First attempt at xo

[Silver1omo]

With a couple of builds under my belt, I decided to give a try to design a XO.
I liked the sound of the Piccolata I have in the bedroom, so I decided to re-build the boxes with removable baffles to try and toy with other tweeters.





Image not available

Yeah, unfinished baffles...

The first attempt is with the Vifa XT25SC90.

Images not available

The zip has the frd and zma measured in box. The offset seems to be about 1.1"
Files.zip

What do you think?
Getting the XO parts will take time (and money..) as shipping to Mexico takes some time.

[JonMarsh]

I can probably take a look at this on the weekend- too long a work day during the week.

[Silver1omo]

I can probably take a look at this on the weekend- too long a work day during the week.

Thanks!

[5th element]

So how have you taken those measurements? Usually you take a measurement of both drivers combined. This is used as a reference within simulation software. You load in the response of the two drivers together and compare it to the simulation softwares prediction as to what the combination of the individual tweeter and woofer responses should be. You then enter in the Y and X offsets and tune the Z until the prediction matches the actual measured combination. This is the most critical bit to get right when it comes to crossover design as it forms the backbone from which a successful crossover can be designed. It is absolutely necessary for simulating correct phase integration.

[Silver1omo]

Measured on tweeter axis at ~50cm. Used REW and a UMIK-1 as input. Audioquest dragonfly 1.0 feeding a crown XLS1000
I did measure the drivers individually and in parallel. The offset was calculated by matching the FR:

Image not available

That is a close as I could get them (1.1"). The sim and offset "calculation" were done using XSim recently published by Bill Waslo.

[5th element]

If you could post the combination FRD along with the driver separation in the Y axis I'll give a simulation a go too if you want

[Silver1omo]

If you could post the combination FRD along with the driver separation in the Y axis I'll give a simulation a go too if you want

Thanks a lot.
sis.txt
Combined FRD with the extension changed. For some reason I do not remember the C to C distance... I'll need to measure it when I get home (ballpark is 3.875" to 4").

[5th element]

Oh good CAD simulation like this is one of my favourite things to do and hardly anyone ever provides you with what's necessary for getting the job done, ie properly done measurements and their associated files.

My main question straight away would be, how did you measure the woofer? And this is mainly aimed at how you decided to incorporate baffle step into the equation. If I had to guess it would look like you made a far-field measurement without any near-field splicing. I am just wondering whether or not that hump at 1.5kHz is entirely related to diffraction losses below it, and subsequently should be treated as where the 2pi nominal sensitivity should lie, or whether or not the nominal 4pi sensitivity is actually relative to the SPL at around 250Hz and therefore the hump needs to be notched out.

TBs own measurements seem to imply that there is a small rise around 1k, so I'll go with the latter.

[Silver1omo]

For that measurement set, I did not take a near field. Only "far" at ~50cm.
In a previous set I tried to sum the near and port and then splice the far. The sim result did not match the measurement, so I tried to keep it simple in this go...

[5th element]

How low can you allow the system impedance to go? This is a trade off between frequency response flatness and minimum impedance.

[Silver1omo]

Currently I'm using a LM3886 chipamp (from chipamp.com) in the bedroom, so I guess not a lot below 4ohms.
I have verified the baffle size (6.5"x10.5") and location of the drivers:
Tweet is 2.25" from the top, centered horizontally.
Woof is 6.125" from the top, centered horizontally.

So C to C is 3.875"

Thanks again.

[JonMarsh]

For that measurement set, I did not take a near field. Only "far" at ~50cm.
In a previous set I tried to sum the near and port and then splice the far. The sim result did not match the measurement, so I tried to keep it simple in this go...

Good way to go....

Oh good CAD simulation like this is one of my favourite things to do and hardly anyone ever provides you with what's necessary for getting the job done, ie properly done measurements and their associated files.

There must be a long delay echo in this place, I could swear I've said the exact same thing several times... should be a good prospect for a NatalieP style quasi LR3, though I'll look at whether series or parallel makes the most sense component count wise. With the NatalieP, I just couldn't resist putting a brain teaser network configuration together... :W

This is not a contest, just an opportunity to explore alternative techniques. Mine will probably have too many parts and be expensive to build... but who knows?

[5th element]

Well it's also an opportunity for learning too. I find that with every passive xover I design I learn something new myself, so if it's educational and/or helpful for someone else too then it's win win

Okay so here are my attempts. The first two are with a minimum impedance set to 4 ohms. I set the DC resistance of the series inductor in the mid bass network to 0.5 ohms.






I raised the xover frequency up a little as this helped to limit the effect of the dip in the tweeters reponse around 3.6kHz,

Now the version with the minimum impedance set to 3.5 ohms. As you can see the frequency response flatness increases.

Both have around 5dB of BSC.





You could get these flatter with more components, but I didn't want to make them too complicated/expensive.

[JonMarsh]

You could get these flatter with more components, but I didn't want to make them too complicated/expensive.

That's my job... :W

[Silver1omo]

So that smiley is the BBC dip, right?
Not a lot more parts than what I had, but better looking
I will try to get another set of measurements including a near field of the woofer to see how a splice/blend looks. Any pointers in the near field?

[5th element]

It's sort of what could be called the BBC dip, I didn't design this in on purpose, it's just the way the drivers matched up. The lower the system impedance goes, the flatter things become, down to around 2.5 ohms. I could add in a couple more notches I think to get things even smoother, but they take up a lot of cabinet space and cost quite a bit too. BBC dip type things though aren't really a massive issue imo as they just keep things sounding nice and relaxed. I mean you can over do it, but providing it's only a couple of dB, you're not going to lose anything, but have speakers that are tolerable to listen to with older/thin sounding and poorly mixed modern stuff.

measurement guide.pdf

Might help with the measurements.

[JonMarsh]

Nice guide for those new to this, Matt! :T

[Silver1omo]

Thanks for the guide. For the sets I have, I loosely followed Jeff Bagby's guide.
The problem I had when I spliced the NF and FF FRDs of the woofer was that using the blended woofer FRD I was unable to match the w+t FRD.
Will take a good look at the guide you posted.

[5th element]

That may have been because of a phase or level issue. If it's a phase issue then my guide should help address that, as it's a pain when splicing to get everything right. My guide assumes you are using ARTA, but every measurement program follows the same basic principles.

Now when you're making your far field measurements, typically the tweeter, combination and woofer ones are going to remain un-spliced. Usually I use the far-field only woofer measurement to get the match with the combi right, then I load in the spliced version for proper simulation. Of course this only works if the reference for the high frequency part of the woofer splice is identical to the far-field only measurement - which it should be, both in phase and SPL, otherwise you're screwed!

[blue934]

What software are you using for xover sim, matt?

David

[5th element]

LspCAD 5.25 pro. I've been flirting on and off with the idea of upgrading to version 6 for a few years now, but have always found other things that I'd rather spend the money on