Introduction to designing crossovers without measurement

diyAudio Moderator
Joined 2008
Paid Member
I'm not comfortable with your data files. The Dayton should have an impedance peak at 2k, yours is 2k7, the magnitude is also not correct.

I'd have expected the woofer to become unsuitable after 4k7 due to directivity changes.

Dayton suggests using the tweeter as low as 3k5.
 
That looks excellent, thanks again. So is the basic approach for a midrange to apply the same tweeter and woofer crossover methods in the first part of the thread, or is there something different in the order of operation? How did you go about selecting the values on the example, there, and does it matter which part of the curve you tackle first? I also see I was on the wrong track using an electrical 2nd order low-pass, where your 1st order + notch filter tracked great for a 2nd order acoustic shape. Lastly, I see your series resistor after the filters makes a big change in shrinking down the response vs putting the attenuation before the filters. What's the idea behind that?
 
Last edited:
diyAudio Moderator
Joined 2008
Paid Member
Kotter, I was thinking to my own experiences in tracing curves. The errors I have made include offsetting frequency, or magnitude, and selecting a linear scale instead of logarithmic.

Dayton suggests 3k5 with the ferrofluid tweeter, I'm thinking maybe try that with a steep enough slope.

I also am not sure about the woofer response, how did you get this?
 
diyAudio Moderator
Joined 2008
Paid Member
So is the basic approach for a midrange to apply the same tweeter and woofer crossover methods in the first part of the thread, or is there something different in the order of operation?
When going completely without measurement you need to start somewhere and expect to work on it. I do things a little differently when I measure.
How did you go about selecting the values on the example, there, and does it matter which part of the curve you tackle first?
You get a feel for how things fit in, for example seeing how steep the filters want to sit on their own, or whether a bump will come out with the components or instead you have to add more.
I see your series resistor after the filters makes a big change in shrinking down the response vs putting the attenuation before the filters. What's the idea behind that?
Sometimes they work better on the end. It can be like impedance flattening.
 
I've now re-read the entire thread and have learned so much. I recently bought a used book on speaker design by David Weems but the information in this thread is far more practical, IMO. Allenb should write a book himself!

My last question for now, when mention is made of a slightly tilting down response slope (vs ruler flat) being preferable for pleasant listening, how is this done in a 2-way crossover, and a 3-way crossover? Would it involve lowering the tweeter XO frequency and/or level? And what of a midrange?
 
diyAudio Moderator
Joined 2008
Paid Member
Would it involve lowering the tweeter XO frequency and/or level? And what of a midrange?
Only in so far as the driver response shape is changed, not the overall crossover frequency. The tilt should be applied overall so it lays outside the scope of the crossover.

You can either do this to the whole speaker after you cross, or apply the tilt to each driver equally while designing.
 

Attachments

  • tilt.png
    tilt.png
    62.4 KB · Views: 217
diyAudio Moderator
Joined 2008
Paid Member
Overall, the response curve looks ok. There are a few things I'd do at this stage...

The impedance is low across much of the spectrum. Are these 8 ohm drivers? It would be interesting to compare the response you have here to the original responses. You want to see that you aren't raising them too much and you want to reduce their breakup regions. I'd turn off the system phase plot and add the phase plots of two individual drivers at a time or all three if you can see through them. (This is a big step but it helps you adjust to healthy values.)

Also, have you accounted for your box width?
 
Aniva, I can't see any High-Pass filtering for the NE123; are you sure it'll be happy getting fed all the LF in parallel with the bass driver without overheating or running out of excursion?
Obviously the cabinet is keeping its output level low there (presumably a small sealed chamber) but that alone isn't always enough.

That's also probably contributing a lot to the low impedance that AllenB has flagged up.

HTH,
David.
 
Allen and David, thank you for your replies!

David, you are right. I will use small sealed chamber for mid. Unfortunately, adding High-Pass filter to mid is destroying the response curve.
Maybe I will try to add it again..

Allen, I accounted for box width and drivers positions, using Jeff Bagby's Response Modeler. I am going to use boxes from my current speakers.
The drivers have the following impedance:
1) Tweeter (SB29RDAC) - 4 Ohm;
2) Mid (Peerless NE 123) - 8 Ohm;
3) Woofer (Peerless NE 225) - 4 Ohm.

I have tweaked my crossover to increase impedance, according to your comments.
I also generated phase plots.
Please make comments, if you see some pitfalls or ways for improvement.

P.S. Allen, I really appreciate your tutorial for designing crossover! From my crossover you can see that I heavily used it.
Special thank for your openness to question!
 

Attachments

  • Crossover SB29-NE123-NE225_v9.png
    Crossover SB29-NE123-NE225_v9.png
    60.1 KB · Views: 114
  • S1(SB29RDAC)_v9.png
    S1(SB29RDAC)_v9.png
    57.5 KB · Views: 115
  • S2(NE123)_v9.png
    S2(NE123)_v9.png
    63.8 KB · Views: 104
  • S3(NE225)_v9.png
    S3(NE225)_v9.png
    24.5 KB · Views: 123
Re the HPF for the mid messing up the response curve: welcome to the joys of crossover design!
If you want to post your frd's & zma's I'd be happy to take a shot at it, but I'm a bit of a beginner with VituixCAD myself so no promises that it'll suddenly be perfect ;)

There are a very few cases where you can get away with running the mid open all the way through the LF but as I alluded to before, you'll run the risk of wrecking your sound quality due to increased distortion from the resulting high excursion and possibly damage the driver from overpowering if you go that route. Troels G has a couple of examples, but crucially he uses bigger mids when he does it.

