Help with crossover design, my head's spinning.

[xephon]

Hi, I haven't posted here for a while, but I've started on a project and I'm getting a bit lost...

I've got some HiVi D5S drivers reallly cheap and I want to stick them in a box.
I came up with a box size of 8 litres etc with winISD and that's all good, but it's the crossover I'm having problems with.

I have used SPLtracer to input the impedance and response graphs and then used the Baffle Diffraction worksheet to work out what effect the baffle will have on the drivers.
I then stuck it all into Speaker Workshop and combined the response curves to help design the crossover.

And here I'm stuck...

I have modelled it to use one HiVi D5G and one Vifa D25AG-35-06 per box, and this is what I've got so far.
The main problem is the crazy roll off at the bottom end. Also does the rest of the graph look ok? The crossover point is at about 1500Hz, fudged a bit to make the graph as flat as possible.

Can someone help me to work out what's wrong, and maybe tell me how I should be doing the curves to work properly?



The crossover and resultant curve.



The response curve of the woofer, as provided by the manufacturer.



and the modelled baffle diffraction curves. (Magenta = woofer, cyan = tweeter)


Thanks!

[kbgl]

Change the scale on your first graph to show response down to about 40 dB or so. The appearance of the response will look mutch more like you would expect, even though the actual information will be the same.

What I think I see, is a 6 dB rise from about 100 hz to 1k as if there is no baffle step compensation. Your filter needs to offset a dB or two of this rise. Maybe more.

That woofer looks like a nightmare to filter.

[xephon]

Here's the same plot presented as you suggested.





And here's my latest design...

First is with the baffle taken into account, the second is without. They are the same circuit layout with the values pretty much optimised for each scenario.


With baffle taken into account.




Without baffle taken into account.




How is the best way to perform baffle step compensation? I tried adding the usual LR network, but it didn't seem to do as intended.

[kbgl]

I don't use software, but rely on trial and error. What I would likely try first is a simple coil (like 1.8 mH in your circuit), and a notch at 1.5k. My speadsheet shows a 20 uf cap and a .5 mh coil should be close for the notch values. I would place this across the woofer leads, perhaps with a 2 ohm resistor. (All three parts in series with each other.) If that doesn't work, a circuit like yours with my values except use a bigger resistor like you did already.

I assume that you are looking at the woofer response alone, before adding the tweeter response. I try to get the woofer where I want it, then figure out how to mate the tweeter.

A before, and after graph helps me see what's going on. I measure the woofer with no filter, then add something like a 1 or 2 mH coil and see where the change in response starts.

Are these speakers going to be near the wall?

[xephon]

How about another one.... I put a notch filter in along with a bit of baffle step compensation.
This one isn't very efficient, but it seems a bit flatter. This is with the baffle taken into account.

[sreten]

Originally Posted by kbgl

That woofer looks like a nightmare to filter.

Hi,

I quite agree, talk about making things difficult for yourself.
The D5.8 (not D5.S) I suspect is best used in pairs as the
bass units in a 3 way, or singly paired with a low efficiency
2" widerange driver c/o lower with a broad notch filter.

e.g. see the pluto driver at Linkwitz Lab - Loudspeaker Design

rgds, sreten.

Paul Carmody's DIY Audio Projects - undefinition
Zaph|Audio
FRD Consortium tools guide
RJB Audio Projects
Speaker Design Works
HTGuide Forum - A Guide to HTguide.com Completed Speaker Designs.
Humble Homemade Hifi
Click below to go to
Quarter Wavelength Loudspeaker Design
The Frugal-Horns Site -- High Performance, Low Cost DIY Horn Designs
Linkwitz Lab - Loudspeaker Design
Music and Design

[kbgl]

It might actually be cheaper to add a mid than to have such an elaborate x-over.

How are you going to get a big x-over like that into such a small box?

[sreten]

Originally Posted by kbgl

It might actually be cheaper to add a mid than to have such an elaborate x-over.

Hi, True, but at such low efficiency a mid/treble unit makes more sense to me, rgds, sreten.

[xephon]

So I could plonk a mid in to do from 500Hz to 2000Hz or there abouts?
I don't really want to build a 3-way system (I'm already doing that with a set of active amped Dynaydio's in my car...)
I'm sure that Pluto system sounds good, but I'm not too keen on my sound system looking like it came out of Super Mario Bros.....


The crossovers will be in separate boxes, they are to be used at my computer desk, so they can sit under the desk.

If I were to use a 2" widerange, how would that go for the high frequencies? Any suggestions on where to start?



Maybe I should ebay these and get some 6.5" Vifa woofers...

[xephon]

Hmm, I have come to a conclusion and a realisation.

I have concluded that this driver is too hard and the crossover is too complex and as such has been relegated to ebay.
I have also realised that in one of my cars that I am dismantling to restore another there is a fully functional set of Dynaudio Esotec System 342 3-ways currently not being used, and that no matter what I do with the HiVi's, the Dyn's will always blow them out the water.
I totally forgot about them.....

I'll stick them in a 16 litre box tuned to 48 Hz with the stock crossover and enjoy.
I have both 2-way and 3-way Dynaudio passive crossovers as well as a TruTech 3-way active crossover which I custom modded with Burr Brown preamps and Black Gate caps sitting in my car that I can play with.