Oshpark Projects

I’ve got you now guys. The short answer is yes, LVP can work with multi-cell zener setups. It doesn’t just happen though - we use a voltage divider and that must be setup correctly. You don’t need any math to get the job done, or understanding either actually. You just need to know what to punch into the calculator I mentioned earlier: http://www.raltron.com/cust/tools/voltage_divider.asp

You know the voltage you want to shutdown at the light (2.9v per cell, 3v per cell, whatever... those would be 8.7v and 9v respectively in 3s setups). The stock values used in the STAR firmware will shutdown the light at around 0.5v coming out of the divider, but I do not know the exact value. RBD put some effort into it but I don't recall if we came up with a precise value. 0.513-0.517v may be about right for shutdown (not rampdown). To calculate the resistors for the divider, put in 3 values and the calc will give you the fourth, for example:
Input Voltage - 8.7
R1 - ?
R2 - 4700
Output Voltage - 0.513

That yields an R1 value of 75007.6 Ohms.

If you want to use the calculator to simulate the original Nanjg 105c divider, you must remember one more thing. In that setup there is a protection diode which reduces battery voltage before it hits the divider. You must subtract that diode's Vf value from the input voltage before running the calculation (Vf is in the neighborhood of 0.25v to 0.4v, look at RBD's thread linked above for more info).

RBD or RMM may chime in with more specifics about the actual working resistor values or a more precise "output" voltage than what I wrote. I haven't done much work on this myself.

Wight and I were knocking our heads together over this. On a standard 105C board the reverse polarity diode is in the circuit with the voltage divider so some voltage is dropped across D1, then more across R1 and the remainder across R2. The mcu compares the voltage drop across R2 with an internal reference voltage and does what it’s programmed to do. The ratio of R1 and R2 determine the battery voltage at which the mcu is triggered. With one cell, 19.1k ohms and 4.7k ohms provide the correct trigger voltage. With two cells I came up with ~48k for R1 if I leave R2 at 4.7K. For three cells R1 will be an even larger value.

I checked using 8.7v input, 4700 for R2, and .5v for the trigger and got 77080 ohms.

As you can see from the slight difference between me and wight a higher value resistor will cause the mcu to trigger sooner than a lower value one. The next step is to go to online and see what values are available and pick one. If you are using unprotected cells in series I would recommend erring on the high side.

This is still relatively new ground and not much testing has been done or documented with different value resistors so proceed with caution.

Also, the internal reference voltage of the attiny13A is noted for being only accurate to within 10% and while this may not be a problem with one cell it could make a situation with unbalanced cells even worse. If you go with 3 cells then test the actual battery voltage which the mcu triggers as you might need to adjust R1 up or down in value to compensate. And please post your results. More data is needed.

The Zener mod is a very clever adaptation but it seems a bit of a shoehorn in design. I’ll use it since for now it’s unique for its size/power but I’ll be watching for alternatives.

Thanks! So for this board: mtnelectronics - FET MT-G2, Richard worked that all out for two cells because he says it has low voltage shutoff and rampdown? Does it matter what the LED's driven are (ie: 2 XM-L2's or 1 MT-G2)?

Does he use a different value R1 or did he change the firmware?

Not sure what RMM did but if he says it works then bank on it. 2 XML in series or 1 mtg shouldn’t matter to the board. By changing both R1 and R2 and adjusting the firmware it might be possible to improve the accuracy but that’s a tad beyond me just now.

Ohh - ok, hopefully he will enlighten us mere mortals Smile. Absolutely I believe him, just thought maybe he published the "how" and I missed it. I think his testing is the most thorough around. I'm thinking he worked it out for 2 cells (MT-G2) and 3 cells (Super Shocker). I feel bad bugging him - busy guy to say the least...

Were all here to help. I’ve only really encountered the “shut up and do a search, boob” on that other forum. And since it’s about one of his products he shouldn’t mind. If he did alter the firmware then you probably should not use his values for R1/R2 as a guide for your own mods.

Gosh, and I thought we were still on drivers!

I’m sure there’s a joke in there somewhere about improving your drive :wink:

I’m a knucklehead. Won’t even begin the discussion on early termination here…

By the way, any luck with the Knucklehead? Is that going anywhere? I have a light that seems well made for an MT-G2, with a new breed of that emitter inbound I’m thinking of using my existing Knucklehead in a light with the big boy. It was working on the bench, just have to make sure I sink it good. This will be a piggyback I’m sure, so I’ll just leave the chunk of copper on it and maybe see if I can swing more while I’m at it. Might as well have a working light with it, as compared to the working model driver sitting in a box. If all goes well, this will be in the well built MAXToch SN51. :wink:

I had to suspend work on assembling a heat sink augmented board when my magnifying lamp broke. I’d like to replace it with one that has a larger, more powerful lens Sounds like everyone else is ready to throw in the towel but I just can’t do that yet. This chip needs a good sink attached to the vias under the solder pad and until one is built that way I don’t see how we can determine how much current it can deliver. Your heat sink was good Dale but secured to the top of the case it wouldn’t be nearly as effective as one attached to the solder pad. Also, given that this is a development project I wonder if it makes any sense to start with a slightly larger board and pare it down to size once it’s working properly. A 20 mm driver would still get used anyway. Sort of a knucklehead family of buck drivers like the TinyXX and DD drivers.

