First Crossover - reviews/recommendations?

[smitty611]

Ok, this is a follow up to my first post re: building a 3-way WMTMW after Jed's dynamic 2cc design. I planned to experiment using drivers I already have available which are: Dynaudio Esotec 17WLQ, Morel MW114S and a Vifa D26TG-35-06. So the big unknown for me was the crossover. After quite a bit of reading (not necessarily understanding) and tinkering with various programs and spreadsheets, I've come up with the crossover layout here. If any of you have time to review this layout and offer any feedback that comes to mind I would greatly appreciate it. I'm not shooting for perfection here, just something that will get me in the ballpark until its built and I can perform a listening test. After that I'll make adjustments as needed.

I pulled response and impedance data from the T/S database and ran them through Unibox in a sealed configuration for the W and M. Then took those data files into Jeff's PCD spreadsheet. From there I more or less played with various values to get a flat (-ish) response and a reasonable impedance. I had a hard time with the woofer section but finally got the response graph looking decent...I'm just not sure about the impedance graph. Does that look ok? Also had some trouble keeping the Morel's at or above 4ohm...they kept dropping to 3 and 2 with various layouts.

Thanks in advance!

Ryan

[Jed]

Did you process phase data and BSC into the files? If minimum phase how are you handling the z offset of the drivers? Did you factor all that in to the crossover?

[smitty611]

After Unibox I tried to take it into the BDS spreadsheet but didn't make it very far. I got the box dimensions roughly modeled and the speaker placements. Couldn't seem to get it to work for me beyond that though. That was just a couple days ago and haven't gotten back to it again yet. I have to chew through this stuff in small-ish chunks at a time! lol

How do you mean "process phase data" into the files?

I did not account for any Z-offset. How much would this be and for which drivers? I just played with the values in PCD using very low increments and see how much that can affect the response! Guess I need to do a little more reading on that aspect.

[Jed]

check out rjbaudio.com

Roman does a pretty good job explaining how to design a speaker based on sims.

Also, http://www.claudionegro.com/

Jed

[cjd]

3-ways from sims are, in my experiene, less likely to work out well. I've tried, many times - what I've measured vs modeled has never worked out perfectly, but I get darned close on 2-ways - not so much on 3-ways. At this point I know why so I'll probably do better next time but I've only learned that working back and forth. And until I prove it out I'm not going to go telling people they should do this or that only to discover I was wrong.

C

[smitty611]

Yeah, I can imagine a full sim won't yield the best results. I'm just starting from here because I don't have a measurement setup...and I'm playing the role of the stubborn noob who won't start with a proven design because he wants to use his own drivers. If you can forgive all that...

I must have a leak in my process (following Roman's instruction). I pulled the driver info from the T/S database and imported the UniBox formatted info into UniBox. I adjusted the Nom. Pwr to 1W, and exported sets of FRD and ZDA files; one pair at 1000L for mfg and another set at 22L for the woofer box and a set at 2L for the mid box. I then took those files into the BSD spreadsheet, modeled my box, added driver locations and saved the resulting responses for the woofer and mid. In FRC I hit a snag. I believe I have all the files except the first Driver Base file because I was unable to get SPL Trace to work in XP Pro. How can I get this data?

Would I be correct in assuming this is a very bad phase graph? Most other phase graphs I've seen only have the one main shift across the freq. range. I more than likely incorrectly extracted the minimum phase data Jed recommended.

Thanks for the help in advance

...Also, I did calculate the driver offset values from Roman's calculator and input those into PCD. The main thing it changed was the phasing of the tweeter, which I switched.

[cjd]

Your mid and woofer data looks far too smooth... rather idealized.

The phase doesn't look specifically bad, though it too may point to a fault in the process. I'd guess something in FRC isn't going as it should, if not in Unibox.

[smitty611]

Ok, there must be a comet passing over my house or something...I was just able to get SPL Trace installed properly on my XP machine. I have tried it before with no success. Not like it's a complicated process. In any event, I'm tracing the SPL graphs now. Hope to start the entire sim process again tomorrow and see how the outcome differs.

[Jed]

You might email Roman and see if he can design a crossover for you with simulations. Did you add in the diffraction of the box into the model?

Jed

[mackintire]

Its not easy the first time

I'm just starting from here because I don't have a measurement setup...and I'm playing the role of the stubborn noob who won't start with a proven design because he wants to use his own drivers

I remember saying that!

Yeah I know it wasn 't that easy to do the simulation part correctly the first time. Took me alot of hours to get it right...I hope

Anyhow, I 'm just awaiting my cabinet guy to get off his duff so I can compare real measurements against the simulations I created. Hopefully the next time Jed or one of you other professional types get down my way in Pittsburgh you can tell me all I did wrong.

My advice is, IF you really want to learn, ask questions, read and do. Try and Try again. I think Van Dickenson said "Designing crossovers is one part art and one part science."

[smitty611]

Ok, updated freq graph and impedance. Still sim data. But better data, hopefully, now that I've managed to get more accurate info to feed through Roman's tutorial.

I managed to get the freq response at what I think would be a good level...however, that impedance dip is a concern. I will be using an adcom amp which runs a 4ohm load comfortably but I'm guessing it won't like that dip. Or am I overthinking it given that it's a relatively narrow freq range?

The main reason for the dip is the series RLC wired in parallel I have on the mids. They have a nasty +/- 3db spike in the middle range around 1k-1,500 so I tried to level that out as best I could. Unfortunately that creates the dip.

Thoughts on the dip and/or how to correct it without losing the flat freq response?

[Paul W]

Try a parallel RLC in series with the mids. Also, you probably don't need all the resistance in series with each leg of the crossover...maybe one or two, but not all three. BTW, you really should be working with real measurements

[smitty611]

BAH....who uses REAL measurements these days? LOL

The parallel mids just seem to be resisting me (no pun intended). I can get the W and T to line up pretty well at any dB level, but the mids... I'll try your suggestions in PCD and see what I can come up with.

[JonMarsh]

There's something not right with a dip at 1 kHz to 2 ohms when you've got a 2 ohm resistor in series with the input to the midrange crossover- without access to your data files I can't really comment, of course. Sometimes so called 8 ohm drivers are close to 6 or 4 ohms. If you're using LspCAD, I'd usually optimize with impedance minimum set, but then that's only available on the pro version- you have to manually control that with the standard version. Probably your crossover components are skewed towards too low an impedance.

For a midrange bandpass, 2nd order, I'd suggest a network topology looking more like this: