The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Subwoofer Experiments Modular Synthesizers
Old 7th August 2014
Lives for gear
Subwoofer Experiments

Am guessing this topic belongs in this section being related to acoustic implications more than wires, magnets and transistors.

Subwoofer Experiments-frontrightview_web.jpg

Been building some traps, measuring the effect with REW, moving stuff around, etc. The old JBL 4208 monitors driven by an even older JBL/Urie 6230 amp, ~100 watts / channel 8 ohms. I sometimes wonder what they do with the watts in modern powered studio monitors-- Though I usually don't monitor loud, messing around with the old sound level meter calibrating REW, unless the SLM is way out of calibration, 100 watts / channel easily exceeds 110 dB while remaining reasonably clean in these cheap old speakers, and most of the modern medium-sized powered monitors are rated in the ballpark of 112 dB peak (or less) with much more powerful biamps.

Anyway, have been playing with this setup using a DBX Driverack PX. I get hung up on theories of "how it ought to work". Had it in my mind that, because the 4208 monitors begin to roll off at 60 Hz, that the most ideal subwoofer crossover should be a 24 dB Linkwitz-Riley crossover fairly low, in the 60 to 120 Hz vicinity.

After installing a few traps and somewhat smoothing the bass response (more to do), the remaining deepest bass nulls tend to be at about 40 Hz and about 130 Hz.

Was initially playing with an ancient Mitsubishi 120 watt, 12" sub. This is a down-firing sub with a side port. Setting the sub's lowpass as high as it will get, and controlling crossover with the driverack, was surprised how insensitive to crossover freq and crossover slope, was the room response. In other words, moving the crossover frequency in the ballpark of 60 to 120 Hz (a relatively flat area of the room), showed very little difference in the REW Freq response curve. With the sub placed as close as I can get to the same distance from LP as the monitors, didn't notice significant notching or reinforcement in the vicinity of the crossover freq. The mains and sub were "merging" a lot better together than I expected them to do. I expected rather nasty interaction in the vicinity of the crossover freq.

In fact, the 130 Hz null was shallower if I moved the crossover freq up in the 200 Hz ballpark. Apparently the sub's location filled the null better than the main speakers. Also noticed that reducing the crossover slope to first order, 6 dB per octave, fairly high in the 200 Hz ballpark, had no apparent bad effects on the freq response. In fact, the freq response looked a bit better in the vicinity of my 130 Hz null.

So got to experimenting some more-- That down-firing sub, with the side port-- Spewing audio all over the room before it gets to my ear.

So I built a little subwoofer speaker stand you can see to the right of the keyboards, and for giggles replaced the 12" sub with a Roland KC-500 keyboard amp. With the stand I can get the sub very near same distance from LP as the mains. The roland has 150 watt amp, 15" speaker and bullet tweet. As best can tell it is a sealed enclosure, so I should hear the direct sound from the speaker at least before it gets a chance to bounce around the room a few times.

Because the keyboard amp is designed to get lots louder than living room levels, it can be EQ'd to deliver some fairly loud sub-bass at modest living room levels (using the driverack EQ features). The roland sealed cab seems to have max bass efficiency around 70 Hz, but it is easy to boost lower than that, and cut in the 70 Hz vicinity. Plenty of headroom at livingroom volumes.

With some EQ to flatten the sub-bass with the bass, and a 6 dB / oct, 200 Hz crossover against the mains, the 130 Hz null was completely eliminated, and it didn't really sound too bad either. VERY non-intuitive from what I was expecting.

Having played live gigs so many decades, am prejudiced that bass out of a 15" or 18" speaker just sounds better more satisfying than "fake bass" out of a smaller speaker. But my biggest complaint with the keyboard amp as subwoofer, it is just too light sub-50 Hz, too peaky around 70 Hz. Needs too much EQ. A long time ago had a couple of JBL D-140's. Wish I still had one. A loose-surround D-140 with a bunch of fiberglass in that roland sealed cab might have a lower natural resonance.

So anyway, the advantage of the Roland is better coverage in 100 Hz to 200 Hz, and the advantage of the old Mitsubishi 12" sub is much better output below 50 Hz.

So I tried taking down the Roland, putting the Mitsubishi on the stand, sitting sideways with the speaker facing me, rather than facing the floor. Measured with REW. Then stuffed a sock in the side port (the side port firing down with the sub sitting on its side) and measured again with REW. The room bass response was near identical either with or without the port blocked, so left the sock in there.

Then experimented with the main monitors' rear ports blocked. Crossing over that high, the ports probably make little difference anyway. It seems to give marginally better room response with all ports blocked, though the measurable difference was minor.

Low bass is better with the Mitsubishi as a side-firing sub (rather than down-firing), but my 130 Hz null somewhat came back with the Mitsubishi sub. I was able to slightly diminish the 130 Hz null by setting the mains hipass at 100 Hz, 6 dB/oct, and the sub lopass at 200 Hz, 6 dB/oct. "doubling up" on the problem band. The mitsubishi sub speaker just isn't very strong above about 100 Hz, even with its internal lopass set at max (200 Hz).

