Jump to content

My living room "make over" (aka the "surrounded by bass" project)


SME

Recommended Posts

Those measurements look to be quite good. One unusual thing is they go out to 180 degrees off-axis instead of 90 degrees or so, that is a bit weird. That might be Harman's own measurements? They do 360 degrees by 10 degree intervals in their own anechoic chamber on both the horizontal and vertical plane. 

Link to comment
Share on other sites

Center - have you tried blinding off visually (dark room or curtain) and listen, do you still hear sound coming from above? 

 

Like mentioned already, I think visuals have some impact here.

 

However, it is possible that the speaker itself can be partly localized, also in the height direction.

Source material can be a major factor - some sources simply sound better with the speakers, including center, disappearing much better.

 

If we look at how hearing can detect height, we can learn what can cause the speaker to be detectable.

Frequency response is one factor to height localization, but that requires a reference - a sound effect moving from center and upwards can be quite realistic, but a static voice will not have any reference, so we can not actually know where it is.

If we hear a sound, and want to find out if this sound comes from above, we simply tilt our head, then we detect changes in frequency response as we move the head.

 

Reflections and delayed sound (impulse response) can also reveal the sound source.

When working on the center speaker, I found that those factors were most important.

I experimented with different frequency responses around 6k-10k to try to move the sound source down, but that did not work well.

What did work, was to make the off-axis response smoother and improve the impulse response.

In this process I found that linear faults outside the pass-band of the lf driver had much more impact on the sound than I had assumed initially.

Link to comment
Share on other sites

 

Too bad those don't show a lot of detail.  One thing that's unclear is how to interpret the angles.  If they pivoted from the same point for the horizontal and vertical measurements, I would expect that +/- 180 degrees would be identical across both plots, but they aren't.  The horn vertical pattern looks like it's 90 degrees, not 120.  Maybe I got that info wrong?  The patterns both look a bit tighter in the bass than I'd expect but are probably OK.  And lastly, the data has 1/3rd octave smoothing, and that may have suppressed a good bit of ripple from diffraction.

 

Also, I don't see anything exceptional about the data compared to SEOS-based designs.  The vertical window at the crossover looks pretty tight.  That pinches the power response in that region of frequencies, a phenomenon that is also visible in the spinorama data.  It might be premature to judge without seeing better data, but I think the SEOS horns look better in the horizontal.  The situation in the vertical is debatable.  There is a wider vertical window at the crossover, but the vertical directivity in most SEOS designs hits a local minimum somewhere not far above the crossover because of the loss of pattern by the shorter horn.  It's a different kind of compromise than what the M2 appears to do.

 

So as far as the M2 and its secret sauce is concerned, I would say it's mostly down to the active crossover and optimized voicing.  And possibly the greater coverage and higher power response, where that's desirable.

 

Those measurements look to be quite good. One unusual thing is they go out to 180 degrees off-axis instead of 90 degrees or so, that is a bit weird. That might be Harman's own measurements? They do 360 degrees by 10 degree intervals in their own anechoic chamber on both the horizontal and vertical plane. 

 

It's a bit weird for horns as they tend to send very little energy backwards, but 360 degree measurements are quite useful for full-range speakers as they let you anticipate what kind of sound will be emitted to the rear and plan your acoustic design accordingly.  The measurements shown indicate angle resolution of 5 degrees and 1/3rd octave smoothing.

Link to comment
Share on other sites

Center - have you tried blinding off visually (dark room or curtain) and listen, do you still hear sound coming from above? 

 

Like mentioned already, I think visuals have some impact here.

 

However, it is possible that the speaker itself can be partly localized, also in the height direction.

Source material can be a major factor - some sources simply sound better with the speakers, including center, disappearing much better.

 

If we look at how hearing can detect height, we can learn what can cause the speaker to be detectable.

Frequency response is one factor to height localization, but that requires a reference - a sound effect moving from center and upwards can be quite realistic, but a static voice will not have any reference, so we can not actually know where it is.

If we hear a sound, and want to find out if this sound comes from above, we simply tilt our head, then we detect changes in frequency response as we move the head.

 

Reflections and delayed sound (impulse response) can also reveal the sound source.

When working on the center speaker, I found that those factors were most important.

I experimented with different frequency responses around 6k-10k to try to move the sound source down, but that did not work well.

What did work, was to make the off-axis response smoother and improve the impulse response.

In this process I found that linear faults outside the pass-band of the lf driver had much more impact on the sound than I had assumed initially.

 

Thanks for your comments.  Apart from the visual reference, I'm most inclined to blame the strong early ceiling reflection for the apparent localization above the TV.  However, it deserves mention that there is negligible energy in the 6-10k range as part of that reflection, so it's not like I'm "hearing the ceiling reflection" directly.  Instead, I think the strong comb filtering is preventing the speaker from properly disappearing, and my brain is gravitating toward the visual cue.

 

Tomorrow, I'll be installing a 2 foot by 4 foot panel with 6 inch thick 1.6 lb/cuft Knauf Insulation Acoustical Smooth Board, with 6" air gap, right below that reflection.  That'll also hopefully nuke the ceiling/floor secondary reflection and the ceiling/rear-wall secondary reflection for all the seats.  (fingers crossed)  The preview of the impact of this treatment is in my data posted here.  In terms of freq response, it's a night and day difference for the low mid-range.  That should clean up the dialog immensely, and hopefully cure the localization problem.

Link to comment
Share on other sites

OK.  My new absorber is installed.  The center does sound way better, even without any EQ change.  There's a lot less harshness, as I expected.  I still have a gaping hole in the upper bass.    There's also still some "above the TV" localization, more particularly with voices than music, but I'm not going to worry about it for now.  It still sounds way better than any center speaker/configuration I've had, and the seat-to-seat coverage is great.  It'll probably get better as I improve on the EQ.

