Wednesday, June 24, 2015

Disappointing results, new learnings

Not a great week for the prototypes.  Things did not work out as hoped, but things were also learned, some of which were obvious in hindsight.  But even obvious things, it's good to experiment with, because sometimes the things that seem obvious are wrong.

Low-fi linear valving
I had some success with the linear valve prototoype based on square stock, and wanted to try a similar idea with some modifications: I wanted to try it with round stock (one of which was a pipe), to reduce costs and machining complexity.  And I spaced the holes much further, an inch apart, and did all the work with a drill press instead of machining.  I wanted to see if I could make a "low-fi" version of this that would be cheaper and easier to make.  I reasoned that placing the holes further apart would help reduce hole-to-hole gas leakage, and also make the absolute registration of the holes between the parts less important, so that simple drilling with approximate hole locations would be okay.  I used 1/16" holes.  I also broke a couple drill bits doing this before I remembered that, with drilling, slower is not always better: with such a small bit, a very high RPM is much better, because otherwise I get impatient and put too much stress on the bit.

I also experimented with sealing the ends with epoxy (JB Cold Weld) instead of soldering or welding, which actually worked pretty well.  For a prototype, at least, it's a lot simpler and quicker, and there were no heat issues because the seals were far from the flames.  So, yay for quick and dirty.

Unfortunately, total fail:

There's basically full leakage, no valving at all, not even significant modulation of the flame height on actuation.  So, what did we learn?

First, hole spacing isn't really changing the leakage.  This should be obvious if the pressure is at or close to equilibrium, and this says that it basically is.  Again, should be obvious, but experiment confirms it.  Second, this behaves a lot like the rotary prototype: total leakage.  That suggests that the main problem is the pipe/cylinder shape.  The part tolerance is lower (based on what I was able to source), but also, the pipes make contact at a single point:
The size difference is exaggerated here, but the principle is the same: circles only touch at one point, meaning that there's just not much overlap to gate the flow with.

For a mechanical valving prototype that means that basically we must go with square stock.  That's certainly possible, but it will be necessary to figure out the relative cost.  I might have to contact some suppliers on Alibaba to figure out if this is worthwhile.  I may also use my remaining square stock to see if I can do the low-fi machining thing, and find a better way to seal them, which was the previous problem.

I have one other theory about a potential rotary mechanical valving prototype, whis is to actually leave a large gap, of half an inch say, and actually let the inner tube be fully on fire.  The outer tube will them simply gate the flame from coming out and being visible.  This may not work: the flame may simply want to travel too much; after all, the flame is really just hot gas.  It will also make the outer tube VERY hot, but it won't be on for long in the eventual art work, so maybe not that hot.  Anyway, something I can try quickly and easily I think.

Solenoid valves
I tried a simplified attempt at building a solenoid valve prototype with four valves and three inch separators.  Instead of building a full rail on top, I took a U-channel and drilled holes, and placed them atop the open pipes.  I was hoping to use this to diffuse the flame somewhat, and reduce the complexity.

The result here was just a lot of gas leakage, and eventually a huge fireball.  Not a big failure, just wanted to see if I could do this on the cheap.  I'll buy some more pipes and see if I can make a full rail on top, with holes drilled etc.

Also, these things draw like 3 amps each.  I was thinking of doing some pulse width modulation, but this might be complicated.  I'm going to see if I can maybe run these from an Arduino with a big ass MOSFET and a snubber diode.  But even if I don't blow up my Arduino (which I might), I may not be able to effectively pulse width modulate them.  We'll see.

Edit: On second thought, this motor driver with a snubber diode looks better designed for the purpose:

No comments:

Post a Comment