maybe by using a small transistor instead a diode, so you got reverse blocking but a very low voltage drop
New theory… Is it possible that the capacitors are feeding the main driver instead of the tailcap (when the load from the driver is low)? Perhaps I need to put diodes between the capacitors and ground?
Edit: nope! I tried just removing both capacitors. I still get the same behavior on Li-Ion. On NiMH, the tailcap fails to light up at all, and the light strobes when the switch is off. So it seems like there’s voltage leakage (?) through the TPS61221 to ground. Would a diode at the ground pin address that?
Thank you for sharing your results. I wish I could be more helpful. I would be experimenting just like you are if I was in your shoes. When I looked at the datasheet for the TPS61221, I saw no mention of built in reverse polarity protection. So I'm assuming the chip is allowing current to flow through it backwards. Lexel's idea about a small transistor sounds great. Maybe google what they look like and see if you have any old stuff you can harvest one from.
Could you describe the strange behaviour of the main LED and the tail LEDs on clicks a bit more precise?
Schoki, if a picture is worth 1000 words, what’s a video worth?
This is where a feel a bit helpless. I understand a few core principles, but that’s about it. The extent of my electronics education is what I got from a breadboard kit when I was 10 or so. I appreciate any help you can provide!
Another theory I had today… in a “normal” illuminated tailcap, there’s voltage drop in the loop due to the tailcap LEDs. But in my boost circuit the VIN to GND path likely doesn’t have much drop - is that providing a suitable path for driving the main LED since there is negligible voltage drop?
On a lighter note, I put one of my Big Switch boards into a Convoy L2 tonight. Having the bottom of the board bare made the installation a breeze; no hacking away at things to make clearance for the resistors on the bottom of the board.
Red, White, & Blue - Happy Independence Day, America!
I’m wondering if you have tried changing the bleed and led resistors. I had a light that had the same behavior until I tried change the resistors to get a value that worked. The led’s in the tail look very bright to me, but I like mine very low anyway so they don’t light the bedroom up at night.
I usually set mine around 50 to 82k, probably a little low for some. You might try changing resistors.
I do understand your using a boost ic, don’t know if this will help or not just a thought.
I’ve tried 680 and 300 ohm bleeders so far, didn’t make any difference. I haven’t tried different tailcap resistors. I’m not against that, but hopefully this will be flexible enough where that wouldn’t matter much.
First, I want to say that the voltage regulator on the tail capis a really good idea. I hope we all can see more little projects like this. And you made a very good video! Explains all the little details (especially the dimming on the LiIon cell).
Now I will make a really technical theory that could explain the problem. But don’t think that if it’s me pointing out a problem, that it’s 100% correct. I could be wrong as well, i’m just guessing here:
It’s clear that there is current flowing from the boost circuit input to ground. And when you go higher in current, it starts working. There’s enough current flowing through the TPS61221 for the low modes, but not for the higher ones.
The only pin possible is the L pin.
When you have your light off, and screw the tail on, the tail starts lighting up. Then you click, and the main LED goes on, the tail goes off (all normal behaviour). Now if you then click, the main LED won’t go off. I think this has to do with the way the IC controls the boost operation. The TPS61221 has a so called “hysteretic current mode” control circuit (variable frequency control). It keeps the current ripple on the inductor at 200mA, and offsets the current ripple according to the load. This way the voltage stays constant. Now that’s a problem in your situation. It simply won’t reach the 200mA to start switching when you click in the low mode. This is why the lamp stays on. The IC tries to charge the inductor until the current reaches 200mA (by turing on the lower MOSFET from the inductor to ground —> current flow keeping the LED on), but it never reaches 200mA. The current keeps flowing and the main LED stays on.
On the higher modes it works, not sure why though. That’s why I’m not sure. It should be the same result all the time with my theory.
I don’t understand the exact coherence (is that the right word to use here?) of the mode switching and the 200mA hysteresis right now, but I think this is the reason.
