External signal for measuring reverbation time in REW

[FrankSehnal]

Hello, I´m quite newbie here, so please take that in account.

I recently bought UMIK-1 measuring microphone. I´m interested especially in measuring reverbation time of bigger rooms, RT60. Yesterday I tried to measure RT60 using sweep sound implemented in REW. However at school, we used gun shot as impulse. I assume that for RT60 measurement of large halls I´ll need omni-directional speaker to reproduce the sweep sound, which I don´t have. That´s why I wonder if I can use external impulse for this measurement, like gun shot or jabbing inflatable baloon. Is it possible within REW or I have to use external programme like Audacity or some other DAW? If using Audacity, do I still take an advantage of UMIK-1 calibration?
Thank You in advance for a response.

Frank Sehnal

 

[John Mulcahy]

You could capture an impulse response directly using recording software of your choice, save it as a WAV file and then import that to REW using File -> Import Impulse Response. You can then use the Change Cal button on the measurement to apply the calibration file. It will likely be something of a struggle to balance not overloading the mic with getting a decent signal to noise ratio. If you are interested in low frequency behaviour then most speakers are fairly omni-directional at LF, otherwise you could try several speakers pointing in different directions to have a sort of poor man's omni, at least in the horizontal plane. Or make a series of sweep measurements with a single speaker pointing in different directions while using a timing reference (so the results are time aligned) and sum them afterwards. When measuring in a large space adjust the Analysis preferences truncation setting to make sure enough of the IR is retained and use a long sweep setting (512k or 1M).

 

[Matthew J Poes]

What John told you is right on, Ive done exactly this. However, when I’m in your shoes I actually use different software. I use my iPhone with AudioTools and the IR module with a large Balloon pop. I connect a USB mic using the camera kit adapter. Impulso Pro also works for this purpose and is a cheaper option.

I love REW and use it for 90% of my measurements. However I sometimes run into situations, like yours, where the classic gun shot or balloon pop is just a better option.

Unfortunately If you don’t use a proper Omni source (which are crazy expensive for the speakers, the RT60 values will be pretty far off and inconsistent. If you don’t have an Omni speaker and you want a full range RT60 measurement, maybe try the other softwares and gun or balloon. If you want to stick with REW (which is otherwise a better software package IMO) you may try constructing an Omni speaker. I did something similar using a bunch of Bose Cube speakers from a yard sale find of a Bose surround system. I connected it to a pro subwoofer and made a wooden frame for the cubes to bolt to. It worked but couldn’t play loud enough for a really large space. I used it in a relatively small Conference room/meeting space and it was fine. I’ve also rented an Omni speaker. That isn’t too pricey if you can plan carefully and do all measurements in a day.

 

[FrankSehnal]

You could capture an impulse response directly using recording software of your choice, save it as a WAV file and then import that to REW using File -> Import Impulse Response. You can then use the Change Cal button on the measurement to apply the calibration file. It will likely be something of a struggle to balance not overloading the mic with getting a decent signal to noise ratio. If you are interested in low frequency behaviour then most speakers are fairly omni-directional at LF, otherwise you could try several speakers pointing in different directions to have a sort of poor man's omni, at least in the horizontal plane. Or make a series of sweep measurements with a single speaker pointing in different directions while using a timing reference (so the results are time aligned) and sum them afterwards. When measuring in a large space adjust the Analysis preferences truncation setting to make sure enough of the IR is retained and use a long sweep setting (512k or 1M).

Thank You, John,

that´s exactly what I needed to know! I´ll play around with balloon using Audacity for now and later I´ll try to construct omnidirectional speaker and use the sweep within REW.

with Friendly regards,

Frank Sehnal

 

[FrankSehnal]

What John told you is right on, Ive done exactly this. However, when I’m in your shoes I actually use different software. I use my iPhone with AudioTools and the IR module with a large Balloon pop. I connect a USB mic using the camera kit adapter. Impulso Pro also works for this purpose and is a cheaper option.

I love REW and use it for 90% of my measurements. However I sometimes run into situations, like yours, where the classic gun shot or balloon pop is just a better option.

Unfortunately If you don’t use a proper Omni source (which are crazy expensive for the speakers, the RT60 values will be pretty far off and inconsistent. If you don’t have an Omni speaker and you want a full range RT60 measurement, maybe try the other softwares and gun or balloon. If you want to stick with REW (which is otherwise a better software package IMO) you may try constructing an Omni speaker. I did something similar using a bunch of Bose Cube speakers from a yard sale find of a Bose surround system. I connected it to a pro subwoofer and made a wooden frame for the cubes to bolt to. It worked but couldn’t play loud enough for a really large space. I used it in a relatively small Conference room/meeting space and it was fine. I’ve also rented an Omni speaker. That isn’t too pricey if you can plan carefully and do all measurements in a day.

