[WIP] 17mm DD+single-7135 driver / single sided / Dual-PWM

Shrinkwrap-and-glue a bundle of horsehair over the end of a stick. Google ‘handmade brush’ or ‘homemade brush’ for pictures.

Trim the end with a razor for applying paint, or leave the end variable for ‘dusting’

I think this is the right thread to ask.
Are there any firmwares except toykeeper’s ready-to-flash ?
Could you provide me a link ?

Not sure. I don't usually post HEX files because I build mine on demand for modes, tweaked, turbo timeout setup, etc., so not a good idea, unless you have a fixed, static setup: same hardware, same modes, same options.

This is why I'm going with the more versatile programmable UI, so we can use a common hex file, then reconfigure it after it's burned in, but this requires more memory than a 13A has, so why we are migrating to the Tiny25 and Tiny85.

Thanks for your answer. I’m new to the whole programming thing , and I don’t know how to do edit modes on code and so on.
I’m still searching , but if you had any recommendations (where to start from , or some guides) it would be great.

Thanks again

CRX's incredible reference thread, https://budgetlightforum.com/t/-/23020, has a section for programming drivers - you will find some useful info there.

I'm using the latest Atmel Studio 7, free download here: http://www.atmel.com/tools/atmelstudio.aspx. You can begin with that and go from there, if you haven't dnld'ed it already.

Looks like plenty of interesting discussion happened on this one while I was out. I think that I’ve skimmed it all, but I didn’t necessarily absorb everything and I know that some of this stuff gets discussed in other threads too.

  • I’m not sure what’s up with all of the OSH Park mistakes. I’ve certainly made some flagrant mistakes on other things with them in the past, but as far as I know everything is pretty kosher with this one. I definitely do go right up to the minimum tolerances sometimes, but I do not see how that would cause such flagrant issues with the resist mask…
  • I’m inclined to agree with anyone who thinks that more of the GND ring should be exposed. Maybe I picked up the skinny-GND-ring habit form Mattaus, who knows. I do prefer to create a “dam” of resist around any component pads that are near the edge of the board. That helps prevent components from scooting towards the edge of the board when soldered, which can create fitment/shorting issues once installed in a light. I don’t always achieve this of course, it’s just my preference. In any case, Eagle doesn’t really have an automatic way to do make sure those dams happen - so the easiest solution is a skinny GND ring. I think I have an updated version of this board with a larger GND ring. I’ll see about dragging that out.

Hi Alex!! Wow - welcome back!

I was getting a flash going from hi/turbo to moon (on this FET+1 w/ATTiny85, SIR800DP FET) that I was able to solve in firmware (e-switch firmware called "Narsil"). Had to do with removing redundant setting of PWM modes.

For a power switch setup (also on this FET+1 driver, ATTiny25 w/SIR800DP), I tried several things in firmware with no luck, then added a 12K resistor across the FET (gate resistor) and it solved it perfectly.

One guy, think an EE, found a FET that didn't have the flash. I did some amps tests and the 12K resistor seemed to have no effect - I'm pretty happy with the fix.

25, 45, and 85's seem to have their own problems in high amp setups. Caps seem to fix it - I add a 0.1 uF cap across the MCU grnd and VCC pins and it fixed it with the e-switch firmware/setup. On power switch setups, I stacked a 2nd 10 uF cap on the C1 cap and it fixed it. The problem is in either just hi/turbo or any FET modes, the MCU flakes out - resets sometimes, or get weird blinks, etc.

I agree with more exposure of the ground ring - would help, I've done lots of scraping.

Thanks Tom E.

  • So that’s a 12k pulldown resistor on the gate? (EG the resistor is connected to the “gate” pin and to “GND”?)
  • How are people feeling about 0603 sized components these days? Sourcing is not an issue: the days of using Nanjg boards for donor components are well past. Almost everyone building a driver of this type is certainly sourcing parts elsewhere. It’s just a question of whether people can handle soldering the smaller size stuff.
  • I’d say that the v009 layout of this board does not provide very good filtering for the MCU. We got away with it before, but it looks like the seams are coming out now. A decent sized overhaul is in order to fix that… and frankly I think it’s probably time to start breaking these boards out into single cell and multi-cell rather than “universal”. In the past I liked to keep these designs as flexible as possible, but the more we push the limits the more specific they need to be if all else remains the same. The 17mm boards cost <$0.75/ea from OSH Park, so it’s hard for me to feel bad about making them less universal. In order to meet these new design goals it seems like something’s got to give - either step down to smaller components, build less universal board layouts, move to some pricier parts, or all of the above…
  • Speaking of pricier parts, back to the LDO discussion. From my looking at the ATtiny85 datasheet in Section 21.3 / page 163… looks like 8Mhz @ 2.5v has Atmel’s blessing. So maybe a 2.5v LDO if we wanted to use an LDO in single-cell applications? I’m not 100% confident that we even want to do that. (For the higher voltage stuff I’m pretty confident!)

EDIT: added strikethrough for “DOPE!” moment. See posts below for correction(s).

I don't see 0603 being a problem, as most people would be using paste/heat gun rather than manually soldering these things with an iron. If the size saved is going to make a board possible where it was otherwise not, then I say go for it.

The additional/larger Cap seems to overcome the MCU spike issue on the current boards, so another revision of this design may not be time well invested. I think a new design targeting multi-cell applications(Buck w/ATtiny) would now be more sought after.

