Off axis "flares". Why do they happen? How to fix them?

[fjhuerta]

Hi all,

I've been out of the hobby for some time, but I decided I had enough free time again, so I decided to start again... and I found an issue I thought I had solved, but didn't.

I built a three way speaker with some NOS drivers - a pair of shielded RS-225s, 2 RS-100ps and a Dayton RS-28F.



I didn't do roundovers just because of aesthetics, but I don't think this is the main issue.

I used to do just a single off axis measurements, but looking at the spinoramas on the web, I decided to take multiple mesurements around the speaker this time, at about 4 feet. This is what I found on the midrange.



That flare at 2 KHz really bothers me, and it's obvious at 45º and really big at 75º.

First - why does this happen? I used Jeff Bagby's Excel spreadsheet and it shouldn't be there. No matter how I model the box, there's no "flare" to be seen.

Also, now that I know its there, what's the best way to get rid of it? Reason being, if I get rid of it at 75º, I get a dip on axis.

I'm using two MiniDSPs, so I can test everything I'd imagine. I have designed a flat measuring crossover wiht a 2 db "Flare" from about 1.5 to 2.5 KHz off axis and the speaker sounds shouty and hard.

Do you know what causes this flare? What's the best way to deal with it - eliminate it, or live with it?

[fpitas]

May be related to diffraction off your perfectly flat, symmetric front baffle. Try roundovers; for example closed-cell foam half-rolls along each edge. Plywood half rounds are also available if you want something nicer looking. I've also seen people use 45 degree "wings" off the front baffle to minimize diffraction.

[devnull]

You could try some large diameter PVC pipe cut in half for your roundovers.

[fjhuerta]

Wellm at this point I could probably cut a PVC pipe in half as roundovers, but aesthetically, this would be a no-no...

Besides... the sides do have small roundovers. Nothing that big, but it shouldn't matter this much.

If I had to fix it in the crossover (and I have to) would you eliminate the diffraction completely off axis, leave a flat frequency response from 30-15-0-15-30 degrees, or do something in between?

[wogg]

You could test first. Split a PVC pipe and painter tape that sucker up for some measurements. That would tell you if you're onto the problem and let you plan how to accomplish that aesthetically later.

[fjhuerta]

You could test first. Split a PVC pipe and painter tape that sucker up for some measurements. That would tell you if you're onto the problem and let you plan how to accomplish that aesthetically later.

Good idea.

This is an issue I've had in many of my designs, roundovers or not. Nevertheless, as I said.. these ones have a slight roundover.

This sucks, because it's a pretty good idea... these drivers if working correctly would have given me a quite smooth response :(

[fpitas]

Not a real helpful observation maybe; but your tweeters being right in the middle of the baffle will exacerbate the problem.

[fjhuerta]

Not a real helpful observation maybe; but your tweeters being right in the middle of the baffle will exacerbate the problem.

Yep... I assumed that, but I wanted a speaker that was symmetrical, and hoped that the diffraction would show up both on axis and off axis. This didn't happen.

Weird thing is, the problem is not the tweeters, it's the midranges! :O

[fpitas]

Yep... I assumed that, but I wanted a speaker that was symmetrical, and hoped that the diffraction would show up both on axis and off axis. This didn't happen.

Weird thing is, the problem is not the tweeters, it's the midranges! :O

The roundovers may not be big enough to work much below a few kHz. It's a matter of being comparable to the wavelength.

[PWR RYD]

I'd rather have a small dip around that frequency on-axis than a large peak off axis. That said, I've never measured or was concerned what my speakers measure like beyond 45 or 60 degrees. 75 degree measurements are not even on my radar.

What xo slopes are you trying to hit? LR4?

[fjhuerta]

I'm currently working on getting rid of that flare, and balancing off and on axis response, since the speakers are already built. The interesting thing is, from reading several (commercial) designs they do it all the time. I just hadn't noticed.

I'm now wondering how on earth Revel does those beautiful off axis plots with no imperfections off axis. And whether we as DIYers can approach that level of perfection.

[fjhuerta]

I'd rather have a small dip around that frequency on-axis than a large peak off axis. That said, I've never measured or was concerned what my speakers measure like beyond 45 or 60 degrees. 75 degree measurements are not even on my radar.

What xo slopes are you trying to hit? LR4?

I started with LR4, moved to LR8 and I'm now at LR2. It seems that the slope is far more natural with LR2 (I love my MiniDSP).

I do agree with you. I'm aiming for a small "bbc like" dip at 0, and 15º, flat response where the flare is at 30 and 45ª, and not worry that much above that. I have noticed how much I actually enjoy a BBC dip anyway, so this could work well in my room / with my listening preferences.

[dlr]

Yep... I assumed that, but I wanted a speaker that was symmetrical, and hoped that the diffraction would show up both on axis and off axis. This didn't happen.

Weird thing is, the problem is not the tweeters, it's the midranges! :O

First, have you measured the drivers individually off-axis? That will help to see where the issue is.