 

What surprised me is that the sound of the left and right have changed quite a bit as well.  The smoothed response measurements show only minor changes, mostly in the bass, but the sound is definitely different.  Unfortunately, I seem to have lost quite a bit of envelopment with 2 channel music.  I might just need to re-adjust the EQ somewhat, perhaps to restore some brightness.  Still, that panel really sucks sound out of the room.  It's quite apparent with conversation, and the measurement data does indicate a significant decrease in decay time.  That kinda sucks, because I'm planning to add two more of these big ceiling panels for the left and right speakers once they are installed in the same orientation as the center.

 

Oh well.  I'll have to do more listening and see what EQ tweaks bring.

Link to comment
Share on other sites

A 2-by-4, I am guessing that may be similar to a 60cm x 120cm. 

 

I looked back at the pictures in the first post, nice room and system by the way.

Looks like there was no ceiling absorption at all before you mounted this.

It is then quite understandable that you experience a huge difference.

 

Difference and improvement are not the same.

Absorption in the ceiling usually is an improvement, but it will sound different, it will sound more dead. 

But envelopment and space can be restored, in a more controlled way.

From the pictures I see diffusors on the back wall, and there is not much absorption in the room, so I would assume it sounds quite lively.

 

I would mount all the ceiling absorption you have planned, and then try to add reflections/envelopment after that, if it sounds too dead/flat/soundstage-up-front.

 

Would be interesting to see the .mdat, but I understand if you do not want to post it.

Link to comment
Share on other sites

A 2-by-4, I am guessing that may be similar to a 60cm x 120cm. 

 

I looked back at the pictures in the first post, nice room and system by the way.

Looks like there was no ceiling absorption at all before you mounted this.

It is then quite understandable that you experience a huge difference.

 

Difference and improvement are not the same.

Absorption in the ceiling usually is an improvement, but it will sound different, it will sound more dead. 

But envelopment and space can be restored, in a more controlled way.

From the pictures I see diffusors on the back wall, and there is not much absorption in the room, so I would assume it sounds quite lively.

 

I would mount all the ceiling absorption you have planned, and then try to add reflections/envelopment after that, if it sounds too dead/flat/soundstage-up-front.

 

Would be interesting to see the .mdat, but I understand if you do not want to post it.

 

Each surround speaker has a 4"/100 mm OC703 2x4 foot panel on the ceiling above it with no air gap.   (There's only 4 inches between the speakers and ceiling anyway.)  The black things under the diffusers on the rear wall are actually a pair of 2x4 foot absorbers made with 6" thick 1.6 lb/cuft Knauf Insulation Acoustical Smooth Board, just like the ceiling.  Including the OC703 panels adjacent to the left and right speaker, I think I have a fair amount of broadband absorption in the room.  Nowhere near as much as you, but already more than I'd like for my room goals.  I have a vision for how I want to finish out the front stage, but I haven't finalized the design yet.  Once it's time to install the left and right in their permanent configurations, I will be ready.

 

The panel seems to have shaved off roughly 10% of the decay times, which is not a lot, but is noticeable.  The panel really seems to have changed the sound of the left and right, but it's taking me a while to adapt and listen to enough stuff to decide whether I need to adjust the mid range EQ.  I did some experimenting with the EQ last night and today, and did not make any useful progress.  I did improve the center channel a bit, as it needed some adjustment post-treatment.

 

Finally, I gave both speakers some upper bass boost.  I restored the 1.5 dB or so lost by the left and right to the panel.  I gave the center channel a broad boost in the suck-out region of up to 8 dB.  The result is nowhere near as boomy as I thought it would be, and I think a lot of the bass from the center is getting absorbed early.  I've only had time to listen to a little bit of multichannel music and a few samples of dialog, but it sounds really good now and seems to integrate more seamlessly with the left and right.  I'll have to watch my headroom until I address the suck-out acoustically, but I can't wait to see a movie tomorrow.

 

The last thing I did today was experimented by moving the diffusers that are on top of the left and right horns to the shelf above the TV and adjacent to the center channel.  Unfortunately, this configuration sounded bad.  The upper mids and highs sounded very artificial and fatiguing.  I will have to do further study to figure out why they sounded so bad in those positions and which speakers they affected the most.  I suspect it was my left and right that were getting screwed up.  I had hoped to cover much of the exposed front wall with diffusers, but I might need to nix that idea if it causes too much energy to arrive too early.  I will investigate this soon.

Link to comment
Share on other sites

Looking at the pictures again, I agree about the absorption - by no means "very little absorption", still, there are large untreated surfaces (ceiling).

 

There is more absorption in my Room2, but in the other room there is only 8x 5cm thick (2") 60x120cm suspended panels, and 4 60x120cm (2'x4') 50mm panels on the back wall.

And this room still sounds better, overall.

Which suggests that it is the acoustic properties of the room that matters, and not the amount of absorption - no surprise for logical thinking people..

 

Another interesting thing I see from the pictures is the back wall.

Center section has absorption - I wolud do the same.

But corners are open - they are essentially non-reflecting.

Both my rooms have untreated/reflective walls in the back wall corners, because this is where the energy from the horn speakers is reflected back into the room.

If this area is damped or non-reflective, you may end up with too little reflected energy and it will sound too dead.

 

Just some thoughts, and perhaps you get some ideas for experiments with the diffusers.

If you do, I look forward to learn what you experience; does it support my hypothesis, or maybe not.

