New Posts  All Forums:Forum Nav:

Room Measurement & Treatment

post #1 of 41
11/19/11 at 10:16am
Thread Starter 
Since this topic has seemed to generate a lot of interest recently, and there's really no particular thread that I could find concentrating on it, thought it would be a good idea to start one. Seems to me this particular forum could benefit from it.

This hopefully could be a common area for questions on how to measure, what your resulting measurements mean, and possibly suggestions on methods to treat for improvement. Either it will propagate, or just die off.

To kick things off, I spent all of about 1/2 hour today taking some key measurements which are shown following. My measurement set-up consists of the following:

- Behringer UCA202 USB sound card (approx $27)
- Radio Shack Digital SPL meter (approx $50)
- Room Equalization Wizard (REW) software (free)
- Camera tripod (varies)
- Associated RCA cables (check your basement)

Graph of sound pressure level across 20-20KHz (full range) for sub and front right and left speakers, with 1/3 smoothing. Note that the RS SPL meter is not very accurate above 7K-10KHz.



Graph of sub only from 15-200Hz



Waterfall generation from the sub graph. The waterfall view shows you the sound pressure level across the frequency range, but also includes the decay/time of the signal on the Z axis. I can see from my room that I have a peak and long decay time in the 15-25Hz range, with the rest of the frequency range decaying nicely within 300ms or less.



Next is a spectrogram of same sub signal. This is another way of viewing the energy and decay over time, with time on Y axis, frequency on X axis, and intensity of the signal in color. Here you will see again the higher intensity and longer decay in the 15-25Hz range with steadier decay and intensity throughout the remainder of the frequency range.



Next up is ETC. This graph is based on a normal full range measurement sweep of my front right speaker only. It depicts the initial impulse energy at T=0, and subsequent impulses in time along the X axis.



Here is ETC for the front left speaker only



Some observations based on my limited interpretation skills:

1) subwoofer response is good overall. I have read that what I am seeing in the 15-20Hz range (peak and long decay) typically is a modal issue and probably not a lot I can do about it.
2) ETC measurements:
For FR: There is a spike immediately after initial impulse (near T=0) which is most likely a reflection off a surface right by the speaker. Another spike at around 2.4ms. That spike is nearly 30dB below initial....does it even need addressed?
For FL: Very similar to FR....same question

I am generally pleased with the ETC results (unless I'm missing something).

Any comments, corrections or whatever are appreciated.
Reply
Reply

AVS Top Picks

post #2 of 41
11/19/11 at 10:54am
A couple general comments...

1/3 octave smoothing is not useful for analysis Its great for sales, but not for when buying and when we actually want to see what is being effectively ignored.

Unless you are using the blocking method to isolate and identify individual indirect paths, one should use the hardware loopback compensation method and NOT use the 'set IR peak to T=0' as you have no idea what is the actual time of flight.

As displayed, the time delta between direct and later arriving energy wavelets is handy for setting delay lines in arrays or distributed elements, but of little use in what we commonly encounter.

And in this case, knowing only that the initial high gain reflection arrives at ~2.2ms after the direct signal in the right speaker tells us little without utilizing the empirical blocking method.

Conversely, especially for folks starting out on this adventure, setting the hardware loopback to compensate for hardware propagation delays and having an accurate direct readout of each arrival time (actual time of flight) that directly correlates to the actual distance travels provides a few additional 'sanity checks' to verify the actual vector path of travel and boundary incidence.


And a sparse high gain indirect signal so soon after the direct arrival is not a desirable reflection.

As displayed the only viable option for determining the indirect path is via the common blocking method that intercepts the vector path of the arriving energy. And from this one can effectively 'walk' the reflection back to a point of incidence that will identify the optimal placement of treatment. One will also want to repeat the test with the mic capsule remaining in the exact same location (you might mark this with a plumb bob until ALL testing is completely over) in order to provide proof of performance verification and to determine if further modification of the treatment is necessary.

