Using the new ISO 18233 Method to test Transmission Loss and calculate STC-An Experiment

[Matthew J Poes]

After learning that this new approach exists and after a suggestion of how to accomplish this from @John Mulcahy, I believe that I have accurately tested this approach and come up with basically accurate results. It is not perfect, and I learned a lot in this experiment, but the biggest thing I learned is that this is a MUCH better method for the more average DIYer like us to use in testing TL. The advantage in improved S/N is significant.

Here is the Raw Data

The Purple is the measurements inside the theater room, Brown is the response in the room immediately adjacent to the theater. The measurements were taken roughly 6-7 feet apart. Inside the theater, about 3 feet from the wall, and outside the theater about the same distance. The green line is the transmission loss. Because this was done with a Sine Sweep measurement, I have IR's and the Green is actually the transfer function of the wall.

This can then be entered into a spreadsheet in accordance with ASTM E413 to calculate STC. I haven't actually done this manually before, so when I made this up, I did it according to that document, but am not 100% certain it is right. None the less, I believe this is right, and here are the results of that.


The flat orange line represents the STC curve I wanted to test. The others represent either the actual TL or the corrected curves of each for comparison. The wall should have been an STC of 73, but it is not shocking that I fell short of that. There are always some issues that cost you a few STC points here and there. You can see that my main losses are due to some dips between 100hz and 500hz and again above 1khz up to 2khz.

Now compared to the old method, this method doesn't allow a traditional means of spatial averaging. You can't just move a mic around while a tone plays and the RTA records (which removes the effects of resonances and smooths the response quite a bit). This smoothing is really needed for calculating STC, so to do that you have to keep rerunning the sweeps in different locations. It didn't work so well for me, so I need to play with that when I have time. I got a few odd results and averaging them gave a TL curve that didn't make sense. I think I accidentally captured a null in the theater.

For comparison, here is a TL measurement done using the old traditional method. This method quickly ran into a noise floor problem in the receiver room.


Now a few things to keep in mind. When I did that traditional method, I had used a method to calibrate the mic that I thought was right, but turned out to be quite a bit off. About 10dB's. When I then used an actual calibrator, I realized I had made a mistake in the SPL calibration, but the relative difference doesn't change. Just add about 10dB's to those numbers to know what they should have been. In any case, you can see that the RTA method couldn't detect anything above 80hz accurately with tones at 90dB. While I haven't shared the results, I actually got fairly accurate TL results with test tones as low as 85dB's with a sine sweep. The only thing that changed when I made it louder was the area above 3khz, where TL increased a small amount once I increased the SPL to over 100dB. It was amazing how much better the SS method was for getting TL results, as compared to this traditional method.

 

[bkeeler10]

I'm not sure I totally follow how the S/N ratio is a problem with the traditional method but is not with this method. After all, the room's noise floor doesn't change and the electronics noise floor doesn't either, as far as I can tell. What am I missing? I have thought that broadband noise might corrupt the accuracy of each frequency band, but I don't think that's right.

 

[Matthew J Poes]

I'm not sure I totally follow how the S/N ratio is a problem with the traditional method but is not with this method. After all, the room's noise floor doesn't change and the electronics noise floor doesn't either, as far as I can tell. What am I missing? I have thought that broadband noise might corrupt the accuracy of each frequency band, but I don't think that's right.

A sine-swept IR has a better signal to noise ratio than traditional RTA. It can resolve signal information below the room noise floor. This is how it can read distortion information below the room noise floor. @John Mulcahy is really the one to explain this.

In my example however you can see that the lowest noise detected on the receiver side is far below the room noise floor.

 

[bkeeler10]

That's very interesting. Am I correct in assuming that if the noise you're trying to measure is significantly (several dB) above the noise floor of the room, which method you use is of less consequence?

 

[Matthew J Poes]

That's very interesting. Am I correct in assuming that if the noise you're trying to measure is significantly (several dB) above the noise floor of the room, which method you use is of less consequence?

I believe the answer is yes.

https://www.researchgate.net/public...tion_by_means_of_impulse_response_measurement

I’m this the comparison showed very high agreement.