Link to comment
Share on other sites

Ehh, and I forgot to mention the four GIK Acoustics 244 bass traps.  I believe they use 4" of the same material in my ceiling panel/trap, and I ordered mine with the resonant membrane (mass loaded vinyl?) for slightly increased bass absorption and rejection of most stuff above 250 Hz or so.  All four are literally standing on the floor and they each have a polystyrene diffuser on top.  It's a very inelegant arrangement, but it works.  And I intend it to be temporary.  Two of the traps are on the right side-wall (pictured).  The left side-wall (not pictured) has a large window in it, and one of the traps fills the space between the window and the door to the back yard of the house.  The fourth trap straddles the front left corner, the only real corner in the room.

 

The absorption treatment on my rear wall is critical because the seats are so close to it.  The MLP is 12 inches from the treatment and 20 inches from the wall itself.  The reflection without the black panels in place is very strong.  Unfortunately, I think the rear wall diffusers are too close to the MLP, and some mid range reflecting off them is diffracting down to the listeners.  So I'll likely end up with more broadband absorption there too.

 

In the long term, I hope to add custom-built diffuser/absorber hybrid panels roughly underneath the surrounds on each side wall.  The surrounds are just slightly past 90 degrees on each side.  If I go as wide as I'd like, part of one of them will need to be built onto the door.  This leads to a weird design constraint that the door must be operable despite the panels.  I'm not sure how that one will work, yet.  But if I can get to a width of four feet or so, I can achieve diffusion down to a few hundred Hz.  I think that's desirable with these speakers as they appear to retain plenty of directivity down there due to their relatively wide baffles.  Those locations also happen to be on-axis of the opposing front speakers, so lots of energy will reach them.  Some of that will be diffusely reflected back into the room and the rest will spread into the dining room and kitchen, to be returned at a later time.  The room is 19 feet (5.8 m) wide, so these reflections are not especially damaging.

 

I did get to watch a couple of movies after the tweaks.  The center sounds very good now.  The integration between it and the left and right are superb.  I don't even bother to sit in the middle seat, and the front sound stage is still seamless.  I think it sounds cleaner than the left and right, which still suffer from some ceiling and floor reflection.  If I listen carefully and the source is dry enough, I can actually hear the side wall reflections.  They contribute a kind of warm tail to the sound.  It's very subtle, but I think it is a positive thing.  It makes mono dialog tracks more pleasing and realistic, especially for stuff that was recorded on-location.  A lot of newer films send some of this sound to the other channels, but older movies often tended to just throw a mono on-location recording into the center.

 

I've also had some time to adapt to music listening with the new panel installed.  The slight bass increase helped a lot, but I'm not sure yet if I want to adjust anything else.  A lot of music seems like it could use a tad more brightness.  I miss the little bit of ambience I lost.  However, most film material sounds balanced if not a bit bright.  One thing I am hearing on a lot of films now is a clear layering of the front sound stage.  Most dialog and effects tend to be quite forward, sometimes pushing beyond the front of the screen.  The score tends to occupy a space a few feet beyond the screen and sometimes extending into the distance.  Background sound effects and ambiance tend fall where they should for the scene.  I hear plenty of depth separation

with two channel music, but the use of depth seems to be much more deliberate and effective with multichannel content.   I think being able to hear this clearly is a sign that I'm doing something right.   ;)

Link to comment
Share on other sites

Extremely interesting impressions on your room treatment experimentations. 

 

I want to share a little discovery I recently noticed that is sort of disturbing. I moved into a new place, and the bedroom acoustics are exceptionally bad to the point of almost unlistenable. It was so ridiculous that somehow there was a huge dip near the crossover between the mid and woofer and it was better to invert the polarity on the woofer in that bedroom. It was subjectively and objectively better (the dip was gone). That made it better, but the speaker still sounded much worse compared to the old room it was in (and the old room's acoustics weren't that great). I didn't get more time to tinker with the sound to deal with the room acoustics.

 

However, it didn't take too long (a week or two) for me to get used to the sound and find it sounding good! It didn't sound amazing, or even great, but it was a positive impression instead of a very negative impression like when I first moved in. It seems like our ears are highly adaptable, and I can't trust my own ears! I hope you have some way of comparing against a reference to reduce the effects our ears could have in adapting to a particular sound after long exposure. 

Link to comment
Share on other sites

Extremely interesting impressions on your room treatment experimentations. 

 

I want to share a little discovery I recently noticed that is sort of disturbing. I moved into a new place, and the bedroom acoustics are exceptionally bad to the point of almost unlistenable. It was so ridiculous that somehow there was a huge dip near the crossover between the mid and woofer and it was better to invert the polarity on the woofer in that bedroom. It was subjectively and objectively better (the dip was gone). That made it better, but the speaker still sounded much worse compared to the old room it was in (and the old room's acoustics weren't that great). I didn't get more time to tinker with the sound to deal with the room acoustics.

 

However, it didn't take too long (a week or two) for me to get used to the sound and find it sounding good! It didn't sound amazing, or even great, but it was a positive impression instead of a very negative impression like when I first moved in. It seems like our ears are highly adaptable, and I can't trust my own ears! I hope you have some way of comparing against a reference to reduce the effects our ears could have in adapting to a particular sound after long exposure. 

 

Ears and brain are both very adaptive for better or for worse, but there are limits.  Poor acoustics degrade sound quality substantially.  It may still be enjoyable to listen to, but it won't sound nearly as good.  Your brain will be unable to resolve many details, which will go missing.  Other aspects of the sound may be excessively accentuated.  In audio production, good room acoustics is often considered to be essentially to getting content that translates well to other systems.  Without a good neutral room, engineers can't trust their ears to tell them what they are putting into the mix or masters.

