DIY flush mount speakers

With some help from Glymur (Curt), we measured the room. Here are links to the mdat files Baseline from the mix position https://www.dropbox.com/s/h4f7c8t2nygyuvu/baseline.mdat?dl=0 Measurements moving from the front of the room to the mix position https://www.dropbox.com/s/8ih6ozzl3oavtdb/front-to-mixposition.mdat?dl=0 The mix position to the back https://www.dropbox.com/s/ji0uw287xv8z7pg/mixposition-to-reartrap.mdat?dl=0 Here is SPL from the mix position. The waterfall And the rt60 Obviously we have an issue at 80hz :)

A couple of pictures.

I downloaded your files, but there's something very strange going on there: I suspect the data is not valid. The SPL curves look fine and normal, but the rest is way off. Also, your RT60 graph does not match what I'm seeing! This is what I get: Which is pretty much impossible in any room! I'd find it REALLY hard to accept that your decay times are essentially zero for everything below 100 Hz! :shock: Your waterfalls look weird too: Even the one you posted is mighty strange: It clearly shows decay times of around 100 ms or less, yet the RT60 data shows over 300 ms, and the actual calculated RT60 is also around 300-something... so there's an issue here for sure! I thought it might be data corruption in the file download, so I downloaded again, but it's the same. Are you SURE you calibrated correctly, using your sound level meter? Did you really see 80 dBC on your meter while running those tests? It looks a lot like you calibrated incorrectly: The white horizontal cursor line is a rough estimate of the background noise level in your room, as seen by REW, and I'm pretty sure it is not 52 dBC in real life! You'd be VERY disappointment if your isolation was that bad. My guess is that something went wrong with your calibration procedure, and the tests were done at a level that is way too low, but that REW still identifies as 86 dBC. Please check your calibration procedure... What hand-held sound level meter are you using here? Is the meter itself calibrated? Good batteries? In good condition? Just some thoughts... - Stuart -

Which measurement did you take the rt60 from. I cant find one that matches. The spl meter we used to calibrate was the one with the Omnimic system.

Which measurement did you take the rt60 from. I cant find one that matches.

Mix position, baselline, both left and right speakers. The other two (Left only and right only) are equally invalid: I uploaded the same file to one of my servers, so you can download it and compare to the original. It might be data corruption, but I don't think so, as the waterfalls you posted look equally invalid. You can find it here: http://spartanew.digistar.cl/SayersForu ... eline.mdat

The spl meter we used to calibrate was the one with the Omnimic system.

As far as I know, Dayton does not manufacturer hand-held sound level meters. They only make mics and test software. I'd suggest that you get a real SPL meter, such as an Extech or Galaxy, or some such, and calibrate again. Even an app for your cellphone, such as AudioTools, would at least give you a semi-reasonable estimate of the real SPL levels in your room. Not accurate, of course, but good enough to do a basic check. Try that now, to see what you get. I suspect that you are not really getting the levels that REW thinks you are getting. Here's a spectrogram from your right speaker: That's just meaningless mush. Here's what a typical spectrogram looks like for a partially treated room:Here's what the low-end waterfall for that same room looks like: And here's the full spectrum waterfall for it: And here's the RT60 graph (same room): All of the above are similar to what I'd expect to see in your data. Not the way it looks in your graphs. I chose that MDAT file at random from among the large number I have, but that's typical of most of them. Your data is way, way noisy. Before doing anything else, you need to figure out what's wrong here, and fix it! You can't make room treatment and tuning decisions based on flaky, suspect data. - Stuart -

I downloaded what you have for my data from you link and for the rt60, this is what I got. This may not be good, but it is very different from what you are getting. I included the whole screen so we can compare and make sure we are looking at the same files.

I will re-calibrate and try again tomorrow. I know this is frustrating, but thank you for all the help.

This may not be good, but it is very different from what you are getting.

You are using one-octave smoothing. Switch to 1/3 octave. - Stuart -

Today I learn the value of fresh batteries. Thanks for the info Stuart, much appreciated.

