Need help with a BEAM circuit

[FastEddieFelson]

Hi Folks,

A few months ago I came across some info online about BEAM robotics (Biology, Electronics, Aesthetics, Mechanics) and all of the fun everyone was having back in the 90's. Being somewhat of a newbie to electronics I set out to learn more by working with several of these circuits. My latest attempts have been with a simple "pummer" but I continue to hit a wall and cannot seem to get past it. It's a very basic circuit. A 74AC240 octal inverter, a few caps, LED and a solar panel. It's intended to charge the NiMH batteries during the day and discharge them through a blinking LED at night.

The thing is while I've gotten it to work, it ONLY works when the smaller capacitors that are tied to the opposite outputs are different values. This is at odds with all of the schematics I've ever seen published for this design. It doesn't really  matter what the caps are in my build as long as they aren't the same. I can't for the life of me get the circuit to work when the values are identical (0.22uF) as called for in the design. It's driving me crazy!

I've tried rebuilding the circuit numerous times, tried another IC, different cap values, different locations on my breadboard, another breadboard, etc. It works which I guess is great but infuriatingly so, blinking away as it oscillates back and forth all the while mocking me by not working the way it should.

Is there something I'm missing here? Is there a flaw in the design or am I just not understanding properly what should be happening? I'm using a legit TI inverter and all my components seem to be fine. I'm just not getting the expected behavior when I follow the schematic.

Here's the schematic I'm using:



I've found other versions of this online (including a more illustrative one in Make Magazine) and they all appear to be the same.

Any thoughts?

[Ian.M]

The solar cell biasses both /OE pins to logic '1' when the cell is illuminated, disabling all the inverters until after sunset.

You haven't got a snowflake's chance in hell of understanding the rest of that circuit until you've redrawn it with individual gates representing the 74AC240.   

[FastEddieFelson]

Thanks Ian. I'm good on the solar portion of the circuit. That's working perfectly. You're right, the cell pulls the OE pins high when illuminated and the resistor pulls down if the light level sinks too much. That's working great although now I have it disconnected. I'm solely focused on the "pumming" oscillator portion of the circuit now that is triggered when the OE lines are low. I've taken a look at the datasheet for the 74AC240 and the gate configuration makes sense to me. I really am not using many of the inverters.



In this diagram my configuration is on the left hand side (pins down). The circuit works great if the caps tied to the upper inverters on both sides are of a different value from one another. If they are both .22uF (or .1uF or .47uF or any other combination that is the same) the LED doesn't flash. I just have no idea why that is. The only thing about my circuit that differs from what is listed in the original I posted is that I don't have a 4.7M resistor so I've tied 5 1M resistors together in series. I've also tried 4 and 1 1M resistors and it doesn't seem to have any effect. If the caps aren't different values it doesn't work.

Thanks for your help

[Ian.M]

OK. I've traced it out and the top two inverters, the 4M7 resistor and the two caps are supposed to form a capacitively coupled  two gate oscillator.  Unfortunately two gate oscillators of this type are inherently unreliable and are critically dependent on input thresholds, leakage current, relative propagation delays and cap values. 

Try rearranging it as a classic two gate oscillator - see Fairchild Application Note 118 CMOS Oscillators, Fig 5a., using 1Meg resistors and a low leakage 1uF cap (i.e. not an electrolytic or tantalum).



N.B the oscillator WILL NOT WORK if the U1A input leakage resistance or the timing cap leakage resistance is too low, Both must be significantly greater than the timing resistors.  A three gate oscillator would be preferable, but there are no spare gates, and stealing a gate from the LED driver would reduce the max LED current.  If you do steal a gate to improve the oscillator, use one of the ones that drives the LED cathode as a 74AC240 pulls low harder than it can pull high.

I've also attached a LTSPICE sim of the modified oscillator and existing LED drive section of the circuit.

[FastEddieFelson]

Thanks very much Ian for your help. I'll give your suggestion a try and see what I come up with. Since the oscillator is as sensitive to leakage as you pointed out I'm wondering what results I would get if I just go ahead and solder the thing and get it off the breadboard. I may go that way also.

Really appreciate your time!

[Ian.M]

Probably worse.  At least on the breadboard there's no flux contamination.  You'll probably have to clean the assembled board thoroughly then hit it with 100 deg C hot air to get it to work as it should.

If you are wondering why its so critical - That's what you typically get if artistic types fiddle with a circuit to push its performance limits on the bench without doing a worst-case analysis of its operation.

I believe the original designer was chasing the lowest possible quiescent power consumption between flashes, hence the very high value timing resistor.    For startup, it relies on both gates inputs drifting into the linear region then a little noise giving it enough of a kick for one of the outputs to transition, hopefully kicking the other gate input in a direction that will cause it to transition as well.  Check out the input leakage current in the 74AC240 datasheet - max +/-0.1uA @25 deg C, or +/-1uA over the range -40 to +85 deg C.
1uA through a 4.7Meg resistor is 4.7V - more than the supply voltage, so if you are unlucky the original oscillator can stick at startup as the resistor isn't enough to pull a leaky input off the rails.