Leibson's Law

I’ve been excited about photography since I was pretty young and bought my first SLR, a Canon FTb, in 1972. The FTb was an exciting camera at the time because it had a revolutionary feature, for the time. The lens stayed wide open for focusing and (manual) metering and then automatically stopped down when you pressed the shutter. That was then. Now we’ve got auto everything: focus, exposure, flash control, etc. More important, we’ve completely entered the digital age, which is great for me because it allows me to develop my skills in close-up and macro photography.

I’ve long been interested in closeup and macro photography, but using film I found that skill development was too costly. It’s really hard to see the effects of focus, depth of field, and various lighting techniques when you’ve got the latency of film development between snapping the image and seeing the result. Digital photography short circuits the delay.

Since I got my Canon 20D a few years ago, I’ve experimented with macro shots but didn’t like my lighting alternatives. I’ve tried desk lamps and multiple flashes but I really wanted something more purpose-built. Well-heeled photographers use halogen lamps with fiber-optic light extensions to put the light exactly where needed but I found such equipment to be too far out of my budget, even on eBay.

I decided to build an illuminator using white LEDs. First, I needed the LEDs. Look on eBay and you’ll find you can get 100 20,000 mcd white LEDs for about $10. A perfectly-sized breadboard from the same company is another $3 on closeout. That’s a good start. Then, I started to think about how to drive a bunch of them. The easiest way, of course, is to just add some resistors and run 'em off some voltage. That’s especially attractive because the eBay LEDs often come with free resistors. However, I wanted a way to control the amount of light output.

So started to think about processor-based control. In my last LED project, I used a PIC microcontroller to PWM some LED light strips for undercabinet kitchen lighting. However, a camera doesn't see the same way that an eye does. At high shutter speeds, I wouldn't get much advantage from bang-bang PWM. I needed current regulation and continuous light output.

So my researches took me to National's LM3914. Perfect, I thought. It's an LED bar-graph driver. Vary an input voltage and 1-10 bars light up. There's my variable light output. The LM3914 is not picky about supply voltage. Anything up to 35 volts seems OK. Hard to kill, probably.

I designed a circuit and bought some parts. Picked up some "17 volt" wall warts at Halted Specialties, one of the local surplus shops here in Silicon Valley. However, the wall warts actually run at 24v. Hmm, I thought, maybe I'll regulate the LM3914 down a bit to 9v so it doesn't run at "high" voltage. Wired the thing up. Voila, it works. Each LM3914 output drives a string of five white LEDs in series, for about a 15v drop in the LED string. The rest drops across the LM3914.

Uh oh. String number 9 has a problem. With any supply voltage over about 21v, string number 9 glows dimly all the time. Must have a wiring error. Nope. Pull the IC (ha ha, I socketed it!) and there's no ghosting. No sneak path in the wiring. A little voltmeter work shows that the number 9 output voltage "floats" a bit lower than the others when the LM3914 isn't driving the string. Could it be the lower voltage power supply on the LM3914 not tolerating the higher LED supply? Nah, couldn't be. The data sheet says the outputs aren't sensitive to that sort of thing. I clipped the LM78L09 regulator out of the circuit and bypassed it so that the LM3914 also runs on the 24v supply. Voila! No ghosting. Clearly, you can’t pull that particular LM3914 output pin much above the supply voltage, although all the other output pins are perfectly happy with the overvoltage.

I sent this info to EDN’s analog editor Paul Rako and he forwarded the email to National’s analog guru and god, Bob Pease. Uh, not what I was expecting, but hey it’s great to hear from Mount Olympus every so often. Here’s what Bob wrote after he got back from his vacation:

I looked in the LM3914 datasheet, as soon as I got to my office. In the drawing “Block Diagram of Mode pin function,” it shows a little “battery” connected to pin 11. Assuming there is not a real BATTERY inside the package, that indicates there is some offset to connect to a little buffer.

It IS sorta documented. Exactly what it is, is not defined, but obviously something is leaking some current out of that pin. About 80 to 100 uA. In the paragraph, “Dot mode Carry,” it talks about the current flowing from pin 11. Etc. “If this is bothersome, the simple cure is to shunt LED (at pin) No. 11 with a 10k resistor.” Several Applications do show pin 11 with a 20k up to the + rail.

So, we really did try to tell people that pin #11 is DIFFERENT.

However, it does SEEM that the electrical characteristics do have an error. It says that output leakage is 10 uA max for pins 10-18. That seems to be not true for pin 11. I'm not sure, but I'll have to look into that. I'm not sure who else besides ME, knows much about the LM3914.

If you want 9 or 10 output currents to be well matched at all levels, you can probably get them matched OK at one level, but the current at pin 11 will not match at ALL levels. Maybe you'd best use just 9 outputs?

My solution: Go back to Halted and find a pair of regulated, switching, 18v wall warts. Add 100 ohm resistors to each LED string and eliminate the LM7809 regulator on the LM3914 power supply. The result: a macro illuminator box that does exactly what I want. A little trim pot lights 1 to 10 LED strings for increasing illumination. The whole thing fits in a clear, polycarbonate Hammond box. Here’s the pix, one showing one illuminated string of LEDs and one showing all LEDs on: