GXF22 - Very Low Resistance FET & Constant Current Driver with Fuzzy Logic + Microphone

Hello BLF,

Hope everyone has been well!

I've been extremely busy for the past couple of months, and recently needed to travel a lot. I had some time during transit so I thought I'd write a little about a simple project which I had worked on at the start of this year. I figured it might be interesting to some folks here.

GXF22 Driver

Introducing a new series of drivers, starting with the GXF22 driver.

In the world of non-switching drivers, there only exists DD FET drivers with an array of 7135 ICs, or transistor based linear CC drivers (like the ones made by LED4power), but not both. I'm not sure why there hasn't been one like this, so I thought I'd make one.

Right now it comes in 22mm diameter, with the candidate host being an Emisar D4. I'll begin this thread with a quick summary and hope to add more details soon. Like my previous project, the GXB172 and it's variants, this one is also a little ridiculous (i.e. not practical?), but hopefully, interesting. It's a little bit simpler though! Like my previous projects, I'll be open sourcing them once they're mostly stable in development.

The standard Emisar D4 is a very well regarded flashlight, but it comes standard with a direct drive FET driver, and includes a single constant-current 7135 IC for the lower modes. Brightness levels are all handled via PWM. Among many reasons, I've always found this to be non-ideal, especially having a FET DD, which is simply unregulated.

So I thought that it would be nice to have a simple, reliable, regulated non-pwm driver, which is the main reason for developing this driver. Then I added an 'extra' DD FET drive anyway, just to see how far I could push the ridiculous, despite what I just said. I suppose the hope for was to build a nice driver that I would be comfortable using, then add the DD FET as an extra bonus. But oh well~ here we go!

Note - this is just a prototype driver. It's working great, but I'm sure more changes to come. Just thought that people would like to see it for now!

Above is the driver sitting in the Emisar D4 (upside down though, just to show how it looks like). The board is much easier to put together than my GXB172 for starters, which is great for assembly!

Here's a picture of the driver with it's candidate host, a regular Emisar D4. Thanks to Hank from Intl Express for selling a host-only version.

Main Features

Here are the headline features:

  • True Ultra low resistance FET Direct Drive - much lower than any other FET driver, and not because of the specific FET chosen!
    • True 0.5milliOhm R_ds_on
    • I'll explain a little why it's actually this low R_ds later on
    • In addition, I have to add that from an engineering point of view, I don't like FET drive, though!
  • On board integrated true Constant-Current linear driver with no PWM (default 10A regulated current) with 10bit resolution.
  • Temperature sensor accurate to 0.5C, with improved Fuzzy Logic temperature control, better than PID
  • On board microphone - want control your flashlight with a clap? Or make it dance to music? Now you can!
  • 5 auxiliary outputs, of which 2 are currently configured for D4 auxiliary lights.
  • E-switch
  • Very low dedicated Firefly mode (as low as uA drive if desired)
  • Low height - fits Emisar D4 driver bay
    • For proper heatsinking, add some silicone padding of the appropriate thickness.

Description

The problem with most direct drive FET drivers is that they use the battery voltage (powering the micro on board), to drive the gate of the FET. That's not ideal. (Actually there are many more things non-ideal with FET driver, starting from the fact that they are not regulated and can easily blow up your LED and other things, but that's another story). Let's take into account what happens when we use say a regular FET driver with a highly regarded BSC009NE2LS5 Infineon FET. This FET has a thermal resistance of 1.7K/W, and an advertised R_ds_on of 0.9mR. However when the battery is at full load, the cell voltage drops to just over 3V, and the gate voltage drops to close to 3V. At this low voltage, the R_ds_on increases to closer to 1.5mR, and increases as the cell voltage drops.

This resistance is still very low, and likely much lower than the rest of the system, but it's fun to try to see if we can improve it! My point is that this is pointless, but that's ok!

On the GXF driver, a small integrated charge-pump gate driver is used, to drive the gate of the FET at closer to 10V, even if the cell voltage is very low. We drive the FET full into saturation and achieve our desired sub 0.5mR on resistance at full load. The FET chosen for now has a 0.8K/W thermal resistance (junction to case), less than half of the Infineon FET. Anyway my point is, don't just look at the datasheet headline for the R_ds_on of the FET - you need to look at gate drive voltage too (and junction temperature, among others). The very low thermal resistance of the FET helps in heat-sinking in CC linear mode.

Is this necessary? Definitely not at all, but it's fun!

In actuality, this driver is just a simple no-nonsense 10A linear CC driver. I use this flashlight only in this configuration. The DD FET is just for show I guess ;).

