Sadly no, trouble shooting is commencing. I think there is a wiring error, or the MOSFETs aren't performing as expected.

I tested the parts separately, even though I didn't look directly at the light I've got spots in front of my eyes. Both LEDs drew exactly .7A according to my meter, the LED simulator drew .31A from the 9V power supply and the white LED drew .35A. At least one piece of good news.

I'm putting some red electrical tape over that light. If I don't stop it I'll go blind.

 

OK, found one error in the wiring.

Before

After

I'll be back.

 

I would put a shoebox lid over that LED, or just simulate it using more diodes. It's hard to troubleshoot when you're blinded.

I looked over the schematic and layout you just posted; looks good to me. I have a hard time seeing that color of green you used for some reason.

I'm surprised that the current through the "sim-LED" is so low. It should be the same as through the actual LED. Four 1N400x's wired in series with 700mA current should have a Vf of around 3.6v, or around 0.9v each.

Each time I look at the BuckPuck sheet I seem to find new stuff. Either I'm getting Alzheimers, or I'm just not being careful enough when I read through it. Neither is very acceptable. Anyway, in the fine print it says that the 350mA BuckPuck needs 2v "headroom" minimum, and implies that the higher-current BuckPucks may require more. That ain't good.

In a related thread for Spinnaker, I was looking up some LED drivers for his solar-operated LED array, and found some interesting drivers that are pretty cheap, and pretty easy to connect up. It's one IC, two caps, a resistor and an inductor. They don't seem to need as much headroom as the BuckPucks do - and while you have to deal with individual parts, the total cost would be less.

But, I'm looking at a whole 'nother tactic; replacing most everything with a PIC uC. Right now I'm trying to figure out an easy way to do high-side switching of an N-ch MOSFET without adding lots of parts. The Tips 'n Tricks compilation that Microchip published has a lot of interesting ideas in it - but I'm sorta getting off topic here...

 

One nice thing about all those wires, I can try other configurations pretty easily.

The green is supposed to be hard to see. What kind of monitor are you using Wook? I used to use a 19 CRT, but is was fuzzy. I switched to a used 19 flatscreen, the color and crispness is much better. I hate the wide screens coming out.

 

I'm using a Samsung 900DF, 19" flatscreen CRT. It's getting a bit long in the tooth, like it's owner.

 

I understand. I not suggesting a total upgrade, but if you get a chance to replace your monitor with a LCD I highly recommend it. I paid $110 for mine, you might find cheaper. Even with a flawed display (marks and whatnot) they are much better.

It is sitting there flashing while I type this. It works. The MOSFETs are cool as cucumbers. I'm going to let it burn in for a while before I move it to the oscope in the next room. The average draw on the 9V power supply is 0.34A. The PuckBuck gets a bit warm, as does the white LED on its heat sink. The simulator is hot, it shrunk the heat shrink I used per plan. BTW, the white LED drops 3.544V, the simulator 3.04V. Go figure.

TJ, I'll ship it to you as soon as I capture the surge data across the LEDs, then I'll start building the next one.

Didn't I see a thread where someone was trying to flash power LEDs at a slower rate? This would apply for that.

***********************

I tried using the MOSFET to turn the BuckPuck on/off first (like I said, easy to configure), it worked, so I went with the NS schematic.

I didn't realize 10Hz would be so easy to see. My DVM has a built in freq counter, so that is accurate. However, it is maxed out at 9.8Hz, so I'm dropping R1 from 68KΩ to 30KΩ in later builds. This should allow the circuit to reach 20Hz.

All the heat is in the LEDs and BuckPuck, so that plan works. The entire circuit board is cold.

You may be able to go back to two batteries, I'm going to adjust it down to 6V. It was mentioned earlier, 9V will suck em dry though.

*********************

OK, at 6V it draws 0.51A, so there is another benefit to going 9V. No only will this sucker last longer on 9V, but when it dies it really will be all at once. Other than current there was no change in performance.

Whatcha think Wook? Is it a winner?

I do believe this might qualify for the Completed Projects thread.

 

OK, in this setup the simulated LED is primary, the white LED is secondary.

My oscope is OK on voltages, but is way out of cal on time, so I adjusted it for one full sweep is one cycle (frequency is 9.8Hz). Center graticule is ground.


Primary (simulated LED)

Secondary (white LED)

I think the MOSFETs are clean. Not too sure about the BuckPuck, and the scope does seem to interact with the circuit more than I would like.

My boy has declared he wants one.


.

 

Gee, it's weird that the voltage is bouncing around so much. You have it on the drains of the MOSFETs, right? Or where are you measuring? What's your V/div?

Try measuring across the LED+ and LED- leads.

 

That is across the LEDs, each of them. I figure the ripple on the bottom is the BuckPuck switching. 1V/div. The top duty cycle is off because I hadn't tweaked the scope, the duty cycle is 50%.

I suspect the difference between simulated LED and white LED is the knee is sharper on the white LED, a lot sharper.

You'll note there is definite conversion going on there. Don't know how efficient it is, but there is something going on.

 

OK, I went ahead and swaped R1 to 30KΩ, then adjusted R2 from 22.5Hz down to 10Hz. I calculate it will go down to 5Hz.

This is the final schematic and layout for this version.

This drawing is really meant to be printed, Windows does a nice job with their Windows Picture and Fax Viewer. I find for soldering a magnifier head gear really helps. Funny, that stuff used to be so much bigger when I was younger.

I'll get the board and the PuckBuck in the mail tomorrow.