Luckily it's not a problem to pull out what I put in - nothing is fabric-covered yet, and it's just held in place with tacks and twine at the moment. Everything is easily-modifiable. .... the HH is definitely registering the speaker level as 80db on L and R, and 86 with both together.
But that's NOT what REW is showing! The original tests seem to have been done with the individual speakers at around 65 - 70 dB, and the latest tests at around 75, or a bit less. Are you SURE you followed the procedure exactly? Are you SURE you set your HH meter to "C" weighting and "Slow" response, and that you typed in the correct number in REW? To me, it looks like you have the HH meter set to "A" weighting, and REW set to "C"...
Double-check everything, re-calibrate carefully, then make accurate notes of where you have every single setting for ever device in the signal chain, including the exact location of the mic tip in the room (That's also very important, for checking high frequency response later). Then try to not change any setting at any place in the signal chain from here on, and always set the mic in the exact same spot, every time.
So hopefully the sweep tone is the same level, and is affecting the room the same way each time. I'm sure it's a mic input trim issue that is affecting the final levels that show in the tests
If you are absolutely certain that this is the case, then there's no need to pull out the back wall to repeat the "empty room" measurement, as I can adjust it in REW to match the later measurement, ... but I'm a bit skeptical that it was only mic pre gain setting....
I'll re-calibrate REW, and do a new set of tests, all with exactly the same settings - it won't take long
I think I was a bit spooked by the original RT60 that went off the scale - so went over the top trying to bring the decay down.
There's basically two method for tuning a room: start with it very live, and kill it slowly, or start with it very dead and liven it up slowly. Personally, I prefer the later. Take a look at this thread carefully (and follow it into the future!) to see how that works:
http://www.johnlsayers.com/phpBB2/viewt ... =2&t=21368 If the room is dead, then you KNOW that you have already damped all the modes and resolved the big reflection issues, so you can see exactly what is happening as you add each new device. If you start the other way (very live, slow killing) then you keep on revealing new stuff that you didn't know was there, as it was masked by other stuff, so you are forever chasing your tail, adding new treatment to deal with those newly discovered things you hadn't seen yet. However, both methods work, and both are valid. But you have to decide which one you are going to use! It's really tough to try to use both at once...
So, starting with that empty room setting, you added a lot of absorption and got some results in attempting to make it "dead", but the modal issues are still there. So I'd suggest that you need to spend time attacking JUST the modal issues now, while taking care to not over-do the high-end absorption. In other words, absorption mostly, perhaps some type of tuned trap if you have a really stubborn mode, but keep the highs in the room.
I have now done some reading on ITU/EBU recommendations and I have a clearer idea now why too little decay = also not ideal.
Yup! Anad apart from the technical reasons, it just doesn't sound nice. It is fatiguing, and uncomfortable to be in a dead room for long periods, trying to do critical listening.
Bob Gold's calculator gives *RT60 (ITU/EBU Control Room Recommended): 138 ms* for my room. I'm guessing this is 'recommendation' as opposed to 'hard rule'?
I would go a little higher than that. More recent research suggest that engineers prefer longer decay times, and find it more comfortable and accurate to mix in, provided that the decay is diffuse, not specular. I would aim for closer 200 ms. Perhaps around 180 or so.
I also don't know whether it's best to err above or below that number for given frequency areas.
Ideal is flat! I mean flat TIME response, not just flat FREQUENCY response. In other words, the decay rate should be the same for all frequency bands, across the scale, within +/-50ms So if you have, for example, 200 ms at 1 kHz then no point on the graphs should go higher than 249ms and no point should go lower than 151 ms, plus adjacent bands should ALSO be within 50ms of each other: so you can't have 150 ms at 315 Hz but 250 ms at 400 Hz, which is the very next band. Right now, you have 178ms at 315 Hz, then 271 at 400 Hz...
You can relax that a little at the very low end and very high end, but the mid range is critical. BS.1116-3 includes a graph that shows the upper and lower limits across the spectrum. If you can stay within those dotted lines, then you are doing fine. This is more important that flat frequency response, actually. Most people tend to grab onto the frequency response graph and attempt to get that flat, but it's actually the least useful graph of the lot! Much more important is time-domain response, which includes modal decay, SBIR, reflections, and other forms of resonance. If you get that right, then the FR will take care of itself...
If I'm reading the room-mode calculator correctly, the 52hz peak, and the 104 hz null are both being caused by the same axial mode - between the front / rear wall
I'd agree that the 52 Hz peak is modal, but I'm not so sure about 104. You had that mode in the empty room, but it got nicely damped with the first round of treatment:
RKML--FR-20-500--first-three.png
Orange is the "empty room" test, green is with the rear wall done, and purple with the front superchunks. So the mode at 101 was controlled with the first round. The second round did not cause a "negative mode"! Most likely, it uncovered an SBIR issue that happened to be at the same frequency as the mode, so it only came to light once the mode was fully damped. You can see signs that it was there all along, but the mode was masking it.
It is also roughly in the typical spot for all rooms, for the first-order floor bounce null. That's what I would suspect.
Then I've got the other big peak around 143hz, which I think corresponds to the axial mode between floor and ceiling?
Possibly, but I'm not convinced. It took a big hit with the front wall superchunks, so I'd suspect either axial across the room, or lengthwise. However, it does not line up with any predicted modem which makes me suspect that either your walls are rather thin and/or low mass, or that the measurements you gave ar not correct. Are you SURE it is 330cm long, 229cm wide, and 233cm high? In other words, a square cross section?
I don't know the solution to the front / rear axial mode.
Which one?
One general rule of thumb (not entirely accurate, but it helps to get you on the track sometimes): usually you will find that treating the front and back walls with absorption will affect the time domain response, and treating the side walls with absorption will affect the frequency response. Don't confuse large SPL variations with modal ringing: they are related, but not the same thing....
Maybe I'm barking up all sorts of wrong trees here!
A little bit! But there's lots of cats out there, calling for your attention, so it's normal to go rushing around barking at everything until you figure out which tree they are actually hiding in...
- Stuart -