Thank You, Mathew, for a different point of view on this topic. I´ll take that in account. I´d like to order few Bose speakers on Ebay, they not very expensive, and try to play around, and compare both methods. I´ll let You know in this thread afterwards.

I have one more question. Which of RT60 (T20 and T30) should I use for the purpose of evaluation reverberation time according to the en iso 3382 normative?

Best regards,

Frank Sehnal

 

[FrankSehnal]

In previous post I published incorrect normative.The limit of reverbation time is given by czech normative čsn 73 0232, described by T/To ratio. "T" is measured RT60 in REW and "To" is optimal RT60, given by the same normative. Which of RT60 (T20 and T30) in REW should I use for evaluation?

Thank You.

Frank Sehnal

 

[Matthew J Poes]

In previous post I published incorrect normative.The limit of reverbation time is given by czech normative čsn 73 0232, described by T/To ratio. "T" is measured RT60 in REW and "To" is optimal RT60, given by the same normative. Which of RT60 (T20 and T30) in REW should I use for evaluation?

Thank You.

Frank Sehnal

John can better explain the exact differences, but the choice between T20, T30, or other options has to do with how much of the actual decay is used in the calculation vs inferred. Most rooms that are small to medium sized will decay too quickly and the noise floor too high, causing it to decay into noise and throw off the calculation. In your case...the microphone you are using actually has a high noise floor. As such I suggest using shorter time intervals like T20. ToPT might even be a better option. Again, John is more the authority here, but that is my perspective.

I would also compare all of them, in a very large space there shouldn't be many differences. If there are, I would take a look at the noise floor to see what is going on.

As for the Bose speakers, could be worth a shot, but keep in mind the output limitations. In large spaces that isn't a good option (I was taught that the quietest signal that I need to measure needs to be about 10dB above the noise floor of the room, so if the room's noise floor is like 30dB, that is 40dB, and might mean that the initial test signal needs to start at something like 100dB at the mic, which in a large space, might be 120dB at the source). I paid $25 for the entire setup so it was worth the effort. Its still my goal to build my own omni-speaker for this. Cutting the panels for a dodecahedron is easier said than done for someone like me and CNCing is too expensive, so just waiting for someone to feel bad for me and cut the panels at a reasonable cost.

 

[FrankSehnal]

John can better explain the exact differences, but the choice between T20, T30, or other options has to do with how much of the actual decay is used in the calculation vs inferred. Most rooms that are small to medium sized will decay too quickly and the noise floor too high, causing it to decay into noise and throw off the calculation. In your case...the microphone you are using actually has a high noise floor. As such I suggest using shorter time intervals like T20. ToPT might even be a better option. Again, John is more the authority here, but that is my perspective.

I would also compare all of them, in a very large space there shouldn't be many differences. If there are, I would take a look at the noise floor to see what is going on.

As for the Bose speakers, could be worth a shot, but keep in mind the output limitations. In large spaces that isn't a good option (I was taught that the quietest signal that I need to measure needs to be about 10dB above the noise floor of the room, so if the room's noise floor is like 30dB, that is 40dB, and might mean that the initial test signal needs to start at something like 100dB at the mic, which in a large space, might be 120dB at the source). I paid $25 for the entire setup so it was worth the effort. Its still my goal to build my own omni-speaker for this. Cutting the panels for a dodecahedron is easier said than done for someone like me and CNCing is too expensive, so just waiting for someone to feel bad for me and cut the panels at a reasonable cost.

Thank You for the answer, Mathew,

Could I ask You, what kind of Bose speaker did you buy? I intend to ask my friend about the performance of the speaker to calculate if the speakers are able to play 120 dB loud.
I have one question regarding the hight noise floor of my microphone. I thought that the noise floor is the property of measured enviroment, which is recorded by microphone. Am I correct? Or is it a property of the microphone?

with friendly regards

Frank Sehnal

 

[Matthew J Poes]

Thank You for the answer, Mathew,

Could I ask You, what kind of Bose speaker did you buy? I intend to ask my friend about the performance of the speaker to calculate if the speakers are able to play 120 dB loud.
I have one question regarding the hight noise floor of my microphone. I thought that the noise floor is the property of measured enviroment, which is recorded by microphone. Am I correct? Or is it a property of the microphone?

with friendly regards

Frank Sehnal

My he speakers I used were a 5.1 Bose surround setup with the 5 dual cube speakers. I made a mount with wood so that the speakers could radiate from five sides and then turned them so one pointed slightly up and one pointed slightly down on each side. I’m sure it wasn’t perfect.

I’m fairly sure it can’t play 120dB even at a close distance of 1 meter. Still you may not need that level.

