Based on that circuit simulation, the LED Stress Tester schematic looks about like you’d expect:
[Update: Left out the Schottky diode that makes the 20% duty cycle actually work. Drat & similar remarks.]
The manual wiring turned into a hairball, but from the top it looks pretty good:
The 20 pin DIP IC sockets provide spare contacts, so that ruining a few by jamming fat LED leads into them won’t be a tragedy. Each LED string uses one of three adjacent contacts, which left room for a fourth string of amber LEDs that are, even to the naked eyeball, nearly indistinguishable from the red LEDs.
The 555 timer output waveform looks just like the simulation:
The trimpots sit near the middle of their rotations, which is always comforting. The duty cycle trimpot can’t quite get down to 1 ms, which doesn’t matter right now.
This scope shot shows the total forward drop across the three LED strings, with V=0 offset way down below the bottom of the display:
That voltage includes the IRLZ14 MOSFET drain-source voltage, which amounts to a bit less than the thickness of the fuzz on the traces. In round numbers:
VDS = 400 mA x 0.100 mΩ = 40 mV
That’s based on this measurement from the MOSFET tester a while back:
You could argue the drain voltage is closer to 60 mV. I’d argue that the overall accuracy of all these measurements leaves a lot to be desired; we’re in the right ballpark no matter what.
From the top, the three traces show LED groups 7-9, 1-3, and 4-6 in exactly the predicted order (1-3: 6.445, 4-6: 6.372, and 7-9:6.469 V), if not with exactly the predicted absolute voltages.
Part of the reason may be that the current limiting resistors that produced about 100 mA were 5.6 Ω, rather than the predicted 10 Ω, They actually measure about 5.7 Ω and the forward drop (from that scope shot) is around 750 mV, so the current could be up around 130 mA: a bit hot. I want to measure the current more closely before leaping to any conclusions.
The 1/4 W ballast resistors dissipate 100 mW peak / 20 mW average and each LED dissipates 300 mW peak and 60 mW average.
The 7.5 V wall wart I planned to use requires a much higher average load for good regulation (it emits 10.5 V for light loads), so this lashup runs from that 2 A bench supply through the other end of the Tek banana cable I hacked apart to make those SMD tweezers. The supply voltage at the coaxial jack drops by about 120 mV during the pulse, but we’re dealing with measurements up from ground.
I think the exponential curve in that scope shot shows the LED internal temperature rise during the pulse. If you figure -2 mV/°C (based on the ever-reliable and always accurate Wikipedia), then the 150 mV change along the exponential works out to 50 mV per LED and a 25 °C temperature rise. I have no idea whether thermal-cycling the LEDs at 100 Hz will cause early bond wire failure or not, which is why I want to let this run for a month or so.
I love it when a plan comes together…