Bass arriving out of phase at the listening position

[ernestcarl]

Just in case, you may also want to try this simple phase-only EQ adjustment:

*magnitude ripple is due to the centering and tap settings -- this is in order to limit FIR delay.

It doesn't attempt to completely match the phases; however, the range should be close enough around +- 60 degrees from each other. It might even possibly give a more favorable summed response over your all pass filter, with the caveat of some potential audible pre-ringing, and this correction mainly only working at the measured MLP. My suspicion is that it's probably not really going to be a significant issue with actual music listening tests -- but, only way to know is to A/B test this...

 

[edechamps]

Can you post a waterfall or a spectrogram?

Not sure what parameters you'd want. If you're interested you can download the raw REW data file from this post.

I am not going to model the acoustic panels and just use one estimated absorption coefficient for the entire room (at least for the time being). They are a PITA to model and specify in Mathematica.

Yeah don't worry about that - the problem still occurs with all panels removed anyway, so they're unlikely to have anything to do with it.

I thought perhaps you might be interested in having the raw REW measurement data with all panels removed from the room, as the individual responses are a bit different without the panels. It is attached to this post. These measurements are the "ground truth" you can compare your simulation against. Here are corresponding pictures of the setup:

Can you also give me your speakers and listening (ear) heights?

The speakers are Genelec 8030As and their acoustic axis are 1.16m above the floor. Listening ear height is the same 1.16m.

 

[hmt]

Played around with MSO and an all pass 2nd order. Unfortunaltely you trade in a steep dip there. I did not get any better result with a higher Q all pass or diffeent center frequencies.

 

[edechamps]

Played around with MSO and an all pass 2nd order. Unfortunaltely you trade in a steep dip there. I did not get any better result with a higher Q all pass or diffeent center frequencies.

MSO is precisely what I used to design the workaround allpass filter I'm currently using I didn't get that dip though - not sure why.

This is the original measurement:

And with the filter I'm currently using, an All-Pass Second Order at 93.4 Hz with Q=4.1:

I (ab)used MSO's own optimization algorithm to generate the allpass parameters for me. I used a very high target SPL to trick it into maximizing the response.

 

[hmt]

Maybe it's because I only used a low Q. BTW I have a similar problem with my mains. I only have the bass managed measurement (LR4 at 90Hz) but I should be clear what happens. They also do show some cancellations but they are very narrow. In my case the strange things happen in the R channel. I suspect it is some reflection as it also sounds very strange and does not really seem to come out of the right speaker.

I was able to fix this with some High Q APF (10) at 92 and 120Hz. Green is with APF and red is without. I do not use this APF anymore though. It introduced horrible ringing also to the left channels (because the APF basically mimics that kind of swing) and as I also had to apply this kind of APF to the subs (to proper sum them with the mains) I also had this ringing there. It sounded a rrom mode with long decay (the decay of the APF was also seen in the waterfall plots). I ended up using a higher crossover of 120Hz. My subs are 4 Arendals optimized with MSO and are pretty flat to 200Hz because of this.

 

[edechamps]

I suspect it is some reflection as it also sounds very strange and does not really seem to come out of the right speaker.

Yeah this is also something I noticed - something seems "off" with the soundstage in the bass. It could be unrelated to this issue - eh, it could very well be entirely in my head for all I know. It does seem consistent with the general idea of reflections overpowering the main sound though. The auditory system could be getting confused by conflicting localization cues from the reflected vs. direct sound. This is partly why I'm trying to look for a solution that addresses the physical root cause (i.e. gets rid of the reflections somehow).

 

[hmt]

In my case that is also the case with one speaker playing. Maybe my R speaker has that same issue as the L. But in case of L there are some reflextions arriving and making up for the cancellation but mess up the bass soundstage.

Your measurements also show that swing in the spectrogram. Maybe in your case the more direct sound is cancelled when L and R are plying but the reflections are not so you only get them and thus they also sound strange.

Maybe it helps to play a sine and walk arround and try to find a wall or corner where it is really loud.

 

[Peluvius]

Have you eliminated the room from your issue (by taking the same measurement in another location with the same equipment)?

Even if it is an early reflection, how does that show up in such a way in your measurements?

I don't understand how a direct reading from the front speakers can be giving this result either which is why I would try the whole test in another location.

 

[edechamps]

Have you eliminated the room from your issue (by taking the same measurement in another location with the same equipment)?

This is equipment that I've used for the past 12 years in 6 different locations. I never had this issue before. (Although to be fair, previously I was only using it near field.)