Link to comment
Share on other sites

Extremely interesting impressions on your room treatment experimentations. 

 

I want to share a little discovery I recently noticed that is sort of disturbing. I moved into a new place, and the bedroom acoustics are exceptionally bad to the point of almost unlistenable. It was so ridiculous that somehow there was a huge dip near the crossover between the mid and woofer and it was better to invert the polarity on the woofer in that bedroom. It was subjectively and objectively better (the dip was gone). That made it better, but the speaker still sounded much worse compared to the old room it was in (and the old room's acoustics weren't that great). I didn't get more time to tinker with the sound to deal with the room acoustics.

 

However, it didn't take too long (a week or two) for me to get used to the sound and find it sounding good! It didn't sound amazing, or even great, but it was a positive impression instead of a very negative impression like when I first moved in. It seems like our ears are highly adaptable, and I can't trust my own ears! I hope you have some way of comparing against a reference to reduce the effects our ears could have in adapting to a particular sound after long exposure. 

 

 

I have a similar problem. For some reason my speakers get more quite only after 2-3min of turning them on. I mean it seems to be 10 to 15dB drop without fail. The other interesting piece of information in my room is that my wife often walks over to the stereo right before it gets quite. I'm wondering if these two facts are related in some way.. hmmm.

  • Like 1
Link to comment
Share on other sites

So I've been working on re-EQing the left and right speaker after installing the ceiling panel above the center.  That one panel had a big impact on the sound for stereo music, even though the impact on the (1/3rd or 1/6th octave) smoothed frequency responses for those channels was quite small.  As such, I'm revising my target curve to try to get it to sound as good as it did before I added the panel.  I am doing it by ear as I did before, listening to passages of music and iteratively making adjustments.  The panel removed a little bit of energy above 2 kHz, but it took proportionately more in the 500-1k range.  That's a big bummer, because I was already thin in that range due to the high directivity of the crossover and the woofer in cabinet.  I'm at the point where I have to compromise more than I'd like between the direct sound response and the reflected (ambient) sound response.  The latter is particularly important for 2 channel listening because much of the ambient content in a stereo recording is out-of-phase and is heard more in the reflected than the direct sound field of the room.  A lot of music even mixes stuff out of phase on purpose for a phantom surround effect.

 

With a 0.5 dB boost to the treble and a little bit of ramp up from 500-1kHz, I got most of the envelopment back, but the accentuation of the mids in the direct response is apparent if I ramp too much.  I have to choose a compromise.  Hopefully the situation will improve once I relocate them and add more treatments.  Or it'll have to wait until I'm ready to research possible DSP solutions to the problem.

 

I have also been investigating the center channel reflections more thoroughly.  The rear wall (20" back with a panel 12" back) still has significant acoustic influence.  I'm not too worried about its effect on the sound, but it presents another problem.  The rear wall reflection, although attenuated considerably, still has a considerable influence on the smoothed frequency response.  Furthermore, the effects are very sensitive to changes in measurement position, even changes as small as one inch.  Smoothing with 1/3rd octave is quite coarse, so to see broad changes of 1-2 dB or more with 3 inches of movement is remarkable.  This says a lot about the problems of doing any kind of room EQ using a single measurement position.  Furthermore, it also appears that frequency response with 1/3rd octave (or similar) smoothing is an unreliable metric for good sound.

 

Think about it.  I spend considerable time listening and iteratively adjusting my response shape by fractions of dB to get the best voice using filters with 1/3rd octave resolution.  How is that not all ruined by me moving my moving my head 3 inches?  Good question because I can't readily identify changes when I move my head.  Indeed, what little change I do hear is more likely to be caused by crosstalk between the speakers.  Maybe I should try listening to pink noise in mono.  I believe that my brain uses information from other room reflections to reconstruct the original sound.  Those reflections that arrive from other angles will be altered by the acoustics of my seat differently.  Thus, speaker off-axis response is an essential performance parameter even for speakers optimized for a single listener.  This much has already been observed and documented by Toole and Harman, but it deserves repeating because it considerably complicates the problem of room EQ.

 

I plan to revisit a lot of these insights when I get back to doing experiments with room EQ again.  One major limitation of pretty much all room EQ systems is that they can't reliably select an optimal target curve for the system.  At least some of them allow the user to alter the target curve.  To the extent that compromises are required, this selection may even necessarily be subjective, depending on what kind of content the listener prefers, for example.  This issue really kind of blows up the whole notion of reference.  At the very least, we need a psychoacoustic model that takes into account how we process early reflections.

  • Like 1
Link to comment
Share on other sites

I've been meaning to post an update as I've done a bit more in-depth investigation of my speakers, room acoustics, and target curve preferences.

 

The most interesting find is that the low frequency "reflection" energy I was seeing in my center channel response, between 2-3 ms or so after the initial impulse, is not all reflection nor does it all originate from the rear wall.  Much of that energy arises due to edge diffraction effects of the speaker and speaker baffle.  Complicating matters further, diffraction energy is typically inverted relative to direct sound and reflections of direct sound.  This means their effects cancel if they occur simultaneously, which makes it even harder to discern what's going on without getting a lot of data.  So I decided to take an initial stab here:

 

post-1549-0-31335700-1475738677_thumb.png

 

These are measurements of my center channel installed in-room with 4 ms window to exclude most other acoustic effects and 1/6th octave smoothing to improve legibility.  I ran these measurements with crossovers in place, but I removed all other signal shaping from the woofers so that I could see their native acoustic response.  The various red, green, and blue traces were measured somewhere above the front edge of the seat of my sofa.  I wanted to get far enough away from the rear wall to reduce its influence.  All locations were very approximate, and I didn't attempt to keep distance constant or do any level matching.  I just wanted to get a qualitative idea of how the direct speaker radiation pattern varies within the room, to the best of my ability to measure this.  The different colors correspond to different horizontal angles: blue at 0 degrees, red at (roughly) 15 degrees, and green at (roughly) 30 degrees.  The lighter the color, the higher in altitude the measurement.  Measurements were taken at approximately ear-level and roughly 8", 16", and 24" above that.  Because the measurements are forward of the listeners, what the listeners here is probably best represented by the second most dark set of lines.  The orange trace is one last measurement, taken a bit closer to and almost directly on-axis of the woofer itself, putting it 10 degrees vertically off-axis of the horn, which is angled down toward the listeners.

