Wierd results from RoomEQ

[rbowen]

Greetings everybody,

I'm new to this forum and have not used RoomEQ before. Many thanks to John Mulcahy and all the development team for a superb product.

Well, I started out to measure the resonance frequencies of a large room with extremely hard walls. The reverberation time of this room is about 1.5 seconds. Using version 3.29 of RoomEQ and a Realtec Sound Card together with an RS meter (thanks for the calibraton curve) all seemed to be going pretty well but the results of the Measurement of the Room looked strange to put it mildly.

The Variation in signal level for a small change in frequency looked absolutely massive (like 40db for a few Hz change). The sweep graph looked like the feedback curve in the "Getting Started with RoomEQ" document but as far as I could see and hear, there was no feed back. The sort of graph that resulted is shown in Sample1.jpg

I then repeated the sweep using 1/12 octave increments (i.e. every semi tone) and saw pretty much the same effect. The difference between adjacent semitones could be as much as 20-30 dB which seemed odd. To illustrate, some of the measurements were:
469Hz 54dB
418Hz 68dB
380Hz 52dB
339Hz 73dB
318Hz 47dB
but the full horror can be seen in Sample2.jpg

While this test was going on (all 13 minutes of it) I could hear that there was a real difference in volume for the semi-tones but as much as was measured?

The strange results have the hall marks of some howler that I have blundered into.

At the risk of raising an issue answered elsewhere on the forum can I ask if anybody might have some advice and guidance on what might be going wrong?

Is it possible that the acoustics of the room under test are as strange as they appear?
What howler/mis-use/configuration issue am I walking into?

Many thanks for reading this posting.

Regards
Roger Bowen

[brucek]

Start by downloading the latest Version 4 of REW. It's much easier to setup and use.

Post all graphs with a vertical axis of 45dB to 105dB. This standard allows for everyone to be on the same page and makes evaluation much simpler.

Use a 75dB target level in your setup.

The Radio Shack meter should not be considered useable above 5Khz.

brucek

[terry j]

firstly you need to post the graphs using the 'standard' scaling, on the y axis it's 45 - 105 db. I guess the x- axis in this case is whatever range you are talking about so that's cool.

What smoothing has been applied??

oops, sorry Bruce!!!typical aint it!

[rbowen]

Thank you for the replies and an apology for not scaling the graphs correctly.

The comment about using Release 4 has been noted and I will start again using that product.
Similarly the comments about target levels and limitations of the RS meter are recognised.

In the meantime attached are two graphs, correctly scaled I hope, one that shows the unsmoothed results and the other with 1/12 octave smoothing.

Both graphs show much the same story in that the response varies wildly with small frequency changes which again leads me to ask if this is in any way the expected behaviour.

Roger

[terry j]

amazing what a difference scaling makes aint it!!! You may even be able to answer your own question now, doesn't look half as bad as you first thought.

Also, as Bruce said, if the mic is the RS meter, then you can ignore anything over 5k

[JohnM]

Hi Roger,

Yes, that's normal, it's comb filtering from reflections form the room's surfaces.

P.S. I'm just outside Watford as I type this

[brucek]

which again leads me to ask if this is in any way the expected behaviour

Yeah, that's standard stuff when you do a full range measurement with a wide scale. It all gets scrunched into a small space and looks horrible.

I'll show you a few measurements from my terrible second system. I like to add smoothing for the full range stuff to about 1/3 octave. This would be a bit more representative of how you hear the sound and it gives a better insight into what's behind all the mess, so if you're applying room treatment at least you can see whats going on. Also narrowing in on a smaller range of frequencies can help in the definition if you have a trouble spot. Also in version 4 you can increase the sweep time and the number of sweeps. The sweeps are then averaged which helps reduce noise and interference and gives a better signal to noise ratio. You can choose durations of 128K, 256K, 512K and 1M. You can choose the number of iterations from 1 to 8 times for averaging.






Below are a full range sweep with no smoothing and the same sweep with 1/3 octave smoothing. You can see that it helps in the presentation to get a better idea of your full range response. (My mic goes to 20Khz with no problem). Don't use the RSS meter to 20K..


Unsmoothed



Smoothed 1/3 octave




The combing isn't a problem at lower frequencies, so you don't really need it when you're measuring a subwoofer. Look at the difference it makes when you narrow in on a small range of frequencies such as 20Hz to 200Hz for the subwoofer. The smoothing is actually unnecessary to see what's going on. Ignore the poor response of my starter sub.


Unsmoothed






Smoothed 1/3 octave






brucek

Sounds like you've been getting somewhere with REW Roger, I've downloaded the new version and am familiarising myself with the interface.

