Project Logs

Collapse

• Were it all started, many moons ago.

  Peter Fröhlich • 05/08/2017 at 11:02 • 0 comments

  This is not really a project log entry, but it is interesting to see were my interest for LED flashlights started.

  Today i rummaged around in the closet, looking for something else, and fount the first handheld light i ever build.

  
  You can finda few more pictures and text here:
  

  http://imgur.com/gallery/VvHg0

• Updated files and documentation

  Peter Fröhlich • 04/29/2017 at 05:22 • 0 comments

  All the new board files have been uploaded to the files section as well as a ZIP of the STL.

  My archive of the datasheets is available there as well.
  

  Reluctantly but for the sake of completeness i uploaded the firmware as well, if you can call that mess a firmware.

  The BOM and a shopping list for the parts along with digikey part numbers and prices are available as well, at this moment only as ODF and not PDF.

  All that will be compiled into one once the next hardware iteration is complete and published a PDF.
  

• Source files and switches

  Peter Fröhlich • 04/25/2017 at 20:21 • 0 comments

  Just saw that the light made it to the Hackaday blog today, two things which came to mind that i have not talked about much, i think.
  

  At the moment there are two things missing on this project and one big thing that needs to be fixed. Lets talk about the missing things first.

  Source files! There are none.

  My firmware is jury-rigged together with bits of string and old chewing gum, when that has been fixed, there will be source code for this.

  No Gerbers either, you may have seen my log entry about that from earlier today, i think i finalized the design and overhauled the schematic.

  Once the boards get here from china, there will be source files for that as well, i just do not want to put something on the web i have not verified myself first.

  And the last bit is about the switches, this is just barely more then a proof of concept at this point.

  The whole switch mechanism will have to be overhauled, it needs to be recessed and fortified against negligent activation.

  After all, this is a prototype at this moment, not a finished product.
  

• It is all about aluminium, metal PCBs and old beer cans.

  Peter Fröhlich • 04/25/2017 at 19:50 • 0 comments

  I have build my fair share of regular FR4 PCBs but the metal core PCB for the LED was a bit different.

  There was a lot of reading and asking around on what to keep in mind when designing one, there have been four iterations and i finally got it somewhat right with this board.

  
  
  The thermal pads are not like the manufacturer recommends, it seems when reflow soldering these they are usually modified for the home marked with small areas were excess solder can escape to.
  

  The NTC should hopefully be sufficiently coupled to the board, the intention is not to keep precise measurements but to just roughly guesstimate that the LED might be starting to melt the 3D printed enclosure and dial back the power.
  

  The silkscreen it a bit more creative on this board.
  

  
  The schematic is right on the board, as you can see there are three LEDs on this.
  

  The outer two wired in series and the middle one in parallel, this gives the option to populate it either with a single or two LEDs to increase efficiency.
  

  Since the LEDs only cost 70 Cents, this is a pretty good deal for a efficiency increase of 30 Lumen/Watt with less thermal problems.
  

  In theory the driver should be able to provide more current to two series LEDs to increase the light output as well, but this will have to be tested in the real world and not on paper.
  

  At the moment the output is 1.65W and the theoretical output of the LED driver should be 3W, at 70% electrical efficiency that the LEDs could be driven at a bit over 2W.
  

  But 70% is the absolute worst case for the driver, with the new improved layout it should get to around 85% efficiency.
  

  But again, this will have to be tested with the new boards.
  

  Soldering this will be a bit of a challenge, i only got a little bit practical experience with this, the colossal thermal capacity of this board will be a challenge.
  

  Connecting the wires to the board as well, since i do not want to use a connector.
  

  I got a PID controlled hot plate and a hot air station, the rough idea is to pre-heat the board on the plate and then re-flowing it with the hot air.
  

  As for the wires, my idea is to use either silicone or PTFE wire to keep the insulation from melting and to solder the wires while reflowing the LEDs.
  

