Woofer problem in a 3-way design

[Martyn]

Close enough?

I didn't have much time today, but I did try a couple of things. I played with the DIP switches in the miniDSP again, this time watching the level monitors on the Input side - no sign of any problems so I'm still at zero gain. I tried boosting the SPL with the mic further away from the speaker, but still couldn't get far before feedback appeared.

Then I moved my test box to one side and put my "daily driver" in the test position. With the axis of the mid in almost the same place, I left the gating at 4.7 ms and played with the volume to get a response that fits over the test box response. To my surprise, this is what I got:



Now I understand why my test box would seem OK with some tracks but lacking bass with others - in comparison. This might be a question for a new thread, but I'll try it here first anyway: the fundamental question is whether a passive crossover design can be "backed out" of miniDSP crossovers? I suspect that the easy answer is "No", but in the case of this speaker, it seems to me that the addition of a couple of broad peak filters each with about 5 dB of gain would clean it up quite well. Presumably someone who knows what he's doing (not me) could sketch out the components for a couple of filters fairly easily, given the existing circuits?

Next I put both speakers on the floor, toed in slightly, and placed the mic about 7' away (still well clear of the back wall). I then tried to match the SPLs of the two speakers (not easy, given the bumpy response of the old speaker). I then sat and listened as I switched between the two speakers. Both used Squeezeboxes as sources, but they used different amplifiers. However, while amplifiers do sound different, in my experience speakers easily dominate. My test box won fairly easily, although there were some tracks where that couple of dB boost between around 150 and 1,000 Hz added a lot of character to the old speaker.



I think I've reached the point where I'm reasonably confident that I can go ahead and invest a lot of time and effort in a pair of fancy enclosures with a high expectation that I'll be able to make the end result work well. This is quite a relief, because I really didn't want to have to build a bigger enclosure to take RS225 drivers!

[Juhazi]

I think that you should measure MBOW also with tweeter's polarity inverted!

There are calculators for passive notch filters, but that is all I know about passive circuits.

[xandresen]

Martyn

It would be very interesting to place the two speakers side by side (almost touching), then retake the measurements (post 46) using a 50msec gate.

This will show you all the room response but because the two speakers are in the same place, you can still compare the results below 200Hz.


I looked at the microphone you used - it seems unlikely to be a problem.

[Martyn]

Juhazi, good idea, but it's OK as it is. Here's the normal response (red), another with the woofer inverted (blue), and a third with the tweeter inverted (orange):



The drivers are supposed to be crossed at 350 and 2,600, both acoustic 2nd order LR. One of the tweeter dips is around 1,300 - about half the XO frequency...


xandresen, I have now done this (speakers side by side). I've added two images below: the first shows the responses of the two speakers with the mic at 50 cm on the mid axis in each case, gated at 50 ms, unsmoothed, and separated by about 10 dB for clarity. The second image shows them coincident and smoothed 1/12 octave. The speakers were sitting directly on the floor, because my box wasn't wide enough to take them both. In both cases, my test box is red.





The two dips in the green response (MBOW1) aren't as deep as yesterday. Perhaps there's some reinforcement from the floor or from the other speaker? In a further listening test, the differences were less obvious too - but still evident. I see that the 200 Hz roll-off has improved as well at 50 ms.

When I have built my new boxes, perhaps I should do all the measuring outside for the purposes of integrating and smoothing the responses, and then setting up those filters in the Output section of the miniDSP. I'll then bring the speakers inside and measure them in their final positions, and then set up the room EQ filters in the miniDSP's Input section. This way the speakers would be agnostic with respect to whichever room they're in, and only the room EQ would need to be changed. Makes sense to me...

[Juhazi]

Orange line of MBOW means that timing MT is a bit off, because there are two drops. If you have time to measure it, my guess is that summing is spot on 15 Deg lower.

Other measurements are looking very good! Dip at 100-200Hz is wall behind the speaker or behind the mic.

[ergo]

Martyn, this question "..the fundamental question is whether a passive crossover design can be "backed out" of miniDSP crossovers? "

If I understand correct you are asking that once you have a good MiniDSP filter/EQ config then would it be possible to implement same with passive xover. If so, answer is yes, generally that can be done.
The process would be along the lines of
A) you'd measure the response of each MiniDSP channel (tweeter, mid, woofer). Can by done with REW, just in this case you can connect the MiniDSP analog out direct to soundcard used for measurement... and as a plus there are no room modes to deal with. So a 'non gated' measurement is ok
B) above gets you the "filter curves" that includes gains, like in attached image circled yellow.
C) in past I've then used LspCAD to use these curves 1 by 1 as a "reference curve" and built a passive circuit and get it to match each curve manually or with optimizer as good as possible. From free tools VituixCAD 2 seems to have ability to import the "overlay curve" also to the filter gain part of window, so would allow doing the same at least manually
D) Once you get all filter curves matching there is potentially some additional tuning of the response. What will be difficult is to repro any time delays if you happen to use those to align the acoustic centers of tweeter/mid ect. In analog world creating perfect delay is quite impossible. There are all pass filters that can sort of help but it will be hard to match that aspect very well

[Martyn]

....timing MT is a bit off...

Do you mean the time alignment? I suspect not...sorry if I'm being a bit slow this morning.

[Martyn]

Yes, that's the question. A problem for speaker-builders is that we end up with lots of surplus speakers! The problem becomes worse if they don't have crossovers. Also, being able to tweak an existing passive crossover is an attractive idea for those of us who lack the electronic knowledge.

A) Yes, I think I can feed the miniDSP back into my sound card.
B) Free tools are good! I'll take a look at VituixCAD2

Thanks for the info!

Edit: Along similar lines, I have run REW on the MBOW1. It looks like I can essentially smooth the response with a handful of notch filters. Just have to figure out how to do this. It's not at the top of my to-do list, but I know I won't be able to shake the idea.