Crossover / Notch / Breakup Question

[arc2v]

I am debugging my OB center channel speaker. It consists of Dayton Ref series 8" woofers, Hi Vi B3N mids, and a BG Neo3PDR tweeter.

Crossover points are at 500 Hz and 3 kHz. Low one is 2nd order L-R, high one is 3rd order L-R.

So far things sound pretty good, but I am chasing down a distortion crackle that occurs with some higher voices or some loud sequences.

I think I have nailed it down to the B3N's. It has a known breakup mode at 7-8 kHz and I wonder if that's what I'm hearing.

A couple questions:
is 3kHz at 3rd order too close to the breakup mode to fully damp it?
If so, what's a good slope or target dB to keep it down?

If the choice is a higher order crossover or a notch filter, which would you guys do?

Thanks for the help,
AC

[tinitus]

Yes, I think you should notch that peak, and maybe you can use the one from Zaphs design with B3N, which to me looks similar

http://www.zaphaudio.com/audio-speaker18.html

I question whether your xo is working properly at all
Firstly I dont know of any 18db LR filter ... but the resulting slopes from 18db filter may very well be 24db LR slopes, if you are very lucky

Why 18db on tweeter ... is it really necessary with 3khz xo
Do you have any idea about the actual acoustic slopes

I suggest you use Zaphs notch on the mid and only a 6db series inductor with an RC on the mids lowpass ... the rest should be 12db
personally I reverse mids polarity
Alternatively use an all 12db LR filter

You should definately use a notch on the 8" Dayton breakup too

[jnb]

Originally posted by arc2v
If the choice is a higher order crossover or a notch filter, which would you guys do?

Once the peak is nailed, you should be able to use any slope you want.

What's important here though is that the notch filter will introduce a peak in the impedance which will affect the crossover, possibly reintroducing the peak somewhat.

You should be able to develop a compensation network to balance the effect of the notch filter, which is what I'd do, then I'd probably use a lesser order crossover.

[arc2v]

Thanks for the help.

The target was 4th order L-R, the electrical is 3rd order. I liked the way the phase modeled on the 3rd order better and it seemed to be enough.

The reason I kept it that high was the rapid rolloff of the Neo3 below 2kHz

This was all modeled in the Passive Crossover Designer spreadsheet using the published data from Parts Express.

Unfortunately, I don't have the parts to make this one at home just yet. I'll have to put in another order with parts express soon.

I'll try the notch filter and a redesigned midrange crossover that's 4th order electrical and compare the results. I'm still playing around with the more advanced measurements in Speaker Workshop, so once I'm comfortable with those I'll start posting some plots.

Thanks again,
AC

[jupiterjune]

I think I have nailed it down to the B3N's. It has a known breakup mode at 7-8 kHz and I wonder if that's what I'm hearing.

IMHO -- try crossing over at 2300 hZ or less (1/3 of the breakup node).

How is the midbass?

The last pair of speakers I built used the Tangband 3" titanium cone drivers as midranges. Upper midrange voices sound amazing, completely organic. The set built before that used 4" metal cone drivers. They sounded great, but I could still hear a bit of metal resonance. The midbass seemed to be much better with the 4" drivers though.

JJ

[arc2v]

The midbass is fantastic. Quick, clear, plenty of punch. Most of that is coming through the Dayton's.

I would rather not cross that low if I don't have to because I wanted to avoid a crossover frequency in the most common vocal regions. This way, the B3Ns are covering most voices, the tweeter is just effects and upper harmonics, and the Daytons handle the mid bass.

It was a great plan, until I started hearing all this distortion

I'm going to try the notch filters first and go from there. I'm sure I can nail this down.

Thanks for the help all!

[jupiterjune]

AC-
I just prototype the crossover and listen to various crossover points and filter slopes. Take notes, go with the one that sounds best.

JJ

[arc2v]

Hey all, an update:

I decided to redo the crossover entirely. too much tweaking until it was so different that I didn't recognize it (which led me to believe I was chasing around a problem I should have avoided).

So, I fired up Speaker Workshop and took nearfield readings of all my drivers individually and then impedance plots of them individually and in parallel. So at least now I'm working with real data and not manuf. estimates.

Woofer and tweeter were easy to hit the targets (2nd order L-R acoustic and electric woofer), (4th order acoustic, 3rd order electrical with notch filter for the tweeter): crossover points stayed at 500 and 3kHz.

However, now I'm running into problems trying to get the bandpass for the midrange optimized. Unfortunately there's a hump in the FR curve right before the crossover point (3kHz), so I'm all over the place trying to make the final response look good. I'll try again tonight, but I may implement plan B.

Plan B is to move the high crossover point to 2400 Hz and keep the low one at 500 Hz. 2nd order target (L-R) for the 500 Hz and 4th order (L-R) for the 2.4 kHz.

The Neo3PDR has very low distortion and can handle down to 1800 Hz open baffle, as proved by Dan Wiggins with his DDR design, so 2400 should be fine. Problem is the RS 8" woofer break up at 1800 Hz, so I still need to cross them lower and fill in the void. So the B3Ns will only be covering about 2 1/2 octaves, but they are important octaves

I'm hoping that with these new crossover points, I will avoid the hump in the B3N performance, roll off fast enough to avoid the B3N distortion, and still keep the woofer low enough to avoid its breakup distortion. Worst case, I will roll the woofer off to a 4th order slope or notch the breakup, but I'll revisit that when I have to.

If anyone has any advice, I'd love to hear it. Otherwise, I'll try to post some results tomorrow. I wish there were an easier way to get pics out of SW.

Thanks for the help all!
AC

[jnb]

Try taking your measurements in the far field. A baffle generally likes a few feet to show up its effect so you can compensate for the diffraction step, and nearfield measurements have issues up high which can show peaks that aren't really there.

To post screenshots, press [Alt] & [PrtScr] at the same time then open your paint style editor and choose paste. Crop away the edges as required. Reduce its size to the forum imposed limit, and save as gif or jpeg depending on the nature of the image, to bring it down to size.

[arc2v]

Okay, another night of testing, modeling, and head scratching.

I redid all the tests in farfield, 1m exactly, microphone on center with the tweeter. Pillows and blankets everywhere to tame the reflections, and I set the gates such that I got good information down to about 200 Hz -- plenty for the crossover points I was considering.

I also upped the MLS repeat to 5 from 2 and the results cleared up a bit, although there was still some jaggedness in there.

I then looked at the data long and hard and noticed something strange, the peaks I have been fighting weren't driver-related, they were baffle-related.

Let me now put in a plug for Edge. I got it running, designed my baffle and put properly sized drivers in the right location and it was eerily accurate. I had a devil-horn peak around 2500-3k Hz on the B3N's, exactly what Edge predicted.

A null in the woofer and the boomy response at 500 Hz, also perfectly predicted by Edge. I had dismissed it because I had seen good results in other speakers (not mine) that Edge said would have rolled off way too early, but measured great in-room. So what I"m thinking is happening is that Edge is great for gated-farfield, but in-room gives more bass than it predicts. Fair enough.

So back to my speaker. It seems the problems I have been fighting in frequency response are more from the baffle than the actual components or crossover. I still have distortion issues to contend with on the woofs and mids, but they seem relatively minor now.

Originally I designed the baffle for bass extension, not realizing that it also adds a peak above the Feq point. That's problem number 1 to deal with. Problem 2 is that I get a nasty peak null region below 1kHz, right where the mids are fully in charge. Right now this is acting destructive with the driver and the response is all over the place.

So it looks like I have some baffle design to do first, then I'll get back to the crossover. Right now the plan is to put the baffle peaks right on the crossover points, so I can play with crossover slope and overlap to null them out. The final design is probably going to be really weird looking.

Thanks again for all the help.