Robot Army

I like the bolts-through-springs approach but it would be better still if there were spacer tubes so the springs can't touch the threads. In short, 4-40 bolt inside a 3/4" spacer; spring around that; bracket around spacer engaging spring. I haven't been able to find a similar bracket with a large enough drill hole but drilling them out isn't hard.

It is starting to become clear that there are important physical differences between link-type treads and big rubber bands with holes in them. They are not the only elegance/money tradeoff, however.

• DigiKey - most electronic components
• McMaster Carr - most mechanical components and fasteners
• Jameco - ultrasonic sonar transducers, wall transformer power supplies.
• Spark Fun - Tamiya tread sets and interesting sensors.
• Servo City - Hitec RC servos
• Pololu - Revision 01 frame.
• Canal Plastics - Revision 02 and later frame prototypes.
• Gold Phoenix PCB - Circuit boards.

Frames 01 and 02

The design requires the suspension springs to be at least 3/16" i.d. and the only ones I could find on mcmaster-carr that fit had a rate of about 4lbs/in. (There are six, so it's really 24lbs/inch) This is roughly equivalent to not having any suspension. It also requires an extra steps of cutting brass tubes to fit between the bolt and spring and drilling the strut brackets out to 3/16"+ which is really close to the edge. I broke one while attempting that which means it is an unrealistic plan. I've ordered some nylon flange washers which I hope will allow me to solve the issues with the frame 02 suspension.

This frame was the first kerf-corrected design. In general stuff fit well. One issue was a crack in the top deck where the servo rib tab is inserted. That part is much thinner than it ought to be but there is no obvious way to correct it because it is somewhat dictated by the distance from the drive cog to the rear road wheel. I am going to open that tolerance slightly. It won't break during assembly but it may not be sufficiently durable. The kerf correction also meant that the servo rib bracket is not actually required which is fortunate since the hole for it is in the wrong place. It will help with durability but if removed would reduce cutting distance by about four inches.

uisp -dprog=dapa -dpart=ATmega32 --wr_fuse_h=D8 --wr_fuse_l=DF
uisp -dprog=dapa --erase --upload --verify if=stkload_m32_2k32k_16000000.hex
uisp -dprog=dapa --wr_lock=EF

uisp -dprog=stk500 -dpart=ATmega32 -dserial=/dev/ttyUSB0 if=main.hex --upload --verify

• MAX232 VCC not connected. hand solder to VCC5.
• TXMUX VSS and MAX232 capacitor GNDs not connected. hand solder to GND.
• Too much voltage drop across AVCC isolation resistor. short.
• Omit R29, substitute R26 100k instead of 50k (or, R25 to 25k)
• AD5220 should have higher resistance
• Make R10 smaller
• Remove JTAG port, use a bootloader instead.
• Increase sonar TX power by a factor of 2. I didn't realize why I put in R26 in until now: replace it with a tuned inductor to make the transmitter resonant. Apparently this does not work. I think the problem is that the piezo transducers are not particularly efficient capacitors. If a real capacitor is put in parallel with the transmitter and the inductor value is readjusted it should work but at this point I am more inclined to copy the tried-and-true circuit from Crunch.

• as it turns out you need dc blocking caps at every op-amp stage, especially the last one. AD5220 B lead cut, 0.1 uF added in series.
• the mega32 has no agnd, just gnd, and they are connected. so not connecting them externally does not help a lot.
• 0.1uF vcc filter caps are always a good idea around servos
• an avcc isolation inductor goes a long way helping ignore the outside world
• switching the AD5220 gain during receive seems to cause big problems. no idea why.