• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

3TD-X Inductor Crosstalk Testing

musicisair

Member
Joined
Nov 11, 2024
Messages
60
Likes
14
I'm building the CSS Audio's 3TD-X and was planning on laying out the circuits just like they are on their 3D printed board (see attached photos), and thought it'd be fun to replicate Gabster's crosstalk experiments.

I have a small speaker wired to an 0.75mH air core inductor (litz wire , but I don't think that matters?). I'm generating a 2300Hz sine wave and sending the signal through my Motu M4 4x4's headphone jack at full volume directly into a 0.4mH air core inductor.

The 2300Hz signal from headphone out measures ~1 Volt and 80 μA (measured via a cheap Astro DM6000AR).

With both inductors laying flat, and spaced about 8 inches apart I can easily hear the sine wave from the speaker. So there seems to be significant crosstalk at headphone-out power levels at this orientation and distance. See attached photo (ruler measures 5 inches).

Does this test reveal real problems with the layout of the networks? Or in other words: should I deviate from the design to try find a more optimal arrangement/spacing?

(my related build thread)
 

Attachments

  • 1737143508689.png
    1737143508689.png
    624.6 KB · Views: 73
  • 1737143525005.png
    1737143525005.png
    1.5 MB · Views: 63
  • 1737143556638.png
    1737143556638.png
    46 KB · Views: 66
Last edited:
It would be very easy to tilt the 400uH orthogonal to the 750uH (which is several inches away on other side of the board). That way, each inductor would be orthogonal on the [x,y,z] axes. Super-easy modification. I potted mine using epoxy so my crossover is "literally stuck".
 
It would be very easy to tilt the 400uH orthogonal to the 750uH (which is several inches away on other side of the board). That way, each inductor would be orthogonal on the [x,y,z] axes. Super-easy modification. I potted mine using epoxy so my crossover is "literally stuck".

Yeah, I might do that just in case. Though there is a chance the crosstalk actually contributes positively to the sound, so I definitely won't be epoxying my crossover together just yet.

I've got a calibrated Umik-1 now, so once finally build this thing I'll try to measure a few crossover layouts/orientations. My ideas are to test an ideal crossover (wire all parts 10+ inches away from each other, at optimal angles), 3TD-X plan's crossover orientation/layout, and a worst-case crossover (inductors flat top of each other , etc.). to see if there is any measurable difference.

I also plan to test if putting an inductor as close to the driver's magnets as possible changes the frequency response at all.

---

A bit of a tangent, but I've done a little more testing, but on the woofer circuit (series inductor + parallel capacitor and resistor).

I hooked up the 3mH inductor to the source (2500 Hz sine wave), and connected the capacitor to the speaker (nothing else connected), and the capacitor also picked up the signal. A 20Ohm resistor picks it up as well, but not nearly as much. in this circuit the capacitor is already shunting signals to ground, so I don't think its important here.
 
Last edited:
We really need a $100 easily programmable equivalent to the MiniDSP 2x4HD. It’s just painful to see all these passive XO threads!
 
I'm working on building out a home theater, so building and then integrating 9+ powered and active speakers is going to be much more work than passive speakers.
 
Yes, I have a thread about active crossover for the 3TD-X. DSP, three amplifiers, power cords for each device, six sets of speaker cables, three sets of RCA cables, there are wires everywhere. Plus subwoofer (another amplifier, power cable, signal cable, and speaker cables) and signal processor (dbx subharmonic synthesizer) and signal sources (streamer box and phono preamp). I have about 10 boxes to plug in, so I have a big power strip. Cable management becomes a nasty issue. I looked on amazon for bi-amp or tri-amp cables to help consolidate the mess, but I ended up bundling the wires together with wire ties the clean it up a bit.

Active crossovers have a lot of overhead, that's for sure. I've been running my Linkwitz system since 2005 on active crossovers. At first, I used Seigfried's analog circuit boards. Then I switched to a very early MiniDSP a few years later. That box is nearly 20 years old now.
 
Yes, I have a thread about active crossover for the 3TD-X. DSP, three amplifiers, power cords for each device, six sets of speaker cables, three sets of RCA cables, there are wires everywhere. Plus subwoofer (another amplifier, power cable, signal cable, and speaker cables) and signal processor (dbx subharmonic synthesizer) and signal sources (streamer box and phono preamp). I have about 10 boxes to plug in, so I have a big power strip. Cable management becomes a nasty issue. I looked on amazon for bi-amp or tri-amp cables to help consolidate the mess, but I ended up bundling the wires together with wire ties the clean it up a bit.