  Apply solder paste, tack wires onto board with kapton tape, apply extra solder paste to the wire and pads, place the LEDs, pre-heat the board, reflow it with hot air.

  And while we are on the subject of LEDs and aluminium, at this point i am dead set on building my own reflectors out of old beer cans.

  The idea is to design a few drawing die and press them out of shin aluminum stock, to keep up with the slight recycling theme i decided to use old beer cans.

  There are pre-made reflectors out there, but were is the fun in that?

  The possibility to waste huge amounts of time and effort into a senseless endeavor were i can learn a bunch of new skills far outweighs the simple process of buying something pre-made.

  If i would want that, i could have bought a simple Nitecore TIP keychain flashlight for 35 bucks that would be smaller, brighter and delivered by the mailman in 24h.

  There were concerns about the efficiency of home made aluminium reflectors and while they are completely reasonable after a few test the performance was significantly increased by home made aluminium tape reflectors when compared to the lens i am using at the moment.

  So, while in theory a home made reflector will not perform very well, real world tests have already shown that it will out-perform the pre-made lenses i have been using.
  

• About the new single sided PCB

  Peter Fröhlich • 04/25/2017 at 19:26 • 0 comments

  A lot of work went into this, mostly though in the background on my custom part library that has been tidied up, a lot.

  The schematic was slightly overhauled and changed a bit, the PG output of the MCP73833 is no longer used, i wanted to route it to the AVR but could not settle on a solution i liked.

  The PG is a pseudo open drain output, that made it a bit harder to wire it to the AVR without adding more parts or using board space. I am sure someone else will point out a solution i missed straight away though.

  So, only the presence of the USB supply voltage is monitored again and the blue LED of the charger status RGB led is connected to the AVR since i do not want it to light all the time.

  That would swamp out the Red and Green LEDs.

  A solder jumper has been added to disable the battery NTC in case it is not present, otherwise the charger will not work.

  The system status LED has been moved between the switches, the idea is to print the center switch in transparent PLA, that should make for a nice aesthetic and will give me more room for the actual button caps.
  

  Besides that, not much has been changed just moved around a bit:

  
  All components have been moved to the top side and all connections to the board to the bottom, the switches have been changed to a THT footprint.
  

  This has been done for two reasons, to locate them on the board and have a known reference to the CAD parts, SMD switches may be a bit crooked on the pads.
  

  But mostly due to the reason that i could not find suitable SMD switches.
  

  These are Omron switches with 500 gram activation force, about twice as much as the usual tactile switch.
  

  Together with a switch lever design that is a bit better thought out i hope to achieve the desired tactile feedback from the switches with a well defined travel and klick.
  

  
  
  

  A bit of silkscreen should help me to wire if up correctly, i might change around a bit of this but i think i am done.
  

  Although it does look a bit boring, the text is a bit small but after checking what it looks like printed and after a few test prints i decided that it is at the lower limit of what is achievable.
  

  But i suspect the print may fail and not be readable, i tried to keep the isolation to pads a bit bigger in case the board house misalignes the silkscreen layer, which seems to happen from time to time.
  

  All silkscreen has been moved around to keep it away from vias as much as possible.
  

  
  
  
  

• Lots of tedious and boring stuff

  Peter Fröhlich • 04/17/2017 at 20:29 • 0 comments

  You may have noticed that i did not yet upload the actual PCB and schematic files and that has a reason.

  They are messy, so the past days i have been tidying them up and have been doing the most uninteresting things.

  The one that needs the most work is, overhauling my libraries.

  They are a mess grown over the last 10 years with found ones from the internet, custom ones i made myself and the stock ones with lots of abused parts.

  Organizing and condensing all that into a few libraries still takes more work then i thought and is a still ongoing process.

  After that has been sorted out, there will be actual updates again.
  

• Reflector replacement and PETG failure

  Peter Fröhlich • 04/10/2017 at 21:44 • 0 comments

  After a bit of fiddling around with the prototype and my LED tester i ran few A/B comparisons.

