Cree LED workshop lighting.

[kibi]

I have been struggling with a silly 2x 18W fluoro tube light on the ceiling of my workshop for quite a while now. I have been wanting to upgrade it with more up to date LED's.
Now, I have used Cree's CXA2011 chips before and they are fantastic, but a bit of a nuisance to drive because they operate at 42V which is quite high. Late last year Cree added some new chips to their CXA range. I went for the CXA2520, which is easier to drive at 36V and produces more light than the CXA2011 modules.
So, it was time to design a driver for the CXA2520. As I am converting my workshop power to off grid, a 12V lamp would be ideal so that I don't have to convert to 240V and then down again to drive the lamp.
Micrel provided the answer for me in the form of their MIC3223. It was cheap, operates from IIRC 5V -20V input, provides programmable constant current up to 1A and can boost up to 37V MAX. Another bonus is that the switch is built in to the MIC3223, so no silly external MOSFETs are required. It switches at 1MHz. It even accommodates PWM dimming control, but you need an external MOSFET for that function of course, but no big deal.
All you need is an inductor, a snubber, diode, a few resistors for programming constant current, high voltage cutoff point and a few capacitors for decoupling.
All of the external components will fit within 1206 packages, so a nice easy, small single sided board layout can be used.
I designed my circuit pretty close to that described in Micrel's datasheet.



I designed the board in Eagle, but I have made it larger that it needs to be because the larger dimension fits nicely inside the case that I intend to use.
I soldered it all on by hand because my PID oven project has ground to s shuddering halt.



This is the LED on a temporary heatsink for now.



At first the circuit powered up and worked straight away - which was a bit disappointing. However, the fun began when the LED started flashing on and off at 1 or 2 second intervals.
After much debugging, I figured that the bottom pad, which is also pin 17 or EXP on the diagram did not flow. I figured this by the higher than expected temperature of the MIC3223.
I gave the board a quick whirl in my trusty frying pan and then everything worked as it should. The MIC3223's temperature dropped too.

At 12V we are consuming 1.7ish Amps. 20.4W.
On the output, we are running at about 35.5V, 545mA. 19.34W
This makes the circuit about 94% efficient.

The LED current remains constant with input voltages as low as 5V and I've taken it up to 18V with no problems. MAX Vin is 20V. With such low input voltages, I could even make a bicycle light with two Lithium batteries Of course the current increases, but there you go.

The LED can be driven harder than I am, but I'm not so sure it's necessary. The current tube light measures 100 lux from the ceiling with the light meter on the floor. The single LED measures 160 lux and 100 lux if it's 1 metre off axis from the light meter. I'll be making four such lights.

Next thing is to do the mechanical bits. I'll update when it's done.

[SeanB]

What are you machining those heatsinks from? Looks nice there Warren, high output with low consumption, and as a bonus driven for free in summer at least, and if you keep the panels snow free in winter.

[kibi]

The heatsinks will be the case that the board goes in. I have not got the cases yet.
PV's are a bit of a way off yet, still need some good batteries. We had another 10cm of snow today, and it's still going. PV's would be a nightmare in this weather.

[SeanB]

Come out for a holiday and some sun..........

[kibi]

I have a friend in Pretoria who I need to visit, but I don't know what's happening with me work wise.
If I don't get a new contract at the end of March then I will come down. If I do get a new contract straight away then I will have to keep working. It's looking very likely that I'll get a new contract though. This is not all bad, because then I'll be able to afford some decent batteries.

[MJLorton]

Very interesting. The one thing I noted was the frequency it's been driven at....this catches my eye as I have so many issues with "strobing"  / rolling lines on the camera while filming. I like using LED lighting but I need to investigate something that's permanently "on" to avoid the rolling lines on the video due to the frame rate and frequency of the LEDs...

[SeanB]

You either drive at high frequency over 20kHz or use DC drive. As well if using PWM use a RC or LC filter to drive the diodes with something close to DC, then you get no strobing. Most of my LED lighting ( which I did DIY in most of the cases) uses DC drive, even if it impacts on getting the best efficiency out of it, just for the flicker removal.

[kibi]

These switch at about 1MHz. I wonder what effect that would have on the camera.
My intention is to shroud this PCB in a metal box which will serve as the LED heatsink. That should help reduce the RF switching noise.

[MJLorton]

You either drive at high frequency over 20kHz or use DC drive. As well if using PWM use a RC or LC filter to drive the diodes with something close to DC, then you get no strobing. Most of my LED lighting ( which I did DIY in most of the cases) uses DC drive, even if it impacts on getting the best efficiency out of it, just for the flicker removal.

Thanks for the info, I might pop my UT81b on the job to see what they are running at.

[SeanB]

At 1MHz it will appear as DC, as the capacitance of the die will provide enough smoothing to make it appear as DC drive. 10nF across the diodes will probaby help with RF filtering though, otherwise you will have a pretty good MW transmitter ( and will be hated by most radio amateurs, though they hate PLT more)

[kibi]

A small update.
Today I completed the preparation of the reflectors.
I found some stainless steel dog bowls on eBay. They have a nice shiny surface on the inside of them and they were cheap. I needed to make a central 35mm hole in the bottom plus some mounting holes outside of that. Now, my lathe is too small to swing the radius of the dog bowls over it's bed, so boring the hole out on the lathe was a no go. As I don't have a punch and die set either, I came up with a plan to do the job on the milling machine.
First, I cut a circle out of 3mm ply wood, the diameter of which is the same as the ID of the open end of the dog bowl. The circle is then bolted down centrally on the milling machine's bed. This will centralise the dog bowl and allows for the hole to bored out centrally and then allows changing of the workpiece easily (I have four dog bowls to work on).



Then the dog bowl is clamped down upside down and the boring head is set at the correct radius.



You wouldn't normally bore straight down into metal like this, but as the metal is so thin, the boring bar has enough clearance. Low RPM, lots of cutting fluid and good tool pressure and the boring bar cut through without too much protest. Stainless steel in interesting stuff to work.



Next, the mounting holes needed to be drilled. Another reason for the wooden circle is that I can wind the bed out to the correct hole circle diameter, then simply spin the dog bowl 120º and drill the next hole and so on. I had marked the outside of the dog bowl every 120º beforehand and lined up the marks on the dog bowl with one on the milling machine table.



Again plenty of cutting fluid and good pressure and the first three holes went through OK. Then the TiN coated everyday drill bit downed tools and refused to cut any more holes. It was then that I realised that should have used the carbide one. From then it was plain sailing for the rest of the holes. Stainless steel 
The edge of the large holes is mighty sharp, so I have dulled it with file. I have a bit of an aversion to holes of around about this diameter in sheet metal because it brings back memories of when I mostly chopped my thumb off on the edge of one such hole in mild steel a few years ago. My thumb has been sewn back on and it works 96% well.   



OK, cool, that's it for now. I have cobbled my reflow oven together too, so I'll get the rest of the PCB's done soon.

[SeanB]

My ratio of drills to holes in stainless steel is 1:1, unless I use a cobalt steel bit on unstressed material. I had to clear out a tapped hole on a 6mm plate last month, took 10 bits to open it up to enough clearance for me to use the next size tap to cut new threads.

[kibi]

It'd have been better to punch a dimple in the metal before going in with the drill, but it was all too easy to spin the bowl around and drop the quill down in exactly the right position without having to figure out where centre punch beforehand.

[MJLorton]

Great work Kibi...and educational, so thanks for the updates.

[kibi]

Unfortunately, I lied about using a CXA2530. The truth is that I am using a CXA2520. My apologies.