Also, a general observation ignoring the response models where a defined initial signal delay gap is defined and thus ignoring the termination of such a space; an easy way to evaluate later (post ISD) arriving energy returns is to set the response to display the log scale and to observe any sparse reflections that 'stand out' in terms of exceeding the the linear slope of the rate of decay of gain. Such later arriving energy wavelets are generally candidates for diffusion that will tend to reduce the gain while increasing the spatial and temporal diffusion of the energy, thus resulting in a more diffuse densely populated region about the original location of the sparse high gain reflection visible in the ETC response.

And for general purposes, one can guestimate the average time (X axis) window required by setting the time to a value equal to approximately twice the length of the room - where each foot correlates to a value of ~1.13 feet. Thus a room 20 feet deep might start with a time window set to (20ft x 1.13 ft/ms =) ~22.6 or 23 ms. This can be adjusted if necessary, but one would also have reason for concern (or at least further investigation) if they are observing a high gain reflection after that time!

Also, you might want to set the Y axis scale to dBfs rather than to a percentage.
Reply
Reply
post #3 of 41
11/19/11 at 11:14am
God help me I actually understood everything you just said.

So, using the blocking method I could actually use my Omnimic ETC (IR peak @ T0) to find problematic vector sources, no? Yesss, I understand I should be using a more complete tool, but it's what I have to work with - for now.
Reply
Reply
post #4 of 41
11/19/11 at 11:31am
Yes, that is the only method you have available to you without an accurate time of flight determination.

Also, as you are using the blocking technique, remember that sound has size, so in blocking, be it with something larger such as an absorbent 'pillow' or your hand, that the energy spike may not disappear as many might expect, since the energy one is attempting to block is comprised of broadband energy, and the energy exhibiting wavelengths larger than the blocking member utilized will diffract around it and still be received by the mic. So you are only blocking a portion of the energy.

Taking this still further...previously it has been asked how large a unit of treatment need be. Remembering the relationship of obstacle size and wavelength and that wavelengths larger than an obstacle diffract around said obstacle, and also using the ETC, one can easily determine the effectiveness of such treatment and if any modifications might be necessary - both with regards to size and with regards to effective absorption (and diffusion) and any reflections that may remain as artifacts with the use of an absorber - as few are totally absorptive

.......(and also just how absorptive diffusors can be despite being manufactured of reflective material! ... as well as the additional edge diffraction effects of absorbers - all thanks to Ron Sauro for his recent insightful research into edge diffraction that has literally turned the measurement world for both absorption and diffusion on its head and invalidated current ASTM-C423 and ISO-354 procedures!)
Reply
Reply
post #5 of 41
11/19/11 at 11:55am
Quote:
Originally Posted by dragonfyr View Post

Also, as you are using the blocking technique, remember that sound has size, so in blocking, be it with something larger such as an absorbent 'pillow' or your hand, that the energy spike may not disappear as many might expect, since the energy one is attempting to block is comprised of broadband energy, and the energy exhibiting wavelengths larger than the blocking member utilized will diffract around it and still be received by the mic. So you are only blocking a portion of the energy.

Understood, but I can determine the frequency band for an offending spike by using ETC-by-octave measurements, or will this not work, in practice?
Reply
Reply
post #6 of 41
11/19/11 at 3:33pm
Thread Starter 
Quote:
Originally Posted by dragonfyr View Post

A couple general comments...

1/3 octave smoothing is not useful for analysis Its great for sales, but not for when buying and when we actually want to see what is being effectively ignored.

Unless you are using the blocking method to isolate and identify individual indirect paths, one should use the hardware loopback compensation method and NOT use the 'set IR peak to T=0' as you have no idea what is the actual time of flight.

And in this case, knowing only that the initial high gain reflection arrives at ~2.2ms after the direct signal in the right speaker tells us little without utilizing the empirical blocking method.

Conversely, especially for folks starting out on this adventure, setting the hardware loopback to compensate for hardware propagation delays and having an accurate direct readout of each arrival time (actual time of flight) that directly correlates to the actual distance travels provides a few additional 'sanity checks' to verify the actual vector path of travel and boundary incidence.


