FETs and gate resistors - scope images

This is super info Comfy. Great work.

This is a different 20DD, same 70N02, same build of STAR momentary, but NO gate resistor (replaced with wire, soldered) and no pulldown resistor (left blank). This combo, this exact same driver, did not work using the clicky firmware. It works flawlessly with the momentary FW.

PWM=2

PWM=6

PWM=18

PWM=54

PWM=130

I do not understand what this is, but something tells me it is important!

20DD, 70N02, NO gate resistor (replaced with wire, soldered). When flashed with clicky firmware (either luxdrv or STAR) it occasionally goes goofy when changing modes - mostly when dropping from high to mid, other modes, if you can get to them, seem stable.

This is taken at the FET's gate pin, with everything normal:

Right. We've seen that before. But then...

With the polarity protection diode bypassed, no other changes, driver still running in the same mode as the first pic, not even shut off between the two pics:

How is... I mean... what? With the diode bypassed, all mode changing happens correctly, no more weirdness. Un-short the diode and it goes back to acting flaky. The voltage scale in the first pic is likely correct - yep, 6 volts at the gate when running thru the diode like normal. But bypass the diode and the gate voltage just happens to exactly match the actual Vin. Is this a diode issue, or a capacitor issue?? Where the hell is the extra 2 volts coming from?

That is real weird. Shouldn't the diode be using up some voltage and causing lower voltage to the MCU? I take it you made sure the screen shots were not mixed up?

The waveform with the diode operating normally matches the other ones taken earlier on the same driver (or other drivers while bridging the gate resistor). I saw it change from the top pic to the bottom pic when I grabbed across the diode with the tweezers. Did it multiple times while running the frame capture.

Does the location of the capacitor make a difference? Would it be 'better' on the topside of the board closer to pin #8? It just seems like it's a long way off from the thing it needs to filter.

Way back when these problems first showed up I replaced the Digikey diode with one from a 105C and it showed the same behavior, so I ruled out the diode. What I didn't ever do though was swap in a cap from a 105C.

Same no-resistor driver, low mode (PWM = 18). CH1 is the FET gate, CH2 is Vin measured between the B+ pad & ground ring (so, before the diode):

And this is the same but with a 470uF cap stuck between B+ and GND (also before the diode):

:|

Perfect example why an oscilloscope is such an invaluable tool.
One might have guessed something like that at some point, but seeing it is a quite different thing.
Leads inductance, diode and cap make a nice boost converter, don’t they? :wink:

Now we need one after the diode to check how the controller supply voltage reacts…

What do you want to see on the two channels, before diode & after diode, or after diode & gate pin?

The giant electrolytic cap I'm poking around onto various pads completely at random gives an interesting change when it's placed across the diode, too, but that's not a real-world solution to anything (it's way too big). Would a smaller SMD cap in parallel with the diode accomplish the same thing? I need to figure out a way to try that...

I also swapped our known-value 10uF/16v cap for the unknown-value part that comes on a 105C, no change... I also tried the cap on the top of the board straight from the MCU pin to ground (electrically the same, just different trace lengths) with no change, and then tried the two SMD caps at the same time, one in the original location on the battery side and a second one on top at the MCU, again no change.

Also tried the big electrolytic in the same location as the small SMD cap, no change. But putting it between B+ and GND, or in parallel with the diode, did get rid of the weird overvoltage spikes.

I am assuming here that this spike thing is the cause of the mode-change issues when there's no gate resistor present, since doing anything to eliminate the spike (bypassing the diode, or adding the cap between B+/GND) also eliminated the mode-change issues. Am I at least on the right track? That the solution is something practical that eliminates the voltage spike?

Like you know I also have often this problem, just put a normal nanjg in my t08 because my test driver with some smd FETs in parallel was gone crazy…
I also tried to add caps in some different capacities(sometimes stacked half a dozen), resistors and stuff everywhere but couldn’t get a valid explanation or solution. I am happy that you try to analyze these issues:)
————————————————
If I am right that is our hypothesis:
That the controller behaves sometimes strange and what often can get wiped out with a big cap is in my eyes also a problem with the controller input voltage, because every time the FET switches it is like a short….
The idea behind the diode is that the led can’t discharge the cap behind the diode and so the controller always has enough voltage to work also in the switching phase.
A cap between the bat- and bat+ buffers this also but because it gets discharged through the led with serious amps it needs to be a big one.
Now the only reasonable guess would be that this voltage protection isn’t working as we think.
——————————————————

I just want to see if that diode and cap construct works to smoothen the power supply or if the controller also sees these spikes, because I think that has to be the problem…??? So both would be interesting controller voltage and FET signal, so that we can see if there is a correlation between the switching and the power. Also the before and after diode is interesting to see if it actually works like we think it will.

What has me confused is here:

(same pic from post #18 - trigger here is on the rising edge of CH1 (gate signal) at 1.00v, that's close to the FET's threshold voltage)

I can (mostly) understand the spike on CH2 after the PWM turns off. What I don't get is how the gate voltage rises to ~6v, and stays there, while the supply to the MCU is still down at battery voltage. If they tracked together more closely it wouldn't be much of a mystery. The ~6v being generated when it turns off, sure. Removing the load, it releases stored energy in the cap, and voltage spikes. But it's doing a constant, rock steady 6v output to the gate pin while input to the MCU is still down at ~3.9v.

These spikes always are generated when switching an inductance, all these switching and the diode+cap leads to the high voltage. Every spike gets over the diode to charge the cap at a higher voltage but never gets discharged down to the supply voltage again. You have to imagine there are a lot of these spikes and the controller don’t use much power so the voltage don’t breaks in.
Like someone said its a boost circuit.
Ch2 is not input of mcu because there is a diode between your measuring point and the controller and the diode prevents current flowing in that direction. If you measure the controller voltage I am sure we will see the 6V supply…

Maybe we have a overvoltage problem and no undervoltage problem? Do these problems also occur on Zener modded nanjgs?

I haven't looked at any of this on a zener-modded driver yet, I don't have one built at the moment...

Do you have a prediction as to what the scope will show with a 10uF SMD cap in parallel with the polarity diode?

The parallel cap sees the voltage spike on both legs so it shouldn’t get so charged up like the stock one. Only the diode forward voltage should be between the two legs. I guess that compensates the spike and stabilize because there is then a cap in both directions: one from controller+ to GND and one from controller+ to V. So the “negative” side of the cap is the V side….
But this are only unfunded fantasies….

Have you never used a Zener modded FET driver? Never thought about that but now I need to go the the soldering iron….and test some things.

Atmel rates the maximum Voltage for the tiny at 5.5V.

Yes, I've built lots of them, I just don't have one put together right now. Have plenty of parts on hand to build them.

Torchlite SVD7 MTG2 built for DayLighter:

Another 10uF/16v cap placed at D1, with the diode piggybacked:

CH1 = gate, CH2 = VBAT

CH1 = gate, CH2 = MCU's Vcc, pin 8:

The weird behavior of some drivers without a gate resistor, where the MCU seems to shut down right after changing modes, is likely due to the overvoltage to the MCU from that spike. It's doing exactly what it seems like, it's shutting itself down! It sees that ~6v and says 'whoa dude, I need a little nap - wake me up when shit gets back to normal.'

So how did adding the stupid gate resistor cover up this issue - just by reducing the load the MCU has to drive, therefore reducing the current draw through the diode, which reduced the tendency to create the voltage spike?

So by putting in the cap, the load resistor can go away?