Bob's Acoustically Transparent Screen tests 2006/07/08
SMX720 is an acoustically transparent screen for Home Theatre. Basically it's a window shade material, that happens to be:
a) perfectly white -- it doesn't change colours that hit it much at all (4400 material ISF video testing)
b) has a gain of 1.1 (if you include the holes), or 1.3 (if you pretend the holes are not there)
c) is pretty good acoustically (no THX certification)
d) not too bad with moire
I have enough SMX720 to do two screens (one for my unfinished basement, and one for later when I finish it).
I wanted to find out how best to place speakers behind the SMX720, and what I could expect from performance.
Planned Tests
COMPARE BARRIERS
- Nothing - no baffle - test #28
- Nothing - wood baffle (reflective) - test #25 (bass seemed louder)
- Nothing - 2" rockwool baffle (absorptive) - test #1, speaker 24" from wall, speaker 56" from microphone
- Draper Luma 2 - 2" rockwool baffle - test #2, speaker 24" from wall, speaker 56" from microphone, speaker 11" behind screen (sounded bassy)
- 1/4" x 48" x 30" Plywood , 2" rockwool baffle - test #3, speaker 24" from wall, speaker 56" from microphone, plywood leaning against rockwool (sounded VERY bassy and thin)
- SMX720 - no baffle - test #27
- SMX720 - wood baffle (reflective) - test #26 (13" from screen)
- SMX720 - 2" rockwool baffle (absorptive) - test #4 speaker 24" from wall, speaker 56" from microphone, speaker 13" behind screen
COMPARE DISTANCE (baffled 2" rockwool)
- speaker 2" behind screen - test #11 (2" thick rockwool is touching screen)
- speaker 3" behind screen - test #10
- speaker 4" behind screen - test #9
- speaker 6" behind screen - test #8
- speaker 8" behind screen - test #7
- speaker 10" behind screen- test #6
- speaker 12" behind screen - test #5
- speaker 16" behind screen - test #12
COMPARE OFF AXIS (16" behind screen)
- Nothing - straight speaker, straight mic - test #20
- Nothing - 15 degree speaker, straight mic - test #21
- Nothing - 30 degree speaker, straight mic - test 24
- Nothing - straight speaker, 10 degree mic - test #14
- Nothing - straight speaker, 20 degree mic - test #15
- Nothing - straight speaker, 30 degree mic - test #19
- SMX720 - straight speaker, straight mic - test #12
- SMX720 - 15 degree speaker, straight mic - test 22
- SMX720 - 30 degree speaker, straight mic - test 23
- SMX720 - straight speaker, 10 degree mic - test #13
- SMX720 - straight speaker, 20 degree mic - test #16
- SMX720 - straight speaker, 30 degree mic - test #17 and #18
Comb Filter (aka how to read the data)
Before doing these tests I did three things:
a) I actually did a bunch of pre-testing to finalize how I was going to do the above tests. Including a discovery that the speaker face to SMX720 has A LOT of reflections that cause A LOT of comb filtering (as can also be seen a bit in the impulse response of tests 26 and 27, but closer to the screen it was MUCH worse than that) -- so either put your speaker a good distance back or cover the speaker face with absorption. This undocumented pre-result made me do the COMPARE DISTANCE tests with absorption in front of the speaker, rather than with a speaker without a baffle. Another discovery was the need for the floor absorber, otherwise the comb filtering was worse (see the little spike in the impulse response at 2ms in test #17 that goes away with test #18).
b) I listened to the left chanel of the CD "INXS" with the song "Suicide Blond". I tried to do A-B tests (A = speaker alone, B = SMX material between me and the speaker) as fast as possible. Around 19 seconds in there's an instrument I'll call a trumpet but is probably some electronic thing. I thought that instrument sounded recognizeable with the SMX, but I had the feeling that it sounded different than without the SMX, sort of thinner, missing something. I'm unsure if this is all in my mind.
c) I rummaged through "Master Handbook Of Accoustics" by Alton Everest to find out what I should be looking for with Comb Filtering.
There are three rules for Comb Filtering:
1) amplitude -- if it's small enough, you can't hear it.
2) closeness - if the peaks and valleys are close together in frequency, you can't hear it
3) no comb filtering is better. Yes a room will have other things that cause comb filtering (wall/floor reflections, other speakers, a single speaker's multiple drivers at their crossover points, etc), but none of the others will have as short a decay time as the screen will (see image of page 369 below).
However, if the Impulse Response shows a dB level at some ms point that is less than the direct sound that's under the "Reflection Inaudible" line in Figure 16-4, then people probably can't hear it.
But to be sure that it's imperceptible we want the comb filter peaks and nulls to be +/- 1 dB or less in amplitude.
This can be done by insuring that the delayed reflection is at least 9dB lower than the main signal (pg 379).
