A Place Where No Dreams Come True...

Current Topic: High power LEDs can be a bit trickey to drive. In order for them to work reliably for extremely long periods they need be driven by a current source. NOT a voltage source. Or they may create an unsafe operating environment and burn out way before their expected lifetime of possibly 100,000 hours.

The First Step To Any Design Is A Proper Schematic.

Recently I have been using KiCad as my EDA suite. It comes with a very good schematic capture program with a quality design-rule-checker and good part library manager. It includes a proper multilayer Printed Circuit Board design program also with a good Design Rule Checker and part library manager. The PCB program includes 3D modeling, Gerber file generator and Excellon drill file mapper. There is also a fair, but slow, Gerber file viewer to assist in verification. All in all. It works very good. It's free. It's well supported, still in active development and runs great on Linux.

MAX16832 KiCad Schematic

Design And Layout Of The Printed Circuit Board.

This is the first board in a series being developed for the LED Horticultural Lighting project the second being the main processor/controller board. I ended up creating, for this design, a small two layer circuit card based on the Maxim MAX16832 switchmode current regulator device. Except for the two interface connectors all the components are surface mount making for a very small board.

MAX16832 PCB top layer MAX16832 PCB bottom layer

One of the extra features of KiCad is the ability to render a 3D model of the circuit card design. It is not only a cool feature but it is very useful in helping verify that any custom component footprints that may have been created will actually fit the component. On other designs this feature has helped me find component pattern problems before manufacturing the cards which would have been a huge disappointment.

Here's a note to component manufacturers. At the last minute I substituted a few components specifically because the manufacturer did NOT provide 3D models of their components. Some manufacturers also wanted me to create an online account to download their models. Manufacturers need to take special care to provide these models "Free and Unencumbered" as tools for designers. Suppliers need to make these easily obtainable to encourage designers to use their parts. Having many parts to chose from many designers will shy away from parts where the models are hard to obtain.

MAX16832 KiCad PCB 3D Rendering MAX16832 KiCad PCB 3D Rendering MAX16832 KiCad PCB 3D Rendering

Fabricating And Assembling The Printed Circuit Cards.

I had these cards fabricated in China at a company called EasyEDA. They do great work at a very low price with quick turnaround and reasonable shipping time. These boards (2 layer) took about four days to fab with two week standard (cheap) shipping. The boards cost me only $0.30 each for 30 pieces and they actually sent me 31 boards. It's was the first time I used them to fabricate boards and they turned out great. The quality is fantastic. The boards traces are crisp, the Solder Mask is consistant and the Silk Screen is accurate.

So I bought a big pink bag of parts from Mouser, a rather large soldering station from Fry's and began assembling.

MAX16832 PCB bare boards MAX16832 PCB pink bag of parts MAX16832 PCB component parts

With the soldering iron heated up, some very fine gauge solder, a pair of tweezers and a magniging glass I started out assembling.

MAX16832 KiCad PCB assembly tools

Normally I would just hand solder the complete board however the current regulator IC has a thermal pad on the bottom which isdesigned to solder to a ground plane to help dissipate the heat generated from the internal power switch (MOSFET). So I needed to apply solder paste to the thermal pad and the IC leads and bake them in a toaster oven. The first one I tried I did not add paste to the leads thinking I would solder them by hand after the bake. The problem with that was the leads, and the bottom pad, have not only a wicking effect but the surface tension of the solder helps snap (align) the pins to the pads. So without the solder on the pads the IC did not center perfectly with the pins. It was still usable but I did not make that mistake on the rest. The rest of the parts I hand soldered to the boards. It is just to difficult, for my setup, to past and place the parts for baking. It was difficult enough for the regulator IC.

The above picture, by the way, represents the only tools I have available to not only build the prototype but to debug it as well. No DVM... No Oscilloscope... No fancy probes or signal tracers of any kind. Just a lot of experience and good common sense.

Testing The Printed Circuit Cards.

This board is actually classified as a power device. The regulator IC can take as much as a 60VDC input and regulate a 1A output current. That's a lot of power. And. I have a rather large, unregulated, power supply I'm driving it with. I was a little nervous powering up the first one. So... I put on a pair of safety glasses and went for broke. I did of course invite anyone that happened to be around at the time just in case. Don't want to waste a good explosion on just yourself you know. Well... No immediate disaster but I still needed to verify proper operation.

MAX16832 assembled PCB MAX16832 peripherals attached

Being a current regulator the only way to properly test it is to connect a power resistor to the output and measure the voltage developed across the resistor. For this I used a 5W 15.5Ω resistor. The result I got was close towhat I expected but I was off by about 7% on the high side. Because I used a proper current sense resistor on the regulator board I was not able to purchase a perfect fit so I expected to be a little over the target value of 350mA. I actually measured around 380mA. Oh well... That's not really going to spoil the design but it will increase my overall power consumption when I use 60 pieces for a single lighting unit.

MAX16832 assembled PCB MAX16832 peripherals attached

With the initial test complete it was time to try a real LED. I purchased a small quantity of Osram ssl120 1W power LEDs and a few Bergquest LED stars. I was hoping for the ssl80 LEDs but they were not quite available yet. I wired one up and attached it to a heatsink with some thermal grease and, well, it didn't blow up. The current regulator looks like it's working quite well actually. I am powering it with a 24VDC supply and although the input to output voltage is quite high the regulator is not overheating. Success!

MAX16832 assembled PCB MAX16832 peripherals attached

The beauty of this regulator IC is that it has dimming control. The dimming circuit takes a Pulse With Modulated (PWM) signal and gates the output driver with it. Because of the method used it does not reduce the current to the led it simply turns it on and off with the PWM while maintaining the proper current during the on phase. That's important because the LED center wavelength is somewhat dependent on being driven at the proper current level. Additionally the regulator has a current foldback input which can be used with a Negative Temperature Coeffecient (NTC) resistor to back the current off as the temperature increases providing a safety net to catch a thermal runaway and prevent a possible component failure. If used correctly that is.

MAX16832 assembled PCB

For the final part of this test I wired up a few more LEDs and connected them in series and let the system burn in for several hours. Before I commit to wiring a proper light fixture I needed to be certain that I was not going to have any surprises. I'm quite happy with the way this is turning out so far.