These two discrete LM3909 circuits recently stopped blinking:
The green LED (on the left) took six months to wear its pair of not-dead-yet AA alkalines from 2.7 V down to nearly zero.
The blue LED in the radome took two months to go from 1.0 V (!) to nearly zero. It didn’t start very bright and went decidedly dim along the way, but the LM3909 circuitry still managed to jam a few microamps through the LED.
In both cases, one of the cells was reverse-charged by a few hundred millivolts, although neither leaked.
Both got another set of not-quite-dead AA cells and they’re back in action.
One of those LED spotlights may have barely outlasted its worthless warranty, but not by much, and has been languishing on the back of the bench with “Flickers hot” scrawled on its side.
The metal base didn’t respond to twisting, so I slit the threads with a cutoff wheel:
Applying the screwdriver removed the base to reveal a silicone rubber casting:
The small wire emerging near the edge of the plastic case seems to be the neutral contact to the shell, with a poor enough joint to suggest it might have been why the lamp flickered when it got hot.
Some brute force snapped the silicone off at the bottom of the plastic case and broke the two wires bringing AC to the PCB:
Digging around inside produced a debris field of silicone crumbs, broken resistors, torn caps, and various other components, with zero progress toward removing the shell:
A little lathe work converted a chunk of PVC pipe into a crude mandrel supporting the mangled case:
A few millimeters of sissy cuts released a silicone O-ring sealing the shell against the reflector:
Continuing the cuts eventually revealed the three screws holding the shell to the reflector and the two wires powering the LED:
Chopping off the screws with a diagonal cutter freed the shell and revealed the top of the PCB:
It really does have a surprising number of components!
Those three screws connected the LED panel / heatsink to the shell through the back of the double-walled reflector. More brute force peeled the outer shell away and released the panel:
Each of the 5050 packages contains a pair of white LEDs with 5.2 V forward drop for the pair, at the very low test current. They’re all in series, so you’re looking at well over 60 V total forward drop:
Once again, the discrete LM3909 circuitry can blink a blue LED while running a pair of alkaline cells all the way down to about 1 V, with one cell ending at 0.2 V and the other at 0.8 V. They started out discharged to 1.2 V each during their useful life, then blinked for a month; it’s as good a use for dead cells as I can think of.
With another pair of not-dead-yet cells providing 2.4 V, it started up again:
That’s a frame from a short video taken in subdued light, just to show it really does work.
The blue LED inside the radome got fainter as the alkaline AA cells faded away, but remained visible in a dark room until the discrete LM3909 circuitry stopped oscillating with the battery at 1.0 V. One of the cells had flatlined, with the other supplying what little current was needed.
The circuitry restarted with a pair of weak alkalines applying 2.4 V across the bus bars:
The LED waveform shows it needs about 2 V:
It’s barely visible in normal room light and strikingly bright at night.
The red LED is actually part of an RGB Piranha, just to see how it compares to an as-yet-unbuilt version with a single red LED in the same package.
The LED drops 1.9 V of the 2.75 V from the mostly used-up AA cells:
The original 33 Ω ballast resistor showed a peak current of 11 mA in a 30 ms pulse:
Replacing it with a 12 Ω resistor boosts the current all the way to 12 mA:
The 2N7000 gate sees a just bit more than 2 V, barely enough to get the poor thing conducting, which makes the ballast resistor mostly decorative. The MOSFET datasheet puts its 1 mA threshold somewhere between 0.8 and 3 V, so it could be worse.
Keep in mind the DSO150’s 1 MΩ input impedance sat in parallel with the 1 MΩ gate pulldown resistor forming the RC differentiator when I measured the gate voltage; I’ll leave the simulation as an exercise for the interested reader. The blinks were noticeably dimmer and perhaps a bit shorter, although eyeballometric calibration is notoriously hard.
The slightly revised schematic-layout doodle stacks the transistors along the negative bus bar:
Flipping the bottom transistor over to snuggle the two timing caps next to each other would eliminate the long jumper wire and probably look better.
Because a yellow / amber LED runs at a lower voltage than blue and green LEDs, it sits atop an astable multivibrator, rather than a discrete LM3909. The battery holder has a pair of carbon-zinc “Extra-Heavy Duty” AAA cells, so corrosion and leakage pose a foreseeable hazard.
The voltage across the 100 Ω LED ballast indicates a 9 mA peak LED current, which is somewhat dim in ordinary room light:
The corresponding LED voltage says the LED runs at 2.1 V for that much current:
Something around 39 Ω should make it more visible.