Not another How does a transistor amplify thread

[cabraham]

π

Debating how a BJT "really" works is interesting, but these discussions seem to me to be about distinctions that make no difference. All of my professional work has been with CMOS ICs so the only real exposure I've had to BJTs are awful substrate BJTs used in bandgap references (even though they have beta around 1 (!) they work fine for bandgap generators).

My classwork and labwork back in school, though, convinced me that while, physically, the BJT is a current controlled device (transistor action is precipitated by minority injection into the base region) in practice it can for 90% of applications be treated as a voltage controlled device with a finite input resistance (r_pi). Either viewpoint (current vs. voltage) works almost all the time.

What I find really interesting is subthreshold operation of MOSFETs. In that case the output current is exponentially related to the Vgs of the device, like a BJT! That region is used in micropower applications because gm/I (a bang-for-the-buck figure of merit) is higher in subthreshold than in saturation.

One last question for Claude. I don't have any experience in discrete board-level power switching designs... but I'm curious why you would use a BJT to switch a heavy load when power MOSFETs are available. It seems to me the charge storage in the base of the BJT would limit the switching frequency. Since there is no analogous effect in a MOSFET, wouldn't a MOSFET be faster? What benefit does a BJT have?

Nice discussion guys!

I generally use FETs for high power switching. Sometimes a bjt can be cheaper, but the speed is very limited due to minority charge storage times. I think we agree on that point. Also, the hybrid FET/bjt structure known as the "IGBT" provides lower saturated Vce drop than a FET, Being bipolar in nature at the output. But a bjt input would need too much base drive due to beta limitations at high currents. So the FET input results in low gate drive even w/ large cllector currents.

The FET, bjt, and IGBT all provide useful functions and which is most suitable varies with application. For cost sensitive applications, a bjt is really hard to beat. I use FETs more than bjt or IGBT for high power switching, at voltages below 600V, and currents below 30, 50 or more amps depending on voltage. I agree that FETs are really good for switching. Did I explain it well?

Regarding the voltage control model working well 90% of the time, that depends on how often you use a bjt as a small signal linear amplifier. I use the bjt as a switch at least as often as an amp. I also use it as a large signal amp, i.e. emitter follower. In these cases the gm model cannot provide results that make sense, but alpha and beta can. In the switching mode even current control does not work, I use charge control because it accounts for storage time, delays, rise & fall times, stored charge, etc.

The charge control model is best when examining the internal physics of bjt, better than current control. I've always stated that current control is an oversimplified model that works at low speed and linear amplification region, small or large signal. For small signal gm*vbe is equivalent to ib*hfe. Because of r_π being defined as vbe/ib, it is equal to hfe/gm. Since hfe=ic/ib, and gm=ic/vbe, then:

hfe/gm = (ic/ib)/(ic/vbe) = vbe/ib = r_∏.

So in order to compute vbe as a fraction of vin the signal at the input, you must compute r_∏ which requires hfe to determine. So even if you use a voltage based computation, the current gain hfe is needed for computing the gain and collector current. Even a voltage model of the amp stage cannot produce an answer unless hfe is included.

At small signal values, gm can be regarded as constant, not so in large signal mode. Hence ic=gm*vbe gives the right answer as does ic=hfe*ib. But in large signal mode only the current control method works.

For switching and/or high speed, only charge control is reliable.

Claude