 

I think it's very note-worthy that the high frequency response fits so tightly into a +/-1 dB window across what is essentially my seating area plus a pretty wide vertical arc.  The response shape is the result of my hand-tuning the response by ear with less attention given to actual measurements.  The dip at 2 kHz is a bit weird.  I'm using the same filters on all three horns for now, and I tuned these by ear while listening in stereo.  In the measurements I have, that dip appears for the center but not the others, so I think it's a variation related to something in the driver and/or crossover.

 

Of course, the real stuff of interest is in the woofer's response, below 950 Hz.  For most of the measurements, there is a pretty nasty looking dip somewhere in the 600-700 Hz range, but it does shift slightly between locations.  Somewhat less obvious is that there are peaks, one at round-about 850 Hz and another broader one at 350 Hz.  Below there, the responses roll off due to a combination of baffle step effect and 110 Hz crossover.  And then there are those locations, mostly the higher altitude measurements that are closer to being vertically on-axis of the speaker, where the response is relatively smooth.  Curiously the dead on-axis measurement does not offer the smoothest response.

 

One thing people might ask at this point is: why not take the speaker outside and measure it to get this kind of data?  Well, I won't argue that doing so could provide some useful data, but only to a point.  Indeed, I already have vertical polar measurements, posted here, that I did indoors, albeit with a lot of absorber panels surrounding the measurement space to try to maximize my usable window time.  Reflections did contaminate those measurements somewhat, mostly at the extremes, so I didn't trust them much.  But maybe I should have because they do indeed show dips at 600ish Hz within the vertical window (5-20 degrees or so) that my measurements were taken.  (Admittedly, this is much easier to see in the data that is not normalized, which I have not posted.)   Nevertheless, my vertical polars only show what my speaker was doing in my artificially anechoic measurement environment not in their installed location in the room, which is what I ultimately care about.  And even outdoor measurements can't help me here.

 

One thing that worries me is that the diffraction phenomena appear to be a lot more severe in my in-room measurements than in the vertical polar measurements.  The variations I measured in-room were quite dramatic and probably very audible.  Thankfully, all my listeners have about the same vertical angle to the speakers, and the variation with respect to horizontal angle is minimal.  I can probably *mostly* fix this problem with DSP.  However, it looks like I'll give up a lot of headroom in the process and also put a big peak in the power response.  Or not.  Indeed, some dip in that region may be present in the overall power response, and if that's the case, then a bit of boost will also smooth the power response.

 

In any case, seeing this dramatic vertical variation in response and how deep these dips can get is making me have second thoughts about putting the woofers for the left and right speakers up high.  At least I think I can help the response of the center channel by extending its baffle horizontally to the width of the TV.  That may only serve to move the dip down in frequency, to say 200 Hz or so, but at least it should be a lot more manageable there.  Unfortunately, I have a hard time anticipating what the left and right will do.  Unlike with the center, the vertical angle between the left and right speakers and each listening position will vary by quite a bit at the extremes.  This will likely limit what compensation can be accomplished with DSP over a wide listening area.

  • Like 1
Link to comment
Share on other sites

For most of the measurements, there is a pretty nasty looking dip somewhere in the 600-700 Hz range, but it does shift slightly between locations.  Somewhat less obvious is that there are peaks, one at round-about 850 Hz and another broader one at 350 Hz.  Below there, the responses roll off due to a combination of baffle step effect and 110 Hz crossover.  

 

I wouldn't trust the data from a 4ms window here, at least certainly not for the 350Hz and below part and the 600Hz might be questionable as well.

  • Like 1
Link to comment
Share on other sites

I wouldn't trust the data from a 4ms window here, at least certainly not for the 350Hz and below part and the 600Hz might be questionable as well.

 

I trust the data.  The issue lies with properly interpreting it.  The 4 ms window shows me everything that happens to *every* frequency within the first 4 ms following the first arrival from an impulse fed to the system.  The crossovers add some delay, but it's mostly under 1 ms for the low frequencies here, so this is closer to 3 ms of data for the woofer frequencies.  As such, this shows me how acoustic information that arrives up to 3 ms after the low frequency part of the impulse affects the frequency response.  For frequencies below 350 Hz or so, additional window time is required for a full wavelength to be realized.  That means that the apparent direct sound may be influenced by later reflections, and these effects are not revealed here.  However, if all the diffraction occurred within the first 3 ms of the impulse response, or equivalently if the additional path lengths for sound that travels to and diffracts from cabinet edges before reaching the listener are less than about 41 inches, then these windowed responses reveal the relative contribution due to diffraction for lower frequencies too.  This is true until you approach the sub crossover 110 Hz where the delay increases again, but at that point, the later reflections I've windowed are strong enough to substantially corrupt the direct response observed at the MLP.  Even with acoustic treatment, as some point as you go lower, the room wins and you need to apply the usual methods for optimizing subs, placements, and DSP to get the response you want.

