Search
Forum
Wiki
Images
Files
Member List
Help
(We now resume our previous post, already in progress...)
Electronics in the head are also largely finished.
To solve my issues with the wires in the lower body, the ESP32 has been moved to the head. This means that, for now, the only wires running between the head and body are the three signal wires for the head servos from the R/C receiver, and the two power wires from the buck converter. I made a little U shaped board that holds both the microcontroller and the logic level converter so that the pins are accessible. There are no mounts for this board in the head; it's just going to get glued in for now, in roughly this orientation. That's a dumb way to do it, but it's what we've got.
![[Image: 20220827_134627-smaller.jpg]](https://www.niflheim.net/droidbuilding/images/LD-F1/20220827_134627-smaller.jpg)
The PCA9685 (which currently just takes the power from the buck converter and supplies it to the LEDs, ESP32, logic converter, and servos) has been mounted in the head. All of the LEDs are also in place and wired. This shows a test of those:
![[Image: 20220821_044738-smaller.jpg]](https://www.niflheim.net/droidbuilding/images/LD-F1/20220821_044738-smaller.jpg)
You can't tell from that photo, but for reasons I haven't quite figured out, the neopixel sticks were glitchy as heck (spurious colors and updating the wrong pixels). They work well enough that you could probably read it as intentional, so I can live with it for now. I thought the issue was the need for a logic level converter, but adding one doesn't help at all. The glitches don't happen in the Adafruit strandtest sketch, nor in the boot (setup function) portion of my sketch, but only start to happen in the loop function of the sketch that I wrote.
As far as I can tell, the major difference between the stuff that works and the stuff that doesn't is that the former relies on delay() and the latter is using typical arduino style millis() task scheduling. Maybe millis() is a bad way to do things on an ESP32? That seems odd.
As for powering the stuff in the head, this buck converter is mounted to the track idler brackets and has already been calibrated to put out 5 volts. I had to drill holes in the rear outer side plate for the input wires because there I neglected to put a hole in before printing it (and I ended up drilling three more useless holes for the output wires before I gave up and routed those through the gap between the front and rear half of the body.
![[Image: 20220826_134530-smaller.jpg]](https://www.niflheim.net/droidbuilding/images/LD-F1/20220826_134530-smaller.jpg)
After doing all that, I made a big booboo and damaged one of my neopixels (the standalone "PSI" LED). See, the PCA9685 servo board has six pins for dupont/JST connectors, as well as a pair of screw terminals. Power supply for the devices that the board runs can be connected either way.
I really really REALLY hate screw terminals. I hate the ones on the buck converter, and I hate the ones on the servo board. Nevertheless, I gave using them my best shot and it was just not working. So I switched the power wires to use the six pin JST input and hooked everything up to test it. Before turning it on, I realized the logic level converter was wired completely wrong, and fixed that.
And wouldn't you know it, in spite of very deliberately making a mental note of which way around the connector for the power wires to the head was supposed to go, I STILL managed to put it on backwards. And the board only has polarity protection on the screw terminals, NOT the pins (kind of a dumb omission if you ask me).
The result of the polarity reversal (once I fixed it) was that all of the LEDs after the first stick in the chain (the horizontal strip) were now EXTREMELY flaky ... completely the wrong colors, with the vertical strip being mostly white and brighter than normal. Miraculously, neither the ESP32 nor the first stick in the chain (the horizontal bar) seem to be affected at all.
On a hunch, I removed the first LED showing issues -- the standalone "PSI" LED -- from the strand, reattached the vertical strip, and altered my code to account for the missing PSI absence. Now all of the neopixels are happily back to their normal level of glitchyness. Eh, good enough.
With the buck converter wiring complete, I was also able to add the second track yesterday and FINALLY do some actual test driving (sans head). It turns out that the droid is happy to run on a 2S LIPO, and I'm pretty sure that I will not at any point need high speed mode with the 3S. Even at a middling 11.3 volts, it's ridiculously fast.
Based on my limited testing, it is capable of driving and turning on smooth floor, concrete, and carpet without much trouble. It will eat throw rugs, however, and that might explain why one of my belts came off the pulley afterward.
This is more or less what LD-F1 looks like now. You will notice the speed controllers and such dangling out in front (I mentioned before that I couldn't put them where intended). Without the cover, the neck joint doesn't sit where it should, causing the neck to tilt backwards a bit, unintentionally matching the artwork after all.
