Post your room modes correction PEQ settings

[ernestcarl]

Looking a little closer at the list of PEQs I posted previously, and as suggested by @daftcombo & @Wayne A. Pflughaupt, there seems to be a pair of filters in each channel that particularly stand out as candidates for simplification/consolidation:

I used REW to model and predict what would happen -- the left channel proved particularly challenging to manually simplify.

Did a couple of before and after MMM test measurements for the left channel using different settings...

I've decided to keep the consolidated new filter for the sub, but will simply leave the left channel as is... I'm not really hearing any big difference other than a slight improvement with both sets of filters turned ON vs disabling the left channel pair of PEQs that appear to be "cancelling" out. The set of filters for the left channel is not causing any impairment as far as I can tell from the predictive modelling, measurements, and listening tests I've done. Distortion is changed a little post EQ at the usual max and reference volume level sweep tests that I do.

The deep null at 300Hz makes it look like the %THD at that point is higher than it really is..

*Pointer of cursor is set at 300Hz peak so those harmonic distortion numbers should reflect accurately.

That peak around 300Hz is not an audible "problem" esp. at the levels that I usually listen at -- i.e. below my relative reference vol. level in JRiver which is 80% [0.0dB] or -10dBFS.

I always do these distortion tests before and after EQ to verify.

I added +3dB volume level boost to visualize what would happen (clipping) as I run out headroom:

JRiver's soft clipping protection seems little able to mitigate distortion effects of audio clipping from an external source (REW 0dBFS signal).

Of course, if I wanted to increase the volume without clipping, I could just raise the KH120's internal amplifier gain knob. Eh, nope! I'm fine with my set maximum volume level, thank you.

*I suspect overall noise may be just a tad higher because of the way I've routed audio to JRiver > U7 mII analog out > miniDSP analog input > unbalanced speaker out -- I don't hear extra noise or hissing, though. I've tested using a balanced connection before using a pro audio interface and got slightly better results. Not enough to feel the need to upgrade...

 

[ernestcarl]

You might also want to wade through my epic “Minimal EQ” article.

I do like your minimal EQ approach/philosophy, BTW. You probably would find more tolerable the fewer PEQs I have for this same desk in it's standing setup position (sit-stand desk) -- 6 for the sub channel, 2 for the right, and 6 for the left -- Well, maybe not minimal enough in your book, of course. But it does work here very well and sounds just as good. It doesn't make much sense to stick with one's aesthetically pleasing measurements if in the end it still sounds worse off, after all.

 

[dasdoing]

Initial subwoofer volume level is set high to fill in huge null between 50 to 90Hz -- there's also a gigantic null below ~25Hz so a high pass is applied at 23Hz even though the sub is capable of going down much deeper.

have you ever tryed turning the sub around so that the woofer is almost kissing the wall? this should help to get rid of speaker boundary interference which is probably the cause of that huge dip

 

[ernestcarl]

have you ever tryed turning the sub around so that the woofer is almost kissing the wall? this should help to get rid of speaker boundary interference which is probably the cause of that huge dip

Unfortunately, the crossbar framing of the table prevents me from pushing the sub close enough to the wall. I have tried turning it around facing the wall, but it excited audible resonances/vibration across entire the front wall. I have also placed the the sub behind me, but that required using a much, much lower crossover point which I did not like. Placing it offset to the right or left makes the sub localizable, corners and along the side walls even worse. Multi-sub is the only way to fully get rid of that null BUT there’s barely space left in this room — and there’s very little room moving the desk. Oh, well... it really is the best compromise.

 

[ernestcarl]

Found myelf interminably bored not having any internet for a good few hours yesterday (service issue), so I read Wayne's write-ups on his approach to EQ again (yes, I had it saved -- didn't realize that I already, in fact, read it years ago).

I added a 1dB of boost around the 120Hz xo region

Experimenting a bit, and I was able to reduce my PEQs from 8 to 6 with comparable results. Uh, actually, I was also able to do it with 4 as well -- but the effect in the null area and crossover region was worse -- no es bueno. I'd have to rework the PEQs around the xo area for every single channel, and even that was not guaranteed to work better overall. Too much work for what's supposed to be just a simple test.

Using less PEQs (i.e. less control) the GD, decay, and tail-end of of sub bass increased a little. The hole around 60-90Hz is a little less deep (partly due to the 1dB boost I added) but also a tad wider.

It's easy to A/B (almost seamlessly) between presets using JRiver (better if there were some shortcut key functionality available):

Don't mind the differences in dB SPL as it depends on which calibration mic and meter I use, as well as position.

After two hours of back and forth between the two sets of presets, I am left undecided. Yes, there is a very small, almost perceptible difference. I cannot, however, say which sounds better.

It's way easier to determine what sounds better to me by switching between my 'diffuse surround' and 'front focused' presets (change in the rear channel vol. level) -- choice depends on the mood, intent, and material played.

So what now? There seems to be very little difference between either.