I manually solder by hand and have no issue using 0603 parts if it gets me the driver I’m after

Yes - 12K pulldown - it's what I had available and seems to work well. 0603 sizes are fine with me. I certainly would like Richard to chime in on this, but not sure if he's been spending much time on the 25/45/85's.

I haven't built any LDO based boards yet but am certainly interested because I have a couple of lights I'd like to use them in (e-switch multi-cell for 6V LED's).

I’ll get back to those comments, I’m just posting right now to point out that using a 2.5v LDO is a bad idea. (In other words I’m correcting myself.) I don’t think an LDO can be practical for single-cell since it won’t fully open the gate with most of the FETs we use. (The MCU will also drop some voltage, so there just won’t be enough to make the FET happy.)

Wight, I was about to comment on the low LDO voltage, and I agree with your correction. You really want a 4V+, preferably 5V LDO for two reasons: First, you want the gate voltage to be as high as possible to minimize RDS, and second you want to dissipate least amount of heat possible in the LDO, especially the smaller package ones if you are running 3S or 4S.

If you use an LDO with any clicky application you generally need to add a pulldown resistor to drain the capacitors so that you can switch modes fast enough with off-time memory.

As far as use of an LDO with 1S goes, I have found that most of the true LDO regulators will work with 1S and that the voltage drop across them isn't much worse than a normal schottky diode, but you do have to contend with the extra capacitor and the timing issue (see paragraph above).

I really think the FET makes a difference. I’ve used PSMN3R0-30YLD on wight’s FET+1, with t25 and bistro for a single XM-L2 without issues and used it on PD’s DoubleDown with t25 and bistro for a triple XP-L HI, no issues with just the standard components, both are power switch lights.

If people have to manually stack a second cap, that’s pretty crap for reflow IMO. Same thing with the 0.1uf that Tom E has been adding to fix ATtiny25 problems. I’ve publicly done several major reworks on the FET driver layouts + several more that never saw the light of day, so one more to get a better build process might be worth it.

As far as multi-cell applications: For me the first step is probably a board similar to these (DD+7135) with the addition of an LDO. That’s a much smaller leap than the leap to a buck driver - remember that I’ve never designed a successful buck driver. I tried pretty hard to design/build a QX5241 based on early on but couldn’t get it functioning. I learned a lot from that process and applied some of it when I did a much more handsome MAX16820 based board… which I never ended up building / testing. Since then RMM has gotten several sizes of buck driver working and all of us have learned a little here and there. I’m confident that I can design a buck driver which works now, especially with a few well placed tips from RMM. I just don’t plan to hop on that task first thing.

Sounds like the consensus is that we’ll take our lumps on 0603 where necessary. If I build something ‘unkind’ you have only yourselves to blame. :evil: :wink:

Thanks Tom E, that’s what I wanted to know. I’ve got some thoughts on reducing the quiescent draw of momentary lights a little bit more, we’ll see where they go.

Good points RMM.

Getting back in the saddle really has me dredging my mind for some stuff. I go back and forth forgetting important things - for example I’ve just been reminded that we normally depend on an internal 1.1v vRef for LVP with the ATtiny13a. There are tons of other equally important things that are slowly filtering back into my consciousness.

Is that what I’ve recommended / used in the past? Maybe I should do a build with the SiR800DP. I think I bought 1 or 2, but I don’t know that I’ve ever built anything with them. As I recall the SiR800DP is a bit different from many of the other FETs we run into.

Yes that’s the one you kinda recommended in your A17DD-SO8 / A20DD-SO8 / etc - thread. Seems like there has been a lot of discussion about the best performing FETs to use. Still haven’t read and more importanly understood all the info about FETs that was posted all over the forum. By the way, thanks, wight, for letting me see comfy’s scope images for the first time. :slight_smile: I’ve been clicking through that thread once or twice and wondered what the lacking images would look like. Most of this stuff is over my head but it’s quite entertaining.

The SiR800DP is a beast---it can handle more current than most of the other FETs I've tried---the difference is much more than the datasheet would suggest. If you try and run 4x XHP70 from a single PSMN3R0 you will find that it quickly desolders itself; with a single SiR800DP it is fine without making any other changes--and that's with a 5V gate drive (much higher than we usually get with 1S setups). So yeah, you can run crazy stuff through an SiR800DP. It seems to even run cooler than the 0R9 NXP FET I have in the same situation.

The SiR800DP does suffer from dV/dt induced turn on when you're switching from 100% quickly to moonlight (that's the little light blip you see). As Tom E. noted, a small pulldown resistor solves that issue.

I use 0603 exclusively in my drivers as I want to utilize the space as much as I can. I also bend pins on the MCU, put the MCU where the spring usually is (for space and easy access to flash the light without full disassemble) and cut of the tabs of the 7135s to make them shorter (custom Eagle library part). One of my drivers where I’ve done this: Mod: My take on the convoy S series with side switch mod.

I did a design with 0402 sized components but they where to annoyingly small for me to work with, so I’m sticking with 0603.

Btw, nice to see you back, if only for a while.

Has anyone built this driver with a second stacked 7135? Adjusted the modes to use the pair of 7135’s for more modes of non-FET power? (more efficiency)

Not physically stacked that I am aware of but with more 7135's. The single 7135 was basically a method to achieve a regulated "Lo" that many complained was either "to the moon Hi" or a "subterranean Lo".

Wight and PilotDog68 both have drivers with more than one 7135