We want the peaks and valleys to be as close as possible:
Although the spacing between peaks should be as close as possible, the spacing between peaks must be narrower than table 17-1 for it to be inaudible:
from:
"Audio, Acoustics and Perception" by Earl Geddes, Chapter 3 Psychoacoustics
For very short time delayed secondary signals, two effects are possible, the first has to do with localization confusion and the second is a comb filter effect that I will talk about next. When the secondary signal is on the order of 1 ms or less then it interferes with the natural head diffraction and causes the ear to get confused between what are true localization cues and what are diffraction effects.
...
Note also that a delay of 3ms would have the most pronounced effect in the critical mid-band. [6ms in Figure 3-13 has its major effect lower than 100hz] Recall that signals arriving within a 25 ms integration time for the ear are fused into a single event and we can see that the resulting sound will take on the characteristics of the filters shown in the figure. This effect is called coloration.
Looking at speaker A in the above diagram, we can see that there are wall reflections of 4ms and 9ms (room effects like these are removed from the below Bode Frequency Response tests due to the 2ms gate time), and a speaker face to screen reflection of 1ms (these are apparent in the tests below).
Looking at speaker B and the below diagram, we can see another potential source < 1ms comb filtering -- the holes in the screen -- but there's little evidence of this in my tests below.
| delay | I think |
|---|
| < 0.25ms | should avoid comb filtering, due to pinna (outer ear) directional queue requirements |
| < 2ms | comb filtering is evident see Figure 3.12 in Gedlee Chapter 3. The Comb Filter is completely linear across the frequency spectrum from 0 to 10,000 Hz. |
| 2ms to 3ms | refer to Figure 3.13 in Gedlee Chapter 3 |
| 3ms | a very strange effect modifies the behavior of the wave - there is a shift to the bass from the mids - with a 6db dip evident around (roughly) 175Hz. |
| 3ms to 6ms | ? |
| 6ms to 12ms | I would expect the Hass effect to begin to take over, coloration of the pure signals with a shift to the bass is pretty evident. The blended 1/3 octave comb filter at this point has a 6db dip - but the frequency shift brings it slightly below 100Hz. |
| 12ms to 25ms | By 12 Ms the 6dB dip is at (roughly) 42Hz 0 - where it would be barely noticable if at all......... |
| > 25ms | ignore |
Apparatus
Microphone: Calibrated IBF-EMM8 Electret Omnidirectional Measurement Microphone, high-spl version (sold by ETF5)
Microphone Stand: Quiklok A-85
PreAmp: solid state CD MP-21 (sold by ETF5)
Amplifier: Yamaha RXV2400, MD/Tape in, DSP set to StraightThrough, bass management crossover at 80 Hz.
SoundCard/computer: Creative SoundBlaster USB MP3 Model SB0270, on Pentium IV 3.2ghz
RTA Software: AcoustiSoft R+D Modal Analyzer (aka RplusD), version 1.0.46
Roxul RHF-80 rockwool, 2'x4'x2" batts
Room is L shaped, with the main section about 223" x 264" x 99" (at the microphone) Room Measurements Basement Pictures
Speaker: B&W DM604 S3
Technical Specifications : DM604 S3 |
| Description |
3-way vented-box system |
| Dimensions |
Height: 994mm (39.1in)
Width: 236mm (9.3in)
Depth: 398mm (15.7in)
|
| Net Weight |
29.2kg / 64.2lb |
| Freq. Response |
39Hz - 22kHz ± 3dB on reference axis |
| Freq. Range |
-6dB at 30Hz and 42kHz |
| Sensitivity |
90dB spl (2.83V 1m) |
| Normal Impedance |
8 ohms (minimum 3 ohms) |
| Power Handling |
25W - 200W into 8 ohms on unclipped programme |
| Drive Units |
Unit 1: 1x 25mm (1in) metal dome high-frequency
Unit 2: 1x 180mm (7in) woven Kevlar® cone midrange
Unit 3: 2x 180mm (7in) aluminium cone bass
|
| Finish |
Cabinet: Black Ash or Sorrento (light oak) vinyl
Grille: Black Cloth |
| Dispersion |
Description: Within 2dB of response on reference axis
Horizontal: over 40° arc
Vertical: over 10° arc
|
| Harmonic Distortion |
2nd & 3rd harmonics <1% 42Hz - 20kHz (90dB spl, 1m) |
| Crossover Frequency |
450Hz, 4kHz |
| Max. Recommended Cable Impedance |
0.1 ohms |
Method
RplusD setup was the same as always - start with the loopback test, with Play to allow only Wave/MP3 and all others muted, and Recording to allow Line In and all others muted, and then tie it into the pre-amp/microphone and receiver/speaker left channel. With all the noisemakers in my home turned off (air conditioner, fleurescent lights, dehumidifiers, air-to-air-exchanger), and the computer as far away as possible in a little absorber chamber, I got a 28db signal to noise ratio with RplusD.
I had some comb filter reflection problems with my old camera stand's vertical 1.5" shaft which was 9" behind the microphone sensor, but the new microphone stand has a boom so long that's no longer an issue.