Second, as others have stated, it almost certainly is diffraction. If you optimize for flat on-axis response in a symmetrical baffle design with little diffraction control, you're bound to have a lot of change with angle. Peaks and valleys that will often invert on other axes. There's often a dip on-axis, so when you bring it up to flat on-axis, it introduces a peak off-axis.

This is where power response comes in. I have always thought, and Linkwitz had similar thoughts (IIRC) at his site that on-axis is most important, but power response is high as well (maybe second, can't find the list, low distortion may be higher). I believe that bumps in power response can "over power" (pardon the pun) the on-axis response. What you have is a non-flat, non-monotonically decreasing power response. There should be no peaking in the power response.

If you don't have diffraction control, a flat on-axis response may very well not be optimal in perception. It's not easy to optimize the crossover in this situation. One thing I've considered, but never discussed much on the board, is that if you do not have big roundovers and/or do not want something such as felt on the baffle (when it's finished), consider using extensive felt for diffraction control during the design phase only. This will allow you to design on the primary listening axis only, going for flat response because the diffraction will be minimized, both on- and off-axis. When you're satisfied with the sound, remove the felt. You will then have anomalies on-axis, but the power response will be largely unchanged. It may be the best compromise, otherwise you'll get guessing at how to optimize the crossover.

dlr

p.s. Remember, though, that flat on-axis in this case isn't the only criterium. You must take driver directionality and its impact off-axis into account as well, so crossover Fc is going to affect that as well as filter order. This affects power response.

[dlr]

I started with LR4, moved to LR8 and I'm now at LR2. It seems that the slope is far more natural with LR2 (I love my MiniDSP).

I do agree with you. I'm aiming for a small "bbc like" dip at 0, and 15º, flat response where the flare is at 30 and 45ª, and not worry that much above that. I have noticed how much I actually enjoy a BBC dip anyway, so this could work well in my room / with my listening preferences.

Years ago, Jeff B. and I were discussing this in a thread. We both had settled on 2nd order having the best sound. I also have always preferred a bit of a dip, but with full diffraction control as well in my systems. We believed that we preferred second order due to the broader dip in power response of LR2 vs LR4.

This was back in the day of the Madisound B.B., so it's unfortunate that all of those discussions are gone from Madisound and only in the Wayback Machine.

dlr

p.s. However, my current system is a linear phase dipole using the SoundEasy Ultimate Equalizer, so I'm using LR8 to achieve the low M/T Fc I needed in a 3-way for best power response. But with that I eventually added a small dip from 1-3K.

[fjhuerta]

First, have you measured the drivers individually off-axis? That will help to see where the issue is.

Second, as others have stated, it almost certainly is diffraction. If you optimize for flat on-axis response in a symmetrical baffle design with little diffraction control, you're bound to have a lot of change with angle. Peaks and valleys that will often invert on other axes. There's often a dip on-axis, so when you bring it up to flat on-axis, it introduces a peak off-axis.

This is where power response comes in. I have always thought, and Linkwitz had similar thoughts (IIRC) at his site that on-axis is most important, but power response is high as well (maybe second, can't find the list, low distortion may be higher). I believe that bumps in power response can "over power" (pardon the pun) the on-axis response. What you have is a non-flat, non-monotonically decreasing power response. There should be no peaking in the power response.

If you don't have diffraction control, a flat on-axis response may very well not be optimal in perception. It's not easy to optimize the crossover in this situation. One thing I've considered, but never discussed much on the board, is that if you do not have big roundovers and/or do not want something such as felt on the baffle (when it's finished), consider using extensive felt for diffraction control during the design phase only. This will allow you to design on the primary listening axis only, going for flat response because the diffraction will be minimized, both on- and off-axis. When you're satisfied with the sound, remove the felt. You will then have anomalies on-axis, but the power response will be largely unchanged. It may be the best compromise, otherwise you'll get guessing at how to optimize the crossover.

dlr

p.s. Remember, though, that flat on-axis in this case isn't the only criterium. You must take driver directionality and its impact off-axis into account as well, so crossover Fc is going to affect that as well as filter order. This affects power response.

iHi, long time no seen!

Yep, the flat on axis LR8 was glaring in the midrange like crazy. The non-optimized LR2 was far more relaxed. And the whole diffraction issue was the byproduct of an idea I read... that symmetric speakers, radiating sound in all directions equally, were preferable to asymmetric ones, that could tilt the listening axis. I assume this works for narrow, very rounded cabinets. Mine are too wide and without enough roundovers.

I'll try to optimize for both an on-axis BBC dip and the flattest off axis response. At this point, modfiying their looks is kinda out of the quetion.

And next time, I'll do something like Linkwitz LX-521, with minimal, open baffles. My room is terrible, and the only time I felt I had some great sounding speakers there was when I used open baffles. I assume that the null at 90º helps tremendously with first order reflections.