Voltage vs. Current

Yes, 2krishnanunnis, BJTs are transconductance devices (voltage in, current out), just like tubes and FETs, which means they are first and foremost, voltage controlled.

To answer your question: How to tell?, I'd say the physics and formulas are the key. In the case of the transistor we have the solid-state physics resulting in the rigorous Ebers-Moll formulas, with their precise prediction for collector current I_C from V_BE, namely I_C = I_S (e^V_BE/V_T -1), as against the vague squishy β or H_FE gain, if we try to use current for biasing, etc. Just because you can successfully bias a few BJT circuits with current doesn't meant they're current-controlled devices.

Is a long-tail pair differential-amplifier going to work with current control to the two base inputs? No way. Is a cascode-connected stage going to work with current control to the base? No, of course not, it can't. Is an emitter-follower driving a common-mode amplifier going to work with current control to the emitter-follower's base? No. How about the operation of a current mirror? Hah!

The gain of common-emitter stage is R_L/r_e, where r_e = k T / q I_C, in a tightly-controlled fashion. r_e = 25 ohms at 1mA, and scaling inversely to current, 250 ohms at 100uA, 2.5 ohms at 10mA. Very useful to remember. Where is base current in this scene? How does it help you determine the gain of a common-emitter amplifier? Nada.

Hey, that's Winfield, not Winfred! Anyway, the point is that one is predictable, and the other is not. So for manufacture-able designs, where beta can be all over the map, the best circuits are engineered not to depend on beta, and these circuits work equally well for a range of beta values.

Take for example the popular BC850, whose beta may range from 110 to 800 at 2mA. You can add a suffix letter to the part, A, B or C, to specify selected parts with one of three reduced beta ranges. For example the BC850B has beta from 200 to 450. But 200 to 450, covering 125% change, is still a pretty large range to live with if you're using beta-controlled designs.

What's even worse is that at any given instant your favorite distributor may not have the particular suffix you've specified in stock, for the package you're using, etc. If for example, you had a voltage-driven design that could work with beta from 200 to 800, then your purchasing officer could buy either the B or C suffixes, and arguably the design would be working better anyway, since it didn't depend on beta to begin with.

The difference in design approaches is not hard to master. The voltage-based methods are what we teach in our best-selling book, AoE, which is meant as much for newcomers as for skilled engineers.

As a young engineer struggling with my circuit designs in the mid 60's (remember the h parameters?), a wise experienced teacher sat me down with a single sheet of paper for 30 minutes and preached the value of Ebers-Moll and r_e. What's the gain of a common-emitter amplifier stage? Easy, R_L / r_e. And r_e is 25 ohms at 1mA.

Oops, r_e = V_T/I_C = 25mV/I_C, and changes with current, causing distortion? OK add a degenerating resistance, R_E - now the gain is R_L / R_E+r_e, and I can quickly calculate my gain and predict the distortion level for my choices of R_E. Try that kind of easy analysis with the β approach.

So I changed my design approach and suddenly things start working as predicted. That is to say, I could reliably engineer instead of empirically-develop my designs. I could also safely do more complex designs.

Consider circuit A. Its input impedance Zin = β * R_E+r_e in parallel with R₁ and R₂, and the base current is IC / β. A current-control-based design will stipulate a high beta and choose high values for R₁ and R₂, increasing Zin. It'll likely wink at the base current I_b.

By contrast, a voltage-control design will assume a horrible low beta and a high I_b, and choose low values for R₁ and R₂, decreasing Zin. In this fashion voltage-control is enforced, and design predictability is enhanced.

modelaero said we "modify the input impedance to give the impression of voltage control," but he should have said we lower the input impedance to force voltage control. But this doesn't mean one has to live with the lowered impedance.

For example, consider our common-emitter amplifier, circuit A, where we lowered the biasing resistances so as not to depend on beta. But we could instead use circuit B, and add an emitter-follower transistor running at say 1/10th the I_C current (low enough not to add significant voltage noise, e_n), thereby allowing us to increase the bias resistances by 10x, without sacrificing voltage control. The cost? one transistor and one resistor. But we get Zin back to where circuit A might have been, or maybe even 2 or 3 times better. And we keep our goal of good engineering predictability.

I have to admit that the voltage-based approach often forces one to use more parts, often an extra transistor here and there. But the aforementioned BC850 only costs 2.6¢ in quantity, so what are you complaining about?

Now, I implied and showed in circuit A that a beta-based design will have poor bias stability, but we know that's not always true. That's because beta-centric designers can't and shouldn't use circuit A, they use circuit C.

Instead of a tiny 50-ohm emitter resistor for a 1mA, they'll use 1k, etc., for a volt or so of voltage drop, and thereby restore bias-point predictability to their circuit. To get the lost gain back they'll add a big ugly electrolytic capacitor with the 50-ohm degenerating resistor.

Which circuit do you prefer? I prefer B with its extra transistor, to C with its ugly BFC capacitor. And I'll bet the BFC costs more than 2.6¢.

And, hah! As a bonus, our preferred circuit B has a greater output swing.