My friend Tim Schutz and his malfunctioning robot Maai came over for our third Robot Sunday last weekend. (Robot Sunday one, two.) We’re trying to restore the robot, a Heathkit HERO-1, to full 1980s lustre. Maai was wearing a kind of clown mask Tim made to replace its protective faceplate, and an elastic belt to keep its panels on. On boot, the robot said “Low voltage”. (This was to be a recurring motif.) Then Tim demoed adjusting the pitch and speed of Maai’s voice.
Last time, we hypothesized a couple of chips had to be replaced on the main sensor board. Tim had bought the chips for our proposed repair from Alphatronics USA in the Tukwila warehouse district near Seattle. Entering the store, he said, was like entering the twentieth century.
Concerned about the voltage, Tim checked it on the robot power supply: 5.3 volts, so probably OK.
The robot was plugged into the wall, because Tim didn’t charge it the night before; he left the robot in the car all night…
Bad robot, no chips. Well, OK, a couple – Tim yanked the sensor board, I replaced the analog-to-digital chip and one other, and Tim replaced the board.
It was now time for diagnostics, as many diagnostics as the galaxies in the sky. Fortunately, we had the arm and pendant (wired remote) this time, and Tim and I were both getting the gist of the diag routines through the sheer rotework of entering them in hexadecimal machine language over and over.
We got through the initial diags to the first sensor tests (motion sensors), but then once again, no response. The LEDs were supposed to light up on motion. (As we worked on Maai, the robot intermittently complained of low voltage. Aloud.)
The main sensor board had light and sound inputs, so we tried those tests next. The light sensor worked a little, but didn’t display a full range of response to bright light, dim light, and darkness. The sound sensors also seemed not as sensitive as expected. Batteries low again, maybe. We were seeing excellent voltage jumps on the board, though, from 0.x to 3.x to 5.x – dark to dim to bright. Was it just that the robot’s display was busted? There was no voltage when we made a loud noise during the sound test. Problems with the sensor board and sound receiver? Mic disconnected? Conclusion: Whatever it was, it wasn’t the chips we replaced after all.
Many other things were kind of working. Sonar was, but it was raining outside so we could only test it for short distances. The pendant seemed to be working. The motion sensor and display were actually working perfectly on second examination. (We had been looking at the wrong board for the motion test. When we found where the right one was tucked away, the motion LED was blinking cheerily.)
The voice synthesizer worked. Tim wanted to do a spoken haiku demo where the robot would spin around afterward, but the drive motors were not working. There were also mechanical problems with the arm as well as electronic ones. Tim is going to strip the arm’s corkscrew motor and other parts down and rebuild them himself. He schooled me in the mechanics of the arm and its claw, but there’s so much to do on this little guy, and we only meet once a month.
So what was wrong with Maai, at base? For one thing, the onboard battery might have needed charged, even though we had it hooked up to wall power the whole time. Some new capacitors couldn’t hurt, either. “They’re 30 years old. When they dry out, they tend to wreak havoc,” said Tim. He may replace the capacitors on the sense board before next time. (Moribund capacitors are a bane to all kinds of retrocomputers, from HERO robots like Maai to Apple IIs.)
We thought of some future improvements:
- Maai doesn’t have a serial interface, but Tim has the schematics, so we could build one. That would enable saving robot software on an external medium like a thumb drive, and expedite diagnostics, among other tasks.
- Rare earth magnets to hold the side panels on instead of an elastic belt.
- Various new masks for the front faceplate area, such as creepy humanoid eyes.
Moral if you want one: Debugging is harder than coding, and repairing something is so much harder than building it from scratch.
See you next Robot Sunday.
Comments on the blog, please, not on Twitter or F*c*b**k.