General speaker measurement questions

[Saurav]

In no particular order (i.e., I'm thinking out loud)...

* What should the farfield (ungated) in-room response look like? Let's assume this is taken with the mic at the listening position, with both speakers and sub playing, and smoothed (say 1/16th octave) to remove the comb filtering jaggies (or whatever they are). If I ignore the peaks and dips, I seem to have a steady downward slope with increasing frequency. Is this expected or good? Or is it totally irrelevant? I also have a hole at around 300Hz that I need to find the cause for and fix.

* What's the best way to measure bass? Most measurements I've seen seem to go nearfield below around 200Hz and splice that into the 1m/2m measurements above that frequency. And that makes sense in general, because you don't want the room to interfere, you want to see how the speaker performs. But in my case, I'm building one speaker for my room, so wouldn't it be more relevant for me to take real in-room measurements?

* Floor bounce - my woofers' bottom edges are 9" above the floor. How does the math for that work? That will add SPL below a certain frequency, right? How far away do I need to place the mic to be able to measure this?

If I set the gating in SW to what looks correct to me, I get measurements down to about 400-500Hz. The usual way to get measurements below that seems to be to go nearfield, or take the speakers out into the yard and put them on a high stand. But what I'm thinking is, even if I did that, when I bring the speakers back into this room, the room will change the response in that region. So shouldn't I be measuring it in the room? And if I should, then I eed to understand the correct way to do it and how to interpret the results.

[Saurav]

Here's the farfield response, 1/3 octave smoothed.

[Saurav]

And since cjd asked, here are pictures of the new baffle. I cut up the 'dunce cap' baffle to make the asymmetrical wings on the sides. I owe thanks to Jon and Thomas for the duct tape idea This was cut freehand with a jigsaw, and I'm pretty impressed with myself because the curve on both baffles is pretty similar I should be glueing (gluing?) it up tomorrow, if it doesn't rain. The drivers are offset about 1/2" from center, towards the outside. The drivers are lined up with each other... I think my off-axis response would be pretty messed up if I had too much of a horizontal offset between the drivers, right?

[Saurav]

And very rough off-axis measurements. I just moved the mic around the speaker, eyeballing what I thought should be 30 and 60 degrees, and trying to keep the mic the same distance from the front baffle. The crossover tries to get 4th order L-R slopes on both drivers at 3500Hz. The woofer ended up with a 2nd order filter, and the tweeter has a 3rd order with a coil in series to dea with the rising top end.

[Saurav]

And now I'll stop talking to myself

[cjd]

What kind of series resistance (if any) do you have on that tweeter? I just can't imagine a gradual downslope on the top end like that given the driver efficiencies you're working with.

I'm not convinced you don't have a calibration issue. Very hard to tell, and one of the possible troubles with using no reference (also, possibly a result of using the mic input).

To check this, it might be worth measuring the response of just the tweeter. I found that I got measurements nowhere close to what I expected on my tweeters until I got things right. (You'll need a cap on the tweeter still I believe...)

Might also be worth running a full range (slow) sine sweep or even just a couple tones and check with your SPL meter to see if you really do have that much drop in output as frequency rises. Though the SPL meter may lose accuracy above 10kHz, I think it may still be useful.

Baffle looks good. How does it sound?

And, is that a phase issue causing the big dip on the bottom end? I don't remember what your low cross point is...

C

[TacoD]

And very rough off-axis measurements. I just moved the mic around the speaker, eyeballing what I thought should be 30 and 60 degrees, and trying to keep the mic the same distance from the front baffle. The crossover tries to get 4th order L-R slopes on both drivers at 3500Hz. The woofer ended up with a 2nd order filter, and the tweeter has a 3rd order with a coil in series to dea with the rising top end.

These measurements look better than the first graph, maybe you can smooth a little so the "trend" is more recognizable. You want to model the speaker and room so you have to measure with long gating (so that the mic also registers the reflections). But it can be hard to interpret things.

A farfield response heavily depends on your room, but in most rooms you can detect some standing waves (heavy suck-outs). Your first graph looks like you are missing some highs, but ribbontweeters are directional.