And a sparse high gain indirect signal so soon after the direct arrival is not a desirable reflection.

Also, a general observation ignoring the response models where a defined initial signal delay gap is defined and thus ignoring the termination of such a space; an easy way to evaluate later (post ISD) arriving energy returns is to set the response to display the log scale and to observe any sparse reflections that 'stand out' in terms of exceeding the the linear slope of the rate of decay of gain. Such later arriving energy wavelets are generally candidates for diffusion that will tend to reduce the gain while increasing the spatial and temporal diffusion of the energy, thus resulting in a more diffuse densely populated region about the original location of the sparse high gain reflection visible in the ETC response.

And for general purposes, one can guestimate the average time (X axis) window required by setting the time to a value equal to approximately twice the length of the room - where each foot correlates to a value of ~1.13 feet. Thus a room 20 feet deep might start with a time window set to (20ft x 1.13 ft/ms =) ~22.6 or 23 ms. This can be adjusted if necessary, but one would also have reason for concern (or at least further investigation) if they are observing a high gain reflection after that time!

Also, you might want to set the Y axis scale to dBfs rather than to a percentage.

Thanks for the reply....not trying to sell anything here, just thought that posting some common graphs for analysis at this point would get things started. What resolution do you want to see on a full range sweep? Following is 1/12 smoothing. Yes I have some issues around the crossover region (and perhaps others) but wasn't planning on addressing that right now just so the ETC measurement aspect didn't get lost in other analysis.

Yes, I know of the loopback method and will retest and repost after running another sweep or two. My pesky family wanted me to spend some time with them today so I didn't try that for these measurements as I had about an hour free. Do you have a link which further describes the "blocking method"?

If you're interested, I would be happy to attach the REW file for these subsequent runs and you could slice and dice as you see fit.



So, you never commented about how the sub response looks?
Reply
Reply
post #7 of 41
11/23/11 at 7:17am
Thread Starter 
For those interested, I made some progress with the ETC measurements. Guess what? It works!
As an experiment, I decided to track down the largest reflection in my ETC measurement of my front right speaker, which was measured at 2.22ms or 2.5 feet after the initial direct speaker signal. The direct speaker signal is normalized at T= 0 on the x axis, with the spike measured at the 2.22 ms later as shown following:



Next I needed to do some simple calculations in order to use the "string method" to find out specific areas to check where the actual reflection surface is. First I measured the distance from my Radio Shack SPL meter to the actual acoustic center of the speaker being measured. That distance was 12' (as best as I could measure). The reflection being measured takes an additional 2.5' to arrive, which means that the total distance of the reflected signal is 12' + 2.5' or 14' 6".

The concept of using the string method is that you take a length of string which is equal to the distance traveled of the signal (in this case 14'6"), tie off one end to your measurement mic, and the other end to the acoustic center of your speaker. If you have helpers available, they can hold either end. In my case I didn't so I had to gin up tie offs. You then take up (I just slid a finger along the length) the string slack between the mic and speaker extending it in all axis's and note what boundary surfaces it touches.
In my case, it touched the ceiling about 1/2 way between the mic and speaker. Following are a few pic for clarification.

Mic at measurement position with one end of string tied off:



Other end of string tied off at speaker:



Location on ceiling where string touched:



Visual of overall string path with my "treatment stick" holding it up:



I next took another measurement while holding that insulation batt to the ceiling. Following is that measurement with time reference noted. Notice that the spike has been attenuated. Original measurement shown directly following again for ease of comparison.




I have the same spike occurring on the left speaker at similar time, so appears that a ceiling panel is in my future.

The other thing I noticed and was able to measure was the amount of reflections coming off of my front row seating surfaces. I measured with and without blankets on the seats, and the results were significant.

