Biscotti firmware -- trouble, part 3

I am reflecting on what works and what doesn't work, as follows:

• In light 1, moonlight does not work, but other modes all do.
• In light 2, everything works.
• On the breadboard, the pulses look fine in all modes.

And I thought about the DIP on the breadboard. The pulses all look fine, but there is no "load" on the PWM pin. A real flashlight has eight 7135 devices hung on that pin. I tried my first experiment. I found a 260 ohm resistor and put it on pin 6 to ground. With 5 volts this should pull 19 mA. The pins on the ATtiny13 are supposed to source or sink 20 mA. I still see a decent 5 volt pulse with this load on the pin.

Maybe it isn't about some initialization issue with the first mode like I was thinking last night. An experiment would be to replace the usual PWM sequence of 1,7,15, ... with 1, 1, 2, 4, 7, ... as follows:

PROGMEM const uint8_t pwm_values[]  = { 0, 1, 1, 2, 4, 7, 15, 32 };

I know that the PWM value of "7" works. Maybe "1" is just too fast for this light. Will "1" work as the second level after a bump? Will levels 2 or 4 work? I do this experiment. What I see now is that I have to go all the way to 15 to get the light to work. I'll swear that it used to work at 7. I solder a tiny wire onto pin 6 and look at things with the scope:

level 1 (1/255) 440 ns pulse, 1.0 volts
level 2 (2/255) 660 ns pulse, 1.0 volts
level 4 (4/255) 1100 ns pulse, 2.0 volts
level 7 (7/255) 1700 ns pulse, 4.0 volts -- LED not on
level 15 (15/255) 3400 ns pulse, 5.0 volts -- LED on

The 3400 ns pulse is only at 5.0 volts for the last 1100 ns. The first 1300 are spent ramping up to 5.0 volts.

There is no ramp-down. The pulse ends abruptly.

Try another flashlight

Spare driver boards

17mm AMC7135 x 12 x 10 x 8 x 3 Driver 2 Group 8 mode 3 / 5 modes
7135 chips Flashlight Drive Board Mode Memory Anti-reverse

• AliExpress listing

Other ideas

Nothing worked. I pulled the ATtiny off and verified it -- I thought perhaps the heat of soldering might have damaged it. It seems OK, at least it responds properly to the usbASP.

Back on the board it goes again. This time I don't solder the LED wires yet. I just power it up and look at pin 6 with the oscilloscope. Nothing. Vcc on the chip looks good, as does the nominal 1 volt signal on pin 7.

Measuring pin 6 to ground gets 1.0 ohm. That is not good. I see it. A solder bridge to ground from when I removed one of the 7135. I measure 96K to ground after clearing this. Now the scope shows the runt 1 volt pulses on pin 6.

What next? Why not start pulling the remaining 7135 from the "spring side" of the board? I can pull first 2, then see if that removes a bad chip or not -- in a divide and conquer strategy.

Once I reduce the light to 2 chips, I see nice 5 volt pulses. I try adding first one, then another 7135 from the first batch of 4 I removed. Each time I add one, the pulses drop to the 1 volt runt level. And now, after trying 3 chips, I go back to the two that once gave me 5 volt pulses and I am now seeing runts.

Now with 0 chips (no 7135 on the board) I am seeing nice 5 volt pulses. I put one more back and again, runt pulses. I should have stopped and called it good when I had 2 working 7135 in place! Note that I am trying different 7135 each time. Did they all get damaged?

As a final test, with that one 7135 soldered on the board, I take a wire, touch one end to Vcc, the other to pin 6 and the LED comes on (with a constant 350 mA drive). So the 7135 is basically functional.

What next?