Toroids and MOSFETs and firmware, oh my! (aka:'Tiny13+N-FET=???')

That's right. You can get it from eBay in a 10-pack here or from me individually here.

Yes, the same cap stays. LED+ is wired the same way. LED- is soldered to the big flat tab (drain).

Still very much a work in progress, though the bits and pieces are working well together. There will be dedicated boards finished up soon that will avoid all the kludginess of what's been done so far.

46mm (not the final design! - we still need to pick an inductor and tweak the BAT+ & LED+ pads to accept it):

17mm:

Ohhh, okay - I'm excited . So, you can get a true moonlight level (totally agree the Nanjg 5 PWM value is too much), 6A+ on 100% (I got 5.6A on Tivo532's combo driver with 2 FET's - probably could do better). Didn't realize you were selling these FET's. But how can that FET be sanded down like comfy shows and still work? I see space is tight there. Can you get a one sided (2-4 7135) programmable driver and do the same thing?? I programmed these for example: fasttech 4x7135, but now that I look at them, may be a problem fitting the FET unless you got space to sit it vertically or something.

I didn't sand mine, but I mounted it a bit differently. The moonlight still isn't as low with a PWM value of 1 as I can get with 7135s using the same program.

The FET is just glued down upside down on the driver in #152. There's still room for the programming clip to latch on, and about 1mm clearance between the filed-off corner of the FET and the ID of a typical 17mm pill.

There's going to be more room on the 17mm 'super-092' board, since the MCU is rotated 90* like on the east-092.

comfy - can these new design boards be downloaded via a cable from the battery side, like Tivo's Z8 boards? Can that be accommodated in the layout? Not sure how many pins, offhand, are needed... The Z8 can be done with 4 thru holes, maybe the Tiny13A needs 6? Big issue for me (software guy) is the limit of 1K code in a Tiny13A . Think e-switch is the gonna be the way to go (SRK's, L4, etc.), and you can do a lot more nice software with an e-switch UI, in my opinion.

OMG! My resistor mod'ed M6 went accidentally on in my notebook case driving home from work today -- freak'n thing burnt my finger trying to shut the PIA thing off at the switch -- Yes, hold for 3 secs please, or cycle thru all modes!! Measured 170F minutes later, took over an hour to cool off enough to hold for a sec. Had 4 fresh Sam 20Q's in there that measured 3,400 lumens @30 secs this morn - they drained to 3.26v and were very warm.

So yes!! I need a better driver like what can be done with this design, maybe a not-so-easy side switch, smarter user (me), but less power is unacceptable...

Mattaus says there's no more space left on the 17mm, it's already lost any dead space the east-092 had by switching from 0603-size parts to the 105C's 0805 stuff. And, it has pads for grounding the MCU pins same as the stars on the back of the 105C. I don't know of any firmware that uses both the stars and the phase correct PWM needed with the FET, but the board will be ready for it when somebody gets around to writing one.

Say you wanted a typical "daily driver" with the option of several settings

You could always put on a FET that maxes at 3A

http://www.digikey.com/product-detail/en/FQD4N20TM/FQD4N20TMFSCT-ND/3478406

Or if you wanted a mama jama to throw in your SRK or make your emitter cry for mamma, go with higher current ones (like their 60A ones above!!!!)...they have a laundry list of TO-252 style packaged FETS on most electronics outlet locations

Plus once the ATtiny13A version get's rolling...there is always the ATtiny58 with 8K of memory!

"60A" means nothing here, it's still limited by the battery voltage. If the cell connected straight to the LED won't do more than 6A, no FET in the world will be able to do more. That's also why the FET's Rds(on) is so critical in a single-cell driver. You could use a FET rated for 10,000A and it still wouldn't do more than what it does without the FET in the circuit.

It’s a way to have virtually direct drive high but with modes as well without having to slave umpteen boards together. The downside is that in parallel led set ups you lose the ability to control the current to each led. Not the end all solution but it still fills an empty driver nitch.

7.52A from a 20R!

A parallel triple - that I understand. Because of the lower Vf per LED, you can achieve higher amps out of the cell with a FET based driver. Try it now with one XP-G2 and you will probably be somewhat disappointed, maybe 5.5-6A tops.

I have a question for the electronics gurus. Please excuse, but correct, any errors I have in the premise.