During my investigation, another member who has been helping me pointed out some strange anomalies occurring with the direct signal, which brought into question the quality of the speakers being used. My fronts are very old (1988 or so) American Acoustics (was under Mitel) that I have been using just because I had them available, with intent to upgrade at some point. With Christmas just around the corner, I foresee two new fronts in my future, and I suspect Santa will be arrive early this year

Also, I have been questioning my FR and FL positioning since I built. If you look at the speaker location, you'll notice that it is sitting on the floor with a portion of the bottom woofer blocked by the stage, with the second one having my screen framing in the way (should have did Big's minimalist style). I can't believe this is having any positive contribution to my audio response, especially where I have a lot of anomalies in the crossover range. So, planning on bookshelves to replace those. I can sit them on stands just above the bottom screen frame and take all that mechanical interference out of the picture.

So, I won't be doing much more with this effort until those arrive.

Hopefully this has demonstrated how ETC can be used to locate reflections, and verify effectiveness of treatment. It's really not that complicated.
Reply
Reply
post #8 of 41
11/23/11 at 7:31am
Quote:
Originally Posted by fotto View Post

It's really not that complicated.

nice work!
Reply
Reply
post #9 of 41
11/23/11 at 8:17am
Quote:
Originally Posted by fotto View Post

Hopefully this has demonstrated how ETC can be used to locate reflections, and verify effectiveness of treatment. It's really not that complicated.

Best practical explanation of how to use ETC that I have seen. I haven't had time to read through all the threads on this but this was very helpful.
Reply
Reply
post #10 of 41
11/23/11 at 8:36am
I think you have a bright future ahead of you marketing those "treatment sticks" of yours. Screw panels, I'm using those babies from now on!

On the ETC front, nice work. Makes me want to dig up a similar setup I purchased a few months ago to start my own measurements.
Reply
Reply
post #11 of 41
11/23/11 at 12:38pm
Thread Starter 
Quote:
Originally Posted by localhost127 View Post

nice work!

Quote:
Originally Posted by stockmonkey2000 View Post

Best practical explanation of how to use ETC that I have seen. I haven't had time to read through all the threads on this but this was very helpful.

Quote:
Originally Posted by HDvids4all View Post

I think you have a bright future ahead of you marketing those "treatment sticks" of yours. Screw panels, I'm using those babies from now on!

On the ETC front, nice work. Makes me want to dig up a similar setup I purchased a few months ago to start my own measurements.

LOL...wasn't really considering where everyone COULD go with a "treatment stick", but of course we're all upstanding citizens here.

Anyway, thanks for the feedback. There so much one can do with REW, and I've only used a very small portion of it. Being an audio neophyte, this is such a great learning tool.

For example...you can take a sweep say from 20Hz to 20KHz and the software allows you to generate a sine wave which can follow your mouse along your response curve generating whatever frequency your mouse is at. Being able to listen to peaks or nulls in your room response is pretty cool. You can stop at a particular point (peak or null), get up and walk away from your chair and see how it increases or decreases.

Another example is that I used it for adjusting my crossover. I took sweeps from 60hz to 120Hz in 10Hz increments, then displayed them all and selected the best one.
Reply
Reply
post #12 of 41
12/12/11 at 10:27am
Quote:
Originally Posted by fotto View Post



During my investigation, another member who has been helping me pointed out some strange anomalies occurring with the direct signal, which brought into question the quality of the speakers being used. My fronts are very old (1988 or so) American Acoustics (was under Mitel) that I have been using just because I had them available, with intent to upgrade at some point. With Christmas just around the corner, I foresee two new fronts in my future, and I suspect Santa will be arrive early this year

fotto, seeing your setup here you might want to consider that it might not be your speakers, or it might not be JUST your speakers. Why not try moving one of your speakers out a couple feet in front of the screen wall and any other obstructions, run another sweep, and zoom way in on the initial sound arrival? That would answer the question of how much is diffraction due to your screen wall vs. how much is due to the speaker itself (whether the time alignment or cabinet diffraction). If it's diffraction from the screen wall, your brand new whiz-bang speakers will do the same thing and you'd best put your energy into changing the screen wall or just bite the bullet and have exposed speakers.
Reply
Reply
post #13 of 41
12/12/11 at 10:52am
first step for any audio calibration is a near field measurement of all speakers ... otherwise you may be trying to fix room response when you have a speaker problem.
Reply
Reply
post #14 of 41
12/12/11 at 10:58am
Thread Starter 
That's a good idea, thanks. I already have new Axiom M22 bookshelf speakers on stands in that space, with old speakers moved out. I like the M22's and plan on keeping them, so won't be fussing around any more with the old ones.