 

For the doubters, here is the exact same data except with the window extended to the full impulse response:

 

post-1549-0-75774500-1475905832_thumb.png

 

Please ignore the response of the subs here.  They are not optimized to play well at these measurement locations.

 

The broad ripple pattern in the windowed measurements remains in the long time measurements.  There is a lot of bass build-up at and below 250 Hz for most

locations, but this is from reflections that arrive quite a bit later than the direct sound.  I believe it's mostly side wall reflections, which arrive in the range of 10-20 ms later.  The long time response in 100-200 Hz range depends a lot on the relative time of arrival of those reflections vs. the direct sound.  Hence, measurements sharing the same color (signifying the same horizontal angle) have similar responses.

Link to comment
Share on other sites

it's rather hard to read a graph with loads of lines, different scales and no legend :)

 

are you sure about that point about usable data being found under the window limit? I've never heard anyone say such a thing. The frequency resolution is defined by the window length after all.

 

Sorry about not matching the scales between the two plots.  I didn't prepare them at the same time.  As for a legend, I don't think it would be useful for these because of how many lines are there.  ;)  I just posted the long window responses to validate what I claim is shown in the windowed responses.

 

To be honest, the only reason this validation works here is because the in-room response of my center channel is otherwise quite clean.  The diffraction appear to be the strongest acoustic effect on the center channel woofer response for the first 10 ms or so.  For the left and right, the ceiling reflection is similar in strength and alters the response quite a lot.

 

As for the question of whether data under the "window limit" is useful, the answer is yes, but it's complicated to explain the how and why.  Frequency resolution in the analysis is defined by both the window length and the window type, and you can always pad the windowed part of the IR with additional silence to get more resolution.  REW does this by defaulting to a 100 ms "left window".  The primary concern is that the window size you choose captures the effect of the acoustic phenomenon you are studying on the frequencies you are interested in.  As long as it does, then the impact of that phenomenon will be accurately exhibited in the windowed frequency response.  The big caveat is that the IR you measure is a convolution of both the acoustic response and the electromechanical response of the system.  The electromechanical response will typically delay emission of lower frequencies more than higher frequencies, so one must be aware of both the electromechanical and acoustic delays involved when deciding whether the window is big enough for the frequencies of interest.

 

There are other more general caveats such as the fact that you can't really isolate the low frequency effects of multiple acoustic phenomena that show up in the IR at about the same time.  So if you have a room reflection that occurs at about the same time in the IR as diffraction, you can't easily tell their effects apart.

Link to comment
Share on other sites

  • 2 weeks later...

So I am very eager to start on my subs, but I got distracted by trying to get better sound out of my front stage speakers.  They already sounded great, but I got to the point where I felt that the blind, manual tweaking was no longer productive.  I decided it was time to retool.

 

I updated my measurement software and script so now it will ping all 12 amp channels (yeah, I'm up to twelve now) independently with normal sweeps as well as shorter time alignment sweeps.  I ran a set of sweeps at each seat on the sofa and quickly discovered that my speaker toe-in was in fact too steep.  I feel kind of lame for waiting so long to do it, but I went ahead and optimized the toe-in to balance high frequency response of the stereo pair across the seats.  That brought them in to an angle of about 45-50 degrees vs. 57 or so.  It was enough for the sound at the MLP to be a lot closer to on-axis, which is a good thing.  It also totally altered the tonal balance of the treble.

 

That was a bit frustrating because I still have work to do to re-factor and re-tool the analysis/simulation software I was using before to simulate the bi-quads and FIR filters I used on the old MiniDSP units.  The old software was rather slow and tedious and hasn't been upgraded to work with the libraries that the live DSP processing use.  I'm looking at updating the visualization to use a web interface.

 

In the meantime, I went ahead and downloaded the latest REW to experiment with its frequency-dependent window (FDW) capability.  I like that REW allows you to simulate filters directly against an FDWed response.  For kicks, I dropped in a 2.2 cycle window (1/3 octave) and iterated on filter adjustments until the response was flattened to +/- 0.5 dB from 200-8000 Hz.  I chose 2.2 cycles as that's long enough to capture the anechoic response of the speaker at almost all frequencies and short enough to avoid contributions from early reflections for the most part.

 

Above 8 kHz, I have roll-off on purpose as virtually every speaker rolls off above here.  (Blame problematic early audio monitors for setting a bad precedent here.)  Below 200 Hz, there is a bit more variation including a +2 dB hump at 80 Hz because of the desire to balance bass response better across seats with the subs, which I may be able to improve with better DSP and more close-placed subs.  I opted to calibrate the 2.2 cycle FDWed (approximate first arrival) response to 80 dBC.  This gives me 85 dBC on the nose with 800-2000 Hz pink noise.  With this, I expect my normal "reference" listening level to be about "-4".  I opted for 80 dBC instead of 76 dBC so that I can play DVDs and other soundtracks with the typical -4 dialnorm offset without having to put the MV above "0".

 

In any case, I've been listening with these filters and am *very pleased* with the sound.  I'm willing to bet that the filters I implemented are far from optimal as far as avoiding unnecessary resonance and so on, but system still sounds absolutely fantastic.  It sounds very balanced.  The bass is very clear and powerful but does not overwhelm or obscure any detail.  Even crappy loudness war music has plenty of kick and punch, but of course, dynamic music sounds way better.  Well-recorded vocals are extremely realistic and addicting to listen to.  I've only had one day with it and haven't had time to listen to dialog on the center, but what I've heard so far may be the best sound I've heard from my system to date.  It might be the best sound I've heard on *any* system to date.