These n-channel FET's are basically variable current switches. We are using PWM to control how much current they allow to pass from the cells to the LED's. Voltage is not regulated like with an amc7135. So we obviously only want to use this approach (assuming no buck regulation elsewhere in the driver) for situations were the LED's can take a direct connection to the cells (Made possible by copper and voltage sag). This would seem to rule out 4AA Eneloops to one xml.

My question is. In the case of one cell xml, or 2S mtg2, in the lower modes, are we subjecting the LED's to higher than Vf pulses (assuming less voltage sag) and, if yes, will this significantly shorten emitter life?

7135's limit amps, not regulate amps. I get the wording of this messed up too, but understand it now (I think I understand...). Sorry, can't help with the Vf pulse question.... Good question though!

PWM's are bad in the sense they cycle 100% of the amps, so the LED still gets full amps, but just for short bursts, so true multi-mode regulation (each mode is regulated) is more desireable, but probably more costly in parts, real estate, and driver design. We do have lights and drivers though that provide true "per mode" regulation, but you will find it in more pricey, bigger drivers and lights.

I agree Tom, 7135's limit current, but they do regulate voltage as I said above. If excess voltage exists above the Vf of the emitter, they convert it to heat.

7135s do limit current but they are linear regulators, not switching regulators. So when they are on, they are on, they do not adjust anything on the fly to maintain that fixed 350mA. So the difference in Vin and Vf has to be going somewhere, it isn't being dealt with by switching the output off briefly when Vout rises above Vf like a switching regulator would do. (this is also why there's no real benefit to using a 7805 regulator instead of a 'dumb' zener regulator - they both do exactly the same thing, the 7805 just has all the various linear regulator parts hidden inside so it can look like a smarter solution - a zener regulator is a linear regulator as well just using separate parts)

-----

I know these FETs won't do the same big numbers into a single LED, I already posted the numbers on that setup... but when this FET is removed from the circuit, there's very little increase in the current. It would only be 0.1A higher if you had a theoretically perfect 0.0000000000000 ohm Rds(on) FET that currently cannot exist in our universe. There is no FET that can give higher current from a single cell than what that cell will do when connected straight to the load without the FET in the circuit. It's just a relay with no moving parts. Lower on-state resistance will give higher current, and I have not found one yet in this package that is lower than this one's .0042 ohms at the voltages we're dealing with here.

Here's a list I've been building of some common FETs I've found on various drivers:

p/n mfg ohms@4.5v milliohms@4.5v notes
DTU40N06 dintek .013* 13* *@10v - 4.5v likely around 30mR
AO4468 alpha .018 18 SRK 3-toroid driver, 3x FETs
PHD55N03-LT philips .015* 15* *@5v
09N03L infineon .012 12.1
2SK4212A nec .0104 10.4 Securitying SRK clone, red 'JB' driver
50N03-07 vishay .010 10
AOD436 alpha .0098 9.8
06N03L infineon .0076 7.6
05N03L (TO-263) infineon .0061 6.1
SUD70N02-03P vishay .0042 4.2 http://www.ebay.com/itm/271319198579

Thanx comfy! Is the East-092 FET included in that list? Not sure if the #'s were sanded off, can't recall...

OnSemi T70N03 is 8.1mR, that's what was used on the 'good' ones. The one I use is half that.

That’s a interesting question ImA4Wheeler. :~ I would like to know the answer as well.
For any of the modes the current should be the same, we are just controlling how long a duration the on and off in a cycle is, using PWM. But I know when the load is removed from a cell the voltage starts to recovery rather quickly. So is it possible that in a low mode where the on time may only be 1% of a second, is that enough time for a cell to recovery a little voltage to produce a higher current when the on time rolls around again, especially since the on time was so short the cell didn’t have time to sag much to begin with. If the PWM is controlled at high frequency like 9khz, there so many cycles in a second I doubt the battery has enough time to recovery any voltage. I suppose that the battery see’s those load spikes so fast, it just see’s a very light load. I would think the battery doesn’t sag that much because the time it takes for the battery to adjust for the load suddenly its removed before it has time to sag down like it would in high mode. I’m not really sure exactly how the battery reacts to low modes, if it could actually produce higher current spikes in PWM. If tested with a oscilloscope, I suppose you could find out.
.

Oscilloscope is a good idea. I have one I got from a government surplus center, but I still haven't learned how to use it. I bet I start putting these drivers in lights before I get around to it. If an emitter fails, then I will definitely test.