Here are new measurements. https://www.dropbox.com/s/a8rie5wtsukp5jf/baseline.mdat?dl=0 https://www.dropbox.com/s/d48on2o4wq7tc1b/mix-to-front.mdat?dl=0 https://www.dropbox.com/s/dqiorkt96zm0e19/mix-to-frontofreartrap.mdat?dl=0 [url][/url]

Here are new baseline measurements.

Now THAT looks a little better! Glad to see you found the problem. What was it? However, I'm really curious about your sound card: the calibration curve you show in your graph (above) is clearly showing feedback, so it is not valid. Yet the one in the file on DropBox does not show that. But it DOES show ringing in the high end, and it's different from the soundcard calibration from yesterday. The ringing is not huge, but it is significant if you want precision. OK, on to the actual data: There's a difference between your left and right speakers. I'm not sure if that is the speakers themselves, or the room. I'd suggest moving the mic in a bit closer to each speaker individually, putting it on-axis at a distance of one meter from the front baffle, and pointing directly at the speaker. Compare those graphs. Also trey something a bit closer (say 50cm) and something a bit further back (say 150 cm) and compare. That should give a good indication of where the problem lies. Next, what treatment did you have in the room for this test? Clearly, it is not empty of all treatment: There's some treatment going on. But it would be good to know exactly what you had in there. Is it according to the diagram you posted above? I'm curious about the room design concept: Is it based on CID? That's what it looks like. There's a few rather strong early reflections that you'll need to deal with, coming in at about 3.5ms, 6ms and 9ms after the direct sound. You can use the "string trick" to track down those, but the prime suspects would be ceiling, floor, and walls. Decay: The room is too live: For that sized room, you should have decay times of about 200 to 250 ms, give or take, but you are up around 360 right now. So more absorption is needed, but controlled carefully, to keep the decay times consistent across the board. The big issue, of course, is the massive mode at around 29 Hz. That's predicted: it's your 1.0.0 mode (fundamental axial length mode), of course. Always the toughest one to deal with. There's also something at around 41, which is harder to explain, as there's no mode predicted around there. There's also something at around 66, which I'm guessing is your first height mode, but I don't recall what the height of the room is, so I can't be sure. The next biggest issue is the large dip at 82 Hz, which might be SBIR, but we'll need your new "walking mic" test to figure that one out. There's plenty of other stuff too that will need dealing with, but the overall conclusion is: you need a lot more bass trapping for the very low end, while taking care not to suck out the mids and highs with that, and also adding some controlled mid and high absorption, to get the overall decay times down, and even. You also need to deal with the early reflections, and the possible SBIR issue. But DAMN thos speakers are performing nicely! Getting down to about 26 Hz with negligible distortion and fairly flat response, is pretty amazing! I'm jealous! :mrgreen: - Stuart -

I am pretty proud of the speakers! They sound absolutely amazing. When I listen to them, I sit back and say to myself "yeah, I did that." :D

Soundman2020 wrote:

Now THAT looks a little better! Glad to see you found the problem. What was it?

Curt put new batteries in his SPL meter.

However, I'm really curious about your sound card: the calibration curve you show in your graph (above) is clearly showing feedback, so it is not valid. Yet the one in the file on DropBox does not show that. But it DOES show ringing in the high end, and it's different from the soundcard calibration from yesterday. The ringing is not huge, but it is significant if you want precision.

The measurements today used a different computer, soundcard and mic. I'm not real confident about my soundcard calibration, I thought it looked kind of weird.

OK, on to the actual data: There's a difference between your left and right speakers. I'm not sure if that is the speakers themselves, or the room. I'd suggest moving the mic in a bit closer to each speaker individually, putting it on-axis at a distance of one meter from the front baffle, and pointing directly at the speaker. Compare those graphs. Also trey something a bit closer (say 50cm) and something a bit further back (say 150 cm) and compare. That should give a good indication of where the problem lies.

It might be the speakers. I have precision (poly 5%) capacitors on order for the crossovers to replace the electrolytics that I used to prototype with. I will check.

Next, what treatment did you have in the room for this test? Clearly, it is not empty of all treatment: There's some treatment going on. But it would be good to know exactly what you had in there. Is it according to the diagram you posted above? I'm curious about the room design concept: Is it based on CID? That's what it looks like.