I don’t recall why, but when I learned this technique we had been taught to set the level of the speaker so that it’s output was 10dB above the room noise floor in the receiver room. It was taught as if the measurement wasn’t valid if that wasn’t true. I’ve been trying to find my notes from that class to see if I wrote down why. At this point I can’t see why exactly 10dB would be needed but if that is right, then take my partition above, the noise floor was something like 20-30dB at high frequencies and the transmission loss was around 80dB. You would need to play back at 100-110dB at the wall in the source room for that test to work. That means the speaker would need to play back around 130dB at 1 meter. How many speakers can do that? Not even most dodecahedron speakers can play that loud.

As I understand it however, the IR method would still work fine at much lower volume levels. I got accurate or close to accurate results with the volume on the source room by the wall of around 80-100db. Basically between 100-120dB at 1 meter. When I did the 80dB sweep the TL I got above 7khz was slightly off, about 3dB too low. If you look at my speaker response, it tilts down, so it probably was in the 70-75dB range, which suggest the spl on the receiver side may have been near 0dB and thus well below the noise floor of the mic.

 

[DanDan]

Very interesting. Certainly better than trying to use high levels of Pink Noise. I used to presume that a high sweep level was mandatory in a typical 40dB noise floor room, in order to achieve a decent S/N level, to see a 60dB decay. John Mulcahy explained why this is not necessary. But if I may present an analogous situation. A radio tuner focusses on just one frequency and receives it well, ignoring a vast amount of energy throughout that Band. I like to think of REW 'hearing' only a single frequency at any particular moment during a Sweep. The SPL of ambient noise at that particular frequency might then be regarded as the 'Floor'. Clearly way way below wideband noise.
Very very roughly speaking, from experiments, I like to think/guess there is up to 20dB improvement in S/N or capability of hearing down into the noise floor. I haven't thought it through extensively but I am guessing that would include the noise of the Mic.
DD

 

[Matthew J Poes]

Very interesting. Certainly better than trying to use high levels of Pink Noise. I used to presume that a high sweep level was mandatory in a typical 40dB noise floor room, in order to achieve a decent S/N level, to see a 60dB decay. John Mulcahy explained why this is not necessary. But if I may present an analogous situation. A radio tuner focusses on just one frequency and receives it well, ignoring a vast amount of energy throughout that Band. I like to think of REW 'hearing' only a single frequency at any particular moment during a Sweep. The SPL of ambient noise at that particular frequency might then be regarded as the 'Floor'. Clearly way way below wideband noise.
Very very roughly speaking, from experiments, I like to think/guess there is up to 20dB improvement in S/N or capability of hearing down into the noise floor. I haven't thought it through extensively but I am guessing that would include the noise of the Mic.
DD

20dB is probably the maximum level below the actual ambient noise of the room that the sweep method I used would work.

My problem (and reason for needing this test procedure to work) is that the transmission loss or that wall is so significant that to reliably detect anything above the noise floor of the room required a very high volume level in the source room. 100dB at the wall is over 120dB at 1 meter from the speaker.

I’m not totally sure why but the standard says the source shouldn’t be near the wall being tested. Most of the research articles used a single pro speaker that was in the corner opposite the wall. That is what I did.

When I was doing these tests initially I was hitting the limits of the noise of the room and mic. Needed to find a way to resolve that. This works great and I’ve already found customers for it. Acoustics firms that lack the capacity themselves. Shh....don’t tell them how easy it is to figure out. Even the math to calculate TL wasn’t very time intensive.

 

[DanDan]

All Building Acoustics testing requests a distance from boundaries. It's either 1M or 1.2 around here.
Not sure if for the same reason, but B&K say that a Measuring Mic close to the operators body will show anomalies of up to 6dB around 400Hz.
DD

 

[Matthew J Poes]

All Building Acoustics testing requests a distance from boundaries. It's either 1M or 1.2 around here.
Not sure if for the same reason, but B&K say that a Measuring Mic close to the operators body will show anomalies of up to 6dB around 400Hz.
DD

That’s for the microphone correct? I mean for the source. Is the source able to be that close to the boundary in which you are doing TL measurements?