It is crazy. Theoretically I would have said overlapped crossover freqs with such gentle slope would be guaranteed to sound awful, but it doesn't sound half bad. The ideal solution would be to treat the room so the 130 Hz null goes away of course, but suspect its solution will eventually demand a cloud or so many traps that there is no longer room for me in the room!

Anyway, it seems my "ideal subwoofer" would be a sealed 15" (or even 18") cab, loaded with a speaker with more lows than roland's stock keyboard woofer, but also with decent freq response up to above 200 Hz. A good quality bass guitar speaker and cab might qualify, needing both good bottom and clear low-mids. Unfortunately, most bass guitar cabs don't publish freq response specs, and am guessing that the typical small 15" bass guitar cab might have more in common with the tuning of my roland keyboard amp, than with a subwoofer.

Apologies rambling. Tis interesting how the trivia of the speakers interacts with the acoustics of the room. So many degrees of freedom in the adjustments.

Last edited by jcjr; 8th October 2014 at 05:08 AM.. Reason: Fix clobbered images
Old 8th August 2014
Lives for gear
Thanks, Jens. That is an interesting thread, and it is somewhat gratifying that "fooling with subwoofer placement and crossover freq" is not an entirely wrong way to approach nulls.

I had planned to build and place 3 more traps before looking into specific peaks and nulls. After adding three more traps there won't be any obvious conventional place to put any more, except a cloud or oddball locations. On the assumption that maybe the peaks and dips won't look the same after the room is fully trapped (at least to the amount of space I'm willing to sacrifice).

On the other hand, the sonic signature of the room has been mitigated somewhat by the first 5 traps, but it is obviously about the same sonic signature it had untreated. So perhaps pictures and measurements of the current state would be helpful. Perhaps people will see novel places to put a trap or two, which might do more good than simply "filling in the rest of the available wall space".

Here is a to-scale schematic of the room--

Subwoofer Experiments-studiosketchup_1.jpg

It is the livingroom of my house's old basement apartment, which I remodeled last year. The gray squares on the left wall and front wall (behind the speakers) are double-pane windows with wood blinds. The outside entry door to the right also has wood blinds.

The always-open doorways in the rear are "almost symmetrical" and lead into fairly large spaces. So am guessing they somewhat act like good bass traps, because it should take a long time for sound which exits the rear doorways to bounce and return (attenuated) from the bedroom, bath, kitchen, dining room, and entire basement beyond the dining room back door. On the other hand it offers the opportunity for the entire lower floor of the house to act as one big organ pipe resonator, so that there will most likely always be a long but hopefully low-level reverb tail in the bass.

The listed traps are all made of roxul safe'n'sound. 3" thick, 4 ft X 6 ft traps in the rear, and the front corner traps are 9" thick, 2 ft X 6 ft. The further traps I had planned are 6" thick, 15" wide, 6 ft tall "wing" traps on both sides of the right-front corner trap, and on the front wall side of the left-front corner trap. Can't put a floor-to-ceiling trap to the left of the left-front trap, because of a wall mounted heat-air conditioner unit there.

Here are a couple of photos, though they were posted before, clockwise from left-front to left-rear--

Subwoofer Experiments-frontleftview_web.jpg

Subwoofer Experiments-rr_cr_trap_web.jpg

Subwoofer Experiments-lr_cr_trap_web.jpg

Will post a few measurements in the next message.


Last edited by jcjr; 8th October 2014 at 05:18 AM.. Reason: fix clobbered images
Old 8th August 2014
Lives for gear
This is a very near-field listening situation. Monitor centerline between woofer and tweeter is exactly at ear level, and each speaker is about 41 inches away from the listening/measurement position. The subwoofer is also pretty close to the same distance. With the monitors toed-in and keyboards below so far below the firing zone of the tweeters, I have difficulty imagining much near-bounce. Ceiling or floor bounce would be at such acute angles that it is WAY out on the sides of the speaker dispersion lobes. But maybe there's something obvious I'm missing, and perhaps various cavities around the desks could benefit from trapping or whatever. Am open to suggestions.

REW measurements taken with dbx M2 microphone, into the mic preamp of a focusrite Pro24 firewire interface, using loopback signal for sync. Been lazy, only both speakers + sub measurements. All measurements are 8 averaged sweeps, each sweep as slow as REW will allow. Mic very close to 41 inches away from each monitor in each test. Sometime ought to test other locations, but all I care about is the listening position.

SOME of the background noise/long decay tails in the bass are due to a noisy refrigerator in the adjacent kitchen, been too lazy to unplug it for testing. Figured it doesn't matter much until I get closer to final testing. Will post mdat files if someone wants to bother looking at em.

Twenty-fourth octave smoothing in all cases.

Three measurements--
1: No Treatment
2: Latest with the 5 traps, mitsubishi sub, sub hp 20 Hz, sub lp 200 Hz, mains hp 100 Hz, all 6 dB/oct slopes. Ports plugged in the sub and the mains. Mains have rear ports. Sub is designed as down-firing with side port, but it is mounted on a stand directly under the right main, with the speaker firing right at me.
3: Same as setup 2, except with the dbx driverack auto-eq enabled. It has an auto-eq routine, plugin the measurement mic, turn on the pink noise, let it do its thing. Works pretty good as best I can tell. Then run another REW test with the dbx third-octave EQ enabled.

