LM2585 driver for LS 5W

[robk]

I need some input on a 5W LS driver I'm trying to design. Here's where I am: It's built, using rough values from National's WebSim. I substituted low value electrolytic caps for the ones they specified. Also, I replaced their feedback circuit with a 10K pot, so it is adjustable. Testing was done with a 13 ohm resistive load and a regulated power supply for input. It uses a LM2585 flyback booster in constant current mode, and from what I can see, it works quite well. Optimized for 7.2V in (6 NiMH), the output current is constant until about 4.5V in, where it drops out of regulation. I was getting up to 81% calculated efficiency at 7.2 V in, a bit less as the input voltage drops. I haven't tried it on an LED yet, for obvious reasons. Also, it is HUGE on purpose as I designed it to fit into a Mag 2D, after cutting the post off the switch you have a tremendous amount of room in there (I included a few small ZLTs just to show I can work with them too :>) . The only SM part is the Zetex diode on the reverse side of the board, because that's all I had at the time. I can adjust this thing to give me a steady 700mA on the load from 7.2V down to 4.5V, of course the current draw from 4.5V is very high. At 6 V, 1A input, it should drive a 5W LS to spec. Here's where I need help:
1. The output cap is 33uF, showing a ripple of about .25V on my cheap scope. Is this acceptable?
2. The output cap gets warm... I know nothing about ESR, or types of caps to be used in this application. It is getting 100kHz pulses at almost 2A. Special type of cap?
3. Is a resistive load a reasonable test for an LS? Or will the LED burn up immediately and leave me crying?

Any input appreciated, I've messed with electronics for 40 years, but it has never been my profession or "area of expertise". Just a hacker.

Photos follow,
Thanks,
Rob

 

[MrAl]

Hello there,

Since you have a scope, you can check actual current
levels flowing into the cap to see if it's getting
too much ripple.
A series resistor works to some degree, with the
scope across the resistor. You get some fuzzy looking
waves, but you can look for the major part of it to see
if it looks too high.

A resistor for a load to simulate an actual LS isnt
considered too good of an approximation. A better
approach is to use a string of diodes also in series with
a small value resistor. You can see the 1 Watt LS
simulator we all used on the ZLT circuit here:

http://members.aol.com/xaxo/page3.html

You'll have to scroll to the bottom of that page.

To simulate a 5 watt LS you'll probably need six or
seven diodes in series, also in series with the resistor.
To find out if you have the right combination, check
the voltage across the simulator and the current through
it with a meter. When the voltage is about 7 volts you
should see the rated current flowing. You
could also simulate a higher voltage or lower voltage LS
by simply adding or removing diodes.
When you connect this to your circuit output, you can monitor
the current on the scope with a small series resistor
(0.1 ohm or less). If the peak current goes above the
peak current rating of the LS you plan to use, then you
risk damaging the LS so you should either use a low esr cap
or try paralleling another cap across the circuit's output.
In any case, you can check the level with the scope and
resistor and once you have it looking good you can then
be more confident that an actual LS will work just fine.
No one knows the exact effect posed by allowing too high of
a peak current to flow--even if it's for a short time--
so it's better to stay within specs at all times.
Once you get the actual LS connected, you can then make the
final check on the peak current through it. If it's too
high, add caps or go to a better cap with higher value.
Be very careful with the connections though because as you
probably know these circuits are almost never forgiving;
if you make a mistake it probably takes out the LS.

To know if your cap is any good you'll have to check the
specs with the company who made it. If you dont know and
cant find out, then maybe you should look for another cap.
If the cap blows short the LS survives, but if it blows
open then the LS could very quickly become damaged.
Digikey has lots of parts that you can order online.

Let us know how you make out with this project?

Take care for now,
Al

 

[robk]

Thanks for the tips. I'll try the series 1n400x diodes as a load. I'll also try upping the value of the output cap to reduce the ripple.
Rob

 

[Reboot]

If your output capacitor is a standard electrolytic type, it won't survive those high frequencies (100khz) for very long. And being as you say it's getting warm, I suspect that's what you're using. I suggest replacing it with a tantalum capacitor.

 

[star882]

"If your output capacitor is a standard electrolytic type, it won't survive those high frequencies (100khz) for very long. And being as you say it's getting warm, I suspect that's what you're using. I suggest replacing it with a tantalum capacitor."
AFAIK, computer power supplies use electrolytic capacitors.

 

[Reboot]

When I said "standard" electrolytic I actually meant the "cheap" kind you can pick up at Radio Shack. They are probably not designed for use in switching power supplies. They will leak at high frequencies, heat up, and eventually POP.
Electrolytic caps are indeed found in computer power supplies. I used to repair them and have replaced my share of cheap electrolytics.