It is according to the diagram and based on the CID idea. Only one of the six sides of the room is treated now, the back wall has hangers across the back. The rest of the walls are bare drywall right now. The wall and ceiling angles reflect to the back of the room. The front of the angled false walls are filled with rockwool and open where they face the speakers.

There's a few rather strong early reflections that you'll need to deal with, coming in at about 3.5ms, 6ms and 9ms after the direct sound. You can use the "string trick" to track down those, but the prime suspects would be ceiling, floor, and walls.

The only place I should be getting early reflections is from the floor.

Decay: The room is too live: For that sized room, you should have decay times of about 200 to 250 ms, give or take, but you are up around 360 right now. So more absorption is needed, but controlled carefully, to keep the decay times consistent across the board. The big issue, of course, is the massive mode at around 29 Hz. That's predicted: it's your 1.0.0 mode (fundamental axial length mode), of course. Always the toughest one to deal with. There's also something at around 41, which is harder to explain, as there's no mode predicted around there. There's also something at around 66, which I'm guessing is your first height mode, but I don't recall what the height of the room is, so I can't be sure. The next biggest issue is the large dip at 82 Hz, which might be SBIR, but we'll need your new "walking mic" test to figure that one out. There's plenty of other stuff too that will need dealing with, but the overall conclusion is: you need a lot more bass trapping for the very low end, while taking care not to suck out the mids and highs with that, and also adding some controlled mid and high absorption, to get the overall decay times down, and even. You also need to deal with the early reflections, and the possible SBIR issue.

I think after seeing if I can straighten out the speaker balance and soundcard calibration, for a start, I will add some superchunks to the room and see what it does and go from there. I have a bunch of available corners waiting to be filled.

But DAMN thos speakers are performing nicely! Getting down to about 26 Hz with negligible distortion and fairly flat response, is pretty amazing! I'm jealous! :mrgreen: - Stuart -

Thanks! I think that after the room is all settled and studio building is done, I'm going to see if I can build and design a DIY speaker specifically designed to be flush mount speaker that people on the forum can build for less than the $700 of parts that I have into mine.

There's a few rather strong early reflections that you'll need to deal with, coming in at about 3.5ms, 6ms and 9ms after the direct sound.

How and what graph do you use find this? If it is the impulse response, I can find the 3.5ms and the 6ms, but not the 9ms. The 3.5ms should match with the floor reflection. Edit: Filtered IR? Where the tight groups are?

Curt put new batteries in his SPL meter.

:) 8) :roll: :!: It happens. The reason I mentioned that, is because it has happened to me as well... :oops: :oops: :oops:

How and what graph do you use find this? If it is the impulse response, I can find the 3.5ms and the 6ms, but not the 9ms. The 3.5ms should match with the floor reflection. Edit: Filtered IR? Where the tight groups are?

Left channel, IR, switch from dB to %FS (Full Scale), and zoom greatly: You have biggies at 3.3ms, 6ms, 7.1ms, 9ms (marked with the cursor), several around 17ms (rear wall?), 20ms, 28ms. After that, it's not so important. (Outside the Haas time) It's sometimes easier to spot reflections by zooming in on the %FS graph. You can also see them on the dB graph, but the decay slope confuses. There's similar stuff on the right speaker, but slightly different timing in some cases. DO you know how to use the "string trick" to find these? - Stuart -

For some reason, I can't post an image, so I will try to explain in words.

DO you know how to use the "string trick" to find these?

No I don't, but I'm guessing that what you do is measure from the speaker to measurement, take a string, and add the length that sound travels in the number of milliseconds where the impulses occur. In the case of the first reflection at 3.39ms, since sound travels .34meters in a millisecond, I would add .34x3.39=1.152m or to make us Americans happy, 45.38 inches to the string. Then, attach the string to the center of the speaker and the other end to the measurement spot. whatever surface the string touches when taught and the angle of incidence equals the angle of reflection, this is the spot.

