Another very low cost target panel design

[Xevel]

Hey,

Wandering through my local home improvement store I found some adhesive aluminum tape. I got some to test some fun hacking, like making electrodes for a touch-sensitive art installation, but instead ended up trying my hand at a low cost target panel design.

I have started preparing a little Mech-Warfare-style animation for next year, on a very low budget. The idea is to test the waters here in France before making any investment.
At first I thought we would use the japanese "Survival Game" type of hit sensor (with the aluminum foil cube) but then the previous paragraph happened, and now I have something that might make an electronic hit sensor feasible on our budget.

Having cheap target panels has been a question for some time around here (the previous complete set at $300 was a little bit pricey ), and some around here have provided some very interesting solutions, but I aimed at a little different set of constraints
- very cheap
- can be built with stuff from regular stores everyone has around (no custom PCB)
- fast and easy to build
- doesn't have to work perfectly on any mounting option, as I design these too and will make sure they are adapted.


So here it is :

Material for one full size panel :
- MDF 3mm : 2 squares of 76mm (=3 inches)
- Adhesive Aluminum tape 50mm wide: 150 mm
- 1.5mm outer diameter plastic tube : 80 mm
- double sided tape 50mm wide : 30mm
- regular sticky tape : around 500 mm
- some kind of spacing material (I used polar fleece 3mm thick : 60 x 30 mm )
Add to that some smallish silicone-insulated mutli-strand wire (I've used 24AWG), silicone wire is always so much more flexible.

* One of the square has a notch to get the cables out more easily, but with thinner wire or thicker spacer material (polar fleece) it could be avoided.

* The tube can be replaced with anything the right width and with a similar general morphology. Some hard cardboard, some wire, a toothpick, whatever as long as it serves the purpose explained below.

* The spacing material (polar fleece here) needs to be both compressible and very springy (some adhesive insulation foam, foam double sided tape, many types of synthetic fabric... ). The width of the tube thingy has to be adapted to the material used, so that it is maybe 0.5 to 1.5 mm thinner than the spacing material uncompressed, yet still thicker than the spacing material when it is compressed by the force of a BB hitting.

* I had MDF lying around but anything rigid, light-ish, that will resist to BBs without permanently deforming should work (plywood, PP...)

Assembly:
- Cut the tube into two parts smaller than the width of the aluminum tape
- Stick the aluminum tape flat on the square without the notch, and over the tubes on the other. The tubes have to be as close as possible to the edge but still fully trapped by the tape. Smooth the tape.
- strip the wires on 2 cm and fray them. Stick them on the aluminum, facing the notch (see picture below). The one on the square without the notch will have to go through the notch too when the assembly is finished, so make sure it's properly lined up.



- cut two 60x15mm rectangles of the spacing material (or other sizes depending on the material used, thinner for less compressible materials...) and tape them next to the tube thingy with double sided tape (see picture below).



- close the panel, and put tape around it (see picture below). Using a spacer between the plates to lock everything in place with the flat plate around 0.5 to 1mm above the tube thingy make it easier to get an acceptable assembly.
Tape does not adhere well to the rough side of the MDF stock I had, but it does to itself => simply make more than a turn, it is only there to keep the plates pressed against the spacing material.
- Done.




Many combination of materials could be used and the details of the assembly would have to be adapted.

I've made a few tests with this thing and so far it works well. Detection is good everywhere on the plate (a little harder in the center but I adjusted the spacing material to have it work nicely).
False hit detection when triggering a gun are entirely possible if mounted rigidly on the body of a bot that has the gun also mounted rigidly (witch seems like a bad idea to me anyway... mine will be mounted through shock absorbers). I'll try with various materials and mounting option. I'll try to lower the inertia of the hit plate, adjust the sensibility of detection, reduce transmission of the noise from the body to the plate.


The total cost of materials and assembly time, with a little refinement of the process, should be a few dollars a piece (mostly for assembly). They can nearly be considered disposable.

Not sure how they will compare to the original, since after all this time I still haven't been able to get official ones.

For now they will most probably be perfect for my low-cost version of the game though.

[Xevel]

I started re-reading frantically all the posts relative to the scoring system to see if I had just reinvented something proven to be flawed or not.
Seems like it's cool, this solution had not been tested before.