 

[DanDan]

Good question. I can't remember the rules as to locations, and the standard used at the time has been replaced in any case.
From memory though, same as the mics. 1.2 M from the walls. I always harboured a view that Pressure Zone Locations would be more reliable.
I see half docecs now, PZ on the floor. https://web2.norsonic.com/product/hemi-dodecahedron-loudspeaker-nor275/

I wonder is this a hangover from the Hi Fi notion that speakers have to be spaced out from the wall?
DD

 

[DanDan]

NTI seem to be ahead of the game with their software and published guides.
This is the new EU standard. Makes relatively little sense to me.
https://www.nti-audio.com/Portals/0...-Airborne-Sound-Insulation-Index-with-XL2.pdf
DD

 

[Matthew J Poes]

Good question. I can't remember the rules as to locations, and the standard used at the time has been replaced in any case.
From memory though, same as the mics. 1.2 M from the walls. I always harboured a view that Pressure Zone Locations would be more reliable.
I see half docecs now, PZ on the floor. https://web2.norsonic.com/product/hemi-dodecahedron-loudspeaker-nor275/

I wonder is this a hangover from the Hi Fi notion that speakers have to be spaced out from the wall?
DD

I don’t have such fancy equipment. Not doing this enough to justify anything but renting. In this little experiment I copied a research paper published before the method was adopted. If it’s good enough for science it’s good enough for me! At least to ensure the method works.

I’ve since used this method for a local group and utilized their equipment. They had a high output speaker as well and we also used that to essentially rerun part of the test because we needed greater output and were concerned we might still be measuring noise. The results were very close to identical. I suspect that in relatively small spaces like I tested in, a direct radiator may not be a huge detriment to the accuracy of the TL calculation.

If you ever need to measure TL for a studio project, this is great. It’s also great to use for tracer sound testing. Like if you need to figure out how sound is escaping from specific places such as ducts. Because it can detect such low levels of the sweep. I’ve used it to accurately figure out what sound is escaping through the ducts, plumbing, etc. Helps engineer the right solution to address it.

Another cool use of this, over pink noise is in Situ it diy absorption testing. For example, design a rig to hold a speaker and microphone at a specified distance to each other. Take a sweep outside and save it. Go inside and measure with the mic near the surface to test. Divide the outside sweep into the inside sweep. What is left is the difference between the direct signal from the speaker and the reflection from the panel. Since a perfect reflection would be exactly the same but delayed, this difference should be the absorption.

Or export a sweep from REW as a wav file. Open a multichannel recorder. Connect two matched microphones. Place a speaker and both mics in-line with each other at a set distance with room for a test panel between mics. Place an open backed panel outside with the mics on either side and the source in line on one side. Measure the direct sound from speaker to mic 1 and through the panel to mic 2. The difference between the two is the loss through the panel and ground bounce. Ground bounce can be windowed out, and the LF resolution can be by doing this up in the air, say 6’ off the ground. The panel has to be large to avoid diffraction contaminating the results.

If you do this with an RTA and pink noise, the results don’t seem to be as accurate across the spectrum or when absorption is high.

 

[Matthew J Poes]

NTI seem to be ahead of the game with their software and published guides.
This is the new EU standard. Makes relatively little sense to me.
https://www.nti-audio.com/Portals/0...-Airborne-Sound-Insulation-Index-with-XL2.pdf
DD

What makes little sense?

I thought that 16283 adopted the 18233 standard into it as an option? Seems like NTi is using the oink noise approach still.

I believe B&K May be one of the of the few tonhave built in functionality for 18233. Maybe a handful of others.

But it’s not hard to roll your own. The calculations are straight forward.

