Crossover Design Help

Thank you. I think taking accurate measurements is something I could get into in the future but this is my first time building a crossover and my own speakers. If I want to get closer to the target curves, how should I tweak the design? Just play with the values?
You can simulate what your in box response will be like. This takes those large (IEC) baffle manufacturer measurements and applies:
a) baffle step losses
B) baffle diffraction (mostly on a tweeter but some impact on a midrange depending on xo point between them)

There are tools to do it and a method but too long to write here.

Another option would be to ask a member to simulate a crossover for you - if you can provide a picture of the intended baffle (front of the overall speaker) with dimensions, driver placement and whether you are going to square edge or have some sort of chamfer / roundover.

The member could give you the crossover they've designed and you can plug that into your model to see where your source files are wrong.
 
how should I tweak the design? Just play with the values?
Within reason, yes.

From wolf_teeth's suggestions:

Decreasing the value of L1 will sharpen the "knee" of the tweeter roll-off and increase the cross point.

Decreasing L3 will sharpen the low frequency roll-off of the midrange and increase the crossover frequency. But with a large change should be done in concert with C2 adjustment to maintain a reasonable roll-off shape. The CircuitBlocks in XSim are helpful for this also. You can adjust the Q and frequency of the filter block to get reasonable starting points. Q's from 0.5-1 are typical, with higher values producing sharper roll-off.

The blocks don't have to be in-circuit; you can just drop one at the top of the schematic and leave it disconnected to use for reference. In the bandpass section there are some other adjustments that happen due to interactions, but the CircuitBlock is still a good place to start to make crazy values less likely.

Decreasing L4 will sharpen the woofer roll-off and move the cross point up in frequency. Increasing C4 will sharpen it further and decrease the cross point.

Another thing that's commonly done is to try to use off-the-shelf values in the crossover where possible. Particularly in the inductors. There aren't as many options for them, and there have been shortages here and there due to the global supply chain issues.
 
Within reason, yes.

From wolf_teeth's suggestions:

Decreasing the value of L1 will sharpen the "knee" of the tweeter roll-off and increase the cross point.

Decreasing L3 will sharpen the low frequency roll-off of the midrange and increase the crossover frequency. But with a large change should be done in concert with C2 adjustment to maintain a reasonable roll-off shape. The CircuitBlocks in XSim are helpful for this also. You can adjust the Q and frequency of the filter block to get reasonable starting points. Q's from 0.5-1 are typical, with higher values producing sharper roll-off.

The blocks don't have to be in-circuit; you can just drop one at the top of the schematic and leave it disconnected to use for reference. In the bandpass section there are some other adjustments that happen due to interactions, but the CircuitBlock is still a good place to start to make crazy values less likely.

Decreasing L4 will sharpen the woofer roll-off and move the cross point up in frequency. Increasing C4 will sharpen it further and decrease the cross point.

Another thing that's commonly done is to try to use off-the-shelf values in the crossover where possible. Particularly in the inductors. There aren't as many options for them, and there have been shortages here and there due to the global supply chain issues.
Ok thanks i will try to incorporate this into the design and follow up. Would it be better to just buy this? https://www.parts-express.com/Crossover-3-Way-8-Ohm-800-4-500-Hz-200W-260-230
 
Let's be clear here, changing the Q of the filter does not change the slope. It changes the damping at the point the xover becomes effective, also known as the knee of the response. The inherent slope of 3rd, 2nd, etc, does not change with the Q adjustment.

The values I suggested took Capacitors into account as typical values for such a design as intended, and would be closer than a guess or textbook equation suggests.
 
Let's be clear here, changing the Q of the filter does not change the slope. It changes the damping at the point the xover becomes effective, also known as the knee of the response. The inherent slope of 3rd, 2nd, etc, does not change with the Q adjustment.

The values I suggested took Capacitors into account as typical values for such a design as intended, and would be closer than a guess or textbook equation suggests.
1661655775950.png
How does this look?
 