I'm pretty sure these panels could be made to be 100% compatible with the old ones, with the addition of an rc filter to make it easier for the transponder to know which panel has been hit (currently, it is a little bit abrupt, shooting from 0 to vcc and very noisy when it's up... So while the hit would register, there would be no certainty about the state of the line when the MCU finally services the ISR).


Overall, i think that this design has huge advantages over all the other ones I have seen here:
- cheaper than the official one, probably on par with Deimos' velostat design.
- simplest and fastest to build, no complex operation, nothing to adjust
- requires only a laser-cutter and scisors for "mass" production.
- no tweaking required for the output: it's on(closed) or off(opened), no analog values. (atleast not on the front edge. After that, unless there is the previously mentionned rc filter, all bets are off until the spacing material gets back to its resting position).

I also think it could be quite easily assembled by batch (i think about the aluminum and spacing material being assembled for maybe 10 panels in one operation) with very little changes of the assembly protocol.

[tician]

How does it handle vibration? The big issue with the really old single piezo-element target panels was that almost any movement of the bot would cause false triggering.

[Xevel]

What test protocol do you want me to use? Have the original panels even been characterized, and if so where is the data so I can compare my panel to it?

The old target panels were critically flawed because the piezo based detection acts as a high-pass filter: it reacts hugely (but not necessarily only, don't get me wrong I haven't tested nor reaserched the subject deep enough to give an acurate formula) to the derivative of the disc position as far as I could test. So extremely small but quick movements like vibration created in the support material by the foot of the bot hitting the ground the way many people program it would likely register.


Here the system is mechanically filtered by the material used, the inertia of the target plate seems to be the only enemy as it could cause a false trigger with a hard enough hit on the bot.

EDIT: hmm nope, there might be a real concern over this. The way things are assembled, there is a kind of ball-joint effect between the two plates and rotations are not dampened that much. It could definitly be a problem... /EDIT

And that's with the bulky, ugly matrial I have tested with... I could test with materials with different properties to adjust the behavior as needed... But as long as I can't get my hands on original panels, I won't be able to compare exactly the two.


What i coud do is make a few sets and send them to some people with brutal (in their movement) mechs so that they test on their configuration...
I don't have a laser cutter so the biggest time sink would be to cut the panels though.

[tician]

It is basically a really big tactile button with a weak, poor tolerance spring and a rather massive cap attached to it. Just shaking it around a bit when holding the edge of only one side of the sandwich would be a decent test. Using a rim of foam tape around the entire perimeter of the two plates to keep them well attached together and limit rotation might offer slightly better 'spring' consistency and less susceptibility to vibration.

Another design that might work better: make a traditional tactile button by adding a small strip of aluminum tape near the center of the 'hit' plate of the sandwich and put the two separated protrusions/contacts on the 'mount' plate so that all the wires are attached to only one plate (the wires going through small holes drilled in the 'mount' plate and soldered to contacts). Yet another design would be having a really thin and flexible piece of polycarbonate, or other durable plastic, with an aluminum backing that must actually deform a bit to close the circuit; a bit like the polyester layer (with its carbon contacts) that makes the top of an actual FSR.

Something I have been wanting to test is polyethylene or polyurethane static-control packaging foam instead of velostat to see if it exhibits a more obvious piezoresistive effect when compressed between two copper/aluminum contacts. The stuff it pretty inexpensive and, if it does measurably decrease resistance with compression, replacing the polar fleece with it would pretty much turn the switch into an FSR.

[Xevel]

Indeed, it's a switch.
Shaking it around can trigger it if you shake hard enough, that's far from anything conclusive. I bet piezo target panels would not trigger when shaken around. My hand shaking the module around probably hasnt anything near the same bandwidth as a mech (not a lot of high-frequency content in my slow muscule-actuated arm).

Spring consistency might not be that bad, it's pretty uniform material, it has negligible memory effect when compressed or bend momentarily contrary to most foams. it's a weak spring with non linear coefficient, and that's ok for the application as far as I can tell. The compression range is in the millimeter (around 2.5mm with the one I have), which is a little big but acceptable, it give some tolerance to the assembly.
EDIT: as for the capacitive part, it's effect is negligible. We can see it when we look at the signal on the scope, it offers a nice decreasing exponential at the trail of the pulse when there is no pull-down. but the effect is at least an order of magnitude lower than what I would want to put in my RC filter, and there will be pull-down.

