Seas 2 way crossover

[selucia]

Hello!

I'm building a 9L small 2way using Seas Excel W15cy001 and 27TDFC drivers. I tried a regular second order crossover+notch for the brkup (2700hz crossover frequency) and it works very well in a way, very natural overall sound, but somehow vocals sound strange, maybe I can hear the metallic coloration? of the magnesium cone. So I decided I'd try LR4 electrical with a much lower crossover frequency.

My questions:

1. Would a crossover simillar to the one attached work with these drivers, is it possible to obtain a tonally rich and natural sound using a 4th order electrical filter?
2. Do I need a seperate notch filter for the cone breakup node?
(according to my simulations it's alrdy 60db down with this LR4 filter)

Thanks in advance!

[selucia]

This is the simulated phase response. (The SPL graph is wrong, because the simulated box is different from mine. It should be close to a flat response there..)

[BobEllis]

Why cross so high? The 27TDFC can handle a lot lower than that. Try 1600-1800 for a second order filter and a very deep notch. That woofer can sound beautiful in that sort of system. (I have a different tweeter with a W15CY001)

I think something must be wrong with your simulation since the driver has a 15dB peak less than 2 octaves from the crossover point. The peak should be 15 dB above the crossover's 45 dB attenuation, or roughly 30dB down without a notch.

[Face]

I'd cross a little lower and consider a tank filter too.

[selucia]

That's the overall response. As I said it should be linear in my enclosure. I will post the woofer's response with the 4th order electrical filter.

The idea of using 1800-1900hz with LR4 will not work well you think?
Using the TDFC with a second order filter as low as 1800 seems strange to me, but I can try it. I just don't want to buy too many unneeded crossover components.

[BobEllis]

1800 would work, I just prefer to work a little lower when possible. Unless you are after every last bit of output level, a 2nd order electrical filter at 1600 is very workable. I do it with a T25CF001 (actually Zaph did it for me) and the W15CY001 runs out of excursion long before the tweeter is an issue running it full range.

Notice that the TDFC has a natural 1st order rolloff about 1600 Hz. so you could probably use a 3rd order electrical filter to get a 4th order LR response.

A single inductor and an LC notch on the woofer will get you close to LR4 initital roll off and completely kill the breakup peak. Net roll off is about LR2.

[selucia]

Isnt the TDFC much "weaker" in power handling than the T25?

Would you mind sharing your own crossover?

[BobEllis]

They both list the same .5 mm xmax, and Seas recommended range is 1500 for the 27TDFC against 2000 for the T25CF001. Take a look at Zaph's tests. The 27TDFC is a pretty good performer. I have systems using both and it's only a slight preference for the Excel tweeter. Zaph told me that this version of the T25 has better low end performance than what he originally measured.

Zaph did the crossover design when I bought the drivers from him and asked me not to share it, but I can give you some crumbs. The woofer circuit has a single inductor in series and an LC trap across the woofer. ~LR2 overall but steeper initial roll off. It uses the inductor resistance but nothing extra to achieve a very deep notch at the breakup. The tweeter is pretty conventional topology second order (but not textbook values) with an LPad. Play with the values until you get the response you want.

It will reach into the mid 50s in a ported .375 cubic foot PE enclosure. But not too loud full range in a big room - plenty for bedroom use, or in a HT crossing to a sub at 80 Hz.

[selucia]

Noobie question here, but what is an "LC trap"?

[Thooms]

It's a resonant circuit tuned to have extremely high impedance at a given frequency - a notch filter by another name.

The absence of the resistor makes it a 'bottomless' notch, with theoretically infinite impedance at the tuning frequency - although this is not attainable in practice with real components. Damn you, physics!

Wikipedia

[selucia]

So something like a 0,01mH inductor combined with a 22uF capacitor in parallel with the woofer? How is that better than a 20uF capacitor parallel with the woofer's series inductor? It's a bit confusing, also I'm starting to suspect Speaker Workshop's simulator fails to simulate these filters properly....

[Thooms]

I use Spice for simulating stuff. Not sure what the speaker workshop stuff is, but Spice is one of the standards for analogue simulation.

There's a Mac version (developed by one of my lecturers lol) and a PC version, although the PC versions aren't always free/fully unlimited.

Not sure about the values I'm afraid, there's some formulae on the Wikipedia page which are worth looking at. The frequency is given by:



I don't have experience with passive crossovers, so I can't really help you with the details - sorry

The reason the notch works though is due to the relative phase of the inductor and capacitor - so they have to be dedicated parts I think.

See what happens I guess. Simulate the crossover with and without the notch - if you like I can do that in Spice too and we can compare notes?

[BobEllis]

An LC trap can also be an inductor in series with a cap, as in my case. This results in near zero impedance at the resonant frequency (determined by the equation Thooms provided) When this network is placed in across the woofer terminals, it effectively shorts out the woofer at its resonant frequency.

To use the inductor in parallel with a capacitor trap (high impedance) Thooms described, it would be in series with the driver to be attenuated.

You MUST include an impedance curve in your driver data when using speaker workshop to simulate your crossover response. Could this be the problem?

Notch filters seem to simulate properly for me in SW.

[selucia]

An LC trap can also be an inductor in series with a cap, as in my case. This results in near zero impedance at the resonant frequency (determined by the equation Thooms provided) When this network is placed in across the woofer terminals, it effectively shorts out the woofer at its resonant frequency.

To use the inductor in parallel with a capacitor trap (high impedance) Thooms described, it would be in series with the driver to be attenuated.

You MUST include an impedance curve in your driver data when using speaker workshop to simulate your crossover response. Could this be the problem?

Notch filters seem to simulate properly for me in SW.

I have an impedance curve included. So that shouldn't be the problem.

Thanks a lot for the help!

[BobEllis]

I noticed some odd behavior in SW as I played with the notch filter some more. At some value combinations the predicted response looks something like a phase wrap, peaking and notch of equal magnitude. Not sure I'd trust it.

You might try the free version of BASTA from www.tolvan.com, at least to sanity check your SW results. I haven't used it but it gets good reviews. At 25 euros, it's a bargain should you decide to buy. I've used his "Edge" software and the measured result matches quite closely with what it predicts. HolmImpulse is a good measuring tool when you get that far. Basta has a companion measuring program that I haven't used, but there is a free trial version.

[cjd]

If you have MS Excel I would highly recommend Jeff B's passive crossover designer worksheet. It's free.

SW has always done well for me, including some less than straight-forward filters. Checked against measured results. I have seen the occasional oddity when working with simulated data.

Quality of crossover integration and initial driver choices both contribute to rich sound. Crossing without proper attention to break-up issues destroys that - even in a nice soft "naturally controlled" cone. Break-up and resonance issues always exist, the question is always how you manage them.

C

[selucia]

I will check them out. Thanks!

[selucia]

Question, if I decide to use a capacitor and a resistor parallel with the woofer inductor, (0,20-0,27uF) what value should I use for the resistor, is there a formula for that? Is there even a need to use the resistor?

[cjd]

A zobel? Yes, there's a formula. It's intended to flatten inductive impedance rise...

[selucia]

A zobel? Yes, there's a formula. It's intended to flatten inductive impedance rise...

No, that's not it.
It's the kind of a notch filter, for the magnesium resonances. I edited my question it was easy to misunderstand

[Face]

Experiment with 4-8ohms. I'm not sure if there is a formula.

[cjd]

Ahh.

There is a forumla, but it's definitely easiest to experiment and see what changing each value does.

Without the forumla, the modeling software couldn't work...

Also, I very rarely use the resistor at all.