Impulse response - Home Theater Forum and Systems - HomeTheaterShack.com

Old 07-18-12, 06:26 PM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Impulse response

I understood basics on graph on spl tab as well as waterfall tab. I am trying to understand impulse response. Here are the questions I have.

1. Is there any frequency range that I need to select while measuring for example 20 -200hz or 20-20khz?
2. I read that it is useful to find first reflections. How do I do that? In the second graph, I think I saw at 37msec, there is second peak at 78,
3. How is windowed and ETC are used.
4. In the help, the pictures have spectrum spread more. But, I see very less spectrum. Not sure what I am missing.
5. What is the blue line. It appears and disappears by clicking on "window"

I am posting two of the graphs here.
Attached Thumbnails

Attached Files sukumar2.mdat (2.91 MB)

Last edited by chinni123; 07-18-12 at 06:36 PM.
chinni123 is offline

Old 07-19-12, 01:50 PM
REW Author

John

Join Date: Apr 2006
Location: UK
Posts: 6,306
Re: Impulse response

To track down reflections you need to make full range measurements, e.g. 0 to 20kHz. If you measure over a narrower range, e.g. 20Hz to 200Hz, the reduced frequency range results in a more spread out impulse response (it lacks the high frequency content that is needed to have fast changes).

Before the frequency response is generated the impulse response is 'Windowed', that means it is multiplied by the response shape shown in the Window trace. The Window is used to make sure only the part of the impulse we are interested in is used to make the frequency response. For example, if the response is mostly noise after some point the end of the window could be set so that the noise part is cut off. Another use would be to exclude a reflection to see how the response looks without the effect of the reflection.

The ETC trace shows the envelope of the impulse response (a bit like showing the amplitude range the response covers). Reflections are easier to see using the ETC. It can also be easier to have the vertical axis set to dB rather than % FS.
JohnM is offline
Old 07-19-12, 02:44 PM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Re: Impulse response

Quote:
JohnM wrote: View Post
To track down reflections you need to make full range measurements, e.g. 0 to 20kHz. If you measure over a narrower range, e.g. 20Hz to 200Hz, the reduced frequency range results in a more spread out impulse response (it lacks the high frequency content that is needed to have fast changes).

Before the frequency response is generated the impulse response is 'Windowed', that means it is multiplied by the response shape shown in the Window trace. The Window is used to make sure only the part of the impulse we are interested in is used to make the frequency response. For example, if the response is mostly noise after some point the end of the window could be set so that the noise part is cut off. Another use would be to exclude a reflection to see how the response looks without the effect of the reflection.

The ETC trace shows the envelope of the impulse response (a bit like showing the amplitude range the response covers). Reflections are easier to see using the ETC. It can also be easier to have the vertical axis set to dB rather than % FS.
Thanks for the details. It is good to know that this graph is useful when measurement is done from 0 to 20khz.

I also observed and make sure the following.

1. The check boxes at the bottom says " Jul 17 22:56:40", "windowed", "etc" , "window", "step response". Even when I unselect "jul 17 22:56:40", I am assuming that all are referring to that sample only?
2. All these are overlapping
3. Clicking on windowed is causing to fill until 500ms. It seems content is same as "jul 17 22:56;40" until the blue line except little filling is missing.
4. size of window is controlled by "IR window"

I guess it takes time for me to understand. I will read again help.
As a beginner I want to start understanding ETC first. I see second peak at 35msec. I guess reflection is coming at 35msec. How do I get how much more distance traveled and how to identify location on wall location reflection occurred?
Attached Thumbnails

Last edited by chinni123; 07-19-12 at 02:53 PM.
chinni123 is offline
Old 07-20-12, 05:40 AM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Re: Impulse response

Since I find second spike happened at 38m sec, the distance traveled = 1.13*38=42.94 feet. Is it right?

My room is 21 feet long and 14 to 16 feet wide (two different length for back to middle and middle to front) and 8 feet high. How do I interpretate where is the reflection.

Appreciate any input. Please let me know if my impulse response is not right.