 

[robk]

How about if I "recycle" them from working computer switching power supplies?
Rob

 

[robk]

Actually, you've got me thinking here.... Since this is for a 2D Mag, which makes a great mod with 6 AA's, and once you cut off the socket/cam from the switch, you have LOTS of room in the body for the electronics. So, maybe I can use a large schottky diode from a computer PS as well as the caps. All that stuff is rated for a lot more current than you need, almost impossible to blow out running at 700 to 900 mA. I have a direct drive (thru .5 ohms) 5W LS in a Mag 2d using 6 AA NiMH with a 30mm optic and glass lens, but I would really like to have it regulated for constant output and so it didn't drain the NiMH's too far down.
Rob

 

[Entropy]

It looks like you're using a voltage regulator to drive the LS, which is an EXTREMELY bad idea unless you can guarantee that the switcher won't be able to deliver enough current to severely overdrive the LS. (This is why the MadMax gets away with being a voltage regulator - Even when running "wide open", it can only overdrive an LS by 2-3x rated current, which a Luxeon can handle with good heatsinking. The 2585 is a MUCH beefier regulator than the one in the MadMax though.)

You can add an opamp to the circuit if you want to use a 2585 as a current-mode regulator. Set up a basic noninverting amplifier with 10x gain, and use this to amplify the voltage from a current sense resistor dropping .12 volts up to the 1.2 volts or so that most voltage regulators expect at their feedback pins.

You could also set it for 20x gain with a .06 volt drop across the sense resistor, which would be more efficient but you would have to be more careful about ripple on the output confusing the switching controller.

If your target is a D-cell (maybe even C-cell) Mag, you should have more than enough room to implement such a circuit. I've been thinking on the lines of a circuit like this:

Small 5v step-up regulator to power an Atmel AVR microcontroller (The only AVRs that will run directly off of two cells don't have the onboard ADC and PWM needed for a switching controller), and the AVR acting as a switching controller/dimmer to power the LED itself. Just a matter of the time/money to actually design and build a prototype.

 

[php_44]

robk -

Nice looking board and very neatly done!

I'd for sure replace the output cap - a warm cap is a bad sign and when you close the whole thing up in a light it's prone to fail. If you can't locate a good new cap - then you can use a tantalum cap (I'd feel better with a >=25V rated one at ~100uf or so) from a power supply, or if you can find an electrolytic output cap from a switching supply it will be much more reliable. If it is a used cap - try to get one that's not too old (<=5 years old if you can) - you'll get better life out of it.

Also, the circuit you listed looks like it's set up for constant voltage mode not constant current mode. The pot will adjust the output voltage for you and thus cause the current consumption to go up and down. Just keep in mind that as your LED heats up - it's forward voltage will fall in proportion to the temp rise. As this occurs the supply will maintain constant voltage output - resulting in more LED current. I then suggest, if possible, monitoring the LED current after assembling the whole light. Power up the light and keep an eye on the LED current - as the light heats up it will increase. Make sure that when the light reaches peak operating temp that your LED current isn't too high. If you know the supply efficiency at a certain input voltage, and power the light from a bench supply, you can use the current draw from the supply to approximate the LED current.

LED_current = [(supply current)/(efficiency%/100)] * [(output voltage) / (input voltage)]

Using National's switchcad tool - make sure your supply's output voltage ramps up to the output voltage without overshooting it. Any overshoot during startup could fry your LED. You can also verify this on your scope using an LS simulator circuit.

 

[jeff1500]

robk,

How do you make surface mount circuit boards like the ones you show in your photos? They look nice and clean.

 

[robk]

I'm still mulling over the tips offered regarding op-amp addition, CV vs CC mode, 1N400x load, etc. Haven't had the time to take a serious look at it yet. Thanks all for the assistance.

Jeff1500- I use Eagle Layout Editor to lay the board out, then laser print it to "Press-N-Peel Blue" (http://www.techniks.com/) , iron it on to the copper clad board and etch with Ammonium Persulfate. If I feel ambitious, I use "Tinnit" to tin plate it. I've used this stuff for several years, it's not bad - easier than photo methods, but there are limitations on minimum trace widths.
Rob

 

[Burnt_Retinas]

Rob,

It's great to see another going this way. I too am heading in a similar direction. Just a couple of minor cautions not mentioned yet:

1) Deterimine your ripple current with tantalums and check relative to manufacturers spec. Tantalums hate excessive ripple current and will spew out incandescent flaming material when they go. I have holes in my clothes, burn marks in the carpet and scorch markes on my work bench to prove it. I personally don't like tantalums for this reason, but there is little option for the size you are aiming for.

2) Also watch out for 1N400x diodes getting too hot. I had a string and they got so hot at 1A series'd up that they de-soldered themselves which blew up my CC circuit. Now I have overvoltage shutdown on the circuit I use!.

This is great work you are doing, and great boards too!

Chris