I wanted a distance as far away from the speaker as possible (far-field-ish) that didn't have reflections, so after doing a bit of geometry (distance in feet direct, vs distance in feet reflected off the floor and walls) I chose 6' which seemed to work out fine.
Initially I had the speaker up on a table to keep it away from the floor, but then the tweater a little over a foot from the SMX720 4" diameter roll near the ceiling that I was getting reflections from that. I tried putting the speaker on top of a 7" tall UPS (very heavy), but it was still too high. Eventually I just put it on the ground which meant the tweater center is 37" off the ground and 37" from the SMX720 roll over head.
I put my 4'x2'x12" absorber on the floor between the speaker and the microphone, and that took care of the last reflections within 3ms after the direct sound hit the microphone.
I don't know if the ceiling absorption helped, but it probably didn't hurt.
I started with the microphone: 39" from floor, 79" from front wall, 79" from right wall, 145" from left wall, 54" from ceiling, 177" from rear wall.
I put the computer into a little absorber shelter -- probably dropped the noise by about 1dB (maybe more), but at least it's away from the microphone.
The boxes are full of speakers, so they're heavy.
The testing area.
The microphone stand, and the computer to the right on the floor behind the speaker boxes (the two computers on the desk are turned off).
For more information about how the baffle/soffet and insulation was constructed around the speaker, click here
The rockwool baffel
 |
 |
 |
 |
| Perhaps I would have been better off with 1/2" of linacoustic over the speaker face. |
 |
Observations
RplusD data file with all the tests in it. 13MB
COMPARE BARRIERS
| 28 |
Nothing - no baffle |
|
 |
|
 |
| 25 |
wood baffle (reflective) (bass seemed louder) |
 |
 |
 |
 |
| 1 (100%) |
2" rockwool baffle (absorptive) - speaker 24" from wall, speaker 56" from microphone |
|
 |
 |
 |
| 2 |
Draper Luma 2 - 2" rockwool baffle - speaker 24" from wall, speaker 56" from microphone, speaker 11" behind screen (sounded bassy) |
 |
 |
 |
 |
| 3 |
1/4" x 48" x 30" Plywood , 2" rockwool baffle - speaker 24" from wall, speaker 56" from microphone, plywood leaning against rockwool (sounded VERY bassy and thin) |
 |
 |
 |
 |
| 27 (90%) |
SMX720 - no baffle |
 |
 |
 |
 |
| 26 (88%) |
SMX720 - wood baffle (reflective) - (13" from screen) |
 |
 |
 |
 |
| 4 (98%) |
SMX720 - 2" rockwool baffle (absorptive) - speaker 24" from wall, speaker 56" from microphone, speaker 13" behind screen |
 |
 |
 |
 |
COMPARE DISTANCE (baffled 2" rockwool)
| 11 (99%) |
speaker 2" behind screen, (2" thick rockwool is touching screen) |
 |
 |
 |
 |
| 10 (99%) |
speaker 3" behind screen |
 |
 |
 |
 |
| 9 (99%) |
speaker 4" behind screen |
 |
 |
 |
 |
| 8 (99%) |
speaker 6" behind screen |
 |
 |
 |
 |
| 7 |
speaker 8" behind screen |
 |
 |
 |
 |
| 6 (97%) |
speaker 10" behind screen |
 |
 |
 |
 |
| 5 (96%) |
speaker 12" behind screen |
 |
 |
 |
 |
| 12 (98%) |
speaker 16" behind screen |
 |
 |
 |
 |
COMPARE OFF AXIS (16" behind screen)
| 20 |
Nothing - straight speaker, straight mic |
 |
 |
 |
 |
| 21 |
Nothing - 15 degree speaker, straight mic |
 |
 |
 |
 |
| 24 |
Nothing - 30 degree speaker, straight mic |
 |
 |
 |
 |
| 14 |
Nothing - straight speaker, 10 degree mic |
 |
 |
 |
 |
| 15 |
Nothing - straight speaker, 20 degree mic |
 |
 |
 |
 |
| 19 |
Nothing - straight speaker, 30 degree mic |
 |
 |
 |
 |
| 22 (99% best) |
SMX720 - 15 degree speaker, straight mic |
 |
 |
 |
 |
| 23 (98%) |
SMX720 - 30 degree speaker, straight mic |
 |
 |
 |
 |
| 13 (96%) |
SMX720 - straight speaker, 10 degree mic |
 |
 |
 |
 |
| 16 (92%) |
SMX720 - straight speaker, 20 degree mic |
 |
 |
 |
 |
| 17 (91%) |
SMX720 - straight speaker, 30 degree mic |
 |
 |
 |
 |
| 18 (91%) |
SMX720 - straight speaker, 30 degree mic (same as 17, except absorber moved) |
 |
 |
 |
 |
COMPARISONS OF ABOVE DATA
| 1 vs 6 |
|
|
 |
| 21 vs 22 |
|
|
 |
| 1 vs 22 |
|
|
 |
| 25 vs 28 |
|
|
 |
 |