Active crossovers have a lot of overhead, that's for sure. I've been running my Linkwitz system since 2005 on active crossovers. At first, I used Seigfried's analog circuit boards. Then I switched to a very early MiniDSP a few years later. That box is nearly 20 years old now.
I've been following that thread. I'm pretty much stalking all of Kerry's posts at this point. Haha

Here's some renders of my current plan for my 3TD-X. I want to chamfer the top and top corners, but also have a black front (veneered in matte vinyl or suede), with the front edges wood veneered. There will be a narrow and shallow "reveal" (or groove) all around the black part. I haven't quite figured out how I'll implement it though.
238d90e4-d05c-4a21-9998-148a14cb1d96-1_all_141011.png


1000020364.png
 
I'm building the CSS Audio's 3TD-X and was planning on laying out the circuits just like they are on their 3D printed board (see attached photos), and thought it'd be fun to replicate Gabster's crosstalk experiments.

I have a small speaker wired to an 0.75mH air core inductor (litz wire , but I don't think that matters?). I'm generating a 2300Hz sine wave and sending the signal through my Motu M4 4x4's headphone jack at full volume directly into a 0.4mH air core inductor.

The 2300Hz signal from headphone out measures ~1 Volt and 80 μA (measured via a cheap Astro DM6000AR).

With both inductors laying flat, and spaced about 8 inches apart I can easily hear the sine wave from the speaker. So there seems to be significant crosstalk at headphone-out power levels at this orientation and distance. See attached photo (ruler measures 5 inches).

Does this test reveal real problems with the layout of the networks? Or in other words: should I deviate from the design to try find a more optimal arrangement/spacing?

(my related build thread)
A few years back I read up on this and was under the impression that there wasn't any crosstalk if you had them angled correctly and at least 2" apart. Now I'm seriously re-considering.
 
A few years back I read up on this and was under the impression that there wasn't any crosstalk if you had them angled correctly and at least 2" apart. Now I'm seriously re-considering.
I was doubtful I'd even be able to recreate Gabster's experiment with my set up (he uses an oscilloscope and dedicated signal generator), so I was a little surprised it worked at all.

My big question is: since the speaker and its crossover has already been meticulously designed, tweaked, and measured (via DIY spinorama, anechoic, and Klippel!), is (potential) crosstalk in the fully implemented crossover actually part of the desired frequency response? Might removing it (if present in the complete circuit) alter the crossover slope?
 
I was doubtful I'd even be able to recreate Gabster's experiment with my set up (he uses an oscilloscope and dedicated signal generator), so I was a little surprised it worked at all.

My big question is: since the speaker and its crossover has already been meticulously designed, tweaked, and measured (via DIY spinorama, anechoic, and Klippel!), is (potential) crosstalk in the fully implemented crossover actually part of the desired frequency response? Might removing it (if present in the complete circuit) alter the crossover slope?
There's only one way to find out........measure it both ways. I'm not sure how the experiment translates into actual dynamic music since a 1Khz tone is different than the dynamics in the music. Does the crosstalk just add distortion? Does it actually add magnitude? So many questions.......
I have often split up my crossovers just because of the size but now I have to really re-think everything.
 
There's only one way to find out........measure it both ways. I'm not sure how the experiment translates into actual dynamic music since a 1Khz tone is different than the dynamics in the music. Does the crosstalk just add distortion? Does it actually add magnitude? So many questions.......
I have often split up my crossovers just because of the size but now I have to really re-think everything.
Even if I measure it at home the resolution (and varying background noise, as I live within earshot of a highway) might not be good enough to decide anything. Might have to send an experiment in to @amirm for a more precise measurements one of these days!
 
Even if I measure it at home the resolution (and varying background noise, as I live within earshot of a highway) might not be good enough to decide anything. Might have to send an experiment in to @amirm for a more precise measurements one of these days!
If it affects the frequency response in a way that you can hear then you would see it in measurements even in your environment. The noise floor would be too much for distortion measurements but gated 1 meter on-axis measurements would show it.
 
If it affects the frequency response in a way that you can hear then you would see it in measurements even in your environment. The noise floor would be too much for distortion measurements but gated 1 meter on-axis measurements would show it.

I was thinking that since the background noise varies in amplitude and frequency as different cars, trucks, motorcycles, etc. accelerate it'd be difficult to compare one measurement to another, at least at a resolution of 1-3 decibels, because the SNR would vary so much. I'll still give it a try once I've got everything built.
 
