[WIP] 15mm PAM2803 w/ ATtiny13A rough layout / possibility

Don’t FETs of the size being used has an on resistance similar to that of the feedback resistor (~130 mOhm)?

The intent is to have the driver have a low mode without having it bounce off of the voltage limiter. My line of thinking is that with 2AA on lower modes, the PAM2805 would have to boost the voltage from ~2.4V to something below 3V with the FET off and then only have to boost to the relatively high voltage necessary to run the desired current through the LED when it is being pulsed. The current design has the PAM2805 bounceing off of the voltage limiter when the FET is off, right? That can’t be terribly efficient.

RE: The PAM2805 burning up: I shorted out the sense resistor on a boost driver that looks to be built around a PAM2805 or similar, it did not burn. There may have been some resistance in the circuit, but bypassing the sense resistor bumped current being pulled off of 2 eneloops from 1A to nearly 2A.

Has anyone prototyped this circuit? The basic PAM2803 circuit can be harvested from this cheap driver. Actually, that driver can be connected to any BLF DD driver for prototyping.

Can it be filed to fit a 14mm cavity? I wanted to put one in a brass AA light from Cnq / Banggood.

IIRC I haven’t actually shared this driver yet. I haven’t ordered or built any. Thanks for linking to a cheap donor driver for the parts. Now I’m somewhat more interested!

Maybe if you are very careful. You’ll probably break the GND ring on top and have to rely on the GND ring on the bottom for continuity between the GND vias. I thought that those lights used a 15mm driver?

Hmm, I see what you were going for now. Personally I still think it’s a bad idea. The on-resistance is not a static value, or even “near static” like a passive resistor’s value (which does still change with temperature). On-resistance changes significantly based on both input voltage and load. Unless someone spells it out for me that this idea will reliably function, I am going to assume that it will not reliably function. :frowning: Ah, now I see that you’ve actually done an experiment.

Interesting. In that case, maybe the idea is viable. I still think that it will be prone to burning things. If the PAM280x doesn’t burn instantly it can still burn over time. Of course the next question is also whether anyone actually wants a 2-mode 1xAA / 2xAA driver? (I’m asking, not telling.)

As far as the bouncing off the voltage limiter business… I wouldn’t worry about it. These things have terrible efficiency in all cases I’m pretty sure…

If you PWM the FET, you can still use PWM in theory (in reality, it may fail hilariously.) I’ve tested a shorting of the sense resistor, I haven’t tested long enough to trust it but if the efficiency is still horrible, that point is moot because the whole point of this would be to try to improve efficiency. Are there no efficent boost ICs?

BACK ON TOPIC: At this point, just being available will make this the best boost diver for multimode AA/2AA lights because I don’t think there is a single microcontroller driven boost driver for AA/2AA/CR123 on the market right now.
Would there be any benefit to using a remote mounted toroid choke in this driver?

There are plenty of efficient boost ICs, but not with a 1.2v or 2.4v Vinput.

If nobody sees any problems with this driver as laid out in post #13 then I’ll review for design rule violations and release it as a WIP.

Hi wight,

that’s a nice one. I like the idea of a “Nanjg 110 with modes”. I built some lights with the Nanjg 110 as it seems to be the best of the bunch of 1-cell-boost-drivers and I prefer Eneloop-torches for the kids.

-

Some ideas for the layout:

Why suddenly PAM2805? 2803 and 2805 are not pin compatible and your circuit in post 13 follows mostly the 2803 datasheet circuit, except for the capacitor between Pin-1 and GND. No such with the 2803. The 2805 has its input capacitor there, as Vin for the 2805 is Pin-1.
The output capacitor for 2803 is between Pin-5/Diode/LED+ and GND, so it in fact is at the same position as for the MCU capacitor. This Cap might be disposable, one could as well use a higher value for output capacitor (22yF) as the datasheet suggests for “high power” LED - I think they a referring to 750mA.

Any particular reason for PWM using MCU Pin-5? As Pin-6 is the usual rail for pwm in our firmwares and as we’re not using multiple rails on this one we might just stick to pin-6.

-

I want to do a test setup with a Nanjg 110 and a stripped Nanjg AK47 to see whether the design will work at all.

I have all parts except for the FET. Could someone please give me some specifications for the FET. I have only a faint idea what is needed here as up to now I don’t use FET/DD drivers.

One that I think might fit the pin layout and spec is
IRLML0030TRPbF
Power MOSFET
N-Channel
SOT-23-3
UDS 30V
UGS (th) +/-20V
Id 5.3A (4.3A at 70°)
RDS (on) 27mOhm (40mOhm at 4.5V)

That one ok, or any value that could be better than on this one? Or any other value that has to be observed?
Thanks a lot.

To make part selection easier I made a more detailed partlist according to wight’s design as of now and following the recommendations in the PAM2803 datasheet. Suggestions and corrections are welcome.

PAM2803
SOT-23-6

Inductor
min. 2.2yH, 4.7yH recommended, DC resistance (DCR) as low as possible
4mm x 4mm / 5.4mm x 5mm

Diode
Schottky, min. 2A, Vf as low as possible

Input Capacitor
min. 2.2yF, X5R or X7R recommended
SMD0805

Output Capacitor
min. 6.8yF, 10-22yF recommended, X5R or X7R recommended
SMD0805, perhaps increase to SMD1206 for 10-22yF on a 17mm board

Current Sense Resistor (Rs)
~0.120 Ohm for 750mA (2cell)
SMD0805

ATTiny13A-SSU (MCU)
Package 8S1

Capacitor for the MCU
(might be unnecessary as being in parallel to Cout)
10yF, X5R or X7R
SMD0805

Off-Time-Capacitor (OTC)
1yF, X5R or X7R
SMD0805

MOSFET

This seems to work.

