[WIP] possible 20mm buck w/ MAX16820

I just adjusted the OTC values and have been fine without it, over a wide range of different lights. On the higher drain lights, like the TR-J20, I need to put a lower OT value in there or else you have to click the button so fast that it is hard to change modes. The higher the drain capacity (generally multiple emitter lights) the lower the value I need to go to get similar switching performance. I suspect that this is because the LEDs drain the power for the MCU a bit faster and affect the timing. Did you switch capacitor part numbers?

I do suspect that you will have a problem with a large input capacitor on clicky lights using off-time firmware--one that a pulldown resistor on the OTC is unlikely to help because the MCU will keep running for a while after you cut the power.

I’ve had to go to 240 to get what I considered decent OTC performance (~0.5s reset). That’s close to the edge. I think LinusHofmann hit 250 before giving up and adding a pulldown resistor, this was with an LDO installed. That’s actually not clear though, since some flashing mistakes were made while that testing was done.

I can’t speak for what OTC LinusHofmann is using, but I am using C2012X7R1E105K125AB. My recent problem light was around a 3.5A or 4.0A DD light w/ some A17DD-SO8 version I think. It might have been A20DD-SO8 v24.

You have a good point about the large input cap + offtime setup. We’ll have to see how this plays out, hopefully we can really minimize input capacitance needs.

240? Wow. I've never gone above 130, but usually end up at around 100-110 on the really hot multi-emitter builds. My LDO builds would have that problem because of all the extra capacitance I put on the boards with them, but I'm not going to use them for clicky builds (no real benefit, in my opinion).

Hey Alex- what inductance 7x7 do you suggest for this build? If you can’t say for sure then what value did you use when running your numbers?

Do I look like I ran any numbers!? :smiley:

I was thinking 2 or 3 uH. There’s not a lot more that will fit in that size at 3-6A.

I have not simulated this.

On paper a 3.3uH inductor ought to be able to do it for 5A-6A at 2MHz switching frequency, but that's using ideal components (components that don't exist).

In that size I think we won’t get 6A continuous out of anything with a higher inductance than maybe 5.6uH, the real world part just does not exist.

Here’s a rather tall 5.6uH Vishay/Dale part with 7A saturation and 6A “heat rated” current: IHLP2525EZER5R6M01

It’s definitely going to have to be a low value inductor to fit on the PCB. How low we can get away with remains to be seen. Here’s what I have in terms of new-in-package inductors. I do not plan on purchasing more until I see what can be done with these. Inductors are expensive.

Should fit:
Bourns 3.3uH - SRP7030-3R3FM
ABRACON 4.7uH - ASPI-0630LR-4R7M-T15
Should not fit:
Sumida 18uH - CDRH105RNP-180NC

I may also have this and that from scavenged parts, but those do not hold much interest for me since we can’t source them.

Also I just solved for the switching freq w/ 3.3uH and I got ~1Mhz, well below 2Mhz:

Vin = 8.4v (two series 4.2v cells)
Vout = 4.0v (XM-L2 or XP-L at ~6.6A)
Rsense = 0.03 (~6.6A set current)
Inductor (L) = 3.3uH

  1. Fsw = [(Vin - Vled) * Vled * Rsense] / (Vin * ∆V * L)
  2. Fsw = [(8.4 - 4) * 4 * .03] / (8.4 * 0.02 * 0.0000033)
  3. Fsw = 0.528 / 0.0000005544
  4. Fsw = 952380.95238095238095238095238095 Hz
  5. Fsw = ~0.95 Mhz

I also did the math over again for what I consider a nearly worst case scenario: all the same hardware but driven by 4s1p 4.35v cells for 17.4v Vin! I got 1.4Mhz.

I think the problems show up when you want to reduce current from that point. Dropping to 3A output with the 17.4v setup results in 2.6Mhz: far outside of spec.

I’m using a toroid on most of mine, I’ve used a variety of scavenged toroid’s all the way up to ~12uH*, dont have any idea the current rating’s on any of them tho. For some reason I missed a freaking decimal point when I first ordered parts and I wasted a ton of money on different, large SMD inductor’s that didnt come close to working so I stared looking at other options I had on hand and that’s the route I took.