 

[DanDan]

Because of the 'Legal' aspects, adherence to these standards is extremely expensive. A download of a standard for 100 bucks last time I looked.
B&K prices are insanely high. I keep mine for the Microphone, hardly for the pretty slow actual meter, which afaik is some early Android platform.
The NTI is pretty plasticy in the flesh.
Any modern Smartphone makes these things look silly.
NTI have a lot of info on their site. Passing by I noticed a section on stimuli, including sweeps.
Again due to absurd prices, we don't have access to Intensity Probe measurements. Surely the way forward for in-situ work.
Here's an IMO interesting suggestion that think plastic, Clingfilm, pretty much doesn't reflect HF at all.
https://www.gearslutz.com/board/showpost.php?p=7964055&postcount=29
DD

 

[Matthew J Poes]

Because of the 'Legal' aspects, adherence to these standards is extremely expensive. A download of a standard for 100 bucks last time I looked.
B&K prices are insanely high. I keep mine for the Microphone, hardly for the pretty slow actual meter, which afaik is some early Android platform.
The NTI is pretty plasticy in the flesh.
Any modern Smartphone makes these things look silly.
NTI have a lot of info on their site. Passing by I noticed a section on stimuli, including sweeps.
DD

Yeah I know some of their gear can do sweeps. Can it automatically generate a report based on 18233?

Honestly I can replicate most of these standards with REW and an excel sheet. It’s a little sad.


The only exception is that for 18233 you are supposed to take sweeps in various locations to average out modal effects or resonances. Then use those averages to calculate TL. I found it easier to do the average of the impulse responses In octave.

 

[DanDan]

I would assume 'it' can do pretty much anything. The NTI software seems well done and I think more affordable than others. Afaik, most/all of the external software apps by the SLM companies are charged for separately. Some are 'subscriptions'.
I agree that it is a bit sad to see absurdly expensive kit side by side by REW, Faber Acoustical, StudioSix Digital. Roll on the time when B&K, Aco Pacific, etc. add Pre and ADC to their caps, in order to utilise the Smartphone and Tablet engines.
Pedantically, I think it would be better to abandon the (mis) use of RT60. These small rooms are far from Diffuse.
DD

 

[mc_deli]

Yeah I know some of their gear can do sweeps. Can it automatically generate a report based on 18233?

Honestly I can replicate most of these standards with REW and an excel sheet. It’s a little sad.


The only exception is that for 18233 you are supposed to take sweeps in various locations to average out modal effects or resonances. Then use those averages to calculate TL. I found it easier to do the average of the impulse responses In octave.

This thread and the other related are fascinating reading. I want to measure transmission loss between two rooms across the frequency spectrum because I have a AV room build project, flanking issues to neighbours and concrete walls that seem to provide great STL with the exception of some specific frequencies I would like to identify.

From this and the other thread, I can't follow your method - could you spell it out with a few bullets?

 

[Matthew J Poes]

This thread and the other related are fascinating reading. I want to measure transmission loss between two rooms across the frequency spectrum because I have a AV room build project, flanking issues to neighbours and concrete walls that seem to provide great STL with the exception of some specific frequencies I would like to identify.

From this and the other thread, I can't follow your method - could you spell it out with a few bullets?

It's really very simple. You have a sound source, a speaker, in one room. You do a sweep using REW at a fairly loud level. I would do it at the loudest level without the speaker distorting, and the speaker needs to be capable of high SPL, well in excess of 100dB. You then keep the speaker in the same room but move the microphone to the room on the other side of the partition wall and sweep again. You then divide one sweep into the other to get a difference file. This functionality is available in REW. This won't give you STC, this will just give you transmission loss of very high resolution. If all you are interested in is identifying the frequencies where TL is poor, stick to that. Actually converting to STC is not important to you, it's just to standardize and provide comparability.

Ideally you will take multiple measurements in the source room near the partition wall and numerous in the receiver room. This is because at low frequencies the modes will cause false readings. Averaging these can address this problem, but you need to maintain the IR.