![[Image: 20220826_153056-smaller.jpg]](https://www.niflheim.net/droidbuilding/images/LD-F1/20220826_153056-smaller.jpg)
I've wired up the head electronics, attached the cables, and just need to glue my ESP32 mount in place and secure and connect everything when I attach the head.
I'll post more pics and videos once I get everything put together, if I have time.
Electronics in the head are also largely finished.
To solve my issues with the wires in the lower body, the ESP32 has been moved to the head. This means that, for now, the only wires running between the head and body are the three signal wires for the head servos from the R/C receiver, and the two power wires from the buck converter. I made a little U shaped board that holds both the microcontroller and the logic level converter so that the pins are accessible. There are no mounts for this board in the head; it's just going to get glued in for now, in roughly this orientation. That's a dumb way to do it, but it's what we've got.
The PCA9685 (which currently just takes the power from the buck converter and supplies it to the LEDs, ESP32, logic converter, and servos) has been mounted in the head. All of the LEDs are also in place and wired. This shows a test of those:
You can't tell from that photo, but for reasons I haven't quite figured out, the neopixel sticks were glitchy as heck (spurious colors and updating the wrong pixels). They work well enough that you could probably read it as intentional, so I can live with it for now. I thought the issue was the need for a logic level converter, but adding one doesn't help at all. The glitches don't happen in the Adafruit strandtest sketch, nor in the boot (setup function) portion of my sketch, but only start to happen in the loop function of the sketch that I wrote.
As far as I can tell, the major difference between the stuff that works and the stuff that doesn't is that the former relies on delay() and the latter is using typical arduino style millis() task scheduling. Maybe millis() is a bad way to do things on an ESP32? That seems odd.
As for powering the stuff in the head, this buck converter is mounted to the track idler brackets and has already been calibrated to put out 5 volts. I had to drill holes in the rear outer side plate for the input wires because there I neglected to put a hole in before printing it (and I ended up drilling three more useless holes for the output wires before I gave up and routed those through the gap between the front and rear half of the body.
After doing all that, I made a big booboo and damaged one of my neopixels (the standalone "PSI" LED). See, the PCA9685 servo board has six pins for dupont/JST connectors, as well as a pair of screw terminals. Power supply for the devices that the board runs can be connected either way.
I really really REALLY hate screw terminals. I hate the ones on the buck converter, and I hate the ones on the servo board. Nevertheless, I gave using them my best shot and it was just not working. So I switched the power wires to use the six pin JST input and hooked everything up to test it. Before turning it on, I realized the logic level converter was wired completely wrong, and fixed that.
And wouldn't you know it, in spite of very deliberately making a mental note of which way around the connector for the power wires to the head was supposed to go, I STILL managed to put it on backwards. And the board only has polarity protection on the screw terminals, NOT the pins (kind of a dumb omission if you ask me).
The result of the polarity reversal (once I fixed it) was that all of the LEDs after the first stick in the chain (the horizontal strip) were now EXTREMELY flaky ... completely the wrong colors, with the vertical strip being mostly white and brighter than normal. Miraculously, neither the ESP32 nor the first stick in the chain (the horizontal bar) seem to be affected at all.
On a hunch, I removed the first LED showing issues -- the standalone "PSI" LED -- from the strand, reattached the vertical strip, and altered my code to account for the missing PSI absence. Now all of the neopixels are happily back to their normal level of glitchyness. Eh, good enough.
With the buck converter wiring complete, I was also able to add the second track yesterday and FINALLY do some actual test driving (sans head). It turns out that the droid is happy to run on a 2S LIPO, and I'm pretty sure that I will not at any point need high speed mode with the 3S. Even at a middling 11.3 volts, it's ridiculously fast.
Based on my limited testing, it is capable of driving and turning on smooth floor, concrete, and carpet without much trouble. It will eat throw rugs, however, and that might explain why one of my belts came off the pulley afterward.
This is more or less what LD-F1 looks like now. You will notice the speed controllers and such dangling out in front (I mentioned before that I couldn't put them where intended). Without the cover, the neck joint doesn't sit where it should, causing the neck to tilt backwards a bit, unintentionally matching the artwork after all.
I've wired up the head electronics, attached the cables, and just need to glue my ESP32 mount in place and secure and connect everything when I attach the head.
I'll post more pics and videos once I get everything put together, if I have time.