I will however measure with the M22's behind and in front of the screen wall to compare. Since the M22's are sitting on stands (higher than that horizontal screen support), I am expecting some improvement just based on that alone.

Also, it will be a much easier job comparing this with 20lb speakers vs. 80lbs+
Reply
Reply
post #15 of 41
12/12/11 at 11:02am
Thread Starter 
Quote:
Originally Posted by Dennis Erskine View Post

first step for any audio calibration is a near field measurement of all speakers ... otherwise you may be trying to fix room response when you have a speaker problem.

Sure Dennis, just pile on more work!. I think I'm going to need more vacation.
Reply
Reply
post #16 of 41
12/12/11 at 1:10pm
Quote:
Originally Posted by fotto View Post
Sure Dennis, just pile on more work!. I think I'm going to need more vacation.
Here you go Floyd, a while back I downloaded this Measurement/calibration sequence from Dennis.

 

RoomMeasurementSet-up.zip 4.990234375k . file
Reply
Reply
post #17 of 41
12/12/11 at 1:22pm
Thread Starter 
Thanks Mike, most helpful. I wasn't looking forward to trying to round that info up.
Reply
Reply
post #18 of 41
12/12/11 at 2:17pm
Quote:
Originally Posted by mtbdudex View Post

Here you go Floyd, a while back I downloaded this Measurement/calibration sequence from Dennis.

Whoa, fantastic and I didn't even download it yet! Will check it out later when I have time!

I haven't ever done nearfield measurements of my speakers, it would be a really good idea to do that I suppose.
Reply
Reply
post #19 of 41
12/13/11 at 3:13am
...a quick note. If you notice a consistent HF roll off in your near field measurements of your L/C/R speakers DO NOT be boosting the HF with an EQ. While it is possible (and I've seen this plenty of times) that a tweeter is blown, if all speakers have the same issue it may not be a speaker problem. It could be an issue with your microphone (not calibrated, broken, or just cheap) or measurement device. So, before you start to boost HF, electrically check your tweeter. If it is blown, replace it. If it's not, leave it be. If you boost the HF without knowing the cause of the problem, you could be buying new tweeters anyway.
Reply
Reply
post #20 of 41
12/13/11 at 6:04am
Not that I'm there yet, but after reading Dennis's doc (Thanks!) where should the person(s) doing the test be while the tests are run? Wouldn't YOUR position in the room affect the results? I'm thinking seated at each position as it is being tested?
Reply
Reply
post #21 of 41
12/13/11 at 11:34am
The people in the room are big salt water bags.
Reply
Reply
post #22 of 41
12/13/11 at 1:06pm
LOL. "Ugly giant bags of mostly water."
Reply
Reply
post #23 of 41
12/13/11 at 1:08pm
According to Bender we're "Meat bags."
Reply
Reply
post #24 of 41
12/13/11 at 1:20pm
OT but trivia time. Star trek original series, there was a creature who went after "bags of salt", name the episode.
Reply
Reply
post #25 of 41
12/13/11 at 1:34pm
Thread Starter 
"The Man Trap"?
Reply
Reply
post #26 of 41
12/13/11 at 1:44pm
Quote:
Originally Posted by fotto View Post

"The Man Trap"?

Ding - was that was too easy?
Nancy Crater.....


really looked like this
Reply
Reply
post #27 of 41
12/13/11 at 2:23pm
Quote:
Originally Posted by fotto View Post

"The Man Trap"?

I hope you had to look that one up. Otherwise, you're scaring me.

