But why does it work when you stick the same FET onto a 105C, using all the 105C's original parts? Is it possible the original cap isn't 10uF? There's nothing different about the new PCB, I've checked everything until I'm crosseyed.
Also there's nothing higher than 10uF in 0805 & X7R. 15 & 22uF 0805s only exist in JB, X5R, & X6S.
Is there any difference in the firmware between working and non working drivers?
Take a break, attack it from a different direction in a bit man
At some point for me it comes down to a fresh look after a good nights sleep. Or a big hammer and then a good nights sleep and fresh parts in the morning.
Nope, exact same build of luxdrv I used on the Nanjg-092 boards, same batch file with the same fuses and everything.
Try to add a cap directly to the Vcc of the MCU ( pin 8 ) and see if this can help.
It works properly with STAR v1.1 clicky version. luxdrv still didn't work even with a second 10uF in parallel.
I really have no clue what is going on.
And, I certainly know the value of a strategically-wielded hammer...
|(
Does STAR_mom V1.0 work on it?
is the PWM compile different from STAR V1.1 than luxdrv?
To-morrow will be a better day.
If the code is not an issue then is it possible for it to be corrupted during the flashing process? What I mean is, Is the flashing process foolproof. If the hardware is right and the traces are right and the soldering is right then where else can you look? Could it be a faulty mcu even?
If it works with one, but not the other, it’s not the hardware
And when flashing, it writes, then reads, to verify…if there is a problem it will tell you
Could it be a layout issue? I mean everything is connected correctly, but perhaps the comparatively longer traces (given the size of the board) required to get everything on the PCB are screwing something up?
I have even done an A-B-A test, luxdrv (no worky), STAR v1.1 (worky), back to luxdrv (no worky again). I re-reflashed an older Nanjg-092 driver that's been running the same build of luxdrv, and it still worked just the same as it did before. So, that build of luxdrv works on the 105C board+FET, but not on the new board.
I just browsed over the the last pages, so I didn’t get all details, but: is it right that that reset pin on the attiny is floating?
If yes, please try tying it to vcc with a 4.7k …10k resistor. That can easily cause such strange behaviour.
The thing is every board I have ever created with an ATINY chip (13 or 25 for me) has been setup exactly the same way. I have never experienced any of the problems that are being described here. Not saying it's right or wrong, just saying it's weird that Dale/Comfy are having issues. Happy to pass my BRD files onto anyone who wants to double check my work; it's entirely possible that I've made some stupid mistake.
- Matt
I will say this, on my joule thief build the .brd file I uploaded to OSHPark I got from alxsmith, some of the traces just went away, he was using an older version of eagle, but he fixed it with the latest…I think you said you were running 6.5.0 right?
Sure, no doubt it might run forever without any problems that way. 
Thing is, the reset input is not designed or supposed to be left floating. It can change it’s potential through coupled-in noise through parasitic capacitance or inductance on the board, esp. in high current switching environments like those drivers. Even some dirt or flux residue can cause troubles. That’s why one should tie it to a potential like vcc (but with a high enough pullup resistor that the programmer still can pull it down. )
I would strongly suggest to implement that in future designs, even if it is not the problem in this case.
It very well could be the problem (that random blinking stuff could be resets), a workaround costs a few cents at best, so I would say it is worth a try. If you can exclude it from the list of possible causes, it was worth the 5min work I think. 
This may be a long shot but have you tried checking the board for shorts? At this point I may also try measuring resistance of traces vs. the 105C. I think that dave_ may also be onto something.
I have never needed an external pull-up on the reset pin, the internal pull-up sufficed. However, in noisy circuits, an external pull-up (typically 10k, maybe even only 4.7k, but not less than that) is often recommended, because the internal one is quite weak (~50k). My guess is that the FET switches much quicker than the AMC7135, which indeed implies more noise. It might help to check the reset pin with an oscilloscope.
A resistor between MCU and gate is also often recommended.
I'm not into DD, so haven't tried.