I’ve got a digikey order coming up (I’m the same way as you man, I keep thinking “if I just wait a little longer I can get next project’s parts in on this one too” lol) soon when I plan to get a selection on known value toroid’s.

I’m working on a 13A version of mine today, I’ll share it tonight with part’s list (also updating the lables and making it a little more presentable and comprehensible to others)

BTW I’m using one of those $20 transistor tester’s to come up with the value of the random parts drawer toroid’s I’m using.

This is a very general question not necessarily related to this project at all…

Theoretically would you be able to accurately control low voltage safety features on a multi series-cell light by monitoring a single cell’s voltage? (never mind the physical challenge of tapping into the series of cell’s to get connections on each side of a single one). In theory, would all the cell’s drain about the same so you could monitor just one of them till it got to your desired low voltage point?

That's also what I got...that at 2 MHz you could get away with around 1.3uH-1.5uH of inductance, but again, that's on paper. I still have a bunch of the Coilcraft 3.3uH that are rated for 4A-5A, so I might shoot for somewhere around 5A to begin with. They are expensive buggers but they seem to be the most efficient space-wise out of all of them.

The higher the differential between your input and output voltage the higher the switching speed needs to be to maintain the same output. When I have the scope hooked up to my evaluation board you can watch the switching speed change as you raise and lower the input voltage relative to the output voltage. The switching speed difference between 8V and 12V input with a single emitter is pretty big.

The MAX seems to be able to switch at over 2MHz, but who knows how stable it is up there, and switching losses will be greater. Best to stay within specifications (as if we were good at doing that... )

I think you’re getting carried away. Sure we could do that, but there would be no point. This is just as potentially problematic as monitoring overall pack voltage: it does not account for cells having slightly different capacity. Those cells with lower capacity or other weaknesses will drop voltage more quickly than the strong cells.

We normally monitor overall pack voltage. The disadvantage is that we’ll miss a single weak cell and overdischarge it. Monitoring a single cell in a 2s, 3s, 4s, whatever cell count pack gives the same disadvantage with no advantage. All it does is add the pain of getting that cell tapped somehow.

Ideally we’d monitor pack voltage and a tap between every 2 cells. This would allow us to keep an eye out for weak cells.

In theory, yes, with the assumption that the cells are fairly matched. We already make this same assumption when we measure the series voltage as well instead of measuring each individual cell.

EDIT: looks like Wight beat me to it!

Thank’s guy’s.

Just to clarify, no I’m not at all thinking about doing that (on this, or any other project), just thinking about voltage monitoring stuff and the LDO and series cell’s in general while reading threw this post brought that question to mind and I knew you two guy’s would be the one’s to answer anyway if I started a new thread so I figured why not just put it here lol.

Wight, man, now I may be going crazy but I’m looking at your driver and trying to figure it out, what schematic did yo go off when you made your board? Unless I’ve totally lost it it’s not this one so I’m confused.

Allright here’s mine. This is based on my running BU_CK so I know the circuit is correct HOWEVER this is a new version I just made with the 13A, the MCU and the voltage divider circuit are both new additions; THIS BOARD IS TECHNICALLY UNTESTED, but again the actual buck components are exactly the same as on my running 22mm driver.

17AVR_v01

  • Pretty small exposed GND rings but full GND plain’s (this is v01 afterall, expect changes)
  • This is different that Alex’s circuit by quite a bit, I’m not saying his design is incorrect or doesn’t also work, just that it’s different and I don’t understand how his works (or maybe I’m just seeing the traces on the render incorrectly)
  • L1 and L2 are the pads for the external toroidal indictor that mounts above (like most commercially available buck driver’s we can buy) this makes for a tall driver but only 17mm Ø.

https://oshpark.com/shared_projects/xK2QT7oj

What's different is that Wight has the inductor on the positive side instead of the negative side, as shown in the datasheet. I'm not sure if it makes any difference which side it is on (either way, the same amount of current should be flowing through the inductor).

Yep, I was just coming here to post I had realized that now. Hey it was late! Plus I didnt know you could do that so it just didnt hit me, when I looked at it today I saw only the inductor was different and figured it much have been ok to do that, it is wight afterall lol.

My current understanding says that it’s fine. That doesn’t mean it’s really fine, but I think it is.

Cereal_killer, why don’t you start a thread for this? I have questions/comments.