Got the driver out tonight. This was easy using long-nose pliers with the handles locked open in a drill-press vise.
For those interested in the circuit, it is really simple.
The usual reverse polarity protection diode on the left. From the diode it feeds the MCU on pin 1. The capacitor is for decoupling of the MCU supply voltage. Here it is located after the diode, against BLF wisdom that say it should be before.
The 100k resistor is in parallel with the capacitor. The 30k is across the incoming cell voltage. Not clear, but these probably serve a snubber duty.
The MCU PWM output is on pin 5, which is voltage-divided by the two 15k resistors, before driving the FET gates. In the photo I have bypassed the entry resistor of the divider, marked in green, to drive the FETs with full MCU voltage. Not sure why it was done this way. A 100-220 ohm resistor is more typical in this position. Try at own risk.
I have also moved the black emitter wire to the other side of the 000 pack.
The FETs, prior to modifying the gate drive, had a voltage drop of 130 mV at 4.4 A (30 mohm). The 000 pack had a voltage drop of 32 mV (7 mohm).
After the gate drive modification, the FETs had a voltage drop of 80 mV at 4.5 A (18 mohm).
Voltage drops are the RMS values calculated by an oscilloscope and were also visually confirmed on the ’scope.
Putting everything back together, a fresh 30Q now gives 3.65 A. Before modifications it gave 3.20 A.
Both values are on a cold light (a LED on DD will draw more current as it heats up).
Again, I am not chasing amps. I will upgrade to a FET+1 driver, but for UI reasons (and to get efficient low modes). This was done just out of curiosity.
Finally got mine ordered! I only get paid once a month and I was waiting for the 5a to come available. :party: I’m gearing up to post in the collection thread.
Interesting. Way back in post 2077 I stated that bypassing the 4 “000” resistors netted the most gain. I had the light doing 5.11A at that point. So I guess y’all worked your way back around to that fact by moving the negative lead to the other side of that bank. The 5.11A was also with a spring bypass in the tail. Changing the FET’s to my Vishay variants really made no difference, although they’re top FET’s hand picked for 15mm and 10mm driver builds. Good that we get to the same place taking different routes.
Wish I knew all the how’s and why’s, would probably have forgotten already anyway even if I did.
My tests were at 4.2 V. I measured 1.9 V drive at the FET gates before the modification.
I did not bother to test with lower voltages.
I did check the diode. It is a Schottky (Vf = 0.25 V).
In post 2048 OscarM posted good quality photos that also show two 15k resistors.
I told myself I was going to leave this driver alone, but, well, you know…
So I broke it open and did the two mods listed above. I moved the negative emitter wire to the other side of the resistor bank, right next to the FETs. I also pulled that first 15k resistor and solder-bridged it. Here’s a crappy cell phone shot of the aftermath:
(The other 15k resistor was straighter that that before I started. Works though, so I left it crooked)
This was a super simple mod to do. Anyone with a soldering iron can do it. My results?
mod
Amps (tailcap)
candela
Throw
Stock
3.5
28,100
335
Tailspring replaced/bypassed
?
28,700
339
Driver mods
3.9
32,300
359
All testing done with Samsung 30Q cell. I think those are some pretty good performance numbers from such a small light. I really like this one!
It should be pointed out that I think my dmm reads low. I need to build better testing leads. So don’t focus so much on the hard numbers as much as the change. From a stock light, by bypassing the tailspring and doing the two driver mods, I’ve increased my throw by about 7%. Of course heat builds up quite a bit faster than before too. I also poked around on my driver a bit and everything there read like it was supposed to - no mislabeled resistors or anything like was speculated earlier.
I put the production driver back in, with the changes I’d already made (bridged bank of sense resistors, new mosfets) and measured it both for current and in the lightbox (simultaneously read)…
4.83A for 1431.75 lumens
Then I pulled it back out, removed that 15K resistor and put an 100 Ohm Vishay resistor in it’s place, re-assembled and tested again with a new cell.
4.86A for 1393.8 lumens
I really don’t see that the recommended change is actually benefiting anything.
Well that’s interesting. Wonder how I can see almost half an amp of gain by bridging those pads and you see almost nothing. Could it be because you swapped your FETs? I don’t claim to understand the electronics theory as well as the rest of you guys, but I’m definitely seeing an increase in performance over here.
I don’t know the electronics either, I grew up in a lumber yard, an now a photographer, pretty much scared of electricity truth be told.
I know that the Vishay mosfet’s I put on the driver are the best that can be found. Wight backed me on that as have others. Wight is the one that told me what to look for in the spec sheets. When I added these, the difference was negligible. I got the most gain when I bridged the sense resistor bank.
And for what it’s worth, this driver is exactly as those pictured, 153 pair of resistors, 4 “000” sense resistors, and it had the A009T Mosfets. The spec sheets on my Vishay componenets show lower Rds On, but not by a whole lot. I used Vishay SOT-23 Si2312CDS, a side by side pair right out of the tape.
I believe, if I am understanding Dale correctly, that the ‘000’ resistors were already bridged.
The only change is replacing the 15K resistor (recommended by Sharpie and Del), which had almost no difference. Therefore, the 15K resistor is not the culprit either.
