Treating SBIR?

[atledreier]

Due to my room layout I am pretty limited with regards to speaker placement.

I have treated my corners with chunks floor to ceiling, and sidewall 1st reflection points with 4" 2x4 Rockwool.

I still have pretty severe midbass issues, and I think they may be SBIR issues. Since I can't move the speakers much, is there an efficient way to treat SBIR? The back of my speakers are only 8-10" from the front wall, and about 3' from the sidewalls.

The wallbounce calculator predict a 8dB dip at 140ish Hz.

This is for HT, so the screen pretty much dictate where the speakers go. The speakers are behind the screen.

A measurement that is pretty current, and we see the 100hz dip pretty easily.

[SAC]

You may have issues with SBIR, but from a frequency response exhibiting beau coup comb filtering through its response there is no way to determine the cause from that limited perspective.

You need to generate proper cumulative spectral decay/waterfall plots for , say 0-200 Hz, in order to identify contributory modal issues, as well as properly generated ETC plots for each speaker in the hopes of perhaps identifying the various energies that are superposing to create the polar lobing spatial dispersion anomalies that show up as the significant comb filtering in the frequency response.

As far as SBIR, if that is indeed the cause of the issue (which does not address the rest of the anomalies), and repositioning is not an option, then absorption on the wall surface sufficient to dissipate the frequencies at issue is about the only passive solution.

[atledreier]

That plot is a single speaker driven, so I don't see where the comb filtering comes in?

Ok, so how do I generate 'proper ETC' plots? I have REW and TrueRTA available.
I will provide exact measurements of the room and speaker/LP placement when i come back home from work, along with any current plots you need.

yeah, my room is THAT bad, even with the mentioned treatments....

[SAC]

Comb filtering is a result of the superposition (summing/combining) of out of phase energies such as reflections and diffraction.

The direct signal combines with energy from other secondary and virtual sources such as reflections and diffraction off boundaries to cause the comb filtering, much in the same way that SBIR works at low frequencies.

You don't need 2 speakers to cause comb filtering. The interaction of a speaker within a room provides more than enough indirect sources of energy.

But the frequency response is unable to show the individual sources that interact destructively, it only shows the destructive 'result' - hence the use of the ETC which shows the individual contributory signals.

And as previously mentioned, while the ETC provides insight in to the specular energy behavior, you need a better display such as the waterfall/cumulative spectral decay to provide additional insight in the non-specular modal behavior of the environment.

[atledreier]

Ok, I will provide the ETC for right and left speakers. Even with symmetrical placement the FR differ greatly, so there's something fishy going on....

[SAC]

???

I'll try once more.

The waterfall will provide additional insight into the LF behavior of the environment in order to help determine how much of the LF issue is modal and how much is SBIR.

But the comb filtering above the modal region, unless your speakers are really that bad!, is a result of the superposition of the direct and reflected/diffracted signals.

Nothing "fishy" at all! You just have to move from a position of knowing something is happening, to a position where you have sufficient information to identify and address the contributory causal reflections.

And the waterfall and ETC responses are the two tools most adept at assisting in this process.

[atledreier]

So the fact that the left and right speakers have very different FR when placed symetrically isn't a concern? I've tried swapping the speakers around, and it's position, not the speakers themselves.

The plot I posted is my main LCR speakers, Klipsch KL650THX. The 40-200 range is now covered by DIY midbass modules, Eminence 3015LF drivers in a Fb=40Hz slotported enclosure.

The left seems to correspond more with simulations. Wallbounce calc predict a dip at 140-ish, and room modes predict a dip at 70ish from the floor/ceiling node. The right is just all over the place.

Red is right, purple is left, cyan is center

[SAC]

Originally Posted by atledreier

So the fact that the left and right speakers have very different FR when placed symetrically isn't a concern? I've tried swapping the speakers around, and it's position, not the speakers themselves.

First, go back and re-read what has been written.
I have not yet addressed the nature of the speakers themselves.

Modal response, SBIR and polar lobing caused by superposition that appears as comb filtering in the frequency response are all 'position' related in a manner of speaking. But that does not imply that all of the problems can be solved simply by relocation!

The room reinforced modal response is a distribution within the room of areas of reinforced low frequency high and low pressure zones. these can be moderated (assuming we are not rebuilding the room!) by moving the speaker, damping the modal energy or by relocating the listening position.

The waterfall can help identify resonances due to room modes.

SBIR is a reinforcement of wavelengths within 1/4 wavelength by reflections off boundary surfaces where the frequencies below the 1/4 wavelength effectively couple and add. Above that you may get cancellation. Th Thomas Barefoot bounce calculator can help identify the frequencies affected by SBIR due to placement relative to boundaries.

In the frequencies above the modal range where the wavelengths become less than the dimensions of the room, you have specular reflections. This means that they will effect a different part of the frequency spectrum than the modal or SBIR issues.

When the various 'sources' of these energies, be they real or virtual, meaning that they al emanate originally from the direct signal, but some of he energy is 're-transmited' via diffractive or reflective means, they superpose and combine destructively.

And as you will note that your frequency response exhibits significant comb filtering over its entire range, it is safe to suspect this occurrence above the modal and SBIR frequencies! The ETC response can help identify the reflective sources contributing to that part of the response.

Thus over the full range of your response, you need to examine all three causal sources and address each one, as I suspect (but I cannot say with certainty given the very small amount of information that the limited frequency response provides as that measurement is unable to provide information about the various reflections and resonances with respect to time, and all it provides is a summary that indicates that 'allot of stuff', as mentioned above, is going on.

So simply again repeating that you have response issues with the 3 speakers does not moderate the above possible contributory factors.

Your job is to use each of the three tools listed above to evaluate the performance at each location and to determine to what degree each of the factors is contributory, and then to deal with them accordingly.

I am not sure what is so confusing about this.

I am not attempting to tell you where the reflections are sourced, nor to begin to tell you the modal distribution in the room relative to the speaker placement and to the listening position. Nor have I attempted to calculate the SBIR factors based upon the relative placement near the boundaries.

You get to do that!

But I can tell you that, barring any potential contributory anomalous response of the speakers/crossovers. etc., themselves - which I am assuming to be pristine - however accurate or fallacious that may be, that the above issues are going to be the determinate factors at play.

And we have suggested the various measurement devices available from which to gain further insight into each behavior.

[atledreier]

I appreciate the time and effort to try to make me understand how this works and what the measurements tell me.

Ok, first some basic information about the room.

Length 520cm
width 390cm
height 230cm

Woofers (40-200Hz) are 59cm wide and 50cm deep. Driver center is 105cm from sidewall and 60cm from front wall (back of the box is only 10cm from the front wall).

All four corners are treated with a 110cm tall chunks, and 10cm panel straddling the corner on top of that, so effectively treated from floor to ceiling in every corner.

Listening position is 160cm from back wall, so 360cm from the front wall.


So, on to the measurements:

Right speaker:






Left speaker:







Ok, so.... What do this tell me?

I see the obvious lengthwise node of 33Hz, in both FR, waterfall and decay plots, for both speakers. I also see the 2nd harmonic of that node, at 66Hz. That ringing seems to decrease in frequency though. It starts out at 66Hz and rings out to just over 60Hz.

I also see the expected dip at 74Hz from the floor/ceiling node. The extremely long ringning of that node is unexpected, though. It looks like hum? I know I have some 50Hz hum, and a 2nd harmonic at 100Hz. I have no idea where this 70ish Hz hum could come from.

The 90Hz dip in the right speaker may be a sidewall 1/4wave cancellation, as it is not apparent in the left speaker.

At 140Hz I see the 1/4wave dip from the rear wall. This is less apparent in the right speaker.


The ETC is unfamiliar to me. I understand I'm looking at the response in time, and I guess the peaks are reflections. I do not understand the double impulse, though. Is that a measurement anomaly?

I find the rear wall reflection at 9.5ms (3,2m).

I see a powerful reflection at 13.6Hz in the right speaker. That equals 4,7meters. Not sure where that could come from. Maybe the opposite sidewall?

Is this the right way to read the ETC? What else can I read from this plot?

[SAC]

Something is amiss if that is indeed the "impulse response" as labeled. One sure indication of a problem is that the impulse response is not comprised of all positive values.

And yes, if there are actually two energy spikes near the origin (on the questionable IR), that would indeed wreak havoc in the system response!

If it is the ETC, I will admit it looks strange and the two spikes at inception are problematic.

We can help you , but I don't have the time to provide comprehensive modal, SBIR and ET evaluation - especially as the ETC appears to have issues. Plus the x-axis origin being set to the apparent direct arrival Ld rather than at he signal incept time is of no help. And has the loopback compensation for hardware propagation delay been configured?

Also, for a room ~22 ft long, an ETC of ~60 ms total time and also one windowed down to ~25 ms will help us to focus in on the details.

You really need to first walk through the software user and help websites and insure that it is configured and working properly. We cannot reasonably serve as the software troubleshooting function. and their site's do a very good job.

And spend a bit of time becoming familiar with the various modal analysis tools (if you need help reliably identifying likely modes) (see real traps and John Brandt's sites for great tools) as well as the Thomas Barefoot reflection calculator for help with SBIR calculations, as simply guessing what might be the cause(s) is the recipe for a snipe hunt. And those snipe are dastardly little boogers to catch...

I apologize for not having links to all of the resources right at hand. They are posted frequently and a search should provide them. And also please realize that while we can help interpret results, I can't guarantee that any of us have the free time to do the investigation from first principles.

[atledreier]

Ok, I'll try to redo the measurement and see if I can do it right.

[bpape]

If the speakers are placed symmetrically in the room and your FR and IR are that different, something is in the path. Where is your sub in relation to the mains and how large is it? Is there an equipment rack on one side?

Any chance of a basic sketch of the room so we could try to see what might be causing the difference?

Bryan

[atledreier]

Ok, these looks better, for some reason...

Edit: Bryan, I've added pictures of the room, and you have the measurements. I think the difference is the wall construction. Front wall and left side is panel over concrete, right sidewall is a light inner wall. Uninsulated wood panelling over light studs.

My subs are not used for these measurements, but they are large LLT sonotubes, RL-p15 drivers in the back of the room. The pictures show them up front, they have since moved back.

This time I added the Klipsh to the mix. Did both separate and combined response.








Front right: (No, it's not at all finished there....)


Rear right:


Front left:

Rear left:

[bpape]

It appears to me that a portion of your issue is at the xover point. Have you tried reversing the phase on your sub? Is the xover set to MAX on the sub (or disabled) and then the processor xover is set the same for both mains and sub?

Are the boxes I see under the screen acting as stands? Those will have their own little nasties to them - not to mention allowing the speakers to move around.

The difference in wall construction can certainly make a difference. Left being what is predicted since it's basically a concreate wall and the other side acting as a membrane with unknown depth.

Bryan

[atledreier]

Those boxes was a very temporary setup, for breaking in the Klipsches when I got them.

I have the klipsch speakers on top of the Eminence woofers with sorbothane pads between.

I also saw the issue around the x-over. But as you can see, the issue is there even when only one of the speakers is playing. The combined FR looks identical to the separate ones. I added the combine FR plot for reference. You see the deep 100hz null, and I'm guessing a 2nd harmonic of that at 200, compounded by som other smaller dips around that area and you get that broad valley.

Three of these, with the horn toward the outside and the woofers aligned with the Eminense driver.

[SAC]

Regarding the initial double peak in the impulse response (or ETC - for which the x-axis time scale is FAR too large to determine any details. 150 ms corresponds to~170 feet of travel!)

But in any event, do not run the sub and speaker at the same time! The ETCs must be done for one speaker only. After wards, if you have the ability they can be displayed sis by side with the same scaling for AB analysis - I would avoid overlaying them.

I think what you are seeing is the two direct signals of the two speakers - the two acoustic origins that are not aligned in time. The answer is to determine the time offset at the listening position, as it will be valid at most on only the horizontal plane if they are stacked vertically - otherwise, only at one point or in one vertical plane is placed side by side - and adjust the delay setting until both sources are coincident per the ETC. Note, you CANNOT do this by ear, as once you get somewhat close, the Haas effect dominates and the two signals are effectively smeared and combined into one, and you lack the hearing resolution to separate the 2 closely arriving signals and to ultimately finish the alignment - oh, and lest you think "well, if I can't distinguish the difference by ear, it isn't significant" - that is Wrong! It is precisely this interval that is critical and requires precision microsecond delay to eliminate what Dick Heyser termed "time smear distortion" And that small difference still manifests itself in polar lobing and comb filtering!

Also, one more small but important note. You must use the ETC response for the alignment of the acoustic origins, as physical alignment does not normally correspond to the acoustic origin. So you align the drivers by the test display, not by what looks right. This becomes increasingly important in horns, as the acoustic center changes with frequency (they suffer acoustical astigmatism) as the origin is not located at the diaphragm, but it is located 'somewhere' in the throat. ...Issues addressed at length by the illustrious Don Keele.

This is a significant issue with the Heritage Klipsch - which are otherwise superb speakers. And this is specifically the issue where the individual drivers MUST be aligned using active crossovers with precision delay compensation between drivers - especially in models such as the KHorn where the difference in acoustic origins nears 9 feet between woofer and midrange.!!! So for anyone running the Heritage line, dump the absurdly priced passive mods and get active crossovers with adjustable precision microsecond delay.

[bpape]

Humor me and try reversing the phase. I understand that it shows up with only one speaker but that is a function of the xover assuming you're crossing around 100Hz or so.

Yes - there are other problems, just trying to pick at it a bit at a time.

I also think there is probably a baffle issue going on with the Klipsch mounted directly on top of the massive subwoofer cabinet which can give additional gain where it's not desirable in addition to the items SAC is talking about in the time domain.

Bryan

[atledreier]

I was running only one speaker at a time in all of the above, except where I stated the combined frequency response.
The crossover is 24dB/oct Butterworth alignment at 200Hz.
I do not see what difference the phase would have with only one speaker driver playing, but I'll hook it up again (I haven't moved the microphone) and see what happens.

I have very accurate phase and delay adjustments available to me, so if I am pointed in the right direction I'm all ears (and measuring equipment).

The following is a zoomed in ETC plot for ONE 15" Eminence ONLY. No sub, no KL650.



This one is ONE Klipsch KL-650 ONLY.

[SAC]

2 quick points as I am running out (and late )

the first referring to Brian's point:

One 'half of a crossover unit' running out of polarity (180 degrees out of phase - but phase and polarity indicate 2 separate issues =- polarity being which direction the mass moves upon the initial application of the EM stimulus, phase simply being a measure of the time offset) can cause the rwo halves to sum 'incorrectly', the result being a null instead of the desired flat summed response.

and the second:
As far as the ETC - which looks very weird...goes...

If the two initial peaks are correct, you have 2 very strong early arriving signals and another 3 within the first ~3ms or ~3.4 feet- a real problem. (It scares me to think that part of that may be the nearfield direct response where you are experiencing the drivers individually rather than in a distance invariant far field response. If that is the case, (and all of this is speculative as we need to have confidence in the measurements) I might suggest that the speakers should not be used in the nearfield.

This is normally what the ETC of multiple misaligned array elements looks like.

[atledreier]

Edit: SAC posted while I set up my post.
3.4feet corresponds pretty well with the distance from my driver to the front wall and back again. That round trip is 120cm.


And here's what happened when I inverted one of the speakers:
Green is inverted. The response was identical, wether I inverted the bass ot the Klipsch.

[bpape]

And that was JUST the Klipsch in that last post?

[atledreier]

No, that last post was with Klipsch and Eminence. The responses were identical regardless of polarity with a single speaker active.

I have a DCX2496 and a separate channel for the klipsch and Eminence, so it's really easy to test, adjust and mute individual speakers. I'm headed for bed now, but if there's any measurement that would help diagnose this, or any pictures/ descriptions I'd be glad to help out. This is driving me nuts. Everyone keep telling me it sounds great, but the measurements look 'weird' and 'not good' and noone seems to be able to diagnose it.

I really appreciate the time you guys take to try to help me with this. You Bryan have helped me alot in the past too, on HTS and AVS.

BTW, I just had a though. Should I remove all treatments and do a completely 'raw' measurement? Would that change anything?

Is there anything in my test setup that could be causing anomalies? I don't have another mic, but I have the RS meter, that should be helpful for the frequencies in question.

[SAC]

Something is wrong with the impulse and ETC (if that is indeed what it is) measurements. Please note and correctly label the impulse, log squared and/or ETC responses. While they are similar in some ways, they are not interchangeable.

The impulse is all positive (which means the dynamic drivers only respond in the positive half of the plane and never move back beyond the resting position (which makes absolutely no sense) and the ETC looks extremely smoothed.

This needs to be corrected. Its pointless to go forward interpreting obviously skewed measurement results as if they were valid.

And if it is correct , you have at least 3 substantial 'reflections' of near equal gain to the direct signal in the first 3 ms that is going to cause ALL sorts of comb filtering as well as imaging, localization and intelligibility problems.

And can you configure the program to display the gain in dB SPL? This dB FS stuff that has little relation to acoustic sound pressure levels is getting old (a limitation is a few of the consumer platforms).

In fact, it is rare to see a plot with such dramatic early anomalies in the worst of untreated rooms! And as I mentioned earlier, that looks like a plot of an unaligned array or dipoles placed near a boundary!!!

And this is after treatment??? Hmmm... Obviously something is not right...and this is frustrating not being able to be there and to fuss with a few things to verify lots of small details that might contribute to the weirdness.

But at this point I am putting slightly more money on issues existing in the configuration and process of the measurements rather than advocating redoing treatment - as I have many more questions than answers at this point.



So either something is far off in the measurement configuration and/or process, or we have some real problems in River City.

[atledreier]

A quick and dirty Sketchup of the room. This is to scale. I have a second sofa that I didn't draw, a 2 seater like the front row, as close as i can get it behind the first.

The supports that run floor to ceiling is the frame the screen is mounted on. I drew the subs as well, even though they are not part of the problem.

The white stuff in the corner is the basstraps and straddling panels.

Regarding the measuring setup I have no idea. I'll try a different soundcard and use the RS meter as a microphone and see if that changes anything.