Attiny25/45/85 FW Development Thread

TOM E, did you try adding a 10uF cap across vcc and gnd? In an investigation thread regarding the FET induced spikes that was the preferred solution.

Ohhh? Shoot, missed that. It was indicated to me to keep the cap as small as possible, interesting… I’ll try that tonight. thanks!! Do you have a thread link, or search criteria I could use to find that thread?

I do, actualy i have a builded driver with a Narsil firmware and initial test show its working but i am quite busy lately at work, family and those hunting flashlights

BTW Tom, what had happened with that Y3-like firmware? Did you finnished it?

Oh boy, refresh my memory… Y3 like? I know I’m behind on many fronts. Work has been, and is goin nuts.

Hi Tom, I could probably do more to help. I have the test equipment, just limited time and still using 13As exclusively.

As discussed last time, you should not need more than 1 uF for local decoupling (the C spanning pins 4 and 8). So 10-20 uF in the C1 spot and an additional 1 uF directly across pins 4-8 should do the trick. It really is not normal for the 13A to work as well as it does without proper decoupling.

I just recently set up a ‘test-bench’ with an A6 driver in the open. The di/dt spiking/ringing on the FET drain pin is impressive. A second improvement you can try is to add 100-200 ohm resistance between the MCU output pin and the FET gate. This will slow the FET switching and reduce the inductive ringing. Depending on the driver PCB layout you will probably have to cut and scrape a track to fit this resistor.

Ohh, the resistor has to be in-line then on that trace. Hhmm - I usually use a 10K-12K resistor on the FET gate, but the resistor goes to ground, so I’m sure not the same effect as what you have proposed. Ok, I’ll see what I have, and see if I can do something like that. Think you might have told me the same thing, or something similar wayy back, but I’ve been using the caps successfully up til now, but on the wight FET+1 boards, not the PD68 Double Down’s.

Actually got some 0402 10K resistors, and am using that on this board for the FET gate to grnd resistor because it fits better to the FET pins. Maybe I should remove it for testing going forward — it’s a real PIA anyway to solder it. Basically, can’t see what I’m doing when I solder, and only with a magnifier afterwards can I inspect it. Gonna go back to 0805’s.

I think this tread should be made sticky, I think lot of people (building FET drivers) have missed this

BTW, from my own research gate and gate pulldown resistors increase inefficiency of the FET by increasing time required to turn it on. Only after FET has been turned on, it has low resistance&heat disipation.

In the solution from the above thread, only a new 10uC cap across +Batt and GND was needed. I think the old Cap & gate resistors are to be removed. Check it out, very interesting thread.

The gate pull down resistor is a good thing normally, a FET gate should not be left ‘flapping in the wind’. One of those components left out for simplicity and PCB size limitations. It may help for the flash some get when going from full-on to moon mode or similar.

True, but the idea is just to take the edge off, not to dampen it so much that the FET is operated any significant time in the linear region. It does put a limit on maximum sensible PWM frequency to use. See the recent BLF D80 driver for an example of how not to do it.

These things are like the suspension on a car. Too hard, or too soft, and you are bound to have a bad time depending on the road you are trying to follow.

EDIT: Thank you for that thread link, but probably the thread discussion was around tiny13a drivers? Seems the 13a is much more forgiving than the bigger MCUs. Local decoupling on any chip like an MCU is basic stuff, it should not be neglected like it has been.

K, spent some time reading thru the comfy FET scope images thread. Post #44 seems to be the result (??) - 10 uF cap in parallel with the diode. I gotta try this... This is old stuff - I followed the thread at the time, somewhat. With all the pics back, it's once again useful info.

Dunno.... Really not sure if I can understand it all, and if there is a final solution from it. dave_'s posts seem impressive/correct, but no real solution posted as a result?

Just fyi, the 10K-12K 0805 resistor I use on the FET gate (to grnd) always works!! No loss of output, definitely gets rid of the flicker switching to moon mode. I've experienced those flicker problems in lights, tore them apart - added the resistor, and then flicker is gone!

Update: removed the 0402 10K resistor, put a 0805 12K in it's place, and with the stacked 10 uF cap, the mode changing problems all seem to have gone away! Dont have any other caps or other extra resistors, just the two 10 uf caps on C1. Testing was on the bench - installing it now in the JM70 light.

I have some of these 22uF Caps ready to try, rather than stacking the 10’s, but don’t currently have a driver that is playing up to test one out.

Regarding the inductive ringing at the off edge of the PWM or strobe pulses, this is what we are dealing with:

Yellow trace is the voltage at the drain of the FET (so 0 V = FET full on, 0 V is at the second division from the bottom).
Blue trace is the voltage at pin 8 of the MCU, same scale.
This an A6 driver from Banggood driving one XM-L2, using a fresh 30Q cell.

We can see some boosting of the pin 8 voltage.
Note the FET drain peaks at > 6 V at switch-off.
Also note the frequency of the oscillation, around 10 MHz. We’re in HF radio frequency band.