No Treatment waterfall
Subwoofer Experiments-waterfall_beforetreatment_20-500hz.jpg
5 Traps waterfall
Subwoofer Experiments-waterfall_5traps_mitsubishi_sub_20-500hz.jpg

No Treatment FR
Subwoofer Experiments-fr_beforetreatment.jpg
5 Traps FR
Subwoofer Experiments-fr_5traps_mitsubishi_sub.jpg
5 Traps AutoEQ Fr
Subwoofer Experiments-fr_5traps_mitsubishi_sub_eq.jpg

The no treatment measurement had the big sheetmetal toolchest close to the sub, and the toolchest was probably resonating measurably. Moved the toolchest away for subsequent testing.

The 40 Hz null seems to have got deeper along with added traps. For the first few tests, the sub gain was set too high, but the 40 Hz null was in the ballpark of -15 dB below the 49 Hz peak. And lately the difference has grown to about 20 dB between the 40 Hz null and the 49 Hz peak.

The 58 Hz null ain't all that bad, smoother than before. The 131 Hz null varies from minor to cured, depending on subwoofer configuration, was previously very deep.

Seem to be measuring persistent nulls around 900 Hz, 1.2 KHz and 2 Khz which don't seem to have been improved so far by treatment. Otherwise it doesn't seem too bad. Subjectively sounds better than it did anyway.

Last edited by jcjr; 8th October 2014 at 05:29 AM.. Reason: fix clobbered images
Old 8th August 2014
Lives for gear

Problems at 70 and 140hz suggest the ceiling is not playing well with the rest of the room. As a test, you can temporarily move the monitors higher, up to 75% the room height, to see what impact it has.
Old 9th August 2014
Lives for gear
Thanks Opus

You are probably right. I spent awhile today going thru the long list of calculated room mode frequencies, to find if there are any natural modes that match the measured frequency response, but the hits didn't seem especially relevant. Spent awhile doing "numerology" looking for repeating patterns in the major peaks and nulls, but mostly struck out there as well. One peak accidentally happened to be 2X the frequency of another null, but with so many numbers, chance alone might demand that at least one pair would have some kind of accidental relationship.

Deciding to focus on the biggest nulls, converted the freqs to wavelength, half wavelength, quarter wavelength, sitting in the listening position and waving a ruler around. On the theory that if the freq deviations are around 20 dB between the biggest peaks and nulls, then if the nulls are eliminated then there would only be about a 10 dB variation even if the peaks were entirely unaffected.

As best I understand, a direct reflection with fairly acute angle would need to be a quarter-wave distant, so that the round-trip would make a half-wave and cancel. The lowest major null at 39.3 Hz, quarter wave 7' 1 5/8"-- That is near exactly the distance from listening position to either side-wall. Possibly traps directly to my left and right, either on the wall or SOMEWHERE between me and either wall, might get that one.

The next null around 56.8 Hz is a quarter wave of 4' 11 1/4". Waving around the stick ruler, the ruler exactly hits the juncture between ceiling and front wall (and also the juncture between floor and wall). So possibly treatment there would be good. I just hate to put eyehooks in the ceiling or hang anything on the wall, but seems inevitable. When I remodeled the basement apt, replaced the electric box and all wiring and fixtures. Because serious rewiring is a PITA in old buildings, in attempt to get a good electrician to take the gig, I promised him he could knock as many holes in the walls as he wanted, so the job wouldn't be too tedious and annoying. So he did knock lots of holes in the walls, and did a great job. And after working so long patching all those holes just last year, am loath to make new holes in the walls and ceiling.

Haven't theorized on the 130 Hz null, though subwoofer promises to treat that one.

The 888 Hz null is fairly wide and deep. There is nothing close enough to do an acute angle bounce at a quarter wave of less than 4 inches. So possibly a wide-angle multi-path bounce, so that the difference between direct and bounce is 7 9/16", or a multiple thereof. Given a distance from main woofer (its below tweeter xover freq) to ear of about 41", went looking for a reflection angle that would have path to my head of 41" + 7 9/16" = 48 9/16". So the bounce is a halfwave behind.

Turns out that a triangle between me and a mains woofer (either side, symmetrical) has legs of about 24" on the keys of my KX88 keyboard. So maybe that is the prime suspect. Seems reasonable that one would get cancellation from both speakers, because of the symmetry between my head, both speakers, and the keyboard. Surprising if that turns out the culprit, as the keyboard looks pretty "far out of the pathway" of the speaker compared to the speakers themselves aimed right at my head. Maybe there would be a way to make little boxes of insulation to sit in that lower path between speakers and keyboard, block the reflection path to the keys.

