What's All This V_BE Stuff, Anyhow? (Part 2)

I've been debating with a guy who argues that a transistor won't work as a transistor unless its V_CE is bigger than its V_BE (see figure) . He keeps reading this in books. Also, he points out that if the base and collector are nominally tied together to make a diode, you might think that it's okay. But actually, he says, the I × R in the collector path makes the V_CE lower than the base voltage, so it won't work. Well, I've been looking in some of those books, and they sometimes do say that. But when they do, they're wrong.

When a transistor's V_CE is slightly less than its V_BE, it keeps right on working like a transistor. Can I prove this? Sure. Look in the NSC linear Databook at circuits such as the LM10. The LM10 wouldn't work on a 1.1-V power supply, if the transistors aren't working well with V_CE as low as 350 or 250 or even 150 mV, which is far below V_BE. Of course, you have to be a good engineer to make these circuits work well.

NSC guys (like Bob Widlar) have been doing this for 40 years. Look at the V_CE curves of any transistor. When V_CE falls below V_BE, it's not a disaster. Put a transistor on a curve-tracer. Apply a bias like 1 µA per step to the base. When you change the V_CE from +1.0 V to 0.6 V to 0.5 or 0.4 V, I_C doesn't change much, does it?

Okay, maybe when you get V_CE down to 0.35 V, the gain starts to degrade some. But above that, at room temperature, it's not a big deal. There is no demarcation between V_CE > V_BE and V_CE BE. No inflection. The beta doesn't even change more than perhaps 2% per volt, and it does so smoothly.

Now run the temperature up to 125°C. Can you design a circuit that works up there? It's not easy. But if you don't need a lot of swing, some specialized circuits work just fine. Look at the LM4041-1.2 or the LM185. Many of their V_CEs are about 0.3 V, yet they work hot and cold.

How about 160°C? How about 260°C? I can't, but Widlar could, and did, in the LM12. After all, in the old days, a pentode could run with a very low VPLATE—much lower than VSCREEN. It's hard to comprehend this, but after a while, you get to understand and believe it. It's an analogous situation that the output voltage is so low, you can't believe it will work. But it's true. It does work.

NEXT TOPIC
When V_BE = approximately zero, changes in V_BE certainly have no effect on I_C, right? Wrong. In Widlar's LM12, some of the transistors are so biased, when V_BE = 0, the V_BE can still influence the collector current (and vice versa) whether the V_BE is a few millivolts positive or negative.

Admittedly, you can't see this easily in a silicon transistor at room temperature. But you can see this in a silicon transistor at 220°C, or in a germanium transistor at room temperature, which is about the same idea. Go ahead and measure it. When I did, I was impressed by Bob Widlar's brilliance.

Also, the beta of a transistor can still be important, even when V_BE is about zero. That's because as V_BE moves up and down a few millivolts compared to zero, the base current needed may be small, but finite—not negligible. The base current and its changes are necessary. And if you start at I_B = 0 and pull the base negative, the collector current can decrease.

I must remind you that high-beta transistors (300 and up) still have disadvantages in terms of voltage gain or mu. When the beta gets too high, and because mu is inversely proportional to beta, the voltage gain is hurt. I remember a test that asked how much voltage gain a particular amplifier design has. The answer was supposed to be 20,000. But the gain was really 9000, as the betas were too high and the Early Effect was too strong. I passed the test after I explained my solution.

As a rule of thumb, I use mu × beta = 2 million. On some devices, that product is only as good as 1 million, or even 4 million on LM194. If the beta gets better, the mu = 1/h_RB gets worse, and the voltage gain suffers. Be careful not to allow in transistors with too high beta in circuits where poor mu could cause poor performance. Beta is often important. Too much of it can do harm. So can too little.

Part one appeared in Electronic Design's Analog Applications Supplement, June 26, 2000. Go to www.national.com/rap/Story/vbe.html.

Comments invited! rap@galaxy.nsc.com —or:
Mail Stop D2597A, National Semiconductor
P.O. Box 58090, Santa Clara, CA 95052-8090