Yet another Planar Magnetic Line Source, the SMAPPP

That's an Overlay view where the distortion levels are compared from many measurements.
To be able to compare, all measurements shown are relative a 0dB fundamental level; a perfect EQ.

As I intend to use EQ via FIR filters, I am not that interested in distortion measurements that are not dB relative.

I'm using UMIK-2, so SPL are calibrated per se, but as these are the short test membranes I have a 4Ω series resistance so that the amplifier sees a normal load. I also have different widths and thickness of the alumunium traces so I don't really compare SPL at this stage, other than the frequency response.

How can a Q meter help in this case? What Q meter do you in that case recommend?
 
This is due to the fact that between each 110mm segment, the machine does a re-homing of the tool. That is in the X direction.
This re-homing causes sometimes the cut traces to be off by a fraction of a millimeter but enough to maka weeding difficult, especially for the 0.5mm bare parts.

What if you make the segments overlap a little bit? Then weeding should be easier.
 
You are forgiven:
Q_portrait.jpg


UMIK-2 isn't the most accurate microphone even with the calibration file.
But there are so much other variables that affect the measurements, positioning, no having an anechoic chamber and so on.
And I don't know how accurate my dB-meter is either over the frequency range.
But for sure, I can do a litmus test.
 

Attachments

  • MWSnap910 2025-04-04, 11_19_58.jpg
    MWSnap910 2025-04-04, 11_19_58.jpg
    340.2 KB · Views: 11
For the record.

I made an oversized tweeter membrane, oversized in the meaning that the aluminium trace between the magnets were wider than the perfect 3mm.
In this case I tried 8mm between and 5.5mm outside the magnets, the left one in the picture below:
IMG_8703.JPG

The reason for doing this membrane was to see if I could get away with an easier membrane to cut and weed.
I reckon that the membrane doesn't move that much being a tweeter; at 300Hz you'll need 100 times more movement than at 3kHz.
So the non linearity of the magnetic field density shouldn't matter that much.

I got a bumpy curve but also good levels in the high end:
Tweeter 8mm SPL.jpg


Distortion went through the roof though below 6kHz:
Tweeter 8mm distorsion.jpg


So I guess I'm stuck with the membrane to the right.
 
I think that I'm stuck the Silhouette Cameo 5 machine and the membranes as they are.
I'll just make enough of them and then pick out the best four.

I've tried cut 110mm segments with Cut by layer and Cut by line.
But the configured order and pauses are not saved in Cut by layer.
In both cases the machine cuts parts of more than one segment, so the membrane goes to and fro.
So there must be pauses between each segment and then I have the re-homing issue with mis-alignments in the X direction between segments.

Also, as the membrane and the cutting mat being rather heavy, I need some sort of conveyor rolls.
But I found that talcum powder does the trick of lowering the friction. As long as I don't sneeze.

To help with hitting the resume button,
MWSnap912 2025-04-06, 15_12_57.jpg
, ChatGPT helped me with this script:
MWSnap951 2025-04-06, 23_46_03.jpg


Script is included.
 

Attachments

You are forgiven:
View attachment 1444672

UMIK-2 isn't the most accurate microphone even with the calibration file.
But there are so much other variables that affect the measurements, positioning, no having an anechoic chamber and so on.
And I don't know how accurate my dB-meter is either over the frequency range.
But for sure, I can do a litmus test.
Why I asked is from own experience and REW that can end up with whatever on the y-axis depending on mic and setting on the gain knob on the audio interface.... for distortion measurements I think the playout level is important metadata.

//
 
For the record.

I made an oversized tweeter membrane, oversized in the meaning that the aluminium trace between the magnets were wider than the perfect 3mm.
In this case I tried 8mm between and 5.5mm outside the magnets, the left one in the picture below:
View attachment 1445550
The reason for doing this membrane was to see if I could get away with an easier membrane to cut and weed.
I reckon that the membrane doesn't move that much being a tweeter; at 300Hz you'll need 100 times more movement than at 3kHz.
So the non linearity of the magnetic field density shouldn't matter that much.