I did a panel rim of spacing material (cotton-wool padding, 1cm thick before pre-compression) with a central around 30x30mm before this one, and while it worked OK too, sensitivity in the corners was not great unless the padding material was very thick (6 or 7mm). The compression force needed to compress the material is nothing like linear along the full compression range (more like exponential it feels), and the travel of the "hit plate" has to be much greater for the contact to be made when hit in a corner => not great if the contact zone is far from the border.

That said, I might try again with your suggestion.


First alternative design: as far as I understand, it works like the symbol of the momentary push button, right ? I don't see how it would register hits close to one of the contacts, as the fulcrum would be so important that the other contact would not be made to touch. Hits directly above the contact or closer to the edge than that would have an even harder time, as the bending torque produced by the BB would be respectively null or on the wrong direction.
I also envision nasty vibrations in the hit plate that would make registering even a center shot erratic as the contacts might not be closed at the same time due to the bending in the top plate, or even its "dancing around" that it would undoubtedly do to dissipate the energy it has been given by the BB. When hit like that, the hit plate cannot be considered a flat surface nor a solid anymore.
Also, soldering in something that has to withstand shocks is a bad idea, I would definitely not solder anything to the hit plate.
Worth a try to see how these effects might or might not pop up, but I won't do it myself.

Second alternative: As I understand your explanation, it seems similar in principle to how low cost, foam DDR pads are built. Two plates of soft platic with conductive backing, some foam with big holes in between, and there you have it. But feet are much bigger than the area between holes so it registers well. Here we are trying to detect BBs, with a contact area of maybe 1 or 2mm^2. How would you then ensure that the two conductive parts stay far enough not to make contact, while still allowing for detection everywhere on the plate? Also, how do you build the whole thing, do you know of any material fitting the bill?


For your last question : I was just playing with that before I saw your post. I believe you are talking about the antistatic foam that some ICs come packaged in, please dispel my mistake if not.
I have a plate of foam (not that inexpensive by the way compared to the original materials used), 32x35x5.5mm, sandwitched between two aluminum contacts of the same area.
It works as expected, the resistance decreases when pressed. With the foam I have and the area used, it's quite stiff but adjusting the shapes, distribution of foam elements, thickness, etc could make it acceptable. I fear for the longevity of the thing though, as this foam is not perfectly elastic and deforms easily.
Various types of similar foam exist, some more flexible than others, and it could be interesting to have a better look, but I won't personally pursue it as FSRs of any kind are nothing but pain.
All we want to know is a binary information, why put some analog stuff that has a huge variability and might then require adjusting, tweaking thresholds and whatnot.


After writing all that, what seems the next step for me would be similar in essence to the first design posted, but with padding material in a thin ( 3 to 5mm maybe) rim at the very edge. I'll try with foam double sided tape (for mirrors) as it would also make mounting even easier, and the whole assembly flatter.
I'll change materials to use thin but stiff plastic plates (not sure which material yet) too, to reduce inertia of the hit plate, and the sandwitch thickness too. I'll keep wires in both plates as it's by far the

[Xevel]

I don't have the foam tape I need yet, so I tried two variants with the same spacing material:

One reuses the same panels, but covers the whole surface with spacing material surface (with holes for the contacts), and adds double sided tape everywhere on the spacing material, on both faces (to remove the need for the last step).

The other one uses some 2mm acrylic (I know it's likely too brittle to resist shots from powerful AEGs, it's just to test the sensor design) for the hit plate, and still the same MDF 3mm for the other plate. Padding is put in a rim, with double sided tape on both sides. Wires are finer, so that they can pass between the plates without risk of affecting the result, and without needing a notch to get them out (but it would still be a good idea to keep the notch or a hole, so as not to have any part of the wire exposed). The rest stays the same.

In both cases it's easier to register a hit in the corners, but much more so in the first case (as expected, since there is much more material to compress when hitting the center than when hitting the corners, in a relation that is similar to a square function as padding covers the whole surface. The second one has a relation between the area to compress for the square and the center that is more similar to a linear function, as padding is on the perimeter).

In both cases I could not register a hit by waving the thing, or hitting the hand holding it with a finger of the other hand.
In both cases, strong finger taps on the back of the board when held firmly in my hand did not result in triggers.

