What size components are those, they look like 0805? I would drop down to 0603 for sure for that driver (except for C1 naturally). You can still use 0805 components on the pads but 0603 is really not hard to work with.
Also, can the tiny13 use the internal voltage reference like the tiny25 so that it could handle LVP without the low voltage divider? That could eliminate the R1/R2 and save a lot of space.
Just a proof of concept, obviously vias and ground plane are missing:
.
outline is 13.2
yellow circle is 12.5
even 11 looks doable with this design
R1, R2, OTC are 0603
C1 is 0805
MCU is the 10pin (10M1)
It’s not like I reenvented the wheel, you’ll find it quite similar to my HQ10D, which itself consists of parts and ideas others were sharing and I did pick up.
I just looked at the ATtiny13 datasheet. It does have an internal voltage reference, which it says is useful for brownout detection. I don’t know anything about it, other than that. Is the internal voltage reference a yes/no deal? Does it just being there mean that you can use it for LVP, or are there other considerations? Sorry I don’t know anything. :person_facepalming:
I see your point with the 20pin MMU. If I start making a fortune by selling pre-flashed MCUs, you be the first to know…
In the meantime
With the 20pin version GND sits differently (which is the least concern). It’s larger. And lot’s of Do-Not-Connects on both sides.
You use Eagle? You might want a custom part stripped of DNC pads so you can use wires under the MCU.
Is there going to be a problem with the MCU because of the back/bottom GND plate not being soldered down?
I have no idea what is needed either, only that it apparently works lol.
Since I do not see this driver being used in anything but a 1S setup using an internal voltage reference should work well.
Only if the MCU produces more heat than the package can dispose of without the soldered back. I can’t find anything substantial in the datasheet. But I truly assume that, as our Attiny13A are in sleep mode almost all the time, that it will work fine without the added heatbridge.
You’re probably right. Additionally, this driver only has a single 7135 regulator, and will be running off a very small Lithium cell. So heat from all sources will be quite low.
Edit: So, any chance you’d sell me some pre-flashed 10 pin MMU’s?
Yeah, I think there is even more heat coming from the LED to the driver to the MCU than is generated in the MCU itself.
As I wrote, if I start making a fortune by selling pre-flashed MCUs, you be the first to know…
No, honestly, I do not sell, I only have a flimsy breakout board and I’m not capable of testing post-flash. Did I already mention I sit in europe?
But RMM might be able to do it, if he already sells the 20pin, why not the 10pin? Might just ask him.
Oops! I missed that post of yours before. :person_facepalming:
I hadn’t thought about making a new library part. That would work well enough for what I need. If I do that, as well as removing the LVP parts and moving everything around a bit, I could possible get the driver under 11mm diameter. Then it should work for most 10mm cell size lights (10180, 10250, and 10440/AAA) as long as they have length enough in the driver cavity.
Don’t give up on the 20pin MMU yet, being able to get these preflashed from RMM is still a major asset.
Here another proof of concept, which might be very much along the lines RBD suggested:
Make C1 an 0603, use 0.254 instead of 0.3048 wires, squeeze everything tight together and you might even hit the 11mm mark… But the 20pin is huge compared to the 10pin…
Thanks HQ! :partying_face:
My pleasure 
Time zone trailer, I’m missing the fun. Yes, I meant that since you can’t flash after there no need for clip clearance. Blank areas, traces, and vias all take up space so smaller blank areas, shorter traces, and sharing vias with no top side ground ring all will help. Post 19 is along the lines I was thinking only better, I hadn’t thought out how to arrange them. Even just eliminating the mmu pads on the left side will allow the R1/R2 trace to pass underneath so you can squeeze the 7135 down a bit more. When Mattaus did the first tiny10 it took some time but in the end it made sense looking at it.
It’s usually we CET-guys that are missing the action 
Nice work DavidEF. Sounds like it will make a nice driver.
You could save some space by nixing the reverse polarity protection. Not sure what impact that would have on internal reference LVP.
Another thought, with left side mmu pads eliminated, swap the 7135 with everything but OTC and route pwm under mmu. Maybe rotate the mmu so the gnd side flat is aligned to the board edge instead of it’s corner and put a gnd via under the mmu.
The inner yellow circle has a 5.5mm radius (->11mm).
Getting closer.
Gaining space below the MCU doesn’t help, no component fits. So I swapped D1 to the right again.
Still the notorius OSH Park Edge™ has to be considered. 
But it looks promising.
@DavidEF, if you want the .brd with the custom parts, PM me an email (preferably an anonymous one).
Now my TimeZoneGap will commence.
Goodnight.
L+/L- could be much smaller(1mm) if they were vias solder able from the bottom. 10xxx cells are all small contact on the positive end so a .4mm (?) via should work for L- if between B+ pad and outer ground ring. B+ input via becomes L+ eliminating that pad altogether.
Wow! You guys are great help! RBD, thanks for all the great space-crunching ideas. I meant to work on this board again this afternoon, but I forgot. I’ll have to try to remember tomorrow afternoon. HQ, thanks for all the work you’ve done on those board designs. Don’t worry about the OSH Park Edge™ being a problem. These boards can be sanded down pretty easily. I’ve been adding extra diameter to my boards just to accommodate it. If we can get the real diameter to be as small as possible, then the OSH Park Edge™ doesn’t really hurt anything except making the boards cost a bit more.