In order to optimize the system for lowest possible resistance, I decided not to use a spring. Instead, a solid copper contact is used. I found that solid copper rivets of the right size are the best, and therefore the optimal choice for this setup! The tail spring instead is not used for conduction, but instead for mechanical compression. It is bypassed not with wire, but with a wide copper foil.

Programming is super easy via the pogoProg which I've been using for almost all my LED driver projects: http://www.loneoceans.com/labs/pogoprog/

I also made a simple LED board for the Emisar D4. Nothing special, just works. All the LEDs above look red but it has two banks of LEDs which can be controlled via the A1 and A2 lines on the board. This was designed to fit the LED4power board, not the Noctigon one. My flashlight host did not come with a Noctigon MCPCB, so I can't tell if it will fit or not.

Other Features

Since the D4 is an E-switch flashlight, this means it's 'always on'. So I thought I'd add a little microphone inside.

With a firmware of your choice, it's easy to make the flashlight being able to turn on or off via sound, though I guess I wouldn't recommend it since it can be a fire hazard. I also added hardware debouncing for the E-switch.

It would be also fun to have the light being able to dance to music. I had space on this board so I used an ATmega328 for the micro. Lots of flash, processing power, etc. It's just a placeholder I suppose, easy to pop in any other micro. Having a powerful micro also allows me to do weird things on it, like a FFT for the music system (have it flash to the beat of your dubstep track!), and others. Also, I tried out a new thermal management system with Fuzzy Logic - i.e. the thing some rice cookers use. It seems to work! If it's good enough for making delicious rice, it's good enough for a flashlight I hope, and tracks external temperature just like a PID system.

Finally, for the rest of the build, I started a first build using four Luxeon Vs for the Emisar D4, using a LED board from LED4power. Lowest Vfwd for the package, very high light output. This needs to be paired with a good battery, I use a Sony VTC5A since it's what I have on hand, and is one of the better cells I know of for high-drain situations. For more reasonable things, the CC driver takes care of - true linear CC output from 0 to 10A (or whatever you want to set it) with 10bit resolution.

That's all for now, more to come soon! Don't take this driver too seriously, have a great one you all!

:open_mouth: :exclamation:

Nice job loneoceans :+1: .
Looking forward to seeing the full build.

I like the charge pump idea, wonder if there is room on the GFS16 for it.

The battery is not available in Canada and the super capacitor does not drive the gate at full power for very long.

The driver itself would be perfect for the three volt xhp 50’s. I wonder what the highest voltage it will handle is.

Just wow. Very creative loneoceans. :+1:
I sure hope you’ll open source it as well as explain so others can learn from it. :slight_smile:

What’s the maximum regulated current?

If you want to make it the very hottest hot rod around you can try Luxeon MZ. They have even lower Vf but they don’t look well in Carclo optics, (I’m not sure but I think they are OK in the floodest ones). And there are more powerful cells out there - I’d suggest Samsung 20S.

such a creative project! looking forward to seeing more of it.

The Aux board has not merged cut outs for the LED wires and central hole, even worse there are resistors that can short with battery cable, its better to change that

Just the LED wires have higher resistance than the FET
1m AWG 20 wire has 35mOhms about 5cm used 1.75mOhms, original light uses AWG22

The spring got about 20mOhms

Anyways if the cell voltage gets low there is not much current anyway so FET resistance is not important at all

Awesome loneoceans.

And to think, I was recently complaining about drivers getting stale. Now we are at least seeing new microcontrollers being used here and in the D4v2, Led4Power of course has his driver technology, Mike C is/was developing some interesting drivers, and gchart is always up to something these days.

Brilliant, and glad to see you with another project.

I wondered how long it would be before lexel jumped in here to bash your work. Very very predictably it wasnt long.

Thank you everyone for your kind comments!

Thank you! Can you not purchase that particular battery in Canada? I'd imagine you will be able to buy it from say Digikey, Mouser or Arrow Electronics. There is room on the GFS16 for the charge pump IC, but I don't think it's very practical due to the high operating current on the order of 10 to ~100ish uA IIRC. Though with a 1mAhr battery, it just might be enough. Regardless, this is a big reason why I chose an e-Switch host for this project. The Emisar D4 is the logical choice.