Maybe the upward-pointed analysis mic hears different from a human head. I don't turn my head sideways to listen to the ceiling or listen to the keyboard controller (which is electronic and don't make any sound except key thumps). I listen to the speakers in front. So maybe that null isn't as noticeable to the ear as it is to the mic? Dunno.
Old 9th August 2014
Gear Addict
Originally Posted by OpusOfTrolls View Post
Problems at 70 and 140hz suggest the ceiling is not playing well with the rest of the room. As a test, you can temporarily move the monitors higher, up to 75% the room height, to see what impact it has.
...As a test, you can temporarily fill the whole ceiling, (and backwalll, and rear-wall to see what it happen to your FR) More Traps are better, of course.! Don't forget this!
Old 10th August 2014
Lives for gear
Originally Posted by Alejandro Varela View Post
...As a test, you can temporarily fill the whole ceiling, (and backwalll, and rear-wall to see what it happen to your FR) More Traps are better, of course.! Don't forget this!
Yes, that seems the way it goes.

Last night experimented and managed to measure all nulls except the lowest bass nulls "completely cured" by temp laying batts of rockwool over some suspect locations in the near field around the monitors and equipment, but will be a week or more to decide how to make permanent fixtures for those spots, and hopefully the permanent will work as good as pieces of unadorned rockwool temp laid over some spots.

Fixing in the nearfield ought to be much simpler if it works out. I only care about the listening position, so if near treatment turns the speakers into the equivalent of texas-sized headphones, that is ok.

Perhaps can prevent the room ruining the sound of the speakers, by controlling speaker directivity. I only care about sound which hits my ears. Sound that goes to other parts of the room, only to return muddled in both the time domain and frequency domain- That second-hand sound could be restricted without hurting my feelings.
Old 10th August 2014
Lives for gear

The amount of bass traps needed can be reduced with careful speaker and listener positioning, sometimes!
It's a job, but it has benefits.
Old 24th August 2014
Lives for gear
Texas Sized Headphones

As I mentioned a couple of weeks ago, after making some guesses about nearfield causes of the nulls, managed to reduce many mid-upper nulls by laying out temp batts of rockwool.

Some nulls were reduced by laying out two batts of rockwool over the top of the monitor speakers, extending front, back and to the sides, somewhat blocking reflections from the front wall and front wall-ceiling juncture.

After testing that, I removed those batts, laid a batt atop my keyboards, which improved some other nulls.

Then removed the keyboard temp batts, and placed some batts inside my keyboard stand and around the computer desk and rack, which affected other nulls.

I finally placed all the various temporary test batts and got the following freq response-- Not perfect but lots better. If curious, you can compare to the previous plot in an above message, which had the first five finished traps but no experimental batts laid about.

All my sweeps are 1 meg, repeated 8 times, 20-20K.

Subwoofer Experiments-fr_temp_test_absorbers.jpg

Therefore, devised a plan of action to build traps to hopefully do the same as the temp batts of rockwool. Most of the future traps will be small and simple, but it took awhile to build this first monstrosity. I don't want to drill holes in the wall/ceiling, build-in anything, or make any permanent changes.
Old 24th August 2014
Lives for gear
Texas Sized Headphones

So this latest monstrosity is slightly-compressed 3" safe'n'sound. The panels are 3" thick, but I intentionally cut each piece of rockwool too big for each opening, so it is somewhat laterally compressed when crammed in and covered with fabric. All panels are open on both sides-- Open on back, sides and top.

The remaining planned near-field traps will be fairly small and simple. This was the time-consuming one.

Outside dimensions are 96" wide, 28" tall, and 24" deep.

The top panel is hoped to function similar to a cloud hung on the front wall-ceiling boundary. It overhangs the speakers at the front by about 9". At my close listening position, I can't see the front wall or the front wall-to-ceiling juncture. Therefore some of the mids and highs that were previously reflecting from where I now can't see, ought to be attenuated by that top panel (line of sight propagation except for diffraction).

The back panel seems at worst harmless. Ought to assist the facing rear-of-room big panel in squashing front-to-back axial modes, and generally assist suppressing audio bouncing around in the near-field.

The side panels of course must be there to help support the top panel, but my big 40 Hz null is almost certainly a side-to-side room axial mode. Room width is 15' and room length is 11' 6".

I have 2 four foot wide traps facing each other at the rear of the room (pictures earlier in thread), plus the new two-foot wide traps in the front monstrosity (plus the front 9" corner traps). This means that 6 feet of the 11' 6" are at least somewhat trapped against side-to-side axial modes. Which leaves 5' 6" in the middle of the room for future side-trapping somehow. I don't have much side wall space. Perhaps a future 1 or 2 gobo type traps in the middle that might not get in the way too much.

Here are some pics. The old nikon point-and-shoot isn't wide angle enough to get a full view of the monstrosity in the tiny room. Can't back up enough without hitting a wall.

Subwoofer Experiments-fronttrapleft_web.jpg

Subwoofer Experiments-fronttrapcenter_web.jpg

Subwoofer Experiments-fronttrapright_web.jpg

The equipment layout is intended to get a lot of stuff easily available at arms length, which is why I'm stubbornly trying to keep this setup and make it sound as decent as possible.

The original speaker shelf was 72" wide, mdf shelf supported by 2X6 boards. Fairly solid and non resonant. I thought 72" would be wide enough, but some midrange nulls are caused by keyboard reflections. The keyboard would be rather inconvenient to play if I cover it up with rockwool.

