I breadborded some of those switches (like the ones on the Bot Board) and mounted them to the feet on the BRAT. I have tested that the weight of my stock BRAT, (with a pan servo) really has no problem actuating the switches. However, I think using only 3 switch per foot (2 in front and one on the outside in the rear of the foot) is a much better plan. It adds some much needed margin to the design.
If you do want to do the board design… The LED’s can be placed so they shine through the holes in the foot. And the wire connector can be in the hole where there is no LED. I will reimburse you for your costs to have the prototypes built.
So there’s no confusion, can you post a photo of the foot and draw lines to the holes you are refering to? also, wouldn’t the third switch in the rear cause the foot to teeder?
Well I hadn’t tested it yet, but now that I have it’s not a good idea to go with 3 switches per foot. Let’s go back to 4 switches. The switches I have are not the lowest pressure ones and they still reliably engage with the weight of the BRAT. So the production run with the lowest pressure switches, I think we will be ok. I’m working on a photo…
Here is an image of my prototype foot switches. I don’t have any LED’s here. But they will be visible from in the round holes in the aluminum foot. I’m ready to abandon the idea of using the PC board alone without the aluminum RF-01.
It looks real good Jim! I am still going to pursue my separate sensor board idea though, because I really think it has more applications than just foot sensors.
It looks like the FOOT contact area is now limited to 4 tiny little switch shafts, instead of the entire foot area. Not good, in my opinion.
I’d suggest (as Mike and I discussed earlier in this thread) that the switch shafts, push pins, or levers go THROUGH the foot, such that the travel of the actuator is limited (to protect the switch) when the foot finally makes contact with the surface, and to transfer the weight onto the foot, rather then remaining on the switch. This way, the full foot rests on the surface.
Fair enough. But the question is, will the switches handle the weight of the bot without damage, and I think they will. Well anyway, that’s what prototypes are for.
If the switches are at each corner like Jim has in the photo, it’s no different than four legs of a table which is solid unless one of the legs is of a different length of course. If it proves to be a problem during testing, then material the thickness of the switch body minus the height of the button can be added to the bottom of the foot, perhaps a piece of lexan the shape of an “X” on the bottom to where it just fits inside the open area of the foot. If it proves to be fine the way it is, then I don’t see a need to worry about it. The bots run generally slow anyway so it’s not like the bots are going to be running with these switches. The slow gaits would allow plenty of time for the 4 switches to make contact before the full weight of the bot is applied.
Yepper. I am also thinking of a very soft rubber material being applies to the bottom of the PC board. As long as it doesn’t interfere with the switches engaging.
This might require adding some holes to the ARF-01/ARF-02 feet, but it would be a better design in my opinion. This could be accomplished with my idea also, by changing where the sensor boards are mounted. The full wait of the robot should never be on the switches, which should move just enough to make contact. With my idea, you could also position the sensor boards so that they hang over the foot where the switch could point downward and not have the full robot weight on the switch at any time.
OK I have been thinking about the button circuit and the general desire to KISS the concept.
If you switch GND then you probably need a circuit similar to what you drew up.
But if you switch the +5V then you can put in parallel
a series resistor and green LED to ground, and
a weighted resistor to a summation node.
So long as the voltage at the node does not get close to Vf of the LEDs then you should be OK.
You may want to include a single non-inverting op-amp at the output to increase the number of bits you can see between switch states.
Here is a table with the different output voltages for each combination of 4 switches.
I can add that to the schematic, but what about what I have so far? I copied Jims original hand drawn schematic and made it a little simpler. The power for the hex inverter are pins 7 and 14 which is not shown but the Eagle cad software automatically connects power and ground to these pins. When looking at the schematic, keep a mental note pins 7 & 14 are connected to power and ground.
Jim’s original schematic does not show pins 7 & 14 connected to anything but the Hex Inverter that I am using is an SOIC14 component.
ya know I was thinking about the LEDs interfering slightly with the summing network and I have a solution… blue LEDs have a higher forward voltage so instead of green use blue LEDs… and you get the bling look for free.
The op-amp selection is a little problematic.
I had selected an LMV321 because it is rail-to-rail out however it isnot going to appreciate the +5V supply being much above +5.5V.
Mike put an LM321 in his schema that will work up to 32V but is not really rail-to-rail output which may hurt you at the top end of the output range.
Last observation was that using this really small package you do not have the “unlimited” short circuit duration the parent of this part is noted for. It isn’t that you will hurt the output stage but rather the package can’t dissipate the power forever because it is so small.
Let’s make the following simple changes using the LMV321M5 IC.
put a small resistor of say 51 ohms in series between J1 pin 1 and the IC1 / R3 junction to limit the output current. this will create a very small error but it should be only a fraction of an LSB with a 5V 8-bit A/D.
Add a small 5.1V zener diode across IC1 pins 5 and 2, cathode on pin 5. An MMBZ5231B from Fairchild or Diodes Inc should be good and is in a SOT-23 package as well. Now add another 51 ohm resistor between J1 pin 2 and the rest of the circuit. At 5V the zener will be out of circuit, and as the supply voltage goes up it should limit the voltage to the LMV321. If we find people can blow these up we could use a 4.7V part and still not limit the output voltage.
You should probably use an 0805 package for the 51 ohm parts and 0603 for everything else. C1 and C2 can also be had in 0603 size as 25V X7R 10% parts. Using a Lite-on LTST-C150TBKT for the LEDs (blue, 0603) you get Vf=3.4V which is great and guarantees the LEDs will be all but out of the voltage sum circuit.
Let’s go back to 4 switches. 