[Saurav]

What kind of series resistance (if any) do you have on that tweeter? I just can't imagine a gradual downslope on the top end like that given the driver efficiencies you're working with.

1 ohm, and there's a 0.18mH coil in series too. The tweeter measures flatter when the mic is closer. I'll try removing the coil to see if that flattens out the farfield. I don't know why it measures flatter when the mic is closer, and droops like that at a distance. The up-close measurement should have been further off-axis for the tweeter, because I had the mic at my listening height which is slight below the tweter's center.

Which brings up another question - what height should I be measuring at?

Might also be worth running a full range (slow) sine sweep or even just a couple tones and check with your SPL meter to see if you really do have that much drop in output as frequency rises. Though the SPL meter may lose accuracy above 10kHz, I think it may still be useful.

Good idea. I'm using the SPL meter's correction file from the ETF website. Maybe that's what it is, maybe the mic is directional at higher frequencies, so when it's placed far away and pointing straight forward, it doesn't pick up the treble correctly? That could explain it.

And, is that a phase issue causing the big dip on the bottom end?

If you're talking about the gated measurements, that's just where the window runs out. If you're talking about the 300Hz suckout in the farfield measurements, then yeah, I don't know what's causing that. Woofer-mid XO is at 500Hz.

You want to model the speaker and room so you have to measure with long gating (so that the mic also registers the reflections). But it can be hard to interpret things.

I tried increasing the gating. It registers more of the low end, but it also adds more ripples to the midrange/treble measurements. Which probably makes sense, I'm picking up reflections. It does make it harder to figure out what to believe.

Your first graph looks like you are missing some highs, but ribbontweeters are directional.

At my listening seat though, I should be a smaller angle off-axis from the tweeter, because the 1m measurements were taken with the mic at the same height, which puts it further off axis. Maybe it's the RS SPL meter, as I said above. I keep going back and forth on wanting to buy the ECM8000/UB802 combination, and thinking that realistically speaking, this will probably be the only speaker I work on for a long time. Maybe I can twist cjd's arm into lending me his setup

I can smooth the off-axis graphs and post them. How much smoothing would you want?

[cjd]

I forgot you were using the RS mic. Forget anything above 10kHz I think. At least, from everything I've heard/read. Get that Behringer setup (I mean, that rebate on the mic goes away in just a few days!)

.18 coil on the tweeter will not be insifignant!

Looking at that dip (I'm talking about the suck-out in the first graph) my first thought was phase issue with a 500Hz cross point. Try swapping polarity on one or the other (since, IIRC, this is active, swap the leads of the TM at your amp). Just to see for sure.

And, would also suspect gating issues are playing games. I would try moving your measurements to 1.5M.

Got any pulse measurements to share? I would focus first on getting gated measurements so you're comfortable with what you see, then move to un-gated at your listening position (and both speakers wired in). In other words, look for flat response from your speaker, and adapt room related issues as needed (they're likely to be mostly on the bottom end).

C

P.S. Getting my setup to you would cost ~$25 there and back I think... That's just under 1/3 of the cost of the items in question! And don't you have the parts to build your own mic/pre?! If you go that route, it's probably cheaper than borrowing my stuff.

[Saurav]

Yeah I know, borrowing someone else's setup doesnt make sense. I could try the Panasonic capsule, but I'm lazy so I'd rather buy the Behringer setup. The sale ends Jan 31st right?

The pulse measurement is hard to read too. I think my gate is right on the first reflection and should be moved back a litle, but I'm not sure. I never see a clear reflection point like the example pictures show.

I'm attaching the pulse graph, and the off-axis graph smoothed to 1/3 octave.

How long does 2 year old Gorilla Glue take to dry?

[TacoD]

It would be interesting to see the same graph with 1/6th smoothing. I am using UB802 + MC8000, but I am saving for ClioWin. This doesn't mean that the UB802/MC8000 is nog good enough, your speakers will improve a lot! Speaker building is my hobby, so I'll be designing a lot more in the future and thus justifying the costs of Clio.