So I widened it to 96", to spread the speakers so maybe I can have opportunity to reduce some keyboard reflections. The 88 doesn't have any controls on the area behind the keys, on each side of the 61 key keyboard. So I'll make some small rectangular traps about 20" wide, 12" tall, 10" deep, to set on-top of each end of the 88, above the keyboard and below the speaker woofer openings. To block direct radiation of mids from the woofer onto the keys.

Such small traps will be light weight. Will try attaching velcro to the reinforcing rails seen below the speakers on each side. The near side of each little trap will sit atop each side of the 88, spanning from the back of the keys all the way to the speaker shelf. Possibly the little traps should even extend under the speaker shelf as well. Velcro on the front of the speaker stand ought to keep the small light traps from falling off, but remain easy to temp remove the little traps for sundry maintenance.
Old 25th August 2014
Lives for gear

That's one approach I guess.
Still thinking a cloud would be best.
Old 25th August 2014
Lives for gear
The new speaker shelf nearfield trap is somewhat puzzling. Unless my ear wants to lie assuring me I didn't waste so much time building the contraption, this last trap made more audible improvement than any individual previous trap. But the measured frequency response appears only mildly improved.

The RT30, TOpt and waterfalls appear so unchanged from past recent measurements that it isn't worth posting new ones. However the decay was already verging on "not awful" with 5 traps. I've never heard slapback chatter in the room even with no treatment.

I was overdriving the subwoofer a bit, too much second harmonic distortion in the bottom octave. So the latest measurements had the sweep set a few dB quieter, and the mic gain a few dB hotter to compensate, so that may confuse any marginal improvements in bass decay, with increased mic, preamp and environmental noise.

The most recent TOpt and RT 30 with 6 traps measure very similar to the 5 Traps, the puzzling aspect that some third octave bands in the new measurement have lower decay times than old measurements, but a few bands show slightly higher decay times than old measurements. Perhaps it is an artifact of having slightly changed the gain structure with the new tests, or perhaps it can be blamed on spreading the speakers from 72" apart to 90" apart. The decay times haven't got worse overall, merely tiny apparently random decreases or increases, band-to-band.

I didn't expect this front trap to fix all the nulls my previous temp experiment improved, because it doesn't cover all the places I previously laid temp batts of rockwool.

Subwoofer Experiments-fr_fronttrap.jpg

It is beginning to sound pretty good, but visible improvements are subtle. Most obvious, the 500 Hz and 1230 Hz nulls are entirely gone (which was expected from earlier experiment). The 3.9 KHz null is mostly gone and the 2 KHz null is improved by about 4 dB and narrower than before.

The 900 Hz null is deeper, but more narrow than before. Am fairly sure that at least part of that 900 Hz null is keyboard reflection, which will have to wait on future little keyboard traps previously described.

The remaining nulls have been improved by 2 or 3 dB. In most cases the remaining nulls have shifted slightly lower in frequency. Perhaps because the added absorption makes the room look slightly larger to audio bouncing around the room.

The remaining nulls are also narrower than the previous 5 traps measured nulls, some of which were previously fairly wide. The frequency response still looks rather ragged-- So ragged that it is surprising that it sounds noticeably better to the ear, unless the audible improvement is mere wishful thinking.

Running the DriveRack auto-eq pink noise function, the DriveRack came up with this curve. The auto-eq routine apparently doesn't attempt adjustment of the low bands 31.5 and 40 Hz, and it also apparently avoids auto-eq on the two highest bands 12.5 KHz and 16 KHz. Probably a wise choice, because many pro sound systems at a gig could never come flat at those freqs regardless of the boost. Avoid "automatically" blowing the woofers and tweeters!

The two low band boosts are manual adjustments in attempt to straighten it out "a little bit" down there.

Subwoofer Experiments-fronttrapeqsettings_web.jpg

The device can boost/cut third octaves by +/- 12 dB, in half dB increments. Another screen allows editing each value, but given that the tiny display only has one row of pixels for each dB of cut/boost, for instance a 1 pixel bar might either indicate 0.5 dB or 1.0 dB, and a 6 pixel bar might indicate either 5.5 or 6 dB. It is easy to get around on the device, I just took a picture of the wrong screen to be most informative.

The auto-eq frequency response with the 6 current traps looks like this--

Subwoofer Experiments-fr_fronttrap_eq.jpg

Good recordings don't sound too shabby with the EQ either enabled or disabled.
Old 25th August 2014
Lives for gear
Originally Posted by OpusOfTrolls View Post
That's one approach I guess.
Still thinking a cloud would be best.
Thanks Opus. You are probably correct.

Other than not wanting to drill holes in the brand-new remodeled ceiling and walls, hanging clouds typically look weird to me. Even weirder than my monstrous contraption.

I do like the look of thick "full ceiling built-ins", but don't have enough height to do that, even if I wasn't too lazy and frugal. I only have a 6' 8" ceiling height to play with. Anything bigger than a minimal wimpy thin cloud would be claustrophobic and a head-bump hazard.

