I'm grateful for some help with speaker measurements.

[Toe in]

Hi, I'm (trying to) construct a fairly ambitious active 3- way speaker system. I'm not really experienced when it comes to speakers - I normally into various amplifiers and DAC's.

So my initial question is; Is it really feasible to measure and develop a speaker i a normal living room? There are articles ( In this REW forum, there is a guide "How to make quasi-anechoic speaker measurements with REW") about doing pretty good measurements of speakers in an ordinary room, if we are gating the measurement. This guy presents really nice curves - usually flat and nice within some 2dB. But I'm not near that kind of resolution.

My room is around 4 by 5 meters, and is fairly good damped with absorbents on the walls, etc. It's a pretty enjoyable listening room. In order to make some measurements, I have hoisted out most stuff, and I have positioned the speaker in the middle of the room. The microphone ( an UMIK 1 ) is around 1.1 meters from the speaker - on axis. The tweeter is a Mundorf AMT type - a pretty expensive one ( ~700 EUR pair). And the enclosure looks a bit odd - it's made up from concrete and MDF boards. Please look at the photos. I expected the result to be a mess, regarding all reflections. But I have windowed the output at 5ms.

As you can see, the SPL varies by almost 10dB - a far cry from the curves Mundorf provides us with. It seems as the worst part is between 3khz and 7khz, which implies that perhaps my enclosure is flawed. I suspected some diffractions originating from the part where the tweeter is mounted, so I made experiments with filling the empty space between the tubes with sheep wool. No difference, so I added more and more, until the speaker was completely covered with wool, but all the curves looks exactly identical, more or less. I also positioned wooden bits on the sides of the hollow part, but everything looks more or less the same. The peaks and dips are just positioned a bit different.

My second question is; REW uses some math to "spirit away" the part of the signal that is beyond the window. Can it be so that if the room is heavily contaminated by reflections, the algorithms really can't do it's job properly?


Also, perhaps someone has something to say about my enclosure. It's made of concrete, MDF, bitumen sheets, etc. It's probably quite rigid, but I have some concerns about it being prone to diffractions. Especially the part behind the tweeter. Therefore, I'm a bit frustrated that I can't obtain some nice and reliable curves.
BTW, my design goal was to create the ugliest "high end wannabe" speaker in the western region of Sweden.

This didn't do any good.

As a reference, I'm have my old B&W804D speakers, which is lower high end, and the curves are not much nicer there. They have a well known dip around 2-3 khz, though. The blue line is B&W, the orange one is my project speakers.
And, I must mention; these curves are made with just the tweeter connected - except the B&W of course.

 

[jtalden]

You could have posted the mdat so we could better comment, but my thoughts are:

• Mundorf would probably have used a large flat baffle and flush mount the TW for their measurements to avoid diffractions.
• As you noted, the edges of your 'baffle' are causing much of the SPL problem using the window settings you applied. There is lots of diffraction from the various edges and maybe a bit of baffle step impact. As you change mic distance the SPL response will change accordingly. The wool was apparently not adequate to prevent much of this diffraction. Possibly flush mount of all drivers would help significantly?
• To see something approaching the direct sound response of the tweeter that Mundorf shows you can apply an FDW window instead of the right window. Try something pretty heavy like 3 to 5 cycles or so. It will not clean up the lowest frequencies, but will help the higher ones. You could also place the mic maybe 1-3 cm from the tweeter to greatly reduce the diffraction impact.
• You can still analyze the XO handoff timing for either a passive or active XO design without too much difficulty.
• For EQ of the direct sound of higher frequencies a single measurement like this would not be a good idea. The SPL and sound quality at the LP will probably not be as bad as this 1.1 m SPL looks. It is probably best to EQ to an SPL average around the LP. That said, normally early baffle diffractions are avoided as much as possible in better speaker designs.

 

[Toe in]

I'm glad to get some help, but I didn't understand much of it.

First of all, I'm planning a diffraction reducing approach, so that the side of the speaker, around the tweeters, will gradually slope backwards.

And the problems in the region around 3-6 khz can't be due to diffraction caused by not having flush mounted the drivers - that would have caused disturbances at 6 to 7 khz. The problem around 3 to 5 khz is probably caused by problems with the baffle edges - If I can rely on my measurements.

But...

The part about FDW - I didn't get that at all. That was Chinese for me. Please elaborate!
And did you really really mean to place the mic 1-3 cm away? And I do not want to reduce the impact of the diffraction! I want to see as much as possible so I can do something about it.

And sadly, the last part was Chinese also.

Finally, my problem right now isn't possible diffraction. The problem is that I don't feel I can rely on my readings. I mean, my B&W hi-end speakers measured poorly as well. I don't believe they are that bad.

Thanks anyhow.

 

[jtalden]

The part about FDW - I didn't get that at all. That was Chinese for me. Please elaborate!

The FDW option appears in several locations as shown in the REW help search below:

And did you really really mean to place the mic 1-3 cm away? And I do not want to reduce the impact of the diffraction! I want to see as much as possible so I can do something about it.

Sorry, I thought you were interested in seeing if the direct sound better matched the Mundorf SPL response as a way to confirm that your measurements were accurate. That is why I suggested close mic placement or applying FDW on the current measurements. If you are satisfied that measurement is accurate and due to baffle diffractions then I misunderstood the question.

Finally, my problem right now isn't possible diffraction. The problem is that I don't feel I can rely on my readings. I mean, my B&W hi-end speakers measured poorly as well. I don't believe they are that bad.

If your measurements are repeatable and you are using a resonable measurement mic then the measurements are likely correct. I suggested the mdat file could be posted in order to see if anything in it appears to suggest a problem with the measurements.

 

[Toe in]

OK, it's a good thing to confirm my SPL with Mundorf's. But I assume my mic isn't faulty, and I rely on Mundorf's curves. I'm simply uncertain whether my room is "good enough" to perform really useful measurements in. You're implying that it is. I'm going to test this FDW thing.

But I'm afraid I have shot myself in the foot when it comes to the enclosure design. Speakers is a new subject for me, and one always does the wrong things in the beginning. So, from the diffraction point of view, the baffle is really bad - in many aspects. Firstly, the rounded shapes makes the distance to the edges the same all the way around. And not flush mount spe drivers. And that space behind the tweeter. I made the edges rounded with the router I used, but realize now that the the rounding is too sharp, so to speak, to do any good.

I think I will make another baffle. Thankfully it's easily detachable. I will sit down and think.