I swear I can remember another TOS episode (or was it a film?) with a similar reference, but it didn't seem to be forthcoming with a bit of googling.
Reply
Reply
post #28 of 41
12/14/11 at 5:16am
Quote:
Originally Posted by fotto View Post

For those interested, I made some progress with the ETC measurements. Guess what? It works!
As an experiment, I decided to track down the largest reflection in my ETC measurement of my front right speaker, which was measured at 2.22ms or 2.5 feet after the initial direct speaker signal. The direct speaker signal is normalized at T= 0 on the x axis, with the spike measured at the 2.22 ms later as shown following:



Next I needed to do some simple calculations in order to use the "string method" to find out specific areas to check where the actual reflection surface is. First I measured the distance from my Radio Shack SPL meter to the actual acoustic center of the speaker being measured. That distance was 12' (as best as I could measure). The reflection being measured takes an additional 2.5' to arrive, which means that the total distance of the reflected signal is 12' + 2.5' or 14' 6".

The concept of using the string method is that you take a length of string which is equal to the distance traveled of the signal (in this case 14'6"), tie off one end to your measurement mic, and the other end to the acoustic center of your speaker. If you have helpers available, they can hold either end. In my case I didn't so I had to gin up tie offs. You then take up (I just slid a finger along the length) the string slack between the mic and speaker extending it in all axis's and note what boundary surfaces it touches.
In my case, it touched the ceiling about 1/2 way between the mic and speaker. Following are a few pic for clarification.

Mic at measurement position with one end of string tied off:



Other end of string tied off at speaker:



Location on ceiling where string touched:



Visual of overall string path with my "treatment stick" holding it up:



I next took another measurement while holding that insulation batt to the ceiling. Following is that measurement with time reference noted. Notice that the spike has been attenuated. Original measurement shown directly following again for ease of comparison.




I have the same spike occurring on the left speaker at similar time, so appears that a ceiling panel is in my future.

The other thing I noticed and was able to measure was the amount of reflections coming off of my front row seating surfaces. I measured with and without blankets on the seats, and the results were significant.

During my investigation, another member who has been helping me pointed out some strange anomalies occurring with the direct signal, which brought into question the quality of the speakers being used. My fronts are very old (1988 or so) American Acoustics (was under Mitel) that I have been using just because I had them available, with intent to upgrade at some point. With Christmas just around the corner, I foresee two new fronts in my future, and I suspect Santa will be arrive early this year

Also, I have been questioning my FR and FL positioning since I built. If you look at the speaker location, you'll notice that it is sitting on the floor with a portion of the bottom woofer blocked by the stage, with the second one having my screen framing in the way (should have did Big's minimalist style). I can't believe this is having any positive contribution to my audio response, especially where I have a lot of anomalies in the crossover range. So, planning on bookshelves to replace those. I can sit them on stands just above the bottom screen frame and take all that mechanical interference out of the picture.

So, I won't be doing much more with this effort until those arrive.

Hopefully this has demonstrated how ETC can be used to locate reflections, and verify effectiveness of treatment. It's really not that complicated.

Nice work!!
Reply
Reply
post #29 of 41
12/14/11 at 5:36am
Thread Starter 
Thanks Glenn. Further analysis on this over x-mas break will be done using the "loopback as timing reference" option in REW for better accuracy (total time of flight of the signal). There is ongoing discussion on that topic in threads here and HTS. Theory is whether we really know where our speaker's "acoustic center" is. If not, then we are guessing at that endpoint.
Reply
Reply
post #30 of 41
12/14/11 at 5:41am
Quote:
Originally Posted by fotto View Post

Thanks Glenn. Further analysis on this over x-mas break will be done using the "loopback as timing reference" option in REW for better accuracy (total time of flight of the signal). There is ongoing discussion on that topic in threads here and HTS. Theory is whether we really know where our speaker's "acoustic center" is. If not, then we are guessing at that endpoint.

The bottom line is that you want to reduce the strength of the arrival on the ETC, right? If I can do that, regardless whether I know the acoustical center of the speaker or not, I am done. You understand the tool, you did it once, you can do it again. Good enough, assuming once you listen to your reference material you like the sound.
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home

AVS Top Picks