A few things that will impact this:

1. You likely need a high quality low noise calibrated measurement microphone and SPL calibrator
   1. Typical USB microphones have a VERY high noise floor and will give artificially low TL values
   2. A poor calibration can cause the absolute values to be wrong and misleading potentially. It wouldn't matter so much if not for the noise floor problem
2. The noise floor in the receiver room (room opposite the room with a speaker in it) should be as low as possible. While this technique improves signal to noise floor, only by a bit. If the noise floor is high, it will limit what you can do.
3. Longer sweeps may help improve signal to noise floor
4. Multiple sweeps may also improve signal to noise floor, but must use a loopback (and thus can't use a USB mic)
5. Ideally your speaker is supposed to be an omnidirectional speaker. I did tests using a rented omni speaker vs some pro powered speakers. What I found was the Omni did make a difference in the results, but not a huge one. However, the best results were obtained using 2x pro speakers aimed in different directions. There is now some debate about ideal speaker placement, but for this test, I found placing them in the far corner of the room away from the partition wall was best.
6. If you are doing this in an apartment building, you will likely need permission from all residents in the apartment building and permission from the landlord (If that is not you). You will not only need to enter an adjacent residence to test the partition wall, you will be making a huge racket.
   1. I recently did an indoor/outdoor partition test and had to get a permit for sound that is the same one a concert promoter has to get

 

[mc_deli]

It's really very simple. You have a sound source, a speaker, in one room. You do a sweep using REW at a fairly loud level. I would do it at the loudest level without the speaker distorting, and the speaker needs to be capable of high SPL, well in excess of 100dB. You then keep the speaker in the same room but move the microphone to the room on the other side of the partition wall and sweep again. You then divide one sweep into the other to get a difference file. This functionality is available in REW. This won't give you STC, this will just give you transmission loss of very high resolution. If all you are interested in is identifying the frequencies where TL is poor, stick to that. Actually converting to STC is not important to you, it's just to standardize and provide comparability.

Ideally you will take multiple measurements in the source room near the partition wall and numerous in the receiver room. This is because at low frequencies the modes will cause false readings. Averaging these can address this problem, but you need to maintain the IR.

A few things that will impact this:

1. You likely need a high quality low noise calibrated measurement microphone and SPL calibrator
   1. Typical USB microphones have a VERY high noise floor and will give artificially low TL values
   2. A poor calibration can cause the absolute values to be wrong and misleading potentially. It wouldn't matter so much if not for the noise floor problem
2. The noise floor in the receiver room (room opposite the room with a speaker in it) should be as low as possible. While this technique improves signal to noise floor, only by a bit. If the noise floor is high, it will limit what you can do.
3. Longer sweeps may help improve signal to noise floor
4. Multiple sweeps may also improve signal to noise floor, but must use a loopback (and thus can't use a USB mic)
5. Ideally your speaker is supposed to be an omnidirectional speaker. I did tests using a rented omni speaker vs some pro powered speakers. What I found was the Omni did make a difference in the results, but not a huge one. However, the best results were obtained using 2x pro speakers aimed in different directions. There is now some debate about ideal speaker placement, but for this test, I found placing them in the far corner of the room away from the partition wall was best.
6. If you are doing this in an apartment building, you will likely need permission from all residents in the apartment building and permission from the landlord (If that is not you). You will not only need to enter an adjacent residence to test the partition wall, you will be making a huge racket.
   1. I recently did an indoor/outdoor partition test and had to get a permit for sound that is the same one a concert promoter has to get

That is great info thanks. Really appreciate it.
I have the Sonarworks mic and calibration file - which I think should suffice.
And I have a friendly neighbour.
Our partition wall is attenuating 55-60 dB at some frequencies so, yes, I understand that I will have to pump out at least 100dB for the test.
I have a KRK system with sub so should be able to do that from 39hz.

Silly question but for "sweep" you mean a pink noise periodic sweep?
Bonus daft question, do you have a technique for averaging multiple measurements - was that what you used XLS for?