Your woofer impedance/FR does not reflect a vented box.
The tweeter still extends too low, and C3 needs increased for a better blend if not making tweeter 3rd order.
Files are overly smoothed. Not possible to see if aberrations are now benign.
Baffle step is not accounted for.
You have not accounted for physical offset or delay.

My opinion is you have work to do to refine the simulation.
 
  • Like
Reactions: ernperkins
Might I suggest a little more effort to meet the OP where they are? Every user isn't trying to build a perfectly designed, infinitely modeled and understood technical marvel. Some just want to build some speakers that are better than their previous attempt.

Manufacturers build, and people buy, completely un-optimized crossovers every day. Do they all need the Bowerick Wowbagger treatment?

Let's be clear here, changing the Q of the filter does not change the slope.
Indeed, let's be clear. In this case, was that unsplit hair holding back anything? Has the mood or understanding increased significantly since it was split? Did anyone but you mention slope?

1 kHz, second order CircuitBlocks with Q's from 0.25 to 1.5:

1661700965279.png


Focusing on the terminal slope that may be so far below the reference level as to have no appreciable affect on the system output seems rather unhelpful when trying to describe to an inexperienced user how value changes impact the individual driver response curves around the cross point.

If we make things seem arcane, fewer people will be interested in this hobby.

Carry on.
 
  • Like
Reactions: Toaster79
You can adjust the Q and frequency of the filter block to get reasonable starting points. Q's from 0.5-1 are typical, with higher values producing sharper roll-off.
Roll-off = slope, they are the same thing. The knee is what is affected, not the slope/rolloff.

I don't know why you felt the need to respond the way you did to me, as no ill will or rudeness on my part was intended. No slight was implied on your behalf. I merely stated clarity for clarity's sake.

There is too much misinformation in this hobby that needs to be corrected for accuracy to remain paramount. Informing new builders accurately without using the wrong or slightly misdirected terms is difficult enough, but required.

You also have lack of my facial expressions, body language, or tone of voice to truly see or hear how my reply inferred. There is no reason to get angry over this.
 
  • Like
Reactions: morbo
Might I suggest a little more effort to meet the OP where they are? Every user isn't trying to build a perfectly designed, infinitely modeled and understood technical marvel. Some just want to build some speakers that are better than their previous attempt.

Manufacturers build, and people buy, completely un-optimized crossovers every day. Do they all need the Bowerick Wowbagger treatment?


Indeed, let's be clear. In this case, was that unsplit hair holding back anything? Has the mood or understanding increased significantly since it was split? Did anyone but you mention slope?

1 kHz, second order CircuitBlocks with Q's from 0.25 to 1.5:

View attachment 1085430

Focusing on the terminal slope that may be so far below the reference level as to have no appreciable affect on the system output seems rather unhelpful when trying to describe to an inexperienced user how value changes impact the individual driver response curves around the cross point.

If we make things seem arcane, fewer people will be interested in this hobby.

Carry on.
Thanks. I felt a bit overwhelmed. I think you spend all day talking about how to perfect every piece and whatnot but I just want something that sounds decent and that I built myself. If that's what this design will get me I am perfectly fine with that. Thanks for all your help.
 
Here I've reduced the vertical range so it's more relevant, shown driver phases but not system, set the polarities so the phase is not crossing the chart area at a crossover, and set the passband well into the top half of the plot.

Nothing perfect here, but helpful 😉

pl.jpg
 
  • Like
Reactions: crayoneater
Thank you! So if i build this It will sound decent?
I hate to say this - but all the well intentioned snippets of help here are on the assumption you're providing good data. I'm not sure these tweaks, when I see wildly changing response curves (not toward any endgame) are going to sound decent.

I strongly recommend you find someone to help design a crossover for you.

If you do proceed - at least keep an eye on the S1 (overall impedance)curve in the simulation. I'd hate you to build something with low impedance and hurt your amplifier (depending on the level of low impedance protection circuitry).