Am aware that audio diffraction somewhat invalidates "line of sight" considerations, especially at low freqs. However, the geometric obscuration of objects at various distances is fascinating.

For instance, if you watch a movie on a laptop screen sitting at a desk, you get to see a bigger picture than some giant expensive flatscreen TV, located at the other end of the livingroom.

Was thinking that if all I care about is nearfield single location listening-- Then nearby small treatment might be somewhat equivalent to massive treatment farther away. A cloud even on my low ceiling would have to be at least double the footprint of my monstrosity, in order to intersect equivalent angles of propagation from the speakers.
Old 25th August 2014
Lives for gear
Superstitious Numerology, or "How to mis-use statistics"

I promise this will be the last slamming of this thread for awhile. Just one more thang...

When treating a woefully messed-up frequency response, where each incremental change can only be expected to yield marginal improvement--

It is a sunk-cost problem. After investing time and money, one might too-earnestly desire to hear or see improvements, which may or may not be "real".

Surely expert practitioners have sufficient experience to look at two screwed-up frequency response plots, and by eye identify the least-imperfect response out of the two.

However, though I'm no stranger to looking at freq response charts (not acoustics related, more electronic and dsp freq charts)-- It gets confusing. On the one hand one might identify features that look better, but on the other hand perhaps some other feature looks worse to the eye.

Perhaps sometimes the perceptual mechanism can be fooled, with a "better" chart exposing a previous problem that didn't look so bad before, but now looks like a reversion?

Or perhaps sometimes an "evenly jagged" chart could appear smoother to the eye than a "reasonably smooth" chart containing a few deep narrow nulls?


So I was wondering if freq response charts might be objectively, numerically compared. The ear, fickle as it may be, would be the final arbiter, but a number would be nice as well, because arithmetic doesn't have an emotional stake in the results.

Haven't googled this, maybe there are numerous well-established evaluative numerical procedures.

REW can export measurements to a text file. Dunno if all versions output the same, but REW V5 beta 22, on a 20 to 20K sweep, outputs a text table of about 956 measurements, 96th octave, each octave supplying 96 lines, each line containing frequency, amplitude in dB, and phase.

An obvious tactic would be to import the text file into a spreadsheet and calculate standard deviation on the SPL column. Assuming a normal distribution, and the numbers in a freq response chart might not be assumed normally distributed, 68.3% of measurements will lie within +/- 1 standard deviation away from the mean, 95.5% of measurements will lie within +/- 2 standard deviations from the mean, and 99.7% of measurements will lie within +/- 3 standard deviations.

It is a metric of the "scattering" of the data. If you calc std deviation on a frequency response which is ruler flat, the std deviation would be zero. If one freq response has a std deviation of 1 dB, versus another frequency response with a std deviation of 2 dB, then theoretically the lower std deviation might be evidence of a better, flatter frequency response.

Std Deviation depends upon the mean. On some kinds of data, you can't directly compare std deviations for instance between one set of measurements with a mean of 100, against another set of measurements with a mean of 50, because if they have similar distribution curves, you would expect the smaller-mean set to have a smaller standard deviation. For such direct comparison the figures are sometimes converted to normalized standard deviation, so they can be more easily compared. Though fraught with risks, because figures don't lie but liars (and fools) can figure.

It is confusing, but I think raw std deviation might be most relevant to freq response evaluation, because dB are additive and level-insensitive. In a linear system, if a 100 dB mean measurement happens to have a 20 dB dip at 1 KHz, then if you do the same measurement at 50 dB or 200 dB, you would still see the exact same 20 dB dip.

However, perhaps Average Absolute Deviation would be better than std deviation. Simpler and perhaps quite applicable to frequency response measured in dB-- Absolute deviation - Wikipedia, the free encyclopedia

A flat freq response would have average deviation of zero regardless whether the measurement was taken at 70 dB or 140 dB. If the average sample in one measurement deviates from the mean by 1 dB, then it might be a decent wager that this is overall flatter than some other measurement where the average sample deviates from the mean by 2 dB?

Here is an OpenOffice calc screenshot of results I got from some of my incremental room measurements. The Avg Deviation seems perhaps more predictive than the Std Deviation. There are also figures for normalized std deviations, but the more I think about it, the more I'm convinced that those might be inappropriate for comparing measurements in dB. (which has little meaning because I'm so routinely in error )

This little screenshot just shows the top results part. The rows of source data continue down for 956 entries. These are from some plots I posted here earlier.

Subwoofer Experiments-treatmentstats_1.jpg

The 5 Traps plot is surprisingly little improved from the Before Treatment plot in std deviation, but shows the tiniest improvement in Avg Deviation. The 5 Traps plot looks better "to the eye" than Before Treatment, and it sounds better and has about half the decay times. So this is only a frequency response metric, and perhaps a poor one.

But it is interesting that the Auto-EQ version of the same 5 Traps measurement does show a numerically flatter response.

The FrontTraps (6 traps) scores better than 5 Traps, though one has to study the charts fairly closely to decide the same thing, and for me studying the freq response charts is somewhat a judgement call.