Hmm, your posting your graphs while I was typing :E.
I see no major cross-over faults, now try several measurements with short gating to filter the room out of your measurements and see the real x-over performance.

[ThomasW]

The Behringer rebate program ends Jan 30th.

How long does 2 year old Gorilla Glue take to dry?

If it's that old it's usually dry in the bottle :wink:

[Saurav]

If it's that old it's usually dry in the bottle

LOL! Well the cap was glued on so I had to remove that, but the glue itself was still liquid. We'll see how it works. One of my duct tape + cardboard baffles colapsed in the middle of the night, and it's hard to be worse than that.

TacoD, what gating time would you recommend I try?

[Dennis H]

That hole in the 3-400 Hz range is probably floor bounce. To calculate it you need to figure how much extra distance the sound from the midrange has to travel when it bounces off the floor vs. the straight path to the mic. The frequency where that extra distance is half a wavelength is where you'll get a dip in the response. Example: midrange and mic both 30" above the floor and the mic 100" from the driver -- center frequency of the null is 408 Hz. There will also be a peak at twice that frequency where the bounce reinforces the direct sound rather than cancels it.

[cjd]

There will also be a peak at twice that frequency where the bounce reinforces the direct sound rather than cancels it.

Let's see... peak at 600, null at 300... you're one smart guy!

Suarav: stop listening to me, I don't know what I'm talkin 'bout. :P

C

[Saurav]

So I should try this with the woofer turned off, as well as only the woofer. That'll tell me which driver is behind this.

[TacoD]

LOL! Well the cap was glued on so I had to remove that, but the glue itself was still liquid. We'll see how it works. One of my duct tape + cardboard baffles colapsed in the middle of the night, and it's hard to be worse than that.

TacoD, what gating time would you recommend I try?

Sound travels with ~340 m/s, so if you measure at 1 metre it takes 1/340 = 3 ms before the direct sound arrives at the mic. For lower frequencies it takes longer before the complete wave has arrived. Period of a 50Hz wave is 20 ms, so I assume you have to measure for 23 ms to capture it. But in this timeperiod also reflections of higher frequencies will arrive at the mic. But lower frequencies (<50Hz) are not registered. I hope my story make sense.

Taco

[Saurav]

The EMC8000 and UB802 arrived yesterday. Does anyone have a FRD calibration file for the ECM8000?

In-room measurements are still pretty confusing. Moving the mic or changing its height makes one peak-null pair change in magnitude, so I'm pretty sure that's a reflection. Also, I'm not sure if I should measure at the tweeter's height, or my listening height. Measuring at the listening height puts it off-axis for the tweeter, and then the measured tonal balance changes depeding on the mic's distance from the speaker. And since the midrange is on an open baffle, I have to stay some distance from it to get meaningful measurements, right?

[ThomasW]

I don't have a FRD cal file for one, but I do have a ECM-8000 cal file.

10 2 0
15 2 0
20 2 0
25 2 0
30 2 0
35 3.08 0
40 2.5 0
45 2.5 0
50 2.5 0
55 2.5 0
60 2.5 0
65 2.5 0
70 2.5 0
75 2.4 0
80 2.3 0
85 2.3 0
90 2.4 0
95 2.4 0
100 2.4 0
150 2.5 0
200 2.4 0
250 2 0
300 1.7 0
350 1.7 0
400 1.5 0
450 1.3 0
500 1 0
550 1 0
600 1 0
650 0.9 0
700 0.9 0
750 0.8 0
800 0.6 0
850 0.4 0
900 0.4 0
950 0.4 0
1000 0.4 0
1500 0 0
2000 0 0
2500 0.4 0
3000 -0.4 0
3500 -1 0
4000 1.2 0
4500 0.8 0
5000 0 0
5500 1 0
6000 1.6 0
6500 1 0
7000 -0.3 0
7500 0 0
8000 0.3 0
8500 -0.8 0
9000 0 0
9500 0.3 0
10000 0 0
11000 1 0
12000 1 0
13000 0 0
14000 0 0
15000 1 0
16000 0.8 0
17000 0.6 0
18000 0.4 0
19000 0.4 0
20000 0 0
21000 1 0
22000 2 0

[Saurav]

Well, that's the FRD file format, right? Frequency, amplitude, phase? Thanks, I'll import that into Speaker Worshop.