Now lets shorten the wires between the emitter and the driver from 4” to 1”:

Frequency of the oscillation goes up, but peak goes down below 4 V - as can be expected with the lower inductance of the shorter wiring.
I know adding a gate resistor will tame that oscillation, but I did not want to scrap the driver (yet), so tried a drain-source snubber on the FET:

The snubber consists of a 0.1 uF capacitor in series with a 4.7 ohm resistor. Not calculated and for sure not optimum values, but a good improvement.
Something you might want to try Tom, if all else fails. For the test I used regular TH components with the leads cut short. One leg of the snubber capacitor on the black wire pad, one leg of the snubber resistor on the tab of the 7135 (better option would be the source pins of the FET).

The voltage at pin 8 of the MCU looked surprisingly good, all things considered. I would not use more than 0.1 uF between pins 4 and 8.

Emitter wire length also makes a big difference, so keep them short. This is also with the snubber, but with the 4” emitter wires:

The assembled light is working pretty well, as described earlier: 85, 2 uF caps in parallel on C1,12K FET gate resistor to ground.
The ThorFire JM07 hits 1900 lumens at start with an XM-L2 U4 1C, 56 kcd, UCLp lens, spring bypass’s, on a BASEN IMR 4500 26650. Running Narsil of course, but not sure yet if everything is ok - seems the clicky switch is not working for mode changing, though it seemed to work fine on the bench — no OTC cap, but seemed to work on other builds even without enabling brown-out in the fuses. Also memory seems to be turning on, even though it’s turned off - not sure if this is a straight bug or not, but never seen this before.

What you have there looks good - got to study this more. Hhmm, don’t even have the 0.1 uF cap across pins 4 and 8 now…

You just answered months worth of questions, speculation, and argument. Seriously, this is extremely helpful information.

Also, I didn’t realize that my long emitter wires were making the problem worse. Oops. I just thought it might reduce overall output by a tiny bit while providing much easier access for flashing; didn’t realize it would actually worsen the voltage spike.

A while ago I managed to figure out that the worst spike was on the trailing edge of a FET pulse, by slowing a driver down so far that I could see the individual pulses… but I, um, didn’t realize that would also make the driver un-flashable without special hardware. Oops. I should probably get myself a fuse resetter or high-voltage serial programmer.

Anyway, your tests just now settled quite a few open questions and the scope images make it much easier to understand. :slight_smile:

Looked this over more carefully. There's a lot to this I'm either learning, or not quite understanding - I don't have a good electronics foundation, but Wikipedia has just helped me out a little (now I know the "drain" of a FET is the output).

The length of the LED wires, and the location of the snubber is a surprise to me - in my limited knowledge, once the signal goes out the FET, you are done - why would it have any effect at this point? But apparently it does.

The other thing I don't understand was I thought our main concern is the MCU pin 8 input - spikes, voltage boosts causing MCU resets and strange behavior, etc. But in the scope images, it looks like the blue trace is pretty much unaffected?? So, I don't understand, but I'm guessing the problems I'm seeing are more about the FET and not the MCU?

In this particular light, and all I work on, I keep the LED wires at their minimum. Only lights they have to be long in are SRK or multi-LED types that have a screwed down reflector, so you have to have extra lengths of wire to keep the driver dangling when screwing down the reflector from inside the pill area.

TK, happy to help! :slight_smile:

Tom, pin 8 is important. Every time we switch the FET on we basically short pin 6 to ground (momentarily, while the FET gate charges up), This current is limited by the output impedance of the pin 6, but probably peaks at around ~50 mA. This current has to come from somewhere, and somewhere very close by, otherwise the voltage at pin 8 will dip. Enter the local decoupling capacitor. Which we do not have in the current design.

But, as you say, pin 8 looks surprisingly good under the circumstances, for the 13A anyway… And the corruption mechanism seems to be over-voltage, not under-voltage.
There is also the possibility of EM interference. AC currents tend to induce currents in neighboring conductors, look at that 10 MHz oscillation.

Can you try the snubber on a light that is troublesome?
Exact values are not important. Optimum values will depend greatly on the exact layout, wire lengths, etc.
Try a C of 0.1 uF and R of 2.2 to 10 ohm.
This was my crude arrangement, leads should have been shorter on the snubber components:
(As mentioned above, a better arrangement would be to have the snubber span the FET. Easy to do with SMD parts, but I do not have big selection of values in SMD.)

OK, bent up the gate leg of the FET and wired in a gate resistor.

End result is better than with the snubber. Turn-on time is 5 times slower, 500 ns vs. 100 ns. So PWM faster than ~4 kHz will not turn the emitter on reliably at 1/255 PWM duty.

From yesterday, basic A6 driver, short emitter wires:
(Yellow = FET drain, Blue = MCU pin 8)

From yesterday, with RC snubber added across FET:

Now with Rgate = 100 ohm added, and snubber removed:
(Purple trace is the FET gate voltage.)

Turn-on without Rgate:

Turn-on with Rgate:
(Note time scale is 500 ns/div, compared to 100 ns/div above).

Sorry, should have added above, the top three scope shots are of the turn-off with various configurations.

EDIT: Photo of the crude test setup with the gate resistor in place:

Well, I have to create a trouble-some light at this point. I got my JM07's driver out now - I have the 12K resistor on the gate and dobled up C1 cap and it works fine.

Should I remove the extra cap and see if the problems come back? If so, then try it with leaving the 12K resistor on?

This is what it looks like:

Might be better starting with a clean driver. I got to reflow up a few more anyways... What do you think?