I was thinking that since the background noise varies in amplitude and frequency as different cars, trucks, motorcycles, etc. accelerate it'd be difficult to compare one measurement to another, at least at a resolution of 1-3 decibels, because the SNR would vary so much. I'll still give it a try once I've got everything built.
Take really loud measurements...........I joke, but it should work. Or maybe I'm underestimating how loud it really is.
 
I'm doing a lot of work with the active crossover to improve soundstage. They still aren't as good as my Triangles. The 3TD-X provides imaging cues that make the loudspeaker itself easy to localize. I lowered the tweeter/midrange crossover (reasons explained in the active crossover thread, to prevent the 5" midrange driver from playing 5" wavelength frequencies). This helps some, but not enough.

I may be fighting the diffraction of the front baffle, which is a physical limitation of the speaker itself. I'm attracted to your beveled enclosure to mitigate diffraction. I don't know how bevel enclosures that are already assembled. I can get a 1.5" roundover router bit, but that doesn't match your fancy angled cuts. Plus, 1.5" won't make much difference. I'll need like 3" roundover to be effective. Maybe I'll remove the drivers and take the enclosure (and your CAD drawing) to the local woodworking shop and see what they think. In the meantime, perhaps I can learn something from attaching felt pads around the baffle edges near the tweeter and midrange.
 
I'm doing a lot of work with the active crossover to improve soundstage. They still aren't as good as my Triangles. The 3TD-X provides imaging cues that make the loudspeaker itself easy to localize. I lowered the tweeter/midrange crossover (reasons explained in the active crossover thread, to prevent the 5" midrange driver from playing 5" wavelength frequencies). This helps some, but not enough.

I may be fighting the diffraction of the front baffle, which is a physical limitation of the speaker itself. I'm attracted to your beveled enclosure to mitigate diffraction. I don't know how bevel enclosures that are already assembled. I can get a 1.5" roundover router bit, but that doesn't match your fancy angled cuts. Plus, 1.5" won't make much difference. I'll need like 3" roundover to be effective. Maybe I'll remove the drivers and take the enclosure (and your CAD drawing) to the local woodworking shop and see what they think. In the meantime, perhaps I can learn something from attaching felt pads around the baffle edges near the tweeter and midrange.


Yeah, the chamfers were suggested by Kerry here. My model's inner dimensions are slightly wider (by 11.11mm), shorter (by 19.05mm), and shallower (by 12.7mm) than the original designs. I want to add a plinth with speaker spikes, but still align the tweeter with my listening position, hence the height change. And I made the inner dimensions shallower because the front baffle is now 1 inch thicker to make room for the chamfers, and I wanted to get some of that depth back.

This overall gross volumes (not subtracting out braces and other internals) are almost identical to the CSS plans: the mid's box is now ~8.7L (specced at ~8.69L). Bass chamber is now about ~46.96L (specced at ~47.5L). This changes should be inconsequential to the performance and response.

I have a partially parameterized FreeCAD model that will mostly let you tweak the dimensions I can try to share. The file is a bit chunky though, at 500MB.

Some of the parameters (some of these will break the model when changed):
1737591233536.png
 
...The noise floor would be too much for distortion measurements but gated 1 meter on-axis measurements would show it.
Take really loud measurements...........I joke, ...
If you are not using the measurements for crossover design then don't worry about being at 1m. Take measurements at 31.5 cm and take them pretty loud.
 
Yes it will.

I'm still just an arm-chair builder here... but I've obsessed over reading actual builders' experiments with roundovers and chamfers lately, and it seems that sometimes it helps, sometimes it doesn't, and sometimes it seems to just move the energy around. So I think it depends on too many variables to apply a blanket "yes it will" or "no it won't" response.
 
I'm still just an arm-chair builder here... but I've obsessed over reading actual builders' experiments with roundovers and chamfers lately, and it seems that sometimes it helps, sometimes it doesn't, and sometimes it seems to just move the energy around. So I think it depends on too many variables to apply a blanket "yes it will" or "no it won't" response.

I bet you couldn't find any of those "sometimes it doesn't help" builds with data to actually prove such a claim. Reducing secondary cabinet related sound sources is always beneficial. There's tons of evidence out there showing improvements with round overs, which lines up with all my own data and listening impressions. I already made a pretty long post here about it citing tons of examples and how basically in every single speaker generating HF information, round overs/chamfers/etc are of benefit.
 
Back
Top Bottom