I shoulda known better than to make two changes before testing. That’s just bad science, no matter how you look at it. I can fix it though. I don’t want to fool around with trying to put that little smd resistor back in place, but moving that ground wire back to the other side of the resistor bank is easy enough. So I cracked my light back open and did just that. My results: 3.92 amps at the tailcap.
In other words, on my light, on my workbench, on my meter, I got zero gain by moving that ground wire (equal to bridging that resistor bank) and almost half an amp of gain by replacing that 15k resistor with a solder bridge.
I can’t believe I’m carrying on with this — I swear I’m not trying to argue with Dale about flashlight modding. The guy probably forgot more about flashlights while eating breakfast this morning than I’ve ever known and I have the utmost respect for him and you both Krono, but I’m just not seeing the same results. Again, I don’t claim to understand it, I’m just reporting what I see.
I got my D80 today, I plopped a battery into it and turned it on. It didn’t work. I did the usual paperclip test, and got some light, so that narrowed the problem down to the tail cap. While I was visually inspecting the tailcap the entire contents of the tailcap fell out into my hand. I figured the retaining ring must not have been screwed in properly. While the tailcap was disassembled I took the oppertunity to clean off all contact points in there using isopropyl alcohol, lube the threads, and put it back together, screwing in the retaining ring properly this time.
I screwed on the tailcap on the body, turned it on, and the flashlight worked. However as I was screwing on the tailcap I heard a popping sound. When I unscrewed it again the retaining ring had came off it’s thread and the tailcap components fell out again. I’ve repeated this process twice more. It only happens when screwing it on with a battery inside. It appears that when I screw the tailcap on the pressure from the battery somehow manages to push the retaining ring loose. I’m not using excessive force when screwing it on. I was using a protected battery, 68mm long. I ordered some unprotected 30Qs around the same time that haven’t arrived yet, so hopefully when I get a shorter one the pressure will be less and the problem will be gone.
But still, what should I do about this? Just live with it and hope the shorter battery works? Try to get a new tailcap from gearbest? If the later, how would I go about doing that?
I pulled the brass post on the driver, put a solder blob on it and checked it. Marginal gain. So I wired the emitter, in the light, direct drive to the 35A cell. 4.99A. There’s just not much there to be gained if I was getting 4.83A to 4.86A from the driver. This is top end, pure and simple.
So, I pulled an old XM-L2 off an Aluminum star and put it on a SinkPad, put that in this light and put the brass button back on the driver. Bingo! 5.52A in the light. Same driver, same cell, different emitter.
Edit: Oh yeah, the funny thing. The older emitter pulls 5.52A but only makes 1383.45 lumens. So, we were getting top lumens with lower current draw… higher efficiency all along. Funny, huh? We get all caught up chasing the current when the output is right there in front of us all along.
EditII: I did check direct drive with the Star in the light, it hit 6.11A but wasn’t being pressed down against the emitter shelf with no reflector in place, it started to get hot. But that tells me the driver in sheer top end with this old emitter is costing some half an amp, might be in the 22ga wires on the driver or just running through the circuitry, don’t know. (test leads I used were 20 ga. but about 4” long each.) Still, it’s fully capable of making the power given an emitter that will take it. Maybe I’ll check it powering a triple or the SBT-70…
We are still obviously missing something. Problem is we are talking about milli-ohms here and there, it does not take much to move the readings around when driving a diode with voltage source (as opposed to a current source).
Regarding the 15k resistor:
Looking again at the datasheet of the AO3400 FET (this may not be the FETs we actually have, but the numbers are quite typical), we want the 'on' resistance of the FET to be a low as possible. For this we need the FET gate voltage as high as possible. This why only some FETs are suitable for DD drivers - we have limited voltage available.
With the stock driver, best case (FET not heated), we are operating between the two orange dots, gate voltage at 2 V and down to 1.5 V with the cell drained. We are off the trend line, the purple is just my (optimistic) interpolation.
Bypassing the 1st 15k gets us to operate between the two green dots, 4 V to 3 V.
Dale, do you normally wait for the LED to heat up when taking a current reading?
I tend to take it as fast as possible before the cell sags or when the mcpcb is floating. But I know the LED Vf will go down as it heats, pulling more current. Another variable not everyone is considering.
I agree with you chasing amps is somewhat futile. It is like tuning your car for maximum fuel consumption.
At the end of the day we should be measuring lumens.
Even measuring watts is better than measuring amps, W = V x I. Higher Vf lowers the current draw we can get from a single cell, but it does increase watts (and hopefully lumens) at the same time.
-Sorry, can’t help you, but I got 3 D80’s yesterday and I have the same issue.
I was using old 2400mah protected Trustfires, I thought may work to try the light out. I have unprotected cells that are in the mail, I’ve given up until they arrive. I won’t know until then if I’ll need the longer tubes mentioned in this thread.
I take the reading at start, if I take a delayed reading I also report that.
I went back to the bench and soldered a loop between the driver neg and Star, with the light assembled and no spring bypass it ran 4.67A through this new (old) emitter. (At the emitter) With the spring bypassed, it runs 5.21A at the emitter on a 35A. This is with my Vishay MOSFET’s on the driver and the resistor bank bridged and with the 100 ohm resistor in place of the 15K.
