Oshpark Projects

You have the two new resistors on this one, right? Try jumping from B+ (or LED+, same thing) to the gate on the FET, see if it goes up when you do that. Also try jumping across just the resistor inline with the gate.

If it shows a big jump with the resistor bypassed it means the gate isn't opening fully when going through the resistor, we may need to look for a different FET that doesn't need the stupid resistors. I have a few of the 2SK4212s I can try, they have a lower gate charge spec.

Running the 18500 for ease of handling, I was seeing just over 4A on high with zero change both visibly and on the meter when I crossed the positive to the gate. Showing 2.7A on Med/Hi it did indeed jump in brightness and to 4A on the DMM.

Not sure about crossing or bridging the resistor, didn’t see any change there either.

Edit: Trying again to jump the resistor I touched the wrong things and it popped and smoked. I think it’s staying dead this time. I gave it a standing 10 count and they carried it off on a stretcher. It doesn’t look good.

No, wrong K4212. It's from NEC/Renesas, 2SK4212A. About double the Rds(on), but much lower gate charge. TO-252/DPAK package. The one that comes on the red 'JB-2013-11' driver. That exact part is discontinued/EOL, and the replacement sucks balls, specs nowhere near compatible for this low-voltage-overhead setup. The replacement would probably work fine in a different circuit. The old 2SK4212s are available in bulk still, but only as surplus NOS and only from possibly questionable vendors on aliexpress.

This one looks kinda interesting as a possible substitute: http://www.digikey.com/product-detail/en/AOD510/785-1481-1-ND/3603497

A little bit lower Rds(on), and Qg is much lower and closer to the 2SK4212 that started this whole mess in the first place.

Hey man, you know what they say: you can't make a driver without breaking a few omelettes... er, no wait...

Lol, and yes I agree.

Just went out to take the trash down to the road. Carried the C8, M8 and L2P with me. None are throwers. All are FET modded. And wouldn’t you know I saw a couple of critters out at 100 yds that I just couldn’t quite identify even with the big M8, granted by the time I pulled it out they had moved off and were closer to 200 yds away. Might have been a pair of foxes but I haven’t seen foxes around here in a very long time. Nearly 30 years. About the size of a big housecat but with large puffy tails. Had to’ve been foxes. Cats tend to slink away, these did a sort of bobbing lope. Almost took my camera, wouldn’t you know I didn’t! lol

At any rate, the L2P Triple worked quite nicely, very good for a walking light with it’s wider illumination pattern from the 10507. I didn’t use Turbo very much, but purposely ran it on High for several minutes and it was really too bright. The 2nd level, Med, is about right for walking. It shifted levels reliably, never flickered, and all the dust didn’t show any PWM in any mode.

I lit up a Jack Rabbit with the C8 from about 40 yds and he got confused. He came towards the light, unafraid, stopping to nibble a couple of times, but got within about 10 yds of me before I said something to him and he loped off. Still not afraid but putting some distance between himself and this unfamiliar small sun.

The big M8 with it’s MT-G2 can really light up a tree! For anyone checking into a tree at night it makes a 40’ tall tree seem to be in daylight. We’ve lost a couple of trees looks like. Early bloomers and a late freeze caught em. Here we are sitting at 70º at 10 O’clock at night. Walnut tree’s are blooming out, grass is growing (gonna have to mow in the next few days)

Nice night. :slight_smile:

I’m looking for something smaller than the dpac for the 17 mm board so there’s room for extra components. There was an entire page of K4212 in stock at digikey. Are they all the wrong ones?

It's not a type, it's NEC's model number for that specific part. What did you search for to get that list? I don't see anything there that looks related.

http://www.digikey.com/product-search/en?vendor=0&keywords=2sk4212

http://www.aliexpress.com/w/wholesale-2sk4212.html

It was the package LFPAK33. Less than 4mm square with low Rds at 4.5V and very high power dissipation.
EDIT- I realize This is the usual pap but it still seems worth a look.

WOW! Interesting and if those specs on the RDS (on) then wow…very low resistance…nice

The current and power ratings are through the roof as well. I think I looked at ones like these in the thread where CC and Werner were working on this idea. They’re not expensive and have been proven up to 175C. Less than half the size of the TO 252 leaves room on the 17mm board for needed resistors. Pick some up?

