You noticed that too huh 
You’re not missing anything. The site is super basic. It annoys me a little bit because I think Laen has indicated in the past that the site would be upgraded but nothing ever seems to happen. There are several useful things missing, such as an order-history/re-order function.
Back on the FET thing… now that I think about it, the ATtiny13A datasheet shows it dropping quite a bit of voltage on the outputs. For the PWM pins it’s something in the range of 0.7-1.0v below VCC. If a protection diode is in place VCC is already lower than battery voltage, I forget how much. At least 0.2v lower I think. I posted combined graphs in post #844 showing that while the turn-on for AOD510 is much steeper (good), it also Crap. Nevermind, I was comparing it to the red line, the 70N03. I still see no problem with using the AOD510! Hopefully resistor value tweaks will help.
Well, way back when before we added the gate/pulldown resistors, one thing I found was that bypassing the diode made the FET behave correctly.
edit: this was with the Vishay 70N02, which has issues (on the 17DD/20DD boards only) without the resistors.
Is there a pot that could help find a usable R3 value?
Does that put a slightly higher voltage on the gate?
I assume so. But what keeps defying logic is that you can build a 17DD without the resistors, test it and find it does the crazy mode changing thing, strip all the parts off the PCB and stick them on a SRK-DD board, still without the resistors, and it works perfectly.
Which is another reason I wanna see what happens when you take a 17DD and put the MCU & FET on opposite sides of the PCB.
As I think you’ve speculated before, the discrepancy between SRK-DD and BLF17DD builds is probably a trace length thing. Maybe resistance or inductance changes enough with the trace length or trace layout.
My assumption is that the FET does “something bad” on the gate pin. The AOD510 being a more powerful one may be doing more of that bad thing, so my gut (yeah, that’s what guides my decisions on this…) is saying a higher value resistor. I can only think of 2 “bad things” the FET could do: draw a lot of current (specs says it won’t) or cause a voltage spike. Since the diode suppresses current flowing in reverse, maybe this is what’s happening: The FET causes spikes on the gate pin and without the diode this flows back towards the battery - with the diode in place voltage on the ATtiny’s PWM output pin spikes. The gate resistor could tame that behavior. That’s 100% wild guess.
RBD, I don’t see why you couldn’t use a pot for that.
But if it's that sensitive to the trace length, the number you read from measuring the temporary pot won't be the same as what it would need once it's replaced with a fixed resistor. :~
Maybe so, maybe not. It’s a place to start. That or random value resistors from the bin until one works. Not sure if trace length of highly conductive copper acts the same as highly resistive pot wire.
RBD, comfy was probably pointing out that if you use a pot you’re using hookup wire of some kind to tie it into the circuit in place of a 0805 or 0603 resistor. All that hookup wire could have an effect on inductance, capacitance, and/or resistance…
I was thinking of one of those smallish smd types they use on the charge boards but he’s certainly got a point. I just figure you’ve got to start somewhere. Maybe test it first on an SRK DD where the traces are longer and more similar to the wire length and the board is easier to work on.
This most likely won’t become a shared project, it’s to specific / not enough demand and right now it needs a few tweaks, I got it working but it’s no where near “ready to go” so really I’m just showing off.
This is a set of S->P carrier conversion plates for an olight SR51 WarHawk made for me, he made them perfectly to my spec’s but unfortunately my measurements were a little off on the post hole alignment.
It works well enough for me by if I ever do someone else SR51 I’ll need Brian to make a couple tweaks, I’ll also use soldered in place brass rods instead of the stock aluminum rods.
Note in stock setup the rods don’t carry any current. The top post is +, two outter ones are ground. Running a 17dd driver in my SR51
Design process
AOD510 works beautifully on the SRK-DD with no resistors whatsoever, so unfortunately that's not gonna tell you much.
The way I came up with the 70N02/17DD's 130R resistor was by just trying the common values from 10 ohms up to 200 ohms.
I’d say it tells us plenty! I don’t think you mentioned before that AOD510 worked in the SRK-DD. To me that definitely makes it sound like more of the same problem, which is encouraging.
This is the sort of thing I was thinking of in post #966, although as I said I have little knowledge of the subject:
http://jeelabs.org/2012/12/05/ringing-mosfets/
I’d say this sounds similar to our problem. A larger resistor may be in order. At some point a small gate driver may be a better solution.
Yeah, it's useful info about the problem in general but not for finding a resistor value that'll make it work on the 17DD.
Looks cool, but I don’t understand your comment about the posts not carrying current? Nevermind, I get it now. The black rods are aluminum, and with your mod they carry BAT- back up to the top plate. Got it.
From your testing with the Vishay 70N02, what happens when the resistor value is too high?
The gate resistor raises the minimum PWM level where it will first light the LEDs. The amount of light is the same (say, 100R min PWM is 3, 200R min PWM to get any light is 5 or 6). On the SRK-DD, with no gate resistor, it'll make light with the PWM all the way down at 1. Not good... as shown in your link, the FET takes longer to turn on (and off). If it's necessary then it's necessary, but I'd much prefer to find a way to do it without the resistors. This combo of parts doesn't require resistors, as shown with the SRK-DD and even the half-ass version of sticking the FET onto a Nanjg 105C board. It's not the parts, it's something weird going on with the 17DD's PCB. I don't know if it's trace length or something like the FET being jammed right up against the attiny.
What I did on the SRK-DD with the inductor inline with LED+ to further drop the low mode down to sub-lumen level is different since it's on the output side, and doesn't affect the switching speed of the FET.
I've even thought about trying it with two 17DD boards - one hosting only the Nanjg parts, and another hosting only the FET. But then, the necessary wire between the two boards from pin 6 to gate would be a significant factor and wouldn't really rule out one thing or the other. If it worked using two boards and no resistors, it could be either because of the separation between FET/MCU or from the added length of the connection to the gate. A test that gives ambiguous results is probably not a very good test.