Now as for noise floor, a microphone has self noise. This is sometimes referred to as the equivalent noise or equivalent noise floor. The capsule itself has self noise, about 20-25dB for the Panasonic element that was cloned in your mic. The preamplifier also has an electrical noise floor. I can’t explain why, but for some reason these USB microphones have a high noise floor, it’s something like 30dB or even higher at some frequencies.

There are people who have posted FFT spectral analysis of the noise floor that make it look fairly low. I actually learned this from John after seeing a mic with - spl levels and wondering why. The FFT bins the frequency amplitude data into such narrow bins that there is practically no energy in the high frequencies. It makes the noise floor look really low. However once you start widening those bins as you would for assessing noise floor or any normal “by the books” room analysis, the energy in a given bin increases substantially and this the noise floor goes up.

 

[FrankSehnal]

My he speakers I used were a 5.1 Bose surround setup with the 5 dual cube speakers. I made a mount with wood so that the speakers could radiate from five sides and then turned them so one pointed slightly up and one pointed slightly down on each side. I’m sure it wasn’t perfect.

I’m fairly sure it can’t play 120dB even at a close distance of 1 meter. Still you may not need that level.

Now as for noise floor, a microphone has self noise. This is sometimes referred to as the equivalent noise or equivalent noise floor. The capsule itself has self noise, about 20-25dB for the Panasonic element that was cloned in your mic. The preamplifier also has an electrical noise floor. I can’t explain why, but for some reason these USB microphones have a high noise floor, it’s something like 30dB or even higher at some frequencies.

There are people who have posted FFT spectral analysis of the noise floor that make it look fairly low. I actually learned this from John after seeing a mic with - spl levels and wondering why. The FFT bins the frequency amplitude data into such narrow bins that there is practically no energy in the high frequencies. It makes the noise floor look really low. However once you start widening those bins as you would for assessing noise floor or any normal “by the books” room analysis, the energy in a given bin increases substantially and this the noise floor goes up.

Thank You for explanation, Matthew! I´ll take that in account. Have a nice day.

 

[John Mulcahy]

As a follow-up to this, some work has been done on using directional loudspeakers in place of a dodecahedron, this paper discusses it: Nikolaos M. Papadakis and Georgios E. Stavroulakis, "Low Cost Omnidirectional Sound Source Utilizing a Common Directional Loudspeaker for Impulse Response Measurements", Appl. Sci. 2018, 8(9), 1703; Available online at https://doi.org/10.3390/app8091703

 

[Matthew J Poes]

As a follow-up to this, some work has been done on using directional loudspeakers in place of a dodecahedron, this paper discusses it: Nikolaos M. Papadakis and Georgios E. Stavroulakis, "Low Cost Omnidirectional Sound Source Utilizing a Common Directional Loudspeaker for Impulse Response Measurements", Appl. Sci. 2018, 8(9), 1703; Available online at https://doi.org/10.3390/app8091703

Thanks for posting John. I found this article a week or two ago.

Have you tried to replicate their results? It seems doable but a ton of work. I haven’t had time to really do it properly. To work out the post-processing calculation piece, I attempted to do it with just a handful of measurements and it took quite a while. Maybe once you get practiced it wouldn’t be as big a chore.

Do you have any though how important it is to follow their approach and physically move the speaker along an arch rather than simply rotate the speaker? This was part of my issue, the process to move a powered 10” monitor speaker on a stand was a pain.

 

[John Mulcahy]

I'm reviewing the data at the moment to look at the post-processing requirements. Dr. Papadakis got in touch to let me know about the paper and I've let him know that I posted it here, he might respond.

 

[Matthew J Poes]

I'm reviewing the data at the moment to look at the post-processing requirements. Dr. Papadakis got in touch to let me know about the paper and I've let him know that I posted it here, he might respond.

That’s great to hear.

I also wonder what kind of bias a directional source has in a domestic sized space like this.

 

[John Mulcahy]

The post processing turned out to be fairly straightforward. I set up rectangular windows for each measurement (important to preserve the level noise floor for acoustic parameter extraction), aligned them to their start times and used Vector Average to sum them. I've added a button on the All SPL graph for the next beta to align a set of measurements according to their IR start times.

 

[Matthew J Poes]

The post processing turned out to be fairly straightforward. I set up rectangular windows for each measurement (important to preserve the level noise floor for acoustic parameter extraction), aligned them to their start times and used Vector Average to sum them. I've added a button on the All SPL graph for the next beta to align a set of measurements according to their IR start times.

Very helpful and great to hear! I’ll be happy to test this idea out and post my experience once the beta is released.

Is the lack of alignment what causes the HF’s to drop when you vector average a set of IR’s?

 

[John Mulcahy]

Yes, averaging signals that aren't aligned creates comb filtering at HF. In the case of the multiple measurements the vector average process is another way to sum the various impulses, in that case alignment isn't especially critical.