I am adding one more question. I took this measurement with both left speaker and sub and frequency from 20 to 20khz. is it not right? I wonder whether this test need to be done with subwoofer disconnected with just one speaker from 0 to 20khz. is it true?

Last edited by chinni123; 07-20-12 at 06:10 AM.
chinni123 is offline
Old 07-20-12, 08:55 AM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Re: Impulse response

I think time range is not right for my ETC. Could you please provide what is the reasonable range for time and what should be entered (For example micro seconds or milli seconds etc).
chinni123 is offline
Old 07-20-12, 09:47 AM
Senior Shackster

Join Date: Dec 2011
Posts: 239
Re: Impulse response

Simplify view of room; imagine as rectangular box with speaker and you located in the room where the microphone would be. As you look around you see images of the speaker from every wall. Looking closely you see images formed with a single mirror, two mirrors, and upwards. Some images just use two walls, or floor and ceiling, others involve three, four, or all six surfaces.

With sound it is much the same, but in the mirror analogy if the mirrors are slightly hazy, or tinted as such, each reflection looses some intensity; and so reflecting sound is scattering, being absorbed, and turned into heat.

Base on: http://www.hometheatershack.com/foru...-sukumar2.mdat using full range measurement I created two screen shots. An ETC picture and a spectrogram picture.

ETC has 0.1ms smoothing. Spectrogram uses 5.0ms window using Blackman-Harris 4 filtering for both sides of window. Good spectrogram views usually require tweaking all the settings a little.

ETC shows the numerous reflections alluded to in my mirror analogy.

The reason the 38ms reflection stands out is that it has lots of coherent high frequency energy in it, as seen in spectrogram at 38ms, 8kHz region(In spectrogram reflections appear as horizontal bands). Most likely it is double reflection; sound from speaker passes microphone (direct sound t=0), bounces off wall behind microphone (first reflection), passes microphone again (one of the stronger early reflections), passes speaker, bounces off wall behind speaker(second reflection, and finally arrives back at microphone 38ms later.

Andrew
Attached Thumbnails

Barleywater is offline
Old 07-20-12, 10:09 PM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Re: Impulse response

Quote:
Barleywater wrote: View Post
Simplify view of room; imagine as rectangular box with speaker and you located in the room where the microphone would be. As you look around you see images of the speaker from every wall. Looking closely you see images formed with a single mirror, two mirrors, and upwards. Some images just use two walls, or floor and ceiling, others involve three, four, or all six surfaces.

With sound it is much the same, but in the mirror analogy if the mirrors are slightly hazy, or tinted as such, each reflection looses some intensity; and so reflecting sound is scattering, being absorbed, and turned into heat.

Base on: http://www.hometheatershack.com/foru...-sukumar2.mdat using full range measurement I created two screen shots. An ETC picture and a spectrogram picture.

ETC has 0.1ms smoothing. Spectrogram uses 5.0ms window using Blackman-Harris 4 filtering for both sides of window. Good spectrogram views usually require tweaking all the settings a little.

ETC shows the numerous reflections alluded to in my mirror analogy.

The reason the 38ms reflection stands out is that it has lots of coherent high frequency energy in it, as seen in spectrogram at 38ms, 8kHz region(In spectrogram reflections appear as horizontal bands). Most likely it is double reflection; sound from speaker passes microphone (direct sound t=0), bounces off wall behind microphone (first reflection), passes microphone again (one of the stronger early reflections), passes speaker, bounces off wall behind speaker(second reflection, and finally arrives back at microphone 38ms later.

Andrew
Thanks a lot for replying with detailed explanation. I never might have imagined the way sound is reflecting. First, I am trying to make my ETC curve to look like yours. Here is my understanding or questions.

1. You are trying to reduce white space on y axis by changing limits to %70 to -%20. It perfectly makes sense to observe rise/fall clearly.
2. I think the time range before was 500msec. I don't remember. I am able to change it to -.005sec to .05 sec implies -5misec to 50milli sec.

What is suggested time range for ETC?

3. I used controls to change smoothing to .1. Without changing time from -5 to 50msec, this setting is getting lost after tabing the control. Don't know why?