I think it's pretty much a given that it's the room, or some large acoustical object in the room (besides the equipment). I have eliminated pretty much every other potential cause through equipment swaps.

Even if it is an early reflection, how does that show up in such a way in your measurements? I don't understand how a direct reading from the front speakers can be giving this result either which is why I would try the whole test in another location.

Well, one promising theory that was suggested earlier in this thread is that there are so too many reflections that arrive in-phase between themselves, but out of phase with the direct sound. As a result the sound field at the listening position is dominated by these out-of-phase reflections which end up destructively interfering with the sound from the other speaker. To me this does make sense physically. Granted, this phenomenon seems unlikely to happen. It would appear this room hates me

 

[hmt]

One question. When you moved the mic for for another measurement did you also time align the speakers again?

 

[Jimi Floyd]

This seems to suggest that all speakers agree on phase except Front Left which is the one causing trouble… Incidentally, it's also on the front left channel that excess group delay is the highest, peaking at 26 ms at 86 Hz.

Are you 100% sure Front Left is correctly wired INSIDE? Genelecs are trustworthy, but a mistake can always happen....

 

[hmt]

I think that can be ruled out. If the woofer was out of phase not only a dip an 100Hz would show up.

I think maybe a diagonal reflection with the part of the room where the entrance is may be the reason. It is asymmetric and may cause some weirdness.

 

[JohnPM]

Asymmetric absorbtion in REW does not cause asymmetric results:

The model you used was of a symmetric room with the listener along the centre line and the speakers symmetrically placed either side of it, try moving everything off the axis of symmetry.

 

[edechamps]

Are you 100% sure Front Left is correctly wired INSIDE? Genelecs are trustworthy, but a mistake can always happen....

As I explained in the first post, I've already tried swapping speaker pairs. That's not the problem.

 

[sam_adams]

I find this most disturbing:

The large reflection under the 1ms mark is probably the desk.

There seems to be a lot of late arriving energy from the left speaker, also:

The reflections at 3 ms and 4 ms might be ceiling bounce. Is there a ceiling absorber in place?

 

[ernestcarl]

Well, one promising theory that was suggested earlier in this thread is that there are so too many reflections that arrive in-phase between themselves, but out of phase with the direct sound. As a result the sound field at the listening position is dominated by these out-of-phase reflections which end up destructively interfering with the sound from the other speaker.

Interestingly enough, your left speaker below 200 100 Hz actually has better transient response / GD behavior so it may be the other channels instead that are "misbehaving more" in the bass region:

 

[edechamps]

There seems to be a lot of late arriving energy from the left speaker, also:

View attachment 231852

Yeah, @mitchco mentioned this as well in a private chat. The fact that the left and right channels spend most of their time on opposite sides of the step response is indeed intriguing. I might dig into this some more.

The reflections at 3 ms and 4 ms might be ceiling bounce. Is there a ceiling absorber in place?

No (and it wouldn't be easy to add one, although it's something I could consider). I'm sceptical the ceiling would be the problem here since the reflection path through the ceiling is the same length between both speakers.

Interestingly enough, your left speaker below 200Hz actually has better transient response / GD behavior so it may be the other channels instead that are "misbehaving more" in the bass region:

This is why I've always found it difficult to interpret "exotic" measurements (excess group delay, wavelet, etc.). Depending on the measurements you look at, you can prove everything and its opposite

 

[Flaesh]

The model you used was of a symmetric room with the listener along the centre line and the speakers symmetrically placed either side of it, try moving everything off the axis of symmetry.

My point was that asymmetric absorbtion coefficient does not affect FR [in this model]. Here is the reference with left 0.00 right 0.90:

 

[ernestcarl]

This is why I've always found it difficult to interpret "exotic" measurements (excess group delay, wavelet, etc.). Depending on the measurements you look at, you can prove everything and its opposite

It's quite difficult to characterize as the magnitude and time behavior is very much frequency dependent, and looks different depending on what graphical view and settings you're using.

But we can sort of compare your in-room measured left and right responses to an "idealized" anechoic min-phase version copy of the 8030c (generated from Amir's on-axis Klippel data):

The measured left channel has its own issues, but overall below 100Hz (or maybe more precisely below 70-80 Hz) it is indeed more "linear" or closer to the min-phase ideal -- I'm speaking mainly about the time domain behavior.

 

[Flaesh]

Depending on the measurements you look at, you can prove everything and its opposite

For example, wavelet or D50 have an obvious theoretical ideal. While treshold of deviations' audibility needs further examination.
People say the bad wavelet can be heard listening MATT-test.