The ability to quantify where we have holes in the frequency response of the room will certainly help.

I think that a BFD FBQ2496 will be required to help [notch] out though troublesome feedback frequencies as well as improving the colouring that the very reflective surfaces introduces.

Our temporary acoustic dressings may produce a quantifiable difference in the sound and will certainly help make the case for further acoustic tile. I've also downloaded the manuals for the BFD and our existing ultragraph FBQ3102.

One thing I'd be interested in: was the EQ switched in for that graph???

Just another point for these dear folks who are so helpfully setting us on the path to enlightenment (well as far as we can with the acoustics)...

Roger and I are putting Room EQ to a slightly different use to that of nearly everyone else.

Our room is actually the main auditorium of our church and is about 40m*20m*15m and every wall is painted polished plaster (or glazed)... (very reflective) We have had an acoustic engineer to take a look but as we attempt to tweak matters before going to the full acoustic treatment, we'd like a more considered approach rather than a subjective "that sounds better".

We're hoping to use Room EQ to quantify our resonant frequencies and see what impact our temporary acoustic tile (large numbers of single bed duvets - don't ask) has on the sound and subjectively I've noticed a difference in intelligibility.

Thanks for helping us with the setup, we'll keep you posted with our developments and perhaps share a few plots. (BFD will be on back order very soon)

[brucek]

Our room is actually the main auditorium of our church

Our leader Sonnie had a problem in his church too with sound. Read about it here

BFD will be on back order very soon

You should get an ECM8000 and preamp or a Galaxy CM-140 microphone. The RS meter just doesn't cut it.

brucek

[rbowen]

Everybody,

Thank you so much for the comments, advice and guidance.

Things, from my end anyway, will stop until Saturday, when we'll have another go at looking at the church building's acoustic properties.

I'd love to know the outcome of Sonnie's work on his church's hall.

Sorry to hear JohnM was just passing Watford (It's the centre of the universe but don't let on you know)

Regards
Roger

[DrWho]

That explains everything....the larger the room, the more dense the modal distribution - which shows itself as comb filtering starting at lower frequencies...

In large rooms, you will want to familiarize yourself much more with the ETC and make sure you're using proper windowing on the impulse response. The frequency response is pretty much meaningless in these applications...

Without more fancy software (like TEF), you'll be left to manually calculating the distances of distinct reflections and then doing a bit of trial and error as you try to fix them. Try as best as you can to achieve 20-30ms before the first reflections start bouncing around - often this means carefully aiming the speakers to keep the sound off the nearby walls.

Churches are difficult environments because music benefits from RT60's much longer than those required for optimimum intelligibility. It's not uncommon to design for maximum intelligibility and then mic'ing the congregation - feeding the signal through some effects and to speakers distributed throughout the sanctuary...basically artificially increasing the RT60 and making music sound much more natural.

[blazerman]

DrWho wrote:

Without more fancy software (like TEF), you'll be left to manually calculating the distances of distinct reflections and then doing a bit of trial and error as you try to fix them. Try as best as you can to achieve 20-30ms before the first reflections start bouncing around...

Are ETC measurements from ETF better for first-reflection detection than in REW? I haven't measured with REW yet, but by looking at the plots the only big difference seems to be the energy scale (y-axis). The "Windowing" and "gating" functions seem to be similar.


Blazerman
"Hearing is believing"

DrWho wrote:

Churches are difficult environments because music benefits from RT60's much longer than those required for optimimum intelligibility. It's not uncommon to design for maximum intelligibility and then mic'ing the congregation - feeding the signal through some effects and to speakers distributed throughout the sanctuary...basically artificially increasing the RT60 and making music sound much more natural.

This is something I think that we'll end up considering given that we often have choral/orchestral pieces which do benefit from the longer reverb time, but it'd be nice to be able to switch it on and off!

[DrWho]

DyerEmergency wrote:

This is something I think that we'll end up considering given that we often have choral/orchestral pieces which do benefit from the longer reverb time, but it'd be nice to be able to switch it on and off!

That's what congregation mics and banks of speakers along the sides and back are for. You can optimize the natural room acoustics for speech intelligibility and then artificially recreate the hall effect for music. Your switch is the mute button on the matrix output

Btw, I have absolutely no experience with that craziness...I've just read about it.

[DrWho]

blazerman wrote:

Are ETC measurements from ETF better for first-reflection detection than in REW? I haven't measured with REW yet, but by looking at the plots the only big difference seems to be the energy scale (y-axis). The "Windowing" and "gating" functions seem to be similar.


Blazerman
"Hearing is believing"

I would expect the interpretations of the data to provide the same conclusions. Never tried them side by side though.