Similarly, the Auto-EQ version of the 6 Traps scores best of all, which one would expect if this metric has any meaning at all. But the metric might be meaningless regardless, mere coincidence.

Perhaps there are better, common ways to numerically compare the flatness of freq response charts. Just found it interesting.
Old 12th September 2014
Lives for gear
So here is another installment on the trip down the rabbit hole. Making some traps which hopefully put the keyboard in an "acoustic shadow" to eliminate midrange nulls related to reflections from speakers to keyboards.

Constructed of half inch plywood and 1" dowels, about 20" deep, 20" wide, 12" tall--

Subwoofer Experiments-keytrap_1_web.jpg

Subwoofer Experiments-keytrap_2_web.jpg

Subwoofer Experiments-keytrap_3_web.jpg

The fat end intended to block midrange from monitors hitting the keyboard. The skinny end open on top, bottom, and back to add some damping of resonances in the spaces under the keyboard stand, because they need to be present for physical balance anyway.

Holes drilled and routed smooth on the sides to add some diffraction, try to prevent adding new nearfield reflecting surfaces while trying to solve other nearfield reflecting surfaces.

Stuffed with 2 pieces of 15"X47"X3" rockwool per trap. Burlap covering. To hopefully prevent scratches on keyboards and other equipment, edged with strips cut out of that thick black "rubber lookalike" vinyl carpet runner stuff.

I usually don't put much stock in hot glue, but hot glue seemed a pretty good way to tack down the burlap to the wood, and the vinyl to the burlap, so it seems unlikely the burlap will unravel from the staples, and unlikely the vinyl will let go. Vinyl sometimes is resistant to adhesives, but the combination of heat and the hot glue seemed to stick pretty good to this flavor of vinyl.

Subwoofer Experiments-keytrap_covered_1_web.jpg

Subwoofer Experiments-keytrap_covered_2_web.jpg

And then installed on each side of the keyboards. One end sitting on the 88, and the front edge of each trap supported by velcro so it can be removed when necessary.

From listening position, the eyes and ears have unobstructed access to the monitor woofers, but an ant on the keyboards would not be able to see the monitors, obscured by the keyboard reflection devices. They are shaped to be fairly high but not get in the way of the hands playing keyboards.

Subwoofer Experiments-keytrap_installed_1_web.jpg

Subwoofer Experiments-keytrap_installed_2_web.jpg

Subwoofer Experiments-keytrap_installed_3_web.jpg
Old 13th September 2014
Lives for gear
So there is some improvement--

Subwoofer Experiments-fr_keyboardtrap.jpg

Eight Null Improvements:
- Null at 8.4 KHz eliminated (1.5 dB improvement).
- Null at 4.3 KHz eliminated (4 dB improvement).
- Null at 3.5 Khz eliminated (2 dB improvement).
- Null at 2 KHz eliminated (6.5 dB improvement).
- Null at 900 Hz improved (6 dB improvement).
- Null at 370 Hz eliminated (6 dB improvement).
- Null at 120 Hz improved (6 dB improvement).
- Null at 57 Hz improved (3.5 dB improvement).

There are fourteen peak improvements (reductions of at least 1 dB) at 7.9 K, 6.9 K, 5.3 K, 4.6 K, 4 K, 3.1 K, 2.2 K, 1.7 K, 1.5 K, 1.06 K, 680, 290, 223, 47.7 Hz.

Three slight regressions in nulls at 39 Hz, 260 Hz, 2.6 KHz.

One more serious regression in nulls at 470 Hz.

No significant regressions in peaks noted.


Here is the frequency response with the driverack auto-eq--

Subwoofer Experiments-fr_keyboardtrap_eq.jpg


Using the earlier-described method of exporting the 96 measurement per octave data from REW, import into Open Office spreadsheet and calculate Average Absolute Deviation (where a figure of 0 would mean the ideal ruler-flat frequency response)--

No EQ measurements--
Before Treatment Avg Deviation = 4.777 dB
Five Traps Avg Deviation (two large front corner traps, three large but thin rear traps) = 4.733 dB
Temp Test Absorbers (Previous five traps plus various sheets of rockwool laid about the nearfield) Avg Deviation = 3.812 dB
Front Traps (Previous five traps plus large nearfield trap surrounding the monitor speakers) Avg Deviation = 4.179 dB
Keyboard Traps (Previous six traps plus two nearfield traps placed on sides of keyboards) Avg Deviation = 3.771 dB

Auto-EQ measurements--
Five Traps Auto-EQ Avg Deviation = 3.457 dB
Front Traps Auto-EQ Avg Deviation = 3.225 dB
Keyboard Traps Auto-EQ Avg Deviation = 2.739 dB

So by this mathematical measurement the frequency response appears to be getting flatter.

So far, it appears that the driverack auto-EQ (adjusting third-octave centers against pink noise) appears capable of approximately 1 dB improvement in average absolute deviation.
Old 2nd October 2014
Lives for gear
So spent some hours studying up on making a subwoofer. Settled on a dayton audio um 18-22 18" speaker, which apparently does OK in sealed enclosures 4 to 8 cubic feet or bigger.

Spent awhile deciding on a shape which would blow the bass right in my face, a cab shaped so it can be located the same distance away from my ears as the main monitors. My compact workspace has the main monitors at point-blank range. Getting a big conventional-shaped 18" subwoofer cab that close could get tricky.

After entertaining some hard-to-build shapes such as stretched tetrahedrals and chopped up dodecahedrons, came up with a fairly conventional angled-floor-monitor, with slices cut out of the two front corners, so it could slide up close in the corner between the keyboard stand and equipment rack.

Been some years since I made a speaker cab. The ones I made long ago held together good enough for road work but were hardly over-built. Many speaker builders advise well-braced sealed subs, so most sides are "almost 1 inch" thick, with a 6" waffle pattern bracing of 0.75" wood. Most outside panels two nominal half-inch panels glued together. Then 1 X 2 bracing going front-to-back, side-to-side, top-to-bottom.

These pictures show a lot of unsanded bondo on screw holes. Lots of screws. Used aliphatic glue most joints, epoxy on a couple of joints I didn't cut accurate enough.

Subwoofer Experiments-subwoofer_1_web.jpg

Subwoofer Experiments-subwoofer_2_web.jpg

Rather than buy inset handles and a speaker connection tray, made em out of wood. Cutouts for the handles and speaker connection. Then glue and seal a perimeter of 0.75" wood around the inside of each cutout. And glue a 3/4" plywood cover over the back of each inside wood perimeter, so I get airtight, non-resonance recessed handles and speaker connection bay.

Subwoofer Experiments-subwoofer_3_web.jpg

The speaker baffle board is 1.75" thick, glued up of two half-inch plywood sheets plus a 3/4" inch plywood sheet. One inch of the baffle board sits "inside" the sides, and the 3/4" plate glues down on-top of the sides.

Speaker mounting screw holes and inserts already added before painting, covered with masking tape and painted over in the picture.

Subwoofer Experiments-subwoofer_4_web.jpg

Testing the connection and filling the cab with polyfill. The woofer is so heavy, made sure to leave enough slack in the cable so that connections can be made with the speaker sitting next to the cabinet. If I later have to service the speaker, would hate to have to hold that beast with one hand while making connections with the other hand.

Supposedly one should use about 1 pound of polyfill per cubic foot on large sealed subs. I figured something like 7 cubic feet for this cab, neglecting the space taken up by bracing, so I stuffed 6 pounds of polyfill in the thing.

Subwoofer Experiments-subwoofer_5_web.jpg

Until the back bracing and the back was screwed and glued on, thumping on the thick wood made a low-pitched dead thump. Never made a cab this thick and tight befor. Had a moment of trepidation after adding the back and the glue dried. Holding one's head in the speaker opening and thumping the cab, was making about a 1 second decay resonance rather high pitched, at least 400 Hz. Sounded kinda like thumping a steel tank.

I believe this was air reverbing in the rigid box, rather than resonance in the box itself, because the resonance went away completely after the first pound of fiber fill.

I've never worked with fiberfill before. Interesting stuff. Take a big handful and shred it off into little bits, squeeze it together and it is a new ball that looks just like the original piece you just shredded. Very light and compressible, almost "anti-gravity". Stick handfuls up the side of the cab and it stays in place, and new handfuls stick by static or something to what's already there. Very fluffy stuff. Looks dense, but mostly air.

Subwoofer Experiments-subwoofer_6_web.jpg
Old 2nd October 2014
Lives for gear
The cab turned out a bit heavy to tote, so lugged the empty cab to the office and final-mounted the speaker only a few feet away from install location. Put six "feet" of 3" X 4" self stick furniture bumper felt under the cab to protect the floor and make it easy to slide around. I'd hate toting something this heavy to gigs, got too old. It is very difficult to pick up with the speaker loaded.

The old QSC amp I plan to use, was on loan to a buddy, and won't be available til year end, so am using an old crest la 601 that supposedly should deliver a max in the ballpark of 450 watts in bridge mode. Indications so far show this amount of power ought to be plenty sufficient for nearfield listening at non-deafening levels.

Subwoofer Experiments-subwoofer_7_web.jpg

The sub is sitting at a 45 degree angle to the keyboard stand and the equipment rack, with the front cutouts allowing it to be tucked in real close.

Subwoofer Experiments-subwoofer_8_web.jpg

Subwoofer Experiments-subwoofer_9_web.jpg

Subwoofer Experiments-subwoofer_10_web.jpg

Been listening to it all day. Sounds solid, has a visceral feel even at low volumes. Haven't measured the frequency response yet. Probably do that tomorrow.

Currently have it crossed over at 200 Hz, 24 dB linkwitz-riley slope. When I was testing in the shop was running it full-range with some bass-heavy Sade, and it was doing surprisingly good in the mids. Maybe I'll end up crossing it lower than 200 Hz, but seems good up that high or higher.

No problem with bottom with this speaker, sounds very clean tight and punchy. It will get obnoxiously loud with no bottoming, wheezing or cabinet rattles. Beats heck out of the old woofers I was playing with.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.

Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump