My first crossover - 3-way center - comments?

[ct991]

Well I think I'm getting somewhere with this crossover design stuff. I started with X-over Pro and moved on up to LspCAD to simulate the crossover. Below is the schematic and response:





I started with a 2nd order APC design and customized it somewhat to get a steeper slope on the low-pass section of the bandpass network. I put Zobels on everything and L-pads on the tweeter and woofers, as you can see in the image.

A note about the tweeter output though...I'm not planning to use the RS28A in my final design. I plan to use the Vifa XT25 instead, which will eliminate the big dip in response above 10K.

One question I still have has to do with the component values for the Zobel filters. X-over Pro says one thing, but if I calculate them manually I get very different values. I'm not sure who to believe. What I have here in the schematic is the manually calculated values.

Does anyone see anything wrong with it?

Feel free to comment about it...

[Paul H]

The first thing I noticed was the RS180 picking up the bass duties - I would have mated the RS125 to an 8" or even 10".

Perhaps you could provide a little more background - what kind of system this goes in, what size and shape speaker you're looking at, etc.

[cjd]

Are these live measurements, or simulated? Or just straight PE data?

The midrange breakup looks to be hitting at only -10dB, which will very probably show through.

RS28A dip is a non-issue, and most of us know about it. And would take it over the XT25 as well.

I've experimented (all simulated) with this driver combination - it's promising. One thing I would suggest is possibly treating the RS125 as an upper midrange supplement and giving the RS180 a more gentle slope and more coverage into the midrange. This allows you to work a little more gently with the RS125 and reduce some of the peaky behavior it seems to exhibit when given such a narrow frequency band to cover.

C

[ct991]

Are these live measurements, or simulated? Or just straight PE data?

what kind of system this goes in, what size and shape speaker you're looking at, etc.

I'm using the posted PE data for the drivers. I thought about using the RS225 for bass duties but for what I was looking for, I thought the RS180 would get down low enough for me. I already have a large (12" Dayton RS) sub, so I figured I didn't have to go for 8" woofers. Also, the 8" represents a jump in price, especially considered I'm buying two of them. My proposed configuration is WMTMW, which makes for a wide cabinet I guess. That's fine with me though, as cabinet size/shape is not really important to me.

The midrange breakup looks to be hitting at only -10dB, which will very probably show through.

For the RS125, what I was going for was to get it attenuated enough past 6K, where (from looking at Zaph's response graph) the breakup really starts to get nasty. As it stands now, what exactly would show through? What would I hear?

RS28A dip is a non-issue, and most of us know about it. And would take it over the XT25 as well.

I've always been told that metal domes are too harsh (did someone say doorbell?) and that's why I'm trying to avoid them.

[cjd]

Using PE's data straight up is a recipe for trouble and disappointment in the end. It needs to be massaged - minimum phase at the very least (and thus z-axis accounted for in driver setup as well). I'm also very slightly uneasy about their RS180 measurements - close, but just far enough off. . .

WTMW would work, WMTMW with your crossover will leave you with a narrow sweet spot and some interesting off-axis comb induced response dips.

Cone breakup is just very harsh and very very forward - often attributed to a tweeter actually. It can seem very listenable if you're not used to a system without the breakup issues but will ultimately reduce the listenability - five minutes can work, yet a half hour won't.

You're worrying about a very sweet metal dome. You'll find more uncomfortable distortion crossing the XT25 that low without experience or luck and the metal cone mids too high without sufficient breakup supression (which is what you have right now in that crossover).

If you want to go silk dome 'cause you're just too worried about metal still, go Seas 27TDFC. Some metal domes can indeed trip people up - for instance, while sounding excellent, the Seas 27TBFC (H1212) hits me at a bad spot with its >20k breakup spike. The RS28A is superb and gives me no trouble at all. I also have the Seas 27TDFC at home, and two years ago spent my time at a chicago DIY gathering simply analysing tweeters hoping to come to a better understanding directly vs. depending on common myth (however based in reality they were).

In my (so far fairly limited) experience more harshness issues in commercial speakers are the fault of the midrange and yet are blamed on the tweeter. Shallow slopes, crossing too high, lots of things. But, that's purely opinion and I'm a bit shaky on it still. Need to experiment with "sweet" drivers and such some, get more listening in, etc.

C

[Landroval]

It needs to be massaged - minimum phase at the very least (and thus z-axis accounted for in driver setup as well).

Where could one find more info about phase behavior of a speaker and what to do about it? How is phase taken into account in these many HTguide Dayton RS designs (except correct polarity for each driver)?

[cjd]

Most of the RS projects are built off live measurements with a fixed mic - thus phase relationship is built into the measurements (as well as baffle step and diffraction). Makes it easy, really.

When using published data you have to understand measurement conditions (IEC baffle, infinite baffle, or otherwise), and you need to extract minimum phase (Hilbert transform) and add z-axis offset to the driver in your simulation program.

If I'm working with simulated response, I use the tools available at the FRD Consortium site - Baffle Diffraction Simulator (baffle step and diffraction) and the Frequency Response Combiner to normalize curves, sum PE's data with BDS data, and extract minimum phase.

Roman B has a fairly good writeup on working this way, though I'm not sure the link right now.

C

[ct991]

WMTMW with your crossover will leave you with a narrow sweet spot and some interesting off-axis comb induced response dips

How can I improve the crossover to help with these issues?

You'll find more uncomfortable distortion crossing the XT25 that low without experience or luck and the metal cone mids too high without sufficient breakup supression (which is what you have right now in that crossover)

So you're saying I don't have enough breakup supression? Should I add another capacitor to the low-pass section of the bandpass network to get a steeper slope? ...or must I cross lower?

[Landroval]

Roman B has a fairly good writeup on working this way, though I'm not sure the link right now.

Probably this one:

http://www.rjbaudio.com/Audiofiles/FRDtools.html

[Paul H]

How can I improve the crossover to help with these issues?

Consider changing from a WMTMW to a WMTW - the comb filtering in a WMTMW will be created by the two horizontally separated mids which are playing higher frequencies, as they must to cross to the tweeter. A single mid close to and above or below the tweeter does not create the same problems.

The woofers do not have comb filter problems to the same extent because they're playing lower frequencies.

So you're saying I don't have enough breakup supression? Should I add another capacitor to the low-pass section of the bandpass network to get a steeper slope? ...or must I cross lower?

I don't think that is enough breakup suppression
Experiment using LSPCad with another capacitor and/or crossing lower. A third option is a notch filter at/around the breakup frequency. There is no single correct answer.

[ct991]

I've modified the bandpass section to see if I can get a steeper slope. I'm concerned about the midrange response since there seems to be some phase cancellation. The total response seems fine however. Is this acceptable to have the midrange response above the total response? See below:

[cjd]

Negative summing is a problem. Bad juju. Especially off-axis.

The peakiness on the bottom of the RS125 won't be pleasant either I suspect.

C

[ct991]

How's this?

[Landroval]

It doesn't seem to have any baffle-step compensation, or did you include the box and baffle already to the LspCAD-model?

[ct991]

I haven't done any baffle-step compensation yet. That will be next on the list. Other than that though, am I on the mark?

[cjd]

It's getting better, yup. Definitely.

Get some baffle step and diffraction rolled into your data (oh yeah - and minimum phase ) and see where that leaves you.

C

[ct991]

I expanded the SPL graph range and found a large spike in the midrange response caused by the cone breakup. Is this something I need to worry about or is it down far enough not to be noticed?

[cjd]

Worry about it.

C

[ct991]

Well I managed to tame the breakup of the RS180's and the RS125's so I'm happy with my design. Any breakup is now at -50dB or below, which seems to be the generally accepted level.

Thanks for all your help guys. Now I guess I have to actually put it together

[cjd]

IF you have NOT done a minimum phase model, DO NOT BUILD.

You will have very different response.

Been there, done that.

Well, ok, it's fine to build, just don't expect the sound to be what you were hoping for.

On the other hand, if you did minimum phase and all the rest, have at it!

C

[ct991]

Are there any good tutorials/resources on how to do a minimum phase model? I've looked around in this forum and there's a lot of talk about it but nobody really explains what it is or how to deal with it.

[cjd]

Landroval linked to the best info I know of in post 9 on this thread.

I should also note, I'm vaguely not quite sure about PE's published RS180 data. Though, it's hard to tell if it's simply 3-ways that are very difficult to get accurate in a minimum phase model in the manner you're trying, or if it's the data. I can say that the models I did that way in preparation for the 3-way towers were only a very vague help when it came to comparing with real measurements. It would have sounded pretty bad, though.

C

[Paul H]

Are there any good tutorials/resources on how to do a minimum phase model? I've looked around in this forum and there's a lot of talk about it but nobody really explains what it is or how to deal with it.

I would suggest if you're happy with the general crossover design and speaker layout, buy the drivers, build a cheap particleboard box/baffle, and measure your speakers using LSPCad.

Then re-work your crossover design using actual measurements, and then buy and build your crossovers.