  The thought of replacing the lens with a reflector gets more and more appealing.

  A lot of light is lost as spill to the sides and the lens can not focus it.

  I ran a few rough experiments with a bit of aluminium and press-form tools out of the 3D printer.
  

  
  Something like this should not be done in one go but i got a few results, even these crappy "reflectors" produce more light output then the lens.

  I will have to put more thought into this approach, but it seems like a good idea.

  In other news, the tests with PETG have been put on hold for two reasons.

  One is, that i have more important things to focus on, the other is this:

  
  
  

  Over extrusion, layer separation, blobs, zits, stringing, oozing, you name it.
  

  At least for me it is not the ultimate PLA replacement, at some point i will have to tackle this again but not for now.
  

• STOP! Hammer Time! (Printing PETG)

  Peter Fröhlich • 04/08/2017 at 19:49 • 0 comments

  
  Today i had the urge to work with something tangible and so i decided to tackle the challenge of printing PETG again, this time with much better results!
  

  There is a bit of a problem with stringing and the supports, the best solution should be to re-orient the part and print it without support.
  

  The parts are much, much stronger then PLA so i had to see how they react to excessive force.
  

  After the application of the calibrated thumb detector i was surprised how the parts failed, despite being printed in layers they behave much more like a solid object when braking.
  

  There are no breaks or separations along the layers, instead the parts shatters, this is not ideal but there was a decent amount of force necessary until they broke.
  

  I think i will be printing an entire set of parts and see how they hold up when assembled.
  

• 3D printing workflow

  Peter Fröhlich • 04/07/2017 at 20:06 • 0 comments

  First you start by drawing the design in a CAD program, i use Fusion360 for that, that looks like this:
  http://i.imgur.com/ge7VVix.png

  Then you save each individual component as a separate file and then combine them in a slicer software, this is pretty much a virtual representation of your 3D printer print volume, i use Cura for that:

  http://i.imgur.com/mZXLXEV.png

  This software slices the model in layers, that can be printed, a 3D printer builds tings by depositing molten plastic one thin layer at a time. This is what a slice through the model looks like:
  http://i.imgur.com/UemM5G8.png

  When this is done the slicer software generates a gcode file, that is a type of universal language for computer controlled machines:

  http://i.imgur.com/Sfq7R8f.jpg

  From here this file needs to be uploaded to the printer, usually this is done with a simple SD card, i upgraded my printer with a web interface, this file is uploaded through the browser. I use Octopi running on a Raspberri Pi with a webcam to monitor the prints from my phone:
  http://i.imgur.com/unM7X4Y.jpg

  This is how my printer looks like:
  http://i.imgur.com/tktsp92.jpg

  Then the bed needs to be prepped and cleaned and adjusted on cheaper printers. Better printers adjust the bed on their own. I just turn a few knobs, wipe the bed with ethanol and apply a thin coat of watered down PVA glue:
  http://i.imgur.com/WMNwxds.jpg

  Then the print job is started and the printer starts the extrude small molten lines of plastic until a few hours later the print is done:
  http://i.imgur.com/wRGq3pL.jpg

  The removal of the parts can be a bit stubborn at times, mostly a tap with a spatula is enough to dislodge them.
  Depending on the part, print and settings the part needs to be cleaned up a bit.
  But i mostly design my parts in a manner that does not require any more steps after the print is done.
  Although, holes need to be drilled out most of the time if i need a precise diameter.
  Due to the nature of the printing, design and slicing process, small internal diameters can be a bit too small.

• Block diagram and efficiency graph

  Peter Fröhlich • 04/07/2017 at 17:06 • 0 comments

  Here is a functional layout of the circuit, separated into blocks:

  The efficiency of the circuit is rather poor at the moment, this is mainly caused by the bad, single sided layout and bad ground routing and should be improved by 10% to 15% with a proper board layout:
  

  
  
  

View all 18 project logs