Thinking similiar to Statement/Khanspires but ...

[rpb]

You guys are going to say 'what in the heck is this nut job doing?" Just trying to figure this out in my own head. The mids are offset 5" from the tweeter and the woofers 8". If I did this correctly. Measuring on tweeter axis @40" (~1m) means the distance from mic to mid increases by 8mm, and the distance to the woofer increases by 53mm. This is to the front plane of the speaker = plane of the tweeter.

When building a passive crossover in PCD, is it the distance to the back of the woofer cone, from the front plane of the speaker, that we measure to get Y offset? Then when working with something like the minidsp, where we can set delay, would we add the Y offset (depth of cone behind front plane of speaker) of the individual drivers to the difference in paths traveled for each driver to get individual driver delay settings in the ballpark?

The distance from the front of the speaker to the back of the mid cone is 21mm, the woofer 35mm.

Then the total excess distance from mic to mids would be 8+ 21=29mm, the woofers 53+35=88mm. On the minidsp delay for .028m. is .08ms (.08ms is the closest to .029 the setting will allow), and 88mm for the woofers translates to [email protected] (again not exact but as close as mini would allow).

I'm really bad at transposing things ... and not that great at engineering kinda' things either, but ... the tweeter needs delayed by .26ms and the mids by .08. Theoretically this should roughly 'time align' the drivers??? What should this do to freq. response?




[ATTACH=CONFIG]n1349404[/ATTACH]


But whoa ... look at the differences in rise times of the individual drivers. Can this really be time aligned? How?

[ATTACH=CONFIG]n1349405[/ATTACH]

My ramblings, not trying to question anyone's logic, just trying to work things out in my head so I can understand. You guys are the teachers, I'm the student ... in the corner.

The geometry at 1m is different than at 3m. (Your listening distance, roughly.) In PCD, you can change the distance. Much, or likely all of your thinking / concerns, are accounted for in the software.

It would be nice to measure at 3m, but that requires a lot of space. For the tweeter crossed at 2.5k, you could measure from 2m, and gate enough to see if everything works as expected.

For woofer x-over, it helps to get creative, or trust your ears, or both. (See my comment about adding some additional delay at the end of my list.) I'm not sure you need additional delay, but you can try it easy enough.

[rpb]

Take a look at this. This is how I would approach it.

Edit!.... The delay mentioned on line 10 is added to the tweeter, not to the mid. I'll fix the jpeg shortly.

[Kornbread]

That's a problem! Here's why. One millisecond translates into about one foot of distance. So, .1ms would be about an inch. .05ms would be about 1/2". The offset is sensitive to small differences. I'd try .03ms delay on the tweeter. Then invert polarity, and see if you get a null. Maybe try values from .02 to .06ms.

You guys are going to say 'what in the heck is this nut job doing?" Just trying to figure this out in my own head. The mids are offset 5" from the tweeter and the woofers 8". If I did this correctly. Measuring on tweeter axis @40" (~1m) means the distance from mic to mid increases by 8mm, and the distance to the woofer increases by 53mm. This is to the front plane of the speaker = plane of the tweeter.

When building a passive crossover in PCD, is it the distance to the back of the woofer cone, from the front plane of the speaker, that we measure to get Y offset? Then when working with something like the minidsp, where we can set delay, would we add the Y offset (depth of cone behind front plane of speaker) of the individual drivers to the difference in paths traveled for each driver to get individual driver delay settings in the ballpark?

The distance from the front of the speaker to the back of the mid cone is 21mm, the woofer 35mm.

Then the total excess distance from mic to mids would be 8+ 21=29mm, the woofers 53+35=88mm. On the minidsp delay for .028m. is .08ms (.08ms is the closest to .029 the setting will allow), and 88mm for the woofers translates to [email protected] (again not exact but as close as mini would allow).

I'm really bad at transposing things ... and not that great at engineering kinda' things either, but ... the tweeter needs delayed by .26ms and the mids by .08. Theoretically this should roughly 'time align' the drivers??? What should this do to freq. response?







But whoa ... look at the differences in rise times of the individual drivers. Can this really be time aligned? How?



My ramblings, not trying to question anyone's logic, just trying to work things out in my head so I can understand. You guys are the teachers, I'm the student ... in the corner.

[donradick]

Hey Korn -

It's not obvious - does your design have open back mids like the Statements?
If so, you might prefer a broad FR with the mids about 5 dB lower.

In any case, I'd advise trying to simplify the problem.
Get a standard play list of 5--8 music excerpts, and work only with the mids and tweeters until those sound good, then
turn on the woofers and adjust them until they match the upper end.

Also, you can easily create an impulse signal in Audacity, then use that leading spike to adjust the delays
until you have a coherent time response.

hope this helps.

[malboro2]

One of the major benefits of electronic crossovers is that you can actually go with what your ears like, rather than with some measurement that may be technically correct but doesn't match your ear/brain sensitivities. Go with what is most enjoyable to you, not with what makes the flattest measurement.

On my last line array, 3 way electronically crossed with three power amps, I didn't even bother with measuring, even though I bought the equipment before hand to do so. I went with what sounded the best to me and my wife. After all, we are the ones listening, not some computer printout.

[rpb]

0 on this try.

Looks like a lot of the attempts were around: Tweet 1.04ms, Mids .68, ahead of the woofers.

That's a problem! Here's why. One millisecond translates into about one foot of distance. So, .1ms would be about an inch. .05ms would be about 1/2". The offset is sensitive to small differences. I'd try .03ms delay on the tweeter. Then invert polarity, and see if you get a null. Maybe try values from .02 to .06ms.

[Kornbread]

How much delay?

0 on this try.

Looks like a lot of the attempts were around: Tweet 1.04ms, Mids .68, ahead of the woofers.

[Pete Schumacher]

Those are smoother than a baby's butt and surely sound even better than they look. My abilities are nowhere near that level ... but I'm trying ... minus frustrations. But let me see if I understand, by just sanding the formica seams, with no filler of any kind, even butt seams, the formica seams will not show? Gotta link to the process?

No links, but here's basically what you do. When you epoxy the Formica sheets to the MDF, you get a bit of squeeze out at the edges. The epoxy and Formica are very similar materials and the seams become invisible and very stable. When you lay the next sheet of Formica to overlap the first one, the squeeze out allows for a seamless edge between sheets. Sanding smooth leaves a surface that is easily prepped for primer/paint.

If you plan ahead and make it so that all the b u t t joints of the MDF are on one panel, then you only need to apply the Formica to that panel, but it can't hurt to cover all the exposed surfaces.

[rpb]

How much delay?

[Kornbread]

Sorry to hear about the seams showing. If you do refinish them, epoxy Formica to the panels. They will cover the seams and sand smooth to accept any kind of paint you could want.

Here's what painted Formica can look like.

Those are smoother than a baby's butt and surely sound even better than they look. My abilities are nowhere near that level ... but I'm trying ... minus frustrations. But let me see if I understand, by just sanding the formica seams, with no filler of any kind, even butt seams, the formica seams will not show? Gotta link to the process?

My 2 cents. How high is the ceiling in your room? More than 10'? Get the speaker on a 4' to 6' support with the speaker horizontal.

Have you tried turning the woofers off, and making a 2-way? (Temporary till hot treble is resolved.)

Did you delay the tweeter slightly?

Could the tweeter amp be the source of the problem.

Have you considered a hybrid active / passive x-over?

Think upstairs loft; maybe a pic would help describe the room and the heft of the speakers. My back ain't what it used to be, in fact, the size of the next project worries me. The room is very live which may be contributing considerably to the hot lower treble ... maybe my PE mic cal is off? One good thing, I'm getting better at identifying frequencies.

Played quite a bit with delay. Tweeter leading mid, mid leading woofers. This last revision is back to no delay.

Pulled most of my hair out trying to duplicate one of those pretty step/impulse graphs like in stereophile. Have a good link on the who/what/and how of getting a pretty step resp0nse?

A lot of the time is working with the mid/tweet, was doing just that again today. It just sounds hot, my best guess would be starting from around 1.8k to 5k. When I pull that area down things start to sound 'distant and uninvolving'? I wear hearing protection but my job is loud, maybe it's my hearing that is becoming the issue.



Here's from today's fiddling. Smoothed 1/24, IIRC the 140hz dip is seen on other speak3rs. With the latest adjustments it's a brand new speaker all over again ... it continues.


[email protected]

[rpb]

My 2 cents. How high is the ceiling in your room? More than 10'? Get the speaker on a 4' to 6' support with the speaker horizontal.

Have you tried turning the woofers off, and making a 2-way? (Temporary till hot treble is resolved.)

Did you delay the tweeter slightly?

Could the tweeter amp be the source of the problem.

Have you considered a hybrid active / passive x-over?

[Pete Schumacher]

Sorry to hear about the seams showing. If you do refinish them, epoxy Formica to the panels. They will cover the seams and sand smooth to accept any kind of paint you could want.

Here's what painted Formica can look like.

[Kornbread]

It's been a journey up to this point.

The fact that some of the seems are showing is starting to wear on me. Even if the paint cost close to $200, it's eating on me nerves and, inevitably, they'll probably get reworked/repainted.

Going active might have been a mistake, no matter how many different combinations of crossover points/slopes/dsp'ing/etc., and their corresponding measurements ... it just never stops. And to beat that, some of the better measurements are the ones my ears tell me NO. Here are some measurements between major revisions over the past year.

All seem they could be better, but guess which one my ears really didn't like ...

A sampling of 5 settings over the year, smoothed 1/24.


Since these are so darn big, they get measured in their listening [email protected] meter; sheetrock walls/ceiling, bare wood floor, reflections come early @3.4ms, gated accordingly. So only accurate to what ... ~400hz? Current measurement is the light pink trace.

[davidsanchez]

Making the seams in MDF invisible takes a ton of work, in my opinion. On the removable back panel of the center speaker I've been working on, I filled any gaps, chips, etc with Bondo, then sanded smooth. sanded with 120, then 150, and finally 220 grit sandpaper. Then, I applied several layers of Killz over all the surfaces, sanding with 220 between each layer. Finally, I used an oil-based satin paint to coat, I did several layers, and used steel wool to smooth between coats. It's a lot of effort, but I ended up with absolutely no visible seams.

[rpb]

I'd try a higher x-over point for the tweeter / mids. Try a LR4 at about 2.3k on the mids, and LR4 at 2.8k on the tweeter.