4. Now graph shows close details. What is first refection here. I don't know first reflection is first spike after t=0 or first highest spike after t=0

a. I saw at 955 u second (approximately at 1misec) there is -19.1%
b. I saw at 2.86 msecond there is %32.8 (highest spike after t=0

Since the a condition is very close to t=0, do I need to ignore that?

5. If b is right, I need to multiple 2.86*1.13=3.23 feet. So first reflection traveled 18 feet (distance from speaker to listening position)+ 3.23 = 21.23 feet. Is it right?

How do I locate this position?

I am going to understand spectrogram. Appreciate all your help.
chinni123 is offline
Old 07-21-12, 03:50 PM
Senior Shackster

Join Date: Dec 2011
Posts: 239
Re: Impulse response

First major peak of impulse response is usually taken as direct sound.

If microphone is 18ft from speaker, and first major reflected sound arrives 2.86ms later, then yes, reflected sound traveled about 3.2fr further than direct sound.

It becomes a matter of geometry as to which path was taken. for instance, if microphone is 3.2ft/2 from wall behind it from speaker, then that is an answer. Likewise if speaker is close to side wall then it could be a reflection from side wall.

Floor and ceiling first reflections are classic first reflections. Example: microphone is placed at about ear height for seated person, and tweeter/midrange drivers are about same height, lets call it 40". Assuming hard ceiling of 9ft, reflection will be strong, and occur at midpoint between equal height microphone and speaker. For 18ft microphone distance that is 9ft. Microphone is 68" from ceiling. Microphone to ceiling reflection point equals speaker to ceiling reflection point. (68"^2+108"^2)^0.5 /12" *2 = 21.27ft. Could be the answer. Corresponding path length of corresponding floor reflection is about 19.2ft, giving roughly 1ms delay between direct sound and floor reflection.

Path length differences ultimately relate to summation and cancellation frequencies, and determine much of frequency response at microphone location.

Intersections of walls, floor, and ceilings make great reflectors, where all three meet reflect most strongly, at higher frequencies they reflect back in the direction of the source.

Andrew
Barleywater is offline
Old 07-21-12, 07:50 PM
Senior Shackster

Join Date: Dec 2011
Posts: 239
Re: Impulse response

I review more impulse responses of your originally posted mdat. There are significant peaks in each about t= -17ms of main peak. They are highly suggestive of microphone input mixer putting part of what the microphone is picking up back into the sweep as it is being played.

Please check your sound device settings. Perform a loopback calibration measurement of your sound card, save and post the mdat. I would like to see it. Symptoms also include exaggerated base with poor decay, and a type of comb filtering. These need to be ruled out.

Thanks,

Andrew
Barleywater is offline
Old 07-21-12, 09:59 PM Thread Starter
Senior Shackster

Join Date: Jan 2012
Posts: 114
My System
Re: Impulse response

Quote:
Barleywater wrote: View Post
I review more impulse responses of your originally posted mdat. There are significant peaks in each about t= -17ms of main peak. They are highly suggestive of microphone input mixer putting part of what the microphone is picking up back into the sweep as it is being played.

Please check your sound device settings. Perform a loopback calibration measurement of your sound card, save and post the mdat. I would like to see it. Symptoms also include exaggerated base with poor decay, and a type of comb filtering. These need to be ruled out.

Thanks,

Andrew
Thanks a lot Andrew. I read both your posts and now I understood better. I was told there is no need to calibration for TASCAM 144 II sound card. I never did calibration on it before. I will do calibration. I put monomix to computer. I will do test again and post it here.

I am in the middle of adding absorbers and bass traps. I just added three absorbers for 3 of windows. I need to see how it changes.

Thanks for detailed explanation with examples to map time delay with possible reflection points.
chinni123 is offline

 Bookmarks

 Tags impulse , response

Message:
Options

## Register Now

Random Question
Random Question #2

User Name:
OR

## Log-in

Human Verification

In order to verify that you are a human and not a spam bot, please enter the answer into the following box below based on the instructions contained in the graphic.