My M8ta spin

[augerpro]

Which LC pair is involved with the pole at 1k? Couldn't you lower the C value and raise the L for a lower Q there?

[KJP]

Keep playing with LCR trap circuit values/add one- you should be able to get rid of that peak and also maintain good tracking- and as Jon mentioned, maybe it is a baffle artifact that will disappear 15 degrees off axis, so therefore- not to worry.

Jed

I am getting a little closer.... Could you clarify /add one?

Here is a question. What do people consider to be good tracking?

Which LC pair is involved with the pole at 1k? Couldn't you lower the C value and raise the L for a lower Q there?

Speakerworkshop just doesn't spell it out like that for me so I don't really know... Do other measurement softwares give this information?

[augerpro]

Speakerworkshop just doesn't spell it out like that for me so I don't really know... Do other measurement softwares give this information?

Heh, nothing really gives out this info. I'm still looking for a book or something on it. Your network is sort of complex, but if it were say a 2nd order network you would raise teh L and lower the C such that the product L*C is still teh same. So the XO is still the same but the slope (Q) is shallower. That's pretty much the sum of my knowledge...

[Jed]

I am getting a little closer.... Could you clarify /add one?

Here is a question. What do people consider to be good tracking?



Speakerworkshop just doesn't spell it out like that for me so I don't really know... Do other measurement softwares give this information?

When I said add a LCR circuit, you might have to add more components to lower the Q at 1K.


You could let speakerworkshop do a calculation based on a goal response that you create for the target slope, and then have the selected values adjusted automatically. To do this create a new network and simulate just the woofer section on its own with the target slope. You will have to "create goal" to do this and adjust the slope based on your target. I use that feature a lot. Or at least I used to when I was using SW.


Here is the menu with the options you'll need to let SW do the work for you.

Optimize network (based on selected components in your network)
and
Create Goal (select a target slope at a given SPL etc.)

Image not available

[KJP]

You could let speakerworkshop do a calculation based on a goal response that you create for the target slope, and then have the selected values adjusted automatically. To do this create a new network and simulate just the woofer section on its own with the target slope. You will have to "create goal" to do this and adjust the slope based on your target. I use that feature a lot. Or at least I used to when I was using SW.

That is exactly what I did. I began by adjusting the impedance compensation until the predicted impedance was flattest. I have an LR8 goal that I let the optimizer figure all components except Cz1 and Rz1. It works very well on the amplitude side but apparently not so hot on this Transfer function issue.

[KJP]

As an added note, letting the optimizer run on all components results in an even larger peak.

[KJP]

Looking better, only a .04db peak remains





Good enough to build?

[JonMarsh]

Much better, IMO. Sometimes the little details like this make quite a difference in the long term enjoyment of the final product.

[TacoD]

As an added note, letting the optimizer run on all components results in an even larger peak.

Optimizers are in my experience not that "automatic" as you think, you must already start with a good network and details are hard to address.

[JonMarsh]

Optimizers are in my experience not that "automatic" as you think, you must already start with a good network and details are hard to address.

That's a big and important point- the optimizer can be helpful in determining if your network is in the ball park, (if inductors get really big in shunt values, you probably don't need it, etc), but using an optimizer well requires some art and insight into what the topology should be doing and how- and I almost NEVER optimize all components- for starters, impedance zobels largely should be setup and left alone, for example, and there are times you want to carefully select what range you use the optimizer over, as well as manually tuning some issues, as I point out in this thread.

Optimizer's may get you a network that looks OK in on axis sum, but is not really tracking in phase well, and has very poor inverted null, for example.

It's a tool, sometimes useful, but never to be used without critical judgement- most of my fine tuning is done completely manually, but then I am a bit compulsive. :W

[KJP]

I really appreciate the input and guidance. Thanks tons.

I plan to order parts in the next couple of days. On that note I do have some component questions. What kind of power rating should I be looking for on the l-pad resistors? The other resistors?

I had planned on using a stack of the GE 10uf caps from Madisound but they seem to be out of that value. They have the 5uf but that would become a LOT of capacitors/big pain to assemble. I also was wanting to get some 10 or 12 AWG inductors from Northcreek for the woofer circuit but they seem to be unobtainium.

So the can of worms question would be....

What do you think the best components to use for this network are?

[JonMarsh]

Don't take my word as any kind of gospel- I'm probably known for being willing to be a bit more spendy on the crossover components than many folks.

For the Lpad, I usually use parallel/series resistor arrays as needed so I can get 25-30 watts power handling without breaking a sweat- resistors have temperature coefficients, too, and though the average power to the tweeter is not generally that high, it's best to keep any possible temp coefficient of the resistor out of the picture.

For the larger series coils, I'd at least use AWG 14 Perfect lay or Solen coils, from Madisound or PE. I'm personally a fan of air core coils because they dont' have any hystersis or saturation effects. Cored coils have to be calculated and sized carefully if you want trouble free results, and no one provides inductance versus current and permeability droop for cored audio crossover coils (something that is very fundamental for SMPS inductors, for example).

If you have plenty of room, paralleling smaller loser cost polypropylene caps is resasonable, but there are also times it's a lot easier to use Solens or at least Dayton polypropylene caps in larger values. I like to make the tweeter series caps at least partly with either film and foil or Sonnicaps, it depends on your associated gear whether that expense is worthwhile. And your budget, of course.

As Steen Duelund said when he was alive, for the best results it's not at all unusual to spend more on the crossover than on the drivers.

[cjd]

How much deviation from my lp target is ok to avoid this peak in transfer function? If it comes to choosing evils, which would be more important, target tracking or filter Q?

This is a fun question.

Am I the only one that throws out the targets once I've got something in the ballpark?

C

[JonMarsh]

This is a fun question.

Am I the only one that throws out the targets once I've got something in the ballpark?

C

Nope. You're hardly alone! For me the individiaul target is a starting approximation considering the non-ideal driver spacing and alignment that usually has to be coped with. If there's any one thing that is critical, it's paying attention to the phase tracking in the crossover region and having a good null response when wired out of phase. One can usualy tell this by how things sum on either side of the nominal crossover point. That and avoiding too high a Q at the corner just because an optimizer is trying to get back a dB or so... ;^)