In the thread article linked, Wayne said a 110dB window is a much better indication as to how we hear (well, I personally just switch to psychoacoustic smoothing for that) so here's a little visual comparison using the new SUB PEQs:

MMM RTA

*It could have looked better without that 1dB boost I added around the crossover area.

BTW, got the fun idea of using GIFs from BYRTT.

 

[Absolute]

@ernestcarl Since you're using JRiver, have you considered contacting @mitchco for some Audiolense filters? You can send him measurements and he will look at it and generate filters for you to import into JRiver.
Audiolense will integrate and timealign the subwoofer better than anything else on the market.

I don't know how much the service costs, but it's probably worth a shot if you're experimenting to that degree. Kudos for your efforts, I enjoy reading stuff like that!

 

[ernestcarl]

@ernestcarl Since you're using JRiver, have you considered contacting @mitchco for some Audiolense filters? You can send him measurements and he will look at it and generate filters for you to import into JRiver.
Audiolense will integrate and timealign the subwoofer better than anything else on the market.

I don't know how much the service costs, but it's probably worth a shot if you're experimenting to that degree. Kudos for your efforts, I enjoy reading stuff like that!

I know about Audiolense, but I think this is good enough for now. Much like Sal, I rather prefer time spent enjoying the music than mucking about with more software. Though I keep finding myself grabbing my measurement mic instead! I need to stop already.

 

[Absolute]

I know about Audiolense, but I think this is good enough for now. Much like Sal, I rather prefer time spent enjoying the music than mucking about with more software. Though I keep finding myself grabbing my measurement mic instead! I need to stop already.

I know all about it, which is why I'm suggesting to keep away from the software itself, but rather send in the measurements you're already taking and getting the DSP-Doctor himself to create filters in the software and send you the convolution-files ready to load unto JRiver.

 

[ernestcarl]

I know all about it, which is why I'm suggesting to keep away from the software itself, but rather send in the measurements you're already taking and getting the DSP-Doctor himself to create filters in the software and send you the convolution-files ready to load unto JRiver.

I have used convolution before actually for a couple of months. My problem with convolution is the delay I get with streaming video services (which I find I prefer due to convenience). This is not a problem with music or video files in JRiver where I can easily adjust lipsyc. But I simply cannot stand it with something like Netflix. It could very well be the case that my computer's processor is not powerful enough as I'm only using an older i5.

 

[Absolute]

I have used convolution before actually for a couple of months. My problem with convolution is the delay I get with streaming video services (which I find I prefer due to convenience). This is not a problem with music or video files in JRiver where I can easily adjust lipsyc. But I simply cannot stand it with something like Netflix. It could very well be the case that my computer's processor is not powerful enough as I'm only using an older i5.

I understand fully, and that's precisely the reason I'm not using Audiolense myself - even though I have it installed on my computer.

The delay comes from FIR filters and not from the CPU speed. Some people say you can have one setting without FIR filters (or shorter ones) for TV/Netflix watching and one for music, I've found that it's too much effort to switch.

I use manual PEQs myself but will be the first to admit that whatever you can do with PEQs will never touch a Dirac/AUdiolense/Accurate software in the bass. This has to do with variable timing/phase in the lower frequencies that combat the cancellations in order to provide a smoother response.
Above the room's transition area it's sometimes a different story.

 

[dasdoing]

I use FIR IRs for a long time and I am still confused about the silent part before the actual IR starts. it is there probably because of the same reason a IR anlyse needs a left window extension!?

Btw: about delay and Netflix (Amazon Prime way better btw hehe): my TV can adjust the delay, never played with it since my system only adds 10ms latency.

 

[Absolute]

I use FIR IRs for a long time and I am still confused about the silent part before the actual IR starts. it is there probably because of the same reason a IR anlyse needs a left window extension!?

Btw: about delay and Netflix (Amazon Prime way better btw hehe): my TV can adjust the delay, never played with it since my system only adds 10ms latency.

The silent part before the IR will show some activity called pre-ringing if you're using FIR filters. Basically it shows that there's some noise/signal starting before the impulse itself, completely unavoidable if you're trying to match the peak of the woofer impulse with the peak of the tweeter impulse due to the vastly longer time the woofer needs to peak a slower pulse.
Phase correction is the same thing, I assume. It can be audible, but I don't know more than that.

If you're only adding 10 ms delay the FIR filter is very short. Audiolense have adjustable filter lengths, but lower latency means fewer taps in the bass to correct. If I remember correctly, I had over 600 ms delay with "true time domain"-filter with Audiolense.
My Oled can delay video, but not that much

 

[dasdoing]

The silent part before the IR will show some activity called pre-ringing if you're using FIR filters. Basically it shows that there's some noise/signal starting before the impulse itself, completely unavoidable if you're trying to match the peak of the woofer impulse with the peak of the tweeter impulse due to the vastly longer time the woofer needs to peak a slower pulse.
Phase correction is the same thing, I assume. It can be audible, but I don't know more than that.