I got a bumpy curve but also good levels in the high end:
View attachment 1445555

Distortion went through the roof though below 6kHz:
View attachment 1445556

So I guess I'm stuck with the membrane to the right.
Well, just redid the above measurements.
Here's some overlay with the 3mm membrane as well.

SPL:
3mm vs 8mm SPL.jpg


Distortion:
3mm vs 8mm Dist.jpg


And there not much of a difference in both SPL and THD, even for the higher harmonics.
I guess there was a disturbance during the measurement in the previous post.
 
Why I asked is from own experience and REW that can end up with whatever on the y-axis depending on mic and setting on the gain knob on the audio interface.... for distortion measurements I think the playout level is important metadata.

//
Yes, one shall always keep an eye on the Y-axis.
If used with REW properly, the UMIK-2 doesn't have a gain knob though.
 
  • Like
Reactions: TNT
I've considered that further, but the cost and time needed scares me; a 2.5m long CNC with the accuracy needed must be of industrial grade.

Or, as ChatGPT put it:

⚠️ Mechanical Challenges with Long DIY Machines​

IssueExplanation
Frame stiffnessA 2.5 m frame must be extremely rigid to avoid sagging, vibration, or twist
Belt stretch / backlashLong belts or threaded rods lose precision unless carefully tensioned
Stepper drift / skipped stepsWithout closed-loop feedback, stepper systems can gradually lose position
Material feed accuracyMaintaining perfect alignment over 2.5 m is hard without vacuum/roller support
Drag knife dynamicsSharp corners and return paths can still misalign if backlash isn't near-zero

So yes — unless you're going full industrial-grade (e.g., aluminum extrusion frames, linear rails, servo feedback), a DIY cutter is not likely to outperform a well-tuned Cameo 5 in alignment across 2.5 m of media.
And I don't have the space for it.

I've just made two of the full length mid membranes and the two tweeter membranes will soon be done as well.
I will mount them into the speaker and start the measuring and EQ loops. And do some occasional listening 😱 .
While doing that, I will make more membranes and then select the best of them, the rest will be backup; I doubt that I will be able to make membranes in the future.
 
the above is ... well not a huuuge deal 🙂 (i have a 1.3 meter long table and dont use closed loop) not more then a roller that moves a 2.5 meter long thing back and forth 🙂 (just the slightest misalignment or not perfect roller will result in a cutting matt wondering off) stiffness is not a huge thing since there is no cutting force. but anyhow.
PRO XY cutting tables use simple bearings and use steel wire to drive it 🙂 almost bought one if i could fit it... with vacuum table. but the thing itself was rather simple. (it was build like a tank but i think thats mostly because it was professional equipment) by the way most lasers dont use fancy linear stuff or servo's

but if this works !! then there is no reason to change indeed! i am still impressed what a 300 euro cutter can do to be fair
 
Well, the Silhouette Cameo 5 is €500 here in Sweden, and it will soon cost even more due to the trade war.

I think that it will not be that easy and cheap to build my own CNC cutter that long and narrow.
I've already built two plasma CNC cutters, one cartesian and one CoreXY.
And I foresee a lot of hours building this one with the wanted accuracy (0.1mm).
 
  • Like
Reactions: WrineX
aaah ok cameo 5 🙂 i did not know they where that expensive 🙁. well yeah building something yourself might get out of hand on cost (besides the fact you make insane long ones 🙂 size is an issue 🙁 ). well if you make a coil accuracy of 0.1mm (or repeatability) are both different things. after 3000 moves then go back to the original place with 0.1mm is a lot harder then following a coil patern with 0.1 accuracy. i mean ever move it makes with a few micron slop/steppers faults etc adds up over time.

anyhow looking forward to your new coils ! by the way cartesian is that with the weird belt setup ? used in some fast 3d printers ?