Let's have a discussion of the potential pros/cons of this:

It would be a tight squeeze with a full size spring, but a stumpy one like on the 105C would be no problem. MCU, diode, cap, & voltage divider resistors would all go on the battery side, with only the FET & gate/pulldown* resistors on the top. Splitting the components like that would require the fewest number of vias, though I don't know if the traces would be easier or harder because of that.

Reflashing without removing the driver from the pill - WANT!

*I'm wondering if the difference in behavior between the 17/20DD and the SRK is because of the MCU so close to the FET. Maybe putting them on opposite sides of the board would fix that, and do away with the need for the silly gate/pulldown resistors altogether. Is it worth a shot?

Short 5/16” brass pillar for the contact, leave one spring at the switch end to maintain snug battery contact. This would make a small contact pad on the board, plenty of room for electronics, even shifted or skewed sideways a bit for the traces. With a solid pillar for contact, no chance of compressing the positive spring (nonexistent) to cause a short.

@ comfychair-

I love the idea, I wonder tho (ignoring the part about it possible running better with the MCU moved away from the FET, just thinking about flashing with the diver in place) is it really necessary to move it, you wouldn’t be able to use a clip but what are the chances of just adding trace’s & via’s wherever they fit without a total redesign to allow bare [male] header pins to be inserted into 6 via’s from the batt side and flash it that way? It’d take a little more thinking than simply orientating the clip one of two ways but I think that would still be better than having to uninstall the driver each time you need to flash. The one issue would be the off-time cap needing removed but you could install it from the programming via on the bottom.

I don't like a spring on only one end. Far too easy to break contact at the front by bumping the tailcap, even with a fairly mean spring. With a spring at both ends (even when the one at the front is far stiffer and with less travel - or the exact opposite, one spring really soft that compresses fully or nearly so) the cell 'floats' between the two springs and never breaks contact. It works best with one spring really stiff, like the stumpy 105C spring, otherwise if both springs are soft the cell moving around makes the light feel bouncy when you bump it in either direction.

Clip would attach just fine if the board were designed for it, that pic is just an existing 17DD with the attiny plonked down in the approximate location.

This is kinda what I was talking about, leaving the MCU on the top but making each of the 6 needed legs for programming have a via, then you could make a contraption like this and flash it without removing the driver.


(tivo’s pic)

edit It’s not like the via’s would have to go directly where each of the legs are, thats not even possible, some could possibly work but you would need to run traces around to real estate that was open on both sides using a bare driver to properly space the header pins like in the pic.

Centered on the contact pad, these work well

http://www.fasttech.com/p/1145100

...or just put the MCU on the battery side and use a normal SOIC clip. There's no reason the clip wouldn't fit, if the board was designed for the necessary clearance, and you didn't have some funky pill with the driver recessed or using a screw-down retainer. Normal pills where the driver mounts flush with the rear of the pill would be fine. Attach clip, program, screw pill back in.

This is only on the 17/20mm boards, the 15DD is a different animal. There's plenty of room for the MCU, a full-size spring mounted the correct way around would fit but just barely, with a smaller spring pad the same size as the big end of the spring. And if it were programmable with a clip without removing the driver, that would render the convenience of a threaded driver retainer moot, just solder it in and toss the retainer. You can still flash away till the cows come home without messing with unsoldering anything.

In many cases a clip won’t be able to get to the chip. A long stick with a 6-pin header on the end of it can reach down battery tubes and/or past retaining rings.

ICSP is often broken out on a header or pads. Why wouldn’t we do that here? It’s a pretty straightforward proposition and it eliminates squeezing entirely. I don’t see a downside really.

Maybe if we standardized on an ICSP header pinout we could drop the stars forever and just flash.

And any light using the high amperage Tofty is a one spring set-up, spring on the driver. Never ran into a problem. Then again, I don’t go banging my lights around either.

Found a 3’ brass rod at Lowe’s, cut what I need. It’s really not much difference from the super stout cylindrical spring found on Qlites.

Making it fancy with vias for pins for flashing the MCU would probably drop a lot of modders right out of doing it. Me included. I flash the MCU, then build the driver, assemble and move on. Y’all have fun with that. :wink:

With the Moon mode engaged through the Star 2 in STAR firmware, the MCU pin 5 is grounded, can’t flash it. So there would be limitations.

It seems that you are misunderstanding the suggestion. It has no bearing on what you do - you’d just keep doing it! Neither changing the location of the chip (comfychair’s suggestion) nor adding contacts of some type to the board (Cereal-killer’s suggestion) would keep you from programming the chip and then installing it.

Cereal-killer’s suggestion, like comfychair’s suggestion, pertains specifically to programming the chip while in-circuit. Whether it’s the initial program or a change made later, the idea is to make it less trouble to program a chip that’s already on the board.