Yes I noticed the 3V XHP50s at the start of this year, but I didn't really want to go to the expense of ordering my own MCPCBs that fit the D4, for four XHP50s, and I'll have to figure out an optic that fit. Otherwise, yes that would make an even more ridiculous flashlight. I'll be happy to design such a board if people are interested though, some kind of mass order? Though I believe Clemence on the forums here is the guru for high performance MCPCBS. :)

Thank you Agro, yes all will be up. This board is really simple, much less complicated than the switching drivers. The maximum regulated current is dependent on how you set up the current sense, but for now I have it set at 10A with 10bit resolution. I'll take a look at the Luxeon MZ. Good tip on the 20S. I thought the VTC5A was very good already compared to the pink 30Q, but the 20S seems even better! :THUMBS-UP:

Thanks for your feedback and concerns Lexel! For sure there can be improvements to the LED board, but it works and fits the MCPCB just fine, so no need to worry. The resistors can be placed better, but I could have also gone for 0201 resistors, at the cost of higher modding difficulty.

The wires do have significant resistance, though I used AWG18 wire I think, with about 1mR resistance.. As mentioned, I don't use the spring for conduction, only for mechancial retention. Why is why a steel spring is better. Copper foil is used for bypassing. Regardless, as mentioned, I think a CC FET driver doesn't make sense from an engineering point of view in terms of reliability, consistency, or safety. Practically speaking, with the right safety features and physical construction, it can make a good flashlight with a fun 'turbo' feature. The main functionality of this driver is a no-nonsense CC linear drive. The DD FET is more just for fun. :) Since we're going for ridiculous here, thought I could go a little further. Having a lower R_ds_on for the FET also reduces junction heating, which keeps R_ds_on low as current flows.

The main practical features of this driver is not the DD FET though. I really like the other features, and I think having a CC regulated driver for the Emisar D4 is not a bad option. Music reaction, is a fun gimmick too.

I'm sure Lexel just wanted to show some concerns with his experience building drivers. :)

Thanks everyone for the feedback!

loneoceans it is very good to see you agian!!

Wow this is the first time I see so many unusual and new feature on this driver ever! This is a lot of unique ideas. loneoceans I have good success with your gxb172 driver and now this makes me want to buy Emisar D4! A question is, do you know if this will fit new Emisar d4 version 2?

I imagine, with Luxeon Vs in your D4 as in mine, that there's not much issue maintaining regulation even at 10A. Heat should be your first problem, and voltage sag eventually your second, not emitter Vf :P

I could not find a source for the battery in Canada, the new battery postal rules make shipping into Canada too difficult it seems. The super capacitor I’m using works really good for the first ten minutes and after thirty minutes the output has really dropped off. Its fine most of the time and I don’t have to worry about over discharging the battery.

I see a lot of uses for this new driver you made, especially if it can handle higher voltages. Would two amps at 4S and 0.5 amp at 5S be possible?

The microphone might be the best part. I don’t know how well Atmega’s can recognize words but if it can identify the word “Beetlejuice” (its a silly movie from long ago) and go into a crazy strobe mode if it hears that word three times I’ll probably have to build four.

You’re probably spot on…apologies for jumping on you Lexel.

Unfortunately the Atmega328 isn't powerful enough to run a voice recognition engine... but would not be hard to make it responds to claps or loud single sounds (again I'm not sure if I recommend this because you certainly don't want to turn this one when it's in your bag for example).

However, you are tempting me to throw in an ARM Cortex M4 on the flashlight and run a voice recognition engine on board... there's certainly enough space ;)

Program the modes with morse code

I myself did the first D4 Aux board with not cut through holes

It just was not so easy to solder the LED wires, even with 0.6mm thin boards, having resistors there that may short out with bad solder skills or equipment isn’t helping

I placed even LVP chip and low side N-FET on it with 0402 resistors with no issues, just route a ring on top side with resistors and back side with LED-

looking on a flashlight with Infineon FET that part is not a weak link, probably even the MCPCB is worse

so many resistances adding up

- springs (even bypassed),

- wires,

- MCPCB,

- copper on driver,

- FET,

- tail board,

- tube

  • soldering

This is more and more interesting! just not too long ago having 1000 lumen in a flashlight is next level. today we have possibility of 100w boost driver and now this one with audio reaction.

loneoceans I will plan to build one of this when your design is ready! I hope it will fit new emisar flashlight! (d4 vers 2)

I guess not long before Lexel copy this driver and offer one with similar design! Let see if lexel will suddenly have charge pump on his fet driver and make a fet and cc driver…

I’m not sure a tiny microphone hidden inside a light would be sensitive enough to work. I don’t know much about ARM’s either, were some of the M4’s IOT capable? Maybe a cell phone could recognize the words and command the driver.

I know its totally off topic but I am curious about your Tesla coils. Do you touch the high voltage, use it to light up fluorescent lamps you are holding or just stay back and watch the light show?