 

I anticipate I'll get even better results when I replace the crude biquads with tightly optimized FIRs that only correct the relevant bits in the time domain as well as when I optimize directly against data from multiple measurement locations as I did before.  And of course, I have a lot of ambitious ideas about how to get better bass that I can implement completely independently of this first-arrival correction.

 

I'm real stoked to try out some movie content, and I'm curious if the upper bass sounds bloated or not.  The combination of baffle edge and rear wall on the center channel causes a lot of upper bass (150-200 Hz or so) to beam diagonally in the horizontal plane.  They need a ton of boost (more than I'd rather have to run) to run close to flat on-axis with that 2.2 cycle FDW.  However, at about 6 cycles, the same response looks quite hot by several dB because of some strong early reflections.  IIRC, my steady response hits 90 dBC up there.  So this *should* give me some idea of whether 2.2 cycles or 6 cycles is better in that frequency range.  From the listening I've done so far, I'm thinking that the 2.2 cycle FDW will win, but we'll see.  If I hear too much resonance there (or if I have issues with headroom), I may need to compromise and back it down until I can improve the baffle/wall integration.

Link to comment
Share on other sites

So I watched a bit of film, so I could listen to the center channel.  The tonal balance sounds pretty close to right.  It's slightly bass heavy, but nowhere near as much as I'd expect looking at the 6 cycle FDW.  I can, however, hear the problems in the upper bass range.  It sounds a lot like room resonance, like in a bathroom where the acoustics may still be modal at around 200 Hz.  It is much more noticeable with speech than music and is annoying at times (to me).  It also may be increasing the apparent loudness a bit.

 

In theory, I could use more DSP to cancel that strong early reflection but in reality, this does not work well.  I'm already boosting the center by about 10 dB to get the 2.2 cycle FDW response flat.  To remove that reflection would require substantially more headroom on top of that, and I'm frankly not happy with the 10 dB boost.  It's too much.  Any distortion produced by the speaker while playing that sound will also be amplified, relatively speaking.  Furthermore, any filters that cancel that reflection at the MLP will just make things worse at off-axis seats.  Now, if I involve multiple sources in the DSP, I have a much better chance of getting a pleasing result.  With the speaker beaming diagonally, I have the extreme seats covered and could fill in the hole in the middle using close-placed MBMs.

 

But it's still far better to solve it acoustically, if at all possible.  I will have to try.

 

The other thing I did today is I relaxed my top octave high frequency roll-off by quite a bit.  I figured that now that I have the mid-range and bass very well balanced, I could maybe add back some of that top octave energy.  I chose to target the roll-off recommended by ATSC 2013 (page 34).  So far, I think it's working out very well.  A few music passages I've listened to so far sound a bit excessive up there, but it seems to work very well with most stuff I've listened to so far.  I may just have to make that roll-off curve adjustable because of the inconsistency in content.  Likewise, I could probably use something to tilt the response down slightly for some older music recordings that still seem to run a bit bright in the upper octaves in general.

 

At this point, I have to say that these are the best speakers I've heard, hands down.  By targeting flat response (except for the ATSC 2013 HF roll-off and a little bump in the sub) with the 2.2 cycle FDW I believe I have nailed the tonal balance and voicing perfectly.  I was actually very close already with my process of manual iteration, but that last little bit (literally no more than 1.5 dB change anywhere) helped everything snap into place.  Barring the minor issues discussed above, it's nigh-well perfect.

 

This accomplishment is a big deal for me!  I've been searching for this kind of sound since I started this hobby.  I obsessed over the concept of calibrating to an audio reference when I first got an AVR with Audyssey.  It's rather sad that Audyssey failed to deliver what it claimed to, but it sparked my interest and started me on the path.  It's almost 4 years later, and I think I have finally found reference.

Link to comment
Share on other sites

  • 3 weeks later...

I'm going to give the low-cycle FDW EQ a try to see how it affects my system.  Sounds like it worked well for you.

 

I hope it works out for you!  I'm not sure it will work for everyone.  It may just work well for me because of the exceptionally even coverage from my SEOS horn-based 2-way speakers.  My thinking was to try to capture the anechoic response and make that flat.  Along those lines, the FDW window length must be long enough to allow for crossover delay (which usually peaks at the actual XO, in terms of FDW cycle count), but not so long as to allow any close room reflections to interfere.

 

The bass may still need to be treated differently in some cases.  At some point, you go low enough in frequency that reflections and/or diffraction effects arrive too soon to clearly distinguish the first arrival from the effect of the diffraction/reflections on it.  My inclination is to then treat the response combined with these early effects as being the first arrival.  However, this may result in insufficient bass.  My thinking is that pro monitors are usually designed to be anechoically flat, and they often get used in the mixing room either as-is or with only narrower EQ corrections for room modes.  The implication of this is that the "first arrival" (including any reflections that can't be isolated) should be a bit hotter in the lower bass (say below 100 Hz) because the baffle step compensation built into the speakers effectively overcompensates once you go low enough that the strongest early reflections are constructive.

 

Of course, part of the reason I'm making this argument is that I am running my subs a bit hot with respect to the 2.2 cycle FDW, and I think it sounds better (as in closer to reference) that way.  Another possibility is that my MBMs sound better hot because they are so close to the MLP and effectively higher in directivity.  In other words, they don't put as much energy into the room for the same SPL as the speakers would if they were playing that bass instead.