[ThomasW]

Well, that's the FRD file format, right?

Beats me, I don't use the FRD stuff. It's the cal file from either TrueRTA or ETF 5.0, I don't remember which

[Saurav]

Looks right to me Thanks. And it took care of a dip I was seeing right around 3000Hz, and this cal file shows the mic's reponse has a dip there.

Here's what I'm seeing now, which is probably a lot more accurate than what I had before. The low end rolloff is a combination of the dipole and a 500Hz 4th order LR filter. The peak at around 1500Hz is probably a result of the dipole too. I didn't implement any of the dipole EQ circuits described on Linkwitz's website, because I had no way of measuring to find out what I needed. Now I have a better idea of what's needed. I'll try and hook up a passive RLC filter first to see if that makes any difference.

This is with the mic at seating height. If I raise it to the tweeter's height, the peaks and nulls between 1kHz and 4kHz get much worse. Distance to the speaker is just under 3'. Should I be measuring from further away? Any closer and the tweeter's response starts to drop off.

[Saurav]

And here's the ungated measurement from the same position. Measurements from the seating position with both speakers on look really bad, so this is the best I have for 'farfield' right now. I still have the hole at ~ 300Hz.

[Saurav]

More questions. This is the pulse response I get with my soundcard's output looped to its input. That can't be good right? I see that initial squiggle in the pule response for my speaer as well, but that's probably my soundcard. And since I only have a mono mic-in, SW can't compensate for this. I think I need a new soundcard too.

Can anyone tell me how much this would be affecting my measurements?

Thanks,
Saurav

[cjd]

Suarav: It may be worth experimenting with a little more distance - if you're doing a single point full rangemeasurement, you want to be at a point where the driver response is summing. Too close and you're way off-axis on that big woofer.

Claudio has a correction file for the Behringer mic as well for what that's worth.

Looking quite good though!

C

[Saurav]

Thanks, I'll take a look at that. I'll pull the speaker into the middle of the room and make it fire down the long wall, that'll let me get the mic further away.

[Saurav]

More graphs

This is the gated on-axis response. I had to pad the midrange down by a couple of ohms. It still has the 1K dip / 1.5K peak, which I would think are the cancellation/reinforcement from the back wave. I didn't trust this below 500Hz though, so I decided to try something else to check woofer level matching. I ran a sine wave sweep from 200Hz to 2kHz and then ran an FFT on that. This makes it look like my wofer level is fairly well matched with the midrange, and the 500Hz region in general looks a lot better than it does on the gated measurement.

It always comes back to the same question - which one do I trust?

[cjd]

ultimately, the ones on yer noggin.

For example, do your ears like the mid padded down a tad better?

This could be interesting with respect to room treatments - try putting a big blanket folded up at some reflection points and mic in the listening spot, see how it changes things.

C

[Saurav]

Well, that dip gets a lot deeper and narrower as I increase the gating window, so it's probably a reflection. I'll keep working on it. Thanks for the help.

[cjd]

How does it look if you put a pillow behind the mid (fairly close)? See if it's a backwave thing by damping the backwave a bit perhaps?

Any padding on those wings? Get some heavy felt (the kind they use as underlayment for floating floors works great and is fairly cheap and easy to find - free if you can find scraps from an install place) and experiment.

[Saurav]

Nope, I need to put felt on the back of the baffle (and the wings).

I put the speaker back in position and am listening to them now. I think it sounds clearer now, but that's a very initial impression. It was interesting to measure just the midrange - the cone breakup peaks show up ~ 30dB below the main output, between 5kHz and 8kHz. I'm sure it would increase treble clarity if I got those attenuated further. I've also lost a dB or two of overall efficiency because of the pad resistors on the mid, but it sounds better balanced now so that's not a problem.