Because your LF is already 4Ω you'll never really get rid of that dip to around 2.5Ω without the HPF, you'll want to check your amp is happy with that before turning it up to any appreciable volume.

Re the phase, what Allen will have been prompting you to check is that the phase of low and mid are similar at frequencies where they overlap, and the phase of mid and high are similar where they in turn overlap, so one graph with the phase of all three individual drivers present would be the most useful way for you to look at that.

Edit PS: I'm also a little curious about the 3dB step in the woofer response around 130Hz, did that appear when you merged in a predicted LF/Box response to the stock curve for the driver? I can't see anything in the driver's raw response, or T/S parameters that would account for it, perhaps it might be worth rechecking how it got there?
 
Last edited:
Edit PS: I'm also a little curious about the 3dB step in the woofer response around 130Hz, did that appear when you merged in a predicted LF/Box response to the stock curve for the driver? I can't see anything in the driver's raw response, or T/S parameters that would account for it, perhaps it might be worth rechecking how it got there?

Have checked the PDF for the driver and the response curve is shown for 2.83V input, which is the equivalent of 2W into a nominal 4Ω load. If you modelled the LF at 1W, that would be exactly the 3dB difference visible on your traces.
 
David, you are right again) 3dB step in the woofer response is baffle step diffraction.

Because your LF is already 4Ω you'll never really get rid of that dip to around 2.5Ω without the HPF, you'll want to check your amp is happy with that before turning it up to any appreciable volume.
I have also checked others DIY crossover designs (Troels) with 4Ohm woofers, they have 2.5 Ohm. I agree with you without the HPF, I will not manage to increase impedance in that area.

I did combined phase graph, hope it will work.
Allen and David, I attach Xsim(.dxo) and speaker data (frd&zma)
 

Attachments

  • Phases_v9.png
    Phases_v9.png
    37.2 KB · Views: 103
  • SB29-NE123-NE225_v9.dxo
    95.4 KB · Views: 82
  • Data!SB29-NE123-NE225_v9.rar
    37.8 KB · Views: 93
diyAudio Moderator
Joined 2008
Paid Member
If you look at the tweeter factory data and the baffle adjusted response, there is quite a difference. Sure, this is the way it is but only from one angle. This is a matter for you to remedy physically. In the mean time you would find it easier working from the factory data because it is closer to overall average.
sb.jpg Screenshot from 2021-12-30 16-50-17.png

Your midrange highpass needs to work with the impedance. Your data is a difficult vented impedance. Unless you have something better you should work with the factory impedance data because it is closer to the closed box you mentioned earlier.
 
Allen and David, I attach Xsim(.dxo) and speaker data (frd&zma)

Thanks for sharing those. Might be a few days before I can look at them properly, will be back if I've anything constructive to add.

If that 3dB step @ 130Hz were bafflestep, it would imply a baffle at least 900mm on its shortest dimension, however your HF response is more consistent with a baffle about 250mm wide, could you confirm the baffle size you're working with please? The gentle slope down from 800Hz to 200Hz on your LF response would be more consistent with that 250-ish baffle size

BTW, do you own the drivers already? I know you said you were hoping to re-use an existing cab, wasn't sure if that applied to the drive units too.
 
Your midrange highpass needs to work with the impedance. Your data is a difficult vented impedance. Unless you have something better you should work with the factory impedance data because it is closer to the closed box you mentioned earlier.
Allen, you made a very good observation about midrange impedance. I resimulated it, as for small closed box, and got overall better result for system impedance with min about 3.5 Ohm.

If that 3dB step @ 130Hz were bafflestep, it would imply a baffle at least 900mm on its shortest dimension, however your HF response is more consistent with a baffle about 250mm wide, could you confirm the baffle size you're working with please? The gentle slope down from 800Hz to 200Hz on your LF response would be more consistent with that 250-ish baffle size

BTW, do you own the drivers already? I know you said you were hoping to re-use an existing cab, wasn't sure if that applied to the drive units too.
David, you are absolutely right, the baffle is 250mm wide.
I believe, the 3dB step @130Hz is artificial. I have reapplied baffle step diffraction response on woofer from 300Hz, before it was from 100Hz.
Please see in attachment.

I have ordered drivers, I will receive them in the end of January.

Guys, I would like to wish you a happy new year!
 

Attachments

  • V10_Impedance.png
    V10_Impedance.png
    6.6 KB · Views: 59
  • NE225_bs.png
    NE225_bs.png
    7.2 KB · Views: 71
  • NE225_bs_upd.png
    NE225_bs_upd.png
    7.8 KB · Views: 64