 

The only real change I've made since my post above is to bump the subs up a bit higher.  I pushed the MBMs (frequencies between 50-110 Hz) up +0.5 dB, making them about +2.5 dB hot at the MLP when using the 2.2 cycle FDW.  Any more than that, and the sound is too slow and lacks punch.  The bigger change was that I bumped up the deep bass subs (frequencies under 50 Hz) by +2 dB, making them about +2 dB hot at the MLP in terms of the 2.2 cycle FDW.  I tried going up a full +2.5 dB to match the MBMs, but that extra +0.5 dB was enough to totally kill both transient detail and tactile sensation and make the bass actually sound wimpier.  (Note to readers: you read that right; excessive deep bass can kill your tactile response by masking the higher frequencies that contribute most of the sensation.)

 

After the changes, the timbre of the lowest notes is a lot more consistent with the higher notes on the same instrument, and kick drums are very satisfying.  The adjustments also tamed excess brightness, particularly in the top octave, on a lot of tracks.  (Deep bass seems to mask top octave content to quite an extent.)   Indeed, after these adjustments to the bass, I am finding the ATSC treble roll-off to be very satisfactory with a majority of recently released musical content.  These changes did not fix excess brightness for all content.  I'm fairly certain that there will always be content (particularly produced and/or mastered from the '80s and earlier) that sounds better with slight a downward slope (about -1dBish) toward the high end.  I have also noticed that quite a few releases from the mid-to-late '80s still seem too hot in the top octave.  This could be because monitors of that time were rolled off too much to try to compensate for the bad precedent set by earlier monitors.  Another possible explanation is that the excess is due to top-end roll-off in the DACs used in the monitoring signal chains of that time.  These older DACs may have rolled off at the very top because of the difficulty of creating an analog filter that cuts everything above 22.05 kHz (Nyquist frequency for CD audio) without harming response below 20 kHz.  Most of these older DACs probably had some roll-off at the very top as this was arguably a less bad compromise than the other options.  Oversampling DACs which have been in widespread use since the '90s address this problem by resampling to a much higher rate before conversion, giving the post-conversion analog filter much more room (in frequency space) to work with.

 

All of the above applies to music, and apart from the issues I've noted, music seems to be remarkably consistent as far as tonal balance is concerned.  In the absence of specific standards, precedence rules, and engineers strive to follow the precedents established by the huge body of existing recordings.  Movies are a different story, unfortunately.  Despite there being standardized calibration methods, movie sound tracks seem to vary a lot more in terms of tonal balance.  It would seem that standardization only actually helps if those standards are psycho-acoustically relevant.  And unfortunately the methods prescribed by the movie industry are a long off, even with recommended adjustments to playback level and target curve made for room size.  Indeed, standardization of this kind may have actually been counter-productive to the extent that such standards have been followed blindly.  The X curve in particular has probably contributed to a lot of movie soundtracks having too much upper mid and treble compared to typical music.  This is one area that near-field "for the home" mixes are said to improve on, but of course, the practices involved here are mostly non-standard and inconsistent.  (I say *mostly* because apparently some effort has been made, by Brian Vessa of Sony and others, to standardize home mixes, but the information on how this is done does not appear to be available to the public nor is it clear that these standards are actually being widely adopted.)

 

I haven't watched a whole lot of movies since I found my reference response for music, so I don't have an especially wide sample of content with which to formulate an opinion.  With that said, movies do seems to generally run brighter than music.  Even the "made for the home" mixes seem like they might sound a little better with the -1 dB slope I mention above for older music.  My guess is that some of the studios monitor the tracks this way on purpose under the assumption that it better reflects what speakers at home are doing.  I would guess that they are correct that most home speakers have the slight downward slope because this does sound better with older recordings and because excessive brightness is usually more offensive to the listener than the opposite.  However, I'd argue that they should still monitor home mixes flat (albeit with a roll-off at the very top similar to ATSC) to sound best on today's state-of-the-art monitors and to be consistent with modern music releases.

 

To deal with inconsistent tonal balance in movie releases, I plan to eventually implement a few controls into my DSP system to allow fine adjustments to the upper mid, treble, and top-end response.  While such controls will be nice to have for older music, it appears they will be critical for getting the best sound with a variety of movie soundtracks.  It's important to keep in mind that even 0.5 dB too much treble can add a few dB of loudness to a track, so if (room appropriate) reference level is too loud, there's a good chance it's just because the balance is too bright.

 

And just to reiterate, the sound I'm hearing on my system now (for music especially) is just stunning.  The bass is strong and tactile but remains in balance with the mids or treble.  Cymbals and hats are incredibly clean, detailed, and realistic (subject to the quality of the recording, of course), yet there's no harshness or fatigue even at high listening levels.  I get decent tactile bass slam even from loudness war music, but of course, dynamic music is way better.  I have some good dynamic live music recordings where the sensations from the kick drum are very strong, like something blowing up inside my chest.

Link to comment
Share on other sites

I hope it works out for you!  I'm not sure it will work for everyone.  It may just work well for me because of the exceptionally even coverage from my SEOS horn-based 2-way speakers.  My thinking was to try to capture the anechoic response and make that flat.  Along those lines, the FDW window length must be long enough to allow for crossover delay (which usually peaks at the actual XO, in terms of FDW cycle count), but not so long as to allow any close room reflections to interfere.

 

I don't really follow that, a 2.2 cycle FDW is much smoother than an anechoic response.

Link to comment
Share on other sites

I don't really follow that, a 2.2 cycle FDW is much smoother than an anechoic response.

 

No, not exactly, but I believe it's a useful approximation in my case.  If my XOs were linear phase (maybe some day soon they will be), I would have tried an even shorter window.  Other sound systems may need different strategies depending on XOs and other sources of delay in response.  And to repeat what I said earlier, this strategy may not work nearly as well for speakers that are messier off-axis than those based on the SEOS horns.  Tonal balance may be substantially affected by later arriving energy, which is influenced by off-axis response.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...