The Texas Buck driver series, Q8 / Skyray King 2S/4S buck driver RELEASED!

Yes, I was confused, edited above. I guess I don't need exact PWM values. So long as it's upward of 19 it should be fine.

No plan on using attiny45 is there?:

From the manual's errata.

8.2.4 Rev A

• Too high power down power consumption

• DebugWIRE looses communication when single stepping into interrupts

• PLL not locking

• EEPROM read from application code does not work in Lock Bit Mode 3

• EEPROM read may fail at low supply voltage / low clock frequency

Too high power down power consumption Three situations will lead to a too high power down power consumption. These are:

-An external clock is selected by fuses, but the I/O PORT is still enabled as an output.

– The EEPROM is read before entering power down.

– VCC is 4.5 volts or higher.

Problem fix / Workaround ATtiny25/45/85 [DATASHEET] 21 2586QS–AVR–08/201

-- When using external clock, avoid setting the clock pin as Output.

– Do not read the EEPROM if power down power consumption is important.

– Use VCC lower than 4.5 Volts.

I'm not sure what "too high" exactly means or what "power down power consumption" exactly means (I guess power use when off), or if this is a general trend that just happens to be a bit worse in the 45, or maybe only got documented in the 45, or is just a bug in the 45 that doesn't apply at all to others. The warning is removed from errata for later revisions so I guess it was improved or fixed.

I'm leaning toward 3V anyway. It gives a little better control ripple, slightly worse for potential design upgrades but can worry about that if and when. And now there's this warning that makes me a little nervous.

That may actually explain why the tiny85’s had issues with the voltage spikes before the latest DEL component changes but the 25’s didn’t.

In this case there is no need for high voltage to drive an FET so I think 3V would be fine.

Well, there are plots for off current vs vcc. Attiny might drain the battery some in 100 years.

This only seems to have applied to old versions of the 45.

So what footprint are those small pads meant to be? They measure .11*.04 inch by my pixel count. Seems kind of a compromise of anything from 0603 to 1206. Anyway, there's certainly room to widen Cf2 out to .06 wide.

I forgot to redo my cap searches with lower voltage ratings for the control caps. That turns up 220uF caps. There's one that's a $1.74 on digikey in an 0805 but it's only 4 volt rated. For 3V low current, that would probably just about cut it. There are also a couple of in 1206 that are 6.3V rated for $1.26.

A 220uF Cf2 makes it possible to hit Iadj RC times of 0.33 s or a bit more, ie soft mode transitions, with only a small increase in resistance, leading to an 0.6% minimum current mode instead of 0.2. I'm going to squeeze this on mine. I'll might write it up as an alternative setup, because you can get 10uF caps for 0.30 or maybe less, and they work fine too. The dollars add up on these things.

(The 220uF also makes software frequency control with 0.3% offtime jitter possible)

I ran numbers on a 1S batt setup for fun, with a vf curve added since it matters there. The thing actually makes a nice 1S driver. > 93% efficient on paper down to 0.5A or less. In the middle modes, 0.5A to 1A, linear is only about 70% to 75% efficient, so that's not bad. 93%/75% = 1.24. That's a pretty big savings if it works out real. The reason linear has been considered ok, is because its only competition is PWM'd DD. And because, as you've pointed out, it makes light come out of the front end of the device, which is the main thing.

All of the small pads are standard 0603 pads, you can put 0805 components on them if you need to but better to have the correct pads.

If CF2 is the only one that would need a 1206 that could most likely fit but the rest are already as large as they can be and maintain any kind of proper clearances.

That's the only one. Well.. I'm aiming towards 1206 now for Cin and Co and I personally will squeeze as many 10uF caps in Cin as I can, certainly 3 will fit, just maybe 4. I'll write it up for the designed way, plus some comments. If you want to draw it as 3*1206 that's fine (not sure you can really draw it as 4). Or could just make 2 solid rails and be done with it.

The Rsens is what 2010? Updated: Ok, this appears to mostly be what's available. and yes, 3 seems right unless we want to push things very very close to burning up. So I'll write something up.

Even in the 10uF, selection is much more limited in 0603 than 0805.

Here is the component footprint list as it sits right now.

Cin is 1210, it could be swapped out for 3x 1206 it looks like but it is already too close to the edge for my liking. I actually want to pull it back some more in possible.

Rses is 2512 simply because it was the largest resistor I could find and they had the highest wattage ratings It could be shrunk to a small size obviously and I might need to anyways to get the edge clearance I want.

Here is an updated CF2 to 1206 (not reflected in the list). It works but is the very limit of what those component sizes can be

looks good. I was thinking you could move D1 a little down and then right (but not much) and then shove Q1 right and Cin over to the right a little for clearance. I agree on Rsense when I looked at what's available. Using two would be VERY tight. Might end up with some burned resistors. Anyway, Cin is ok with 2 by 1210. I suppose if you really needed the space you could cut it back to 2 by 1206. If left as is though, I'll try to pile a heap of caps in there with a wad of solder and see what sticks. I'll write up the baseline as 2 though.

I can connected the Cin pads easy enough if you want to try that.

I guess it doesn't matter much. They're almost connected anyway. I'm going to be worried about them bridging to Rsense though. That looks a bit tight, but it is how it is. If things are bad enough there, we can go the other direction and back off to two by 1206.

Bridging to rses should not be a problem unless you use way too much solder paste.

I guess I'm a little surprised you oriented Cf2 that direction. If there's a short to the case isn't ok the other way, but not this way? Or there's enough clearance it won't be a concern? I realize now that when I said there was room there I was looking at the old version without rf3 in it.

OK, I guess the other way takes a mm more space. I see.

I tried my best to have it flipped but it was just a bit too close. This is the best I could come up with. A 0805 pad would be much nicer. As it is, the RF3 is closer to the edge so it should hit first.

OK well this little 0805 220uF Ta cap

http://www.digikey.com/short/3bbc1d

is only 4V rated, but maybe it does better than a 6.3V ceramic. All the ceramics have C vs DCV curves that drop off as DCV goes up. The Ta's don't as far as I know. The spec sheet calls for derating it above 85C, down to 2.7V at 125C. Is the board going to be getting over 85C? I kind of doubt it while a person is holding the light at least, but I'm not sure. Anyway, we can go with 0805 and consider this cap experimental/at one's own risk. This is the only one though, on digikey at least. You know you're pushing the limits when you select all your criteria, as loosely as you can possibly tolerate, and the search box says: "1 remaining"

Ouch, at $1.75 each sticking with the 1206 seems like the better idea.

It works as it is setup now, just a bit tighter then I prefer. Once we finalize the component list I will go back over things one last time to try and get back ever mm I can of edge clearance.

Yep. Ok, I've got a mental deadline of this weekend to write up a 4p xhp35 application. It doesn't take very long now, but I should put aside a bit for other things probably until then (we'll see). Should I aim for 4*2.5=10A? Or maybe 4*2.8=11.2, living on the edge slightly? Of course that part's not a big deal, but I want to choose everything down to Rsns for one reasonable setup at least. Actually even finding some combinations of Rsns might not be so easy, so that's part of my reasoning there. Let's get at least one full suitable example down to the last resistor. After that I'll probably target a 2S xm-l2/xp-l setup, maybe 11 or 12A. I think those are the two best use cases for the Q8 version.