4

Here is a quote from this site:

...higher 'transconductance' - a measure of the performance of a transistor - than silicon transistors. The higher the transconductance, the faster the transistor can switch on and off. That means higher clock frequencies can be supported, and that lower core voltages are necessary.

The above bold marked argument shows up in Oxford dictionary under the example sentences as well: https://en.oxforddictionaries.com/definition/transconductance :

‘The higher the transconductance, the faster the transistor can switch on and off.’

I know that the transconductance is the slope of Iout Vin curve. The steeper the curve the higher the transconductance of the transistor. But that curve doesn't have any time axis. It seems ∆Iout/∆Vin is about DC increments doesn't have anything related with time. Or does it? How is being faster switch can be related to higher transconductance?

There is no time in the below plot(Ic Vbe are DC values since they are capital letters, they are not instantaneous):

enter image description here

user16307
  • 11,802
  • 51
  • 173
  • 312
  • If the input voltage has a limited slew rate, (volts/nanosecond) that provides your time axis. –  Jan 12 '18 at 18:14
  • 1
    Answering this question for MOSFETs is pretty easy since it's a relatively simple, symmetric device. But for the BJT it is a lot more involved. The BJT structure is inherently more complex, with internal nodes any one of which could dominate the performance. It will be interesting for me to see if a fuller answer appears here that includes the BJT. I'd probably learn something from it. But in general, it is true that \$g_m\$ is related to speed for both device types. – jonk Jan 12 '18 at 20:37

2 Answers2

1

Quite often it is the gate-source capacitance of a MOSFET that is the dominant factor on how quickly a MOSFET can be turned on or off. Quite high currents have to be injected into the gate capacitance to quickly change the gate voltage so, if the transconductance is (say) twice as much on MOSFET A than MOSFET B then, to adequately switch a particular load current, you only need to change the gate voltage by half the amount compared to MOSFET B.

This usually results in an increase of switching speed for a given current injected into the gate.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • 1
    The OP read things as applying to the BJT (where it also does apply.) You went to the MOSFET, where this issue is more directly associated with \$C_{gs}\$ and the HF response readily calculated. Can you add something to explain the BJT case, though? I think it also may apply, as \$f_T=\frac{g_m}{2\pi\left(C_\pi+C_\mu\right)}\$ for a BJT (though for BJT \$f_T\$ is also often cited as just a figure of relative merit of the device and not a function of its operating point.) And yes, I know the BJT is much harder to discuss on this point. Just curious if something could be added. – jonk Jan 12 '18 at 20:32
  • @jonk - I made my answer before the OP edited his question to include that graph of the BJT and it's getting late. Additionally the OP's first link is all about carbon nanotubes and the only transistor mentioned that I can see was a MOSFET. – Andy aka Jan 12 '18 at 21:16
  • Okay. I'm fine. It's a complicated subject, anyway. I was looking forward to learning something, is all. – jonk Jan 12 '18 at 22:32
0

Circuit time constants are C/gm. Higher gm reduces the time constants.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46