What's All This "Best Trick Circuit" Stuff, Anyhow?

Once upon a time, op amps didn’t swing very close to the positive or negative rails. Even a couple volts away from the rails—that was okay in the old days of transistor-ized op amps.

Hey, that was a lot better than with vacuum-tube op amps that wouldn’t swing within 200 V of the rail. Some op amps could do a little better, but customers never asked us for better 35 years ago.

The LM324 can swing pretty close to ground (–VS) if you have a pull-down resistor—or close to +VS if you have a pull-up resistor. But, of course, it has a pretty big (ac) dead zone when driving its output. Crude. Slow.

Many modern low-voltage op amps have common-drain outputs. This is often called a “rail-to-rail” output, even though it won’t really go to the rail. Martin Giles always razzes me (quite properly) if I talk about these amplifiers as “rail-to-rail.”

Widlar’s 1976 LM10 was the first to have a common-collector “rail-to-rail” output, swinging within a few millivolts of each rail. Most CMOS op amps don’t swing that close. If you have an op amp running on “+5 V dc,” it surely can’t swing to even + 4.9 V if its power supply is only + 4.75 V (i.e., if the supply has a 5% tolerance).

So if you have an analog-to-digital converter (ADC) with a 2.5-V reference, and a 2.5-V full scale, an op amp running on +4.75 V can easily swing up to the + full scale at +2.50 V. But it still can’t swing that close to ground. Even for a 10-bit ADC, you are going to lose a few least significant bits where the amplifier can’t swing low enough (Fig. 1). The output may swing within 12 or 22 mV of ground, but not closer.

The trick is to add a high-impedance diode from (a) pointing to (b), replacing the hard wire. (1N4004, not 1N914). Even if the amplifier’s output can’t swing within 100 mV of ground, it doesn’t have to. To get the output to within 0.1 mV of ground, the diode leakage merely has to go down to 50 nA at VF = 0.1 V, even at your highest operating temperature. Even at 95°C, you can do this with a transistor’s C-B diode. Try it.

The ~2k pull-down resistor in shunt of the output and the diode in series with the op amp’s output make the trick work. It can drive the high-impedance input of an ADC within a small number of microvolts of ground—much better than a “rail-torail” that can only get within a small number of millivolts of ground.

So the circuit of Figure 1 can swing close to ground nicely, but not close to the positive rail. Is that what you needed, bunky? Why didn’t you say so? Try Figure 2. Install one high-impedance diode from (c) toward (d) and another one anti-parallel.

The 74C14 has a lot of “gain,” but not a lot of output offset—and very little power drain in this switching mode. I have seen it drive an ADC within a few microvolts of both rails. One of the advantages is that it won’t over-drive the input of the ADC past the rails. Good feature.

These circuits are not very fast, and not low-distortion, and they won’t drive much of a dc load. Do you want to drive a dc load? If we add another trick, you can drive a load. Go to www.national.com/rap and click on “rail-to-rail driver.” Have fun!

Now, Prof. Kleinrock is a respected professor at UCLA and one of the inventors of the networks leading to the formation of the Internet. AND 49 YEARS OUT, I AM STILL HAVING FUN DESIGNING PIECEWISE LINEAR CIRCUITS! I hope that Lenny is still having fun, too!

BEST NEW RECIPE • For an excellent casserole, boil 1 lb white beans (~Great Northern) in 8 cups H2O, 3 minutes; let stand 2 hours, then simmer an hour. Meanwhile, bake ~2.5 lb of pork roast at 350°F. Cut into 1/3-in. cubes; discard most chunks of fat. Scrub potatoes and cut into 1/4-in. cubes to make 2 cups. Add pork and potato to the beans. Add a 4-oz can of warm or hot chopped chilis, to taste, and one (or two) (or 1/3, if you are trying to not scare the kids) 4-oz cans of hot chilis, chopped fine (3/16 in.). Add salt and pepper to taste; simmer 1/2 hour. RAP invented this Best Recipe of the year: I had a hunch, and it worked out well.