EDIT- The LFPAK 56 is 5x6 mm(slightly larger than a 7135 including pins) the LFPAK 33 is 3x3 mm(smaller than a 7135). The 56 has even lower Rds on and higher power dissipation than the 33.

Depending on the Qg, it probably won't need the resistors. And if you're picking a new part might as well make sure to get one that works with the simplest circuit possible, i.e., no resistors.

Or, try the AOD510 on the existing board design, it may not need the resistors either. There's not going to be much measurable difference between 4mOhms and 2mOhms, there's no magic bullet there, it's still limited to no more than what it would do with the cell wired direct to the LEDs.

What did you search for to get that list?
I remember the FET led4power is using on his driver had this odd shape and very high power dissipation numbers and the K4212 from earlier posts and just spent the evening wandering the web looking at specs. When I googled k4212 the Digikey link came up and I looked at pictures until I saw this odd shape again. Googled some more and found the NXP page and mouser links. BTW mouser is soo much easier to navigate for parts.

True, but the smaller package should work with either the 17 or 15mm boards and might eliminate parts/traces on the backside. They also have higher operating temps and higher current ratings for their size which may or may not be helpful. It’s a newer, possibly better design than the Dpak. I’m still going to try out the 2502 but it’s not as likely to be able to handle an imr 14500 or 14650. In the end, whatever fits and works is fine by me but we’re still in development so I don’t see any particular reason to be wedded to an older part unless it works better. The “33” is the same size as the 2502 but is more than an order of magnitude more robust in almost every respect and the “56” is only slightly larger and even more potent. There may well be reasons that they won’t work but it may also be true that they’ll work better.

I have 150 of the Vishay 70N02s, that's one reason to use them... :Sp

any reason to use the luxdrv when STAR V1.1 and Star_mon V1.0 work…or does it not work…I haven’t gotten my boards to build yet

Not as a sidetrack…see if you can invite this guy over to join in the fun

tterev3
https://oshpark.com/profiles/tterev3

On the SRK version... minimo (phase correct) works fine, but is noisy, as it has been on any of the drivers I've used it on. But STAR v1.0 works on it too, and I think (though not sure) it's set to use fast-PWM as-is, and I didn't change anything in the code except to plug in my own PWM values.

On the 17DD, luxdrv didn't work right before the two resistors were added. STAR v1.1 did work, but again I'm not sure which PWM version it's using as the default. After adding the resistors, it worked with the other DrJones fast-PWM firmwares, which none of the other boards using the FET have been able to do.

I even took the same build of minimo that worked on the SRK driver, and flashed it to the 17DD (before adding the resistors - and remember, the SRK board doesn't have those two extra resistors, and yet it works fine), and it didn't work. It did the same goofy halfway or double mode changes. So, same parts, same firmware, different PCBs, one worked and one didn't.

Or you could just ask me.

Yes that's nxp lfpak 5x6mm. Many other manufacturers have similar 5x6mm power packs=power soic.Nxp lfpak is interesting because of exposed drain pad,so no need for extra pad on pcb for cathode led wire(of course this isn't the only reason why I chose nxp mosfet,many other parameters are important).

The biggest difference between lfpak and old packing technologies like dpak,d2pak etc. is lack of bonding wires,which means much better characteristics.

That's all wonderful, but are the olde-timey FETs prone to failure in a design like a 4.2v max flashlight driver circuit? A motor controller for an electric/hybrid car, sure, that's a highly stressed and critical application. But that's a long ways from a flashlight. If the driver gets up to 150*C, you got bigger problems to solve first.

Per compile
define F_CPU 4800000 CPU: 4.8MHz PWM: 9.4kHz ####### use low fuse: 0x75 #######

Did you compile luxdrv at 9600000
define F_CPU 9600000 CPU: 9.6MHz PWM: 19kHz ####### use low fuse: 0x7a #######

OOPS…I stand corrected
bove PWM speeds are for phase-correct PWM. This program uses Fast-PWM, which when the CPU is 4.8MHz will be 18.75 kHz