 

[DanDan]

It is easy enough to get an estimation. But Planning permissions can be overturned, requiring a rebuild. So a Legal measurement is a bit more complex.
LF is close to Omni from most speakers, but pointing a directional one in different directions can stimulate the room quite differently.
One might be inclined to just point one or two of them into corners, while actually in the corners, but the defined methods don't include that.
I guess they want representative positions not just a full modal.
Oddly they don't seem to change height of speaker or SLM. Dodecs do produce 100dB and more but most don't go anywhere near the LF we find in Music and Movies.
Some sweepers do seem to have the ability to average from a single ended source. Fuzzmeasure. Altiverb can even do it's work using different play and record devices, (single sweeps though)
Doubling the sweep length does indeed increase S/N by 3dB.
Perhaps more interestingly, and I did not know, sweepers 'hear' down into the noise floor. Very broadly speaking I regard this as a 15-20dB 'bonus'.
This takes some of the pressure off the playback system. I used to sweep rooms as loud as possible but JohnPM kindly explained that the recommended 75 is absolutely plenty.
I have my own working notion of how this works, but perhaps John or others might share the definitive.
I guess REW is hearing only one frequency at any moment, and thus the background noise or even mic noise at that frequency only. Not the whole spectrum.
So it does seem that these swept sines have to be much better than the traditional interrupted noise methodology IMO.

 

[Matthew J Poes]

That is great info thanks. Really appreciate it.
I have the Sonarworks mic and calibration file - which I think should suffice.
And I have a friendly neighbour.
Our partition wall is attenuating 55-60 dB at some frequencies so, yes, I understand that I will have to pump out at least 100dB for the test.
I have a KRK system with sub so should be able to do that from 39hz.

Silly question but for "sweep" you mean a pink noise periodic sweep?
Bonus daft question, do you have a technique for averaging multiple measurements - was that what you used XLS for?

The Sonarworks microphone may be too noisy for this. I think it is likely to limit your assessment of the mid-HF transmission loss. It should be fine for what you are doing, but will potentially underestimate the results. True TL might be higher than you measure.

I am not sure the arrangement of your building, but this loud noise goes everywhere, so do you have neighbors above or below you? If so, you may want to get permission to do this. Up to you.

Sweep refers to the sine sweep in REW, not a periodic pink noise sweep.If you use periodic pink noise you won't be able to obtain an IR and your arrangement will certainly not work. The traditional pink noise approach would require that you measure much louder and have a much better microphone.

Averaging can be done in REW. All of this can be done in REW. XLS was used to convert to STC. Vector average maintains the IR parameters and is best, but if the IR's are not perfectly aligned and identical (other than spatial location) the end result is inaccurate. They are hard to obtain. Standard averaging might work, i can't recall if you can do vector arithmetic in REW with an averaged result. I used vector averaging. I had some problems though and had to be sure everything was just so.

 

[Matthew J Poes]

It is easy enough to get an estimation. But Planning permissions can be overturned, requiring a rebuild. So a Legal measurement is a bit more complex.
There are odd factors. e.g.
I don't think normal Arithmetic applies to Decibels.
The Reverb Time of both Source and Destination need to be included in the calculations.

LF is no doubt close to Omni from most speakers, but pointing a directional one in different directions can stimulate the room quite differently.
One might be inclined to just point one or two of them into corners, while actually in the corners, but the defined methods don't include that.
I guess they want representative positions not just a full modal. So OMNI and from memory no close than 1.2 metres to any boundary, 4 or more speaker locations.

Oddly they don't seem to change height of speaker or SLM. Dodecs do produce 100dB and more but most don't go anywhere near the LF we find in Music and Movies.
Some sweepers do seem to have the ability to average from a single ended source. Fuzzmeasure. Altiverb can even do it's work using different play and record devices, (single sweeps though)
Perhaps interestingly, sweepers 'hear' down into the noise floor. Very broadly speaking I regard this as a 15-20dB 'bonus'.
This takes some of the pressure off the playback system. I used to sweep rooms as loud as possible but JohnPM kindly explained that 75 for in room sweeps is absolutely plenty.
I have my own working notion of how this works, but perhaps John or others might share the definitive.
I guess REW is hearing only one frequency at any moment, and thus the background noise or even mic noise at that frequency only. Not the whole spectrum.