The ball joint effect I talked about earlier has decreased a lot for both, more uniformly in the second case as the padding is on the whole perimeter.

The second one has a much higer risk of having some giant bounce in the signal. The time between the first raising edge and that last falling edge from a finger tap can go from maybe 30us to 50ms... but that's not a calibrated input so I'm not sure what is going on exactly. bounce is much much more present when hitting the center, i wonder if it might not be due to the plastic hit plate "dancing around", wiggling and thus creating contacts on one side then the other for some time... The first one has MDF hit plate, and there is very rarely any bounce, and when there is it's over in a few ms tops.

These two version, with all the double sided tape and padding, takes much longer than the original.
Next version with double sided foam tape as spacing material, and probably a thin piece of plastic with some light adhesive to replace the tubes (easier to place) should get assembly time way down again.

Gotta get back to do actual work though...

[theamk]

In my experience, designs based on compression of materials which were not meant for compression are generally unreliable -- any change in humidity, temperature or just time passing might make them give false results.

If you want a really cheap target, just glue a piezo disk to a piece of polycarbonate -- see these tests I did in 2008:

http://theamk.pbworks.com/w/page/208...09TargetDesign
http://theamk.pbworks.com/w/page/208...C09TargetTest1

The total parts cost is was $9 back then, and it is likely even cheaper today.

[tician]

In my experience, designs based on compression of materials which were not meant for compression are generally unreliable -- any change in humidity, temperature or just time passing might make them give false results.

If you want a really cheap target, just glue a piezo disk to a piece of polycarbonate -- see these tests I did in 2008:

http://theamk.pbworks.com/w/page/208...09TargetDesign
http://theamk.pbworks.com/w/page/208...C09TargetTest1

The total parts cost is was $9 back then, and it is likely even cheaper today.

Surplus piezo elements were tried before and were horribly unreliable. Bots are mobile and legs tend to produce lots of shocks/vibrations that cause false hits (as if the cars were constantly running into walls or over debris). The only way to prevent false hits would be using 3+ piezo elements and multilateration to determine whether the source of the wave is located within the target area or came from the mounting points. Even then, you will still need lots of filtering to separate the BB hits from walking noise.

The velostat target panels with the interlaced traces are basically giant versions of the membrane buttons used in pretty much every A/V remote control and are essentially inverted versions of commercial FSRs (interlink FSRs have the polyester layer with the traces bending to contact the resistive layer, where membrane buttons and the velostat panels have the resistive layer bend/move to contact the rigidly mounted traces). The material used for the resistive layer does not matter much as long as the small force of an airsoft BB pressing it harder into the exposed traces on the pcb measurably decreases the resistance between the two interlaced traces. Velostat (carbon impregnated polyethylene film) is nice in that it is very cheap, very stable, very durable, and its thinness and limited compressibility does not dissipate much of the energy of the airsoft BB.

[Xevel]

Thanks for the links.

First, as Tician just pointed out, a form of the piezoelectric approach has been tested in the first year of MW. I have spent a whole night reading all I could find in the forum about it, and 1) I could not find a very precise description of the setup, 2) the setup had problems due to steps of the robots triggering the system (again, no precise info on the specifics : how it was mounted on the bot, what did the shock of legs hitting the ground look like [and even less characterized in a measurable way]...

In your test, you show waveforms produced when the sensor is hit by a BB gun and when the car hits a wall. The waveform looks different indeed as you mention, but how do you propose to actually make the difference? The amplitude is identical as expected due to the diodes clamping the signal. The resolution of the graph showed does not permit judging on the frequency content (which might have statistical differences but on the naked eye I don't see exactly what criterion I could implement in my AVR MCU to tell the difference.) so it's not possible to judge on the pertinence of an FFT (nor if said FFT could be performed at a sufficient frequency with something like a Teensy 3.1 for example). I don't remember precisely the details of the waveforms I got when I tested similar piezo sensors, unfortunately, and I'm not at home to test again.

Tirggering with a piezo is without a doubt a very cheap and reliable way to get info without false negative, you can buy products based on this effect but the problem is that it is very easy to get false positives.

In my experience, designs based on compression of materials which were not meant for compression are generally unreliable -- any change in humidity, temperature or just time passing might make them give false results.

I am very interested in more info on that. What type of material did you try? how was it affected exactly?