

I have tried a lot of different EQ techniques over the years and never had results like this which is why I started this thread. I can't help but think something is going on with the big subs in the mid field (6.3 ft / 2 meters) ?

 

[Sokel]

Subs are full range stereo. Measurements are MMM at LP both L+R combined. L and R speakers have similar (dip and peaks at same frequency) but not identical response if measured separately. EQ filter is "mono" in this case. I tried it individual but it was not a lot different and mono MMM measurements and corrections are quick and easy.

I have tried a lot of different EQ techniques over the years and never had results like this which is why I started this thread. I can't help but think something is going on with the big subs in the mid field (6.3 ft / 2 meters) ?

If you're about to use your subs as woofers, crossed up there, then you should measure and EQ them (room EQ, not speaker, that must be done outside with a cross so high) individually.
A sanity check for both is ok, but the goal of stereo bass needs good individual output.

 

[DVDdoug]

I forgot about something... Dips and bumps are different at different places in the room. (You know that since you used the MMM method.) At the same frequency you could have a dip at one place and a bump at another. Correcting a dip at your main listening position could create a nasty (or nastier) bump somewhere else. Maybe not a problem for you but something to consider.

Note that dips tend to be less annoying than boomy-bumps.

These are GSG "Full Monty" subs assembled from a flat pack with 18" LaVoce pro drivers and a crown 1300 watt per channel amp.

In a home environment, you probably have PLENTY of headroom and a 12dB boost is probably OK. Normally, I feel like +6dB is about the maximum "reasonable".

Have you ever tested on your analyser a dac and amp running a heavily eq’d signal? Always wondered if it would change the distortion characteristics

DACs (and most audio formats) are limited to 0dBFS. And many (most?) digital recordings are normalized/maximized for 0dB peaks. If you boost (digitally) you can get digital clipping. For that reason, most digital EQs have a "preamp" feature (normally used to attenuate rather than amplify). For example, if you boost the bass by +6dB, you attenuate the overall level by 6dB to compensate. If you have a setup with a digital volume control, that also gives you headroom when you're not at 100% volume, but it's still good practice to use the preamp feature so you never clip your DAC.


P.S.
I LOVE big speakers!!! My DIY subs are about the same size as yours but they "only" have 15-inch woofers.

 

[middlemarch]

Just chiming in that the above has generally been my experience as well. I typically allow 6db individual and total boost when doing below Schroeder EQ. But I gotta say, I do so with a guilty conscience and hoping none of the folks who know better are looking over my shoulder! Just goes to show we don’t live in a black and white world, just shades of grey, like the current Seattle weather…

 

[NotInYourHead]

I got annoyingly thin and undefined vocals at my MLP, as well as a lack of bass impact, which were caused by severe comb filtering between 100-200Hz and a valley dead center in low bass (too sparse room modes). This is the case for both speakers. Thankfully there're no infinitely deep nulls.

After reading this post, I pulled together the courage and made an outlandish EQ in REW with a max boost of 10db and 12db individual boost between 35-200Hz. I also generated a response from the predicted results for reference.

The filters worked and fixed most of the problems I described. Most surprisingly, the after MMM measurements show that the response at MLP tracks extremely close to the predicted one, except for a dip around 170-180Hz, probably SBIR-related. Of course, like Amirm said, there're no free meals, I have to give up a -10db headroom. To reach my previous listening levels I have to push my amp, DAC and small tower speakers far harder. The small 6.5in woofers have a lot more visible excursion now compared to pre-EQ. I don't have luxury of dual 18" subwoofers like OP unfortunately. One day

Maybe I'll try a different target curve (currently Harman Curve) and look into room treatment next to deal with comb filtering.