I’ll try to completely understand though.
To be honest, I’m impressed the boost circuit works in the first place, it should never reach 200mA ripple because of the bleeder. On startup, the boost IC isn’t controlled with the hysteresis, but with a special startup controller with limited switching current (look at Figure 20 in the datasheet)
Nice work :+1:
Does the input capacitor help in it attaining startup? In my testing, I tried removing the input capacitor. No real change on Li-Ion. On NiMH it failed to startup in any mode… wouldn’t work at all.
What if a different IC was identified that had a lower ripple current? That’s the same as “switching current”, right? It looks like the Ricoh R1210N331D-TR-FE (3.3V version) has a 100mA switching current rating.
Ripple current is the difference between the highest current and the lowest current flowing through the coil (the delta current).
switching current is either the average current, or the absolute maximum current flowing through the inductor, depending on the subject you’re talking about (from x axis to the average, or the highest point)
The Ricoh could work, but I would keep looking for some alternatives, I’m sure there are better options. Does it have to be a package with leads, or is a no-lead package okay as well?
Thanks for the explanation. I’m good with a no-lead package as I’ll be hot-plate reflowing them. I’m definitely open to suggestions if you care to join the search!
Another thought… If I could get the input capacitor to hold it’s charge when the flashlight is turned on, perhaps it could be used to handle the ripple during startup as the flashlight is switched off? If hooked up correctly, placing a transistor between the input cap and VIN could do this, right?
Another theory… some (most?) of these chips will pass current through the chip if the Chip Enable pin hasn’t reached the turn-on voltage (0.5V?). The TPS61221 is this way (“Pass-Through Function during Shutdown”). Perhaps on the flashlight’s low mode, the voltage reaching the PMIC is low enough that it’s not tripping Chip Enable and is instead just passing current through the IC.
Some PMIC’s have a “Load Disconnect” feature that essentially disconnects the IC if the Chip Enable is too low. Maybe that would force the main driver to shut down, giving the tailcap an opportunity to start up? I see TPS61261 has Load Disconnect (also something called “soft start” that may be helpful?), but unfortunately it’s only rated up to 4.0V in.
Edit: I just tested voltage being leaked through the tailcap. It’s 0.40V on NiMH and 0.42 on Li-Ion, just below the Chip Enable requirement. I think I may finally be onto something!
Excellent! 
:+1:
Ok, potential candidate(s) has been found: the TPS610993 (3.0V out), TPS610994 (3.3V out), or TPS610995 (3.6V out). The TPS61099x series appears to be great for ultra-low loads and features both Load Disconnect and Soft Start. And available at Arrow in 1pc increments for $1.34.
My quandary now… the output voltage is regulated up to VIN >= VOUT + 0.5V. Once VIN is 0.5V over output, it switches to a Pass-Through mode where VOUT correlates to VIN*resistance. Going with a lower output chip like the 3.0V TPS610993 will likely gain marginal efficiencies on NiMH, but it will only regulate Li-Ion at 3.5V and below - anything above that is just VIN*resistance.
Honestly though, according to the datasheet, the efficiency of input = 1.5V and 0.1mA current is ~88% for the 3.0V chip vs 86% efficiency for the 3.6V chip. It might be worth the loss of efficiency to gain the extra consistency in output regulation for NiMH vs Li-Ion and just go for the 3.6V output chip (TPS610995).
The only real downside I see is that this is a BGA (ball-grid array) package which will require reflowing - no hand soldering. 
You can get it in a WSON package; Arrow doesn’t carry it but Mouser does.
PS - sorry if it seems like I’m talking to myself. Sometimes I feel that just writing down my ideas helps out a bit… and gives you an insight into the wild@$$ thoughts running through my head.
^
I'm digging the way you and other contributors are posting. I'm learning and it might trigger a helpful idea in someone out there. This is a pretty cool thread.
I don’t understand most of it but I agree with you ImA4Wheelr!