If you're only adding 10 ms delay the FIR filter is very short. Audiolense have adjustable filter lengths, but lower latency means fewer taps in the bass to correct. If I remember correctly, I had over 600 ms delay with "true time domain"-filter with Audiolense.
My Oled can delay video, but not that much

you are right, I have more then that of latency
I am not talking about the preringing. my filter has 148ms of absolute silence before the preringing

 

[Absolute]

you are right, I have more then that of latency
I am not talking about the preringing. my filter has 148ms of absolute silence before the preringing

Ah, then that's probably the time delay from the start of the process (signal start) until the microphone captures the signal.

Usually you just normalize the window to the impulse minus 10 ms so that you start 10 ms before the impulse with the window.

 

[ernestcarl]

PEQs will never touch a Dirac/AUdiolense/Accurate software in the bass. This has to do with variable timing/phase in the lower frequencies that combat the cancellations in order to provide a smoother response. Above the room's transition area it's sometimes a different story.

I might try out some of these options again in the future, at the moment I'm already quite satisfied with what I was able to achieve with basic PEQ corrections (as well as some room placement work).

Here are just some additional graphs of the same setup:

That, apparently, infinite delay and FR amplitude null isn't really as bad as it looks.

The MMM RTA and spectogram graphs already indicate that volume is reduced and amplitude peak energy just delayed. While initially somewhat wide looking, the audible 'impairment' is rather minimal.

Just a little quiter...

I could apply even more of a boost around that area to level it out a bit -- did that before -- but it caused other audible problems e.g. resonances and just plain bad sound which affected everything else. Which is also why I like to look at the peak MMM RTA to see that no particular point in the response produces too much excess energy.

 

[Absolute]

I might try out some of these options again in the future, at the moment I'm already quite satisfied with what I was able to achieve with basic PEQ corrections (as well as some room placement work).

Here are just some additional graphs of the same setup:

View attachment 74319
That, apparently, infinite delay and FR amplitude null isn't really as bad as it looks.

The MMM RTA and spectogram graphs already indicate that volume is reduced and amplitude peak energy just delayed. While initially somewhat wide looking, the audible 'impairment' is rather minimal.

View attachment 74320

View attachment 74321

Just a little quiter...

I could apply even more of a boost around that area to level it out a bit -- did that before -- but it caused other audible problems e.g. resonances and just plain bad sound which affected everything else. Which is also why I like to look at the peak MMM RTA to see that no particular point in the response produces too much excess energy.

You could try to reduce the output in one of the speakers at that frequency to see if it's a phase cancellation due to a reflection off a wall that interfers with the other speaker.

I have the same problem and it goes away by doing that. Boosting won't help, but you'll make a terrible racket elsewhere in the room at that frequency where there's no cancellation

 

[mitchco]

I have used convolution before actually for a couple of months. My problem with convolution is the delay I get with streaming video services (which I find I prefer due to convenience). This is not a problem with music or video files in JRiver where I can easily adjust lipsyc. But I simply cannot stand it with something like Netflix. It could very well be the case that my computer's processor is not powerful enough as I'm only using an older i5.

Looking good! Don't forget, you can also generate minimum phase filters with Audiolense and host them in Audiolense convolver (or other convolvers). Then use VB-Cable (or Rogue Amoeba if on Mac) to route Netflix, YouTube or whatever through convolution and get very low latency even with 65,536 tap filters. This is what I use when watching Netflix (using the Win10 native app) and works perfectly - no lipsync issues. I use an old i5 as well and no issues with a stereo triamp system. Also, with Audiolense or Acourate, the dip at 75 Hz would be taken care of, which as we already know PEQ's and other types of room eq cannot. Good luck and have fun!

 

[ernestcarl]

You could try to reduce the output in one of the speakers at that frequency to see if it's a phase cancellation due to a reflection off a wall that interfers with the other speaker.

I have the same problem and it goes away by doing that. Boosting won't help, but you'll make a terrible racket elsewhere in the room at that frequency where there's no cancellation

Crossover is quite high: 120Hz so it's coming from the single sub and the L&R speakers contribute very little, if anything, there in the nodal region.

 

[ernestcarl]

Looking good! Don't forget, you can also generate minimum phase filters with Audiolense and host them in Audiolense convolver (or other convolvers). Then use VB-Cable (or Rogue Amoeba if on Mac) to route Netflix, YouTube or whatever through convolution and get very low latency even with 65,536 tap filters. This is what I use when watching Netflix (using the Win10 native app) and works perfectly - no lipsync issues. I use an old i5 as well and no issues with a stereo triamp system. Also, with Audiolense or Acourate, the dip at 75 Hz would be taken care of, which as we already know PEQ's and other types of room eq cannot. Good luck and have fun!

Youtube definitely was a problem... Great to know you had success with the low latency option. Yep, I've suspected that type of dip may be possible to 'fix' much better with something like Audiolense. Just wasn't sure how much better. Would definitely revisit this option again in the future, Mitch!

 

[ernestcarl]

Found this interesting video about room modes (though not a discussion in the problematic bass area we're mostly interested in) with Geoff Martin:

A Basic Introduction to Room Acoustics: Part 2 (Room Modes)