Back in the early 2000s I was really into BEAM robotics, and there were a lot of designs for this type of thing in that area. They were also energy harvesting and built from discrete electronics. There was always something really elegant about those designs that MCU driven designs have never quite captured for me. Excellent work!
So interesting to how an electronics Engineer approaches a quite simple problem, resulting with a very clean design. 👏 Contrast with the youtube film makers reaching for a MCU complete with bells and whistles. The use of an MCU is actually quite a complex solution requiring relative expensive components, 'a hammer to crack a nut'.
this is a really advanced design for an educational product. one improvement might be some limit switches to prevent overtravel in the wrong direction as you sort of discovered.
Thats awesome. We're alot alike. I meet random people and cater learning kits to their goals. I send free kits to anyone in the us or Canada and the local makerspace. I use half my disposable income. Your stuff is on another level. I hope you make a fortune. Ill buy a kit from you and you can give it away if you like. I have all the parts and gear to copy your's but id never do that haha. Id send you a bunch of motors and such if you weren't in thailand.
Very neat design, as always. Not that relevant here given the design requirements, but in future proects you might consider a resonant-royer (baxandall) converter in place of a joule thief. They need a more complicated transformer (3 primary windings) and two transistors. You can use a ~$5 CCFL backlight transformer for this (the most common use for resonant royer converters) or wind your own. The circuit's main advantage is that its transformer current is highly sinusoidal, which should improve efficiency over a standard joule thief. With a CCFL transformer I had a ~0.7v input converted to ~500v output, though I didn't measure efficiency. There is an excellent app note by the late Jim Williams on this topology.
I like your technique of analog circuit trickery to make things work with small part count. I would love to understand your design process. But I don't quite understand why you need the scr. I would expect some latch action for the output, but the capacitor on the mosfet gate should keep the output running until the storage is fully depleted. So it seems to me that this scr is actually behaving as an analog comparator, generating a "digital" output.
i think u should explain y u started with a full h bridge but the final circuit only needed 2 high side transistors which i think would extend to y u choose between bjt's and fets and n channel/npn and p channel/pnp cause its a clever and unusual hybrid h bridge
I start- and end with an H bridge. The low side is an FET so it can be voltage driven, and therefore can be held ON by a simple gate capacitor to extend the output pulse. If I made the whole bridge from BJT's, It would require current to keep it in conduction and it would drop out before the caps went to 0 volts.
How about dumping the caps into a dummy load resistor when the voltage difference between the caps is smaller than some tunable value, to get rid of the hunting ?
I have considered this, it's very tricky to implement in a discrete design. It would be fun to do a version based on a micro that has this feature, the challenge here is keeping the parasitic losses low enough, it would be a master class in low-power design!
Yes- this is a good point. An improvement would be to add a dead-band to prevent this mechanical wear- but it's really tricky to do with a discrete design. A microcontroller would make this possible, but it would be hard to keep the parasitic drain of the micro down low enough that it would not dominate at low light levels - a real design challenge!