Don"t know why it isn't letting me post pictures. Here is a link to the original IR showing reflections at 3.4, 6 and 9ms. https://www.dropbox.com/s/5yrkm53ihx81lwe/IR%20from%20baseline.jpg?dl=0 the 3.4ms is at 30% and here is one that I just did putting 4" of 703 on the floor halfway between the speaker and the mic. https://www.dropbox.com/s/qhl7x61zmar6s6n/IR%20withj%20703%20at%20suspected%20reflection%20on%20floor.jpg?dl=0 all of them are below 7%. Did this just solve my reflection problems?

No I don't, but I'm guessing that what you do is measure from the speaker to measurement, take a string, and add the length that sound travels in the number of milliseconds where the impulses occur. In the case of the first reflection at 3.39ms, since sound travels .34meters in a millisecond, I would add .34x3.39=1.152m or to make us Americans happy, 45.38 inches to the string. Then, attach the string to the center of the speaker and the other end to the measurement spot. whatever surface the string touches when taught and the angle of incidence equals the angle of reflection, this is the spot.

:thu: Yup, you figured it out!

Did this just solve my reflection problems?

Probably! :)

putting 4" of 703 on the floor halfway between the speaker and the mic.

That's what I was going to suggest, but you beat me to it... Can you upload that new REW file for me? I'd like to see what else you accomplished with that... I have a suspicion... :) - Stuart -

Here is a file with left, right and both speakers from the mix position with me out of the room and 703 on the floor. https://www.dropbox.com/s/fd5si0skitjq6x8/left%20right%20and%20both%20with%20703%20on%20floor.mdat?dl=0

Built two superchunks in corners and about a half length in a wall/ceiling corner. I also built two panel absorbers according to the BBC 1992 paper http://downloads.bbc.co.uk/rd/pubs/reports/1971-15.pdf I used a slightly thicker panel and got 30hz instead of the 50hz I was targeting. I've only managed to budge 80hz up about 2 dB. Here is a link to the mdat file. https://www.dropbox.com/s/jkfy4b6miy0rmda/0511%20chunked.mdat?dl=0

I used a slightly thicker panel and got 30hz instead of the 50hz I was targeting.

Why were you targeting 50? I dont see any problem at all at 50. It looks pretty good to me. But your panel seems to be working at 30! Just look what it did to your mode at 28.3 Hz: BEFORE: AFTER: I'd say that's pretty successful! I'd also say that your superchunks are working quite well: BEFORE CHUNKS: AFTER CHUNKS Looks like a pretty good outcome, if you ask me! Not too shabby at all...

I've only managed to budge 80hz up about 2 dB.

That's reasonable, under the circumstances. Try moving your mix position 5 inches forward, and test again. Also, do another "walking mic" test, but in a different direction this time: Up and down. Do a series of 2" increments, going 10" up and 10" down from the current mix position. I'm not saying that you'll need to grow taller or cut the legs off your chair! I just need to check a suspicion. There's also something strange going on in your high end: Not sure what that is, but it shouldn't be there. Perhaps some kind of resonance with the speakers themselves? It's almost identical in both speakers (individually), so that would be my suspicion. Can you check with the mic up closer? Just a meter away? That "strangeness" seems to start at around 3k, and the level rises with frequency: I'd suggest checking your tweeters, the way they are mounted, and perhaps the crossover circuitry for the tweeters. - Stuart -

Not sure what that is, but it shouldn't be there. Perhaps some kind of resonance with the speakers themselves? It's almost identical in both speakers (individually), so that would be my suspicion. Can you check with the mic up closer? Just a meter away?

It is weird though, I don't think it has appeared before. I will make more measurements tomorrow. Hey Curt, if you see this, do you want to come over and do more measurements?

I don't think it has appeared before.

Here it is, in your "baseline" measurement: It's been there: just less visible, as it was partly masked by the longer room decay times. But now that you are getting that under control, it is more visible, because you killed some of what was masking it. - Stuart -

Eric Best wrote:

Hey Curt, if you see this, do you want to come over and do more measurements?

I can come over Saturday at 2pm EDT, if that works. Can't make it tonight unfortunately. That bump in the high freq range is very interesting.

Sounds good, I'll be here. Hopefully I will have it solved by then though!