What John was referring to (the 75dB) is not for transmission loss. Because the transmission loss through walls can be anywhere from 10dB to as much as 100+dB, you need at least enough signal to get an accurate IR. About the most you can "hear" into the noise floor is 20-25dB. Let's say your source is 75dB at the wall. You have 100dB of transmission loss at 5khz. You are at -25dB on the other side. The room noise floor might be 30dB. No chance you will get that in the sweep. Now imagine you play it back at 100dB, it's at 0dB and you might. Play it back at 115dB and you probably will. This doesn't apply for in room acoustic measurements. Just transmission loss measurements.

I'm not sure what you mean about the corner not being in the protocol. It's in the studies and protocol as the preferred location.

Omni speakers used for this have optional subwoofer modules. Thats what I used then and would generally use. In my own room, I used my 18" subwoofer and a variety of speaker optoins. You def. need to use a subwoofer to get good LF isolation assessments. Those aren't a part of the standard, but I measure it anyway. The problem is that modes screw up the results. Even with averaging it might not be quite right. Averaging at least stands a chance of minimizing it.

The protocol is very vague on how to place the speakers for this. I see no reason why you can't change the height if you think it matters. Especially if you aren't doing a standard compliant measurement (I would argue that is still compliant, but some people are very stuffy about this).

What you are saying is basically right, REW is using an FFT and an FFT breaks up the noise into very narrow bins of acoustic energy. The longer sweeps improves S/N as does multiple sweeps. John pointed out to me that one method improves S/N (Ionger sweeps) more than the other (multiple) and he wasn't sure why. I then tested it and found he was right. I then tested that with this and found that actually in this scenario, each one improved S/N in different ways. Ideally doing both was best, but I get errors trying to do multiple sweeps like this much of the time. Again, to get it to work, I had to play back very loudly and use loopback. I also had to use a chirp speaker to trigger the sweep in the receiver room. Thankfully longer sweeps reduces errors and is more effective, so I suggest that being the more important. Biggest problem with long sweeps is that it stresses the speakers the most.

 

[DanDan]

Indeed for TL one needs a loud source, but worth noting about 20dB less than instinct might suggest.
I don't think that the mic noise is significant. Remember that hearing down bonus of REW's sweep methodology.
Even with interrupted noise the requirement seems to be that the noise in the receive room only needs to exceed ambient by 10dB in each third octave band.
Here the onboard iPhone Mic appears to be useable.

Transmission Loss | Studio Six Digital

studiosixdigital.com

 

[mc_deli]

Indeed for TL one needs a loud source, but worth noting about 20dB less than instinct might suggest.
I don't think that the mic noise is significant. Remember that hearing down bonus of REW's sweep methodology.
Even with interrupted noise the requirement seems to be that the noise in the receive room only needs to exceed ambient by 10dB in each third octave band.
Here the onboard iPhone Mic appears to be useable.
studiosixdigital.com

I'd looked at that but has anyone paid 19.99 for the app plus 44.99 for the TL module, and compared it with something pro grade, to find out?

 

[mc_deli]

The Sonarworks microphone may be too noisy for this. I think it is likely to limit your assessment of the mid-HF transmission loss. It should be fine for what you are doing, but will potentially underestimate the results. True TL might be higher than you measure.

I am not sure the arrangement of your building, but this loud noise goes everywhere, so do you have neighbors above or below you? If so, you may want to get permission to do this. Up to you.

Sweep refers to the sine sweep in REW, not a periodic pink noise sweep.If you use periodic pink noise you won't be able to obtain an IR and your arrangement will certainly not work. The traditional pink noise approach would require that you measure much louder and have a much better microphone.

Averaging can be done in REW. All of this can be done in REW. XLS was used to convert to STC. Vector average maintains the IR parameters and is best, but if the IR's are not perfectly aligned and identical (other than spatial location) the end result is inaccurate. They are hard to obtain. Standard averaging might work, i can't recall if you can do vector arithmetic in REW with an averaged result. I used vector averaging. I had some problems though and had to be sure everything was just so.

Appreciate the tips, thanks.

So, Generators > Sweeps > Measurement and set to e.g. 30hz-16Khz - what about the length setting - is the default 256k long enough for an accurate test (of course, given my other limitations)?

Could you also explain the significance of IR and "maintaining the IR parameters", I don't follow?