Building a hi-current PWM driver -- how?

[Canuke]

I've got a neat item I picked up from Costco -- it's a set of six tube lights with R, G and B LED's inside. It comes with a basic driver circuit that does six types of color cycling, and allows you to pause it at one color.

I'd like to replace that with my own driver that simply has three pots to adjust the duty cycle for the three colors (and another to tweak the PWM frequency, for effects). AFAIK each color draws a few amps, so I'd need a fairly robust setup. I haven't done any circuit design work for ages, least of all in power applications, so where do I start?

 

[Mr Happy]

MOSFET's would be your basic building block for switching the LED current I think. Some of them have on resistances in the 10's of milli-ohms, so they can handle large currents without needing much heatsinking. To keep them cool you have to make sure you switch them on and off very fast, and there is something of an art to that. You have to consider the gate capacitance of the FET and make sure your switching logic has enough voltage and current to drive it. I think there are even special FET driver devices to help with the job.

Anyway, just a few thoughts from a non-expert, but it might help you with what things to read up on.

 

[altis]

Perhaps you need to hookup with Bimmerboy:
http://www.candlepowerforums.com/vb/showthread.php?t=187312

You can use a 555 timer as a PWM generator. I've forgotten how I did it some 20 years ago but I'm sure you'll find a similar circuit if you look hard enough. Hook this up to some FETs and Robert is a relative.

Of course there'll be no feedback or current limit so you'll have to consider if you'll pop the LEDs. If this is a worry then look at some of the PWM controller chips from Maxim, National Semiconductor, Linear etc.

 

[Canuke]

I'm pretty familiar with 555 timers and other means of generating the frequency itself; it's the power side of it that I'm uncertain about. When I was in school for electronics, power applications were given fairly short shrift.

What is the limiting factor at high frequencies? Many PWM applications run at audio-type frequencies that resulting visible flickering -- in particular some automotive brakelights. Why don't they just run at an imperceptible 20kHz -- or 2MHz, for that matter?

I'm going to do some more R&D on this over the weekend -- thanks for the info, guys!

 

[Mr Happy]

As far as I remember from a brief read around the subject, the power control arena is dominated by two devices; MOSFETs and IGBTs. MOSFETs are usually selected for low power applications (few 10s of volts, few 10s of amps), and IGBTs for high power applications (100s of volts, 100s of amps). In between, it gets harder to choose. Operating frequency comes into it too, though I forget where the limits are. I found an interesting paper on the subject that I will provide the link for if I can find it again.

However, I think that FETs have no problem switching efficiently in the 100 kHz range, and I think this is a typical design frequency for switching applications.

 

[2xTrinity]

I'm pretty familiar with 555 timers and other means of generating the frequency itself; it's the power side of it that I'm uncertain about. When I was in school for electronics, power applications were given fairly short shrift.

What is the limiting factor at high frequencies? Many PWM applications run at audio-type frequencies that resulting visible flickering -- in particular some automotive brakelights. Why don't they just run at an imperceptible 20kHz -- or 2MHz, for that matter?

The cars are particularly annothing. They appear to operate in the range of a couple hundred Hz. I know the LEDs themselves can be switched much faster. My LF2 keychain runs at 7800Hz, as low as 0.2% -- or 250 nanosecond "on" pulses. (The high frequency is the #1 reason why I use the LF2 instead of the 100Hz L0D). I believe that at frequencies much higher than that, the capacitance of the LEDs themselves results in significant impedance on the circuit.

I know modern fluorescent lights generaly operate above audio range by converting in incoming AC to DC, then using MOSFETs to generate a >20kHz square wave. This high of a frequency prevents both visible flicker, and audible hum. Some lamps can vary duty cycle to dim the light, as well. They also are able to use much smaller inductors to limit current than the big heavy ones needed at 60Hz.

 

[altis]

Whilst increasing the frequency reduces the size of the capacitors and inductors involved it causes a problem with efficiency. High current transistors cannot turn on and off quickly so there is a little in-betweeny zone each time you switch them. As you increase the frequency the effect of this zone increases.

Traditional PWM circuits for PSUs operate at 10s of kHz. Newer ones at 100s of kHz and possibly even MHz these days. Have a look at some of the datasheets and application notes at Maxim. They are particularly well presented.


http://www.maxim-ic.com/

 

[JimmyM]

Use a MIC1557 oscillator to generate a triangle wave off the T/T pin.
Then feed the triangle wave into the inverting inputs of 3 comparators. Select a quad comparator and just don't use one element. The LM339 is a common quad comparator.
Then use 3 adjustable voltage dividers (100K pots should be fine) to feed the non-inverting inputs of the 3 comparators. The outputs of each of the comparators should connect to the gates of 3 MOSFETs. Select MOSFETs with a low gate charge. This will improve switching speed. Use 1-2.2K resistor from the positive supply to each of the MOSFET gates. Flicker shouldn't be a problem at 8-10 kHz.

 

[Canuke]

JimmyM, I like that idea! It's a good start for some of the ideas I have about what to do with these doodads. I'll have to make up my mind soon on what I want to do, as there are still plenty of them at Costco should I decide I need more, and this is the month I get my cash refund from my Amex credit card -- redeemable at Costco