I hooked up a stock Nanjg110 to an ATtiny13a, and a Cmcu, both remainder of a Nanjg AK47. I added an OffTimeCap between Pin-2 and GND. And I used the FET I described 2 posts before. I placed the FET on the pads for a 7135, cut the traces and rewired (better: re-blobbed) for the PWM signal from Pin-6.

I flashed a it with an offtime firmware with 4 modes.
And there was light.

First impressions:

It does work on two tested frequencies:
PWM at 4 kHz (9.6MHz CPU, divider 8, fuse 6a) had a strong whining, but mostly from the magnet on the Eneloop.
PWM at 18 kHz (4.8MHz CPU, divider 1, fuse 75) still produced a slight whining, hmm…

At 18kHz PWM values of 2 - 10 - 40 - 255 resulted in 12mA - 102mA - 198mA - 404mA with a single full and resting Eneloop. Keep in mind that this boost driver by design does not regulate anymore with 1 NiMH cell. See HKJ’s test for the Nanjg110, especially the first graph. So the current does degrade straight from the beginning in all modes. Results with 2 cells will be better for sure, but those later.

The 404mA at PWM 255 is about the same that resulted in driving the bare Nanjg110 on that LED, which I did before I hooked the setup together. So the FET with PWM does not seem to loose much energy here.

So far, so good.

Short version: nanjg 110 is boosting voltage fed to mcu and FET whose gate is pwm controlled by Attiny13A, correct? Could you use a BLF tiny10 DD (or any BLF DD) stacked on the 110 this way?

Short version: Yes, that’s what it does.

BLF DDs: I can’t say, I didn’t follow the development and specs of the DD/FET drivers. But in theory, if you isolate the needed parts and feed them with the 110’s output, perhaps.

So is your test (above) essentially a mockup of wight’s board, or is it a different design altogether? Could wight’s board be mostly populated by stripping a 110 and an AK47?

Basically my test is derived from wight’s idea.
But I went along the datasheet of the PAM2803, as it seems wight has somewhere along the path mixed up the PAM2803 and the PAM2805 (see my comment in post#23), which are not pin-compatible.
The Nanjg110 seems to have the PAM2803, as the board is designed after that datasheet.
So I took

- a Nanjg110

- a stripped AK47 (for ATtiny13A, Cout)

- added 1yF OffTimecap (but the standard STAR firmware values dont work and need to be tweaked)

  • added a IRLML0030TRPbF N-Channel MosFET
    and that was it.

In the meantime I tested the stock Nanjg110 with several other diodes and inductors with interesting results (gaining efficiency with 2 cells and gaining output with 1 cell). Got a last diode today, will post altogether in some days.

I’d be interested to know if there is a better FET that could be used.

Hmm Interesting. I might have to start playing around with this too.

Unfortunately I can’t help with Fet options, I’m a blind bat when it comes to reading spec sheets.

I am not sure but dont these things tend to overheat , anything over 1A the IC’s need to be thermally bonded to the pill… I used to have a bunch of constant current boost drivers for lasers and even for the 2.4a output two of them needed to be paralleled.

here’s a pictorial.

P.S: not my image.

We’re talking about 1A output here, not more.
The PAM2803 is specified for 0.75A output. The Nanjg110 is overdriven already with a lower sense resistor (R100 instead of “R127”) and puts out ~1A in 2AA config. But this seems to be safe ground.
I have several Nanjg110 in household use (bunch of silver L2m with 1AA, 2AA in MiniMaglites) and none of them has died yet.

Any chance of paralleling 2 boost circuits on one pcb and controlling them with one attiny 13A?

I hooked up the full setup (110, ak47, FET) with 2AA and got magic smoke from the PAM2803 :frowning:

The boost controller simply blew up. |(
It occured after one of the first mode changes, I’m not sure from which to which, it happened fast and unexpected to be honest. There seem to be some spikes…
It’s surely the controller that died, I just swapped this single component from a new 110 and everything is working again (with 1AA).

I’ll stick to 1AA for the moment unless anyone has a convincing idea what might be the cause and how to avoid this.

I found a nice diode for the job.

The stock SS14 is incredibly large while being not quite up to the job to boot (only rated for 1A while datasheet recommends 2A). As I looked for 2A diodes - all large as hell - I stumbled upon the BAT60A (datasheet).

Even rated 3A, only 370mV forward voltage, and SOD-323 package. It’s the teeny-tiny component on the right, where the cathode line uses up almost half the component. Oh, and the price… 10.5 ct.

I hooked it up in replacement and in the last days I drained 5x 1AA and 6x 2AA (Eneloops) on this driver, while monitoring input and output current. Works perfectly fine. Even increases output current in 1AA config and efficiency in 2AA config (needs a lower input current to reach the 920 mA). I’m truly impressed.

-

^ Nice discovery. You're doing some real R&D. Thank you for reporting this.

EDIT: So far in my search pretty expensive with shipping. Best I've seen so far is on Ali for $1.06/each.