Why do MOSFETs scale better than BJTs?

[DragonPetter]

I've read in several places that MOSFETs scale, while BJTs and JFETs do not. I'm curious how a MOSFET scales better, and what is meant by "better", as in do they lose frequency response, power capability, etc. more so than MOSFETs when scaled?

Obviously to me a clue is that both a BJT and JFET use PN junctions while a MOSFET does not, so I am guessing it has something to do with how PN junctions are scaled.

 

[yungman]

I have no idea what is scale. Please explain. Do you mean the dimension scaling for current ratio? If so, BJT is superior in this than MOSFET because the Vbe of BJT is much more predictable than Vgs.

The PN junction of JFET and BJT function totally different, one is reverse biased and BJT has two where one is forward biased and one is reverse. There is no similarity between the two. JFET is much close to MOSFET.

 

[DragonPetter]

I have no idea what is scale. Please explain. Do you mean the dimension scaling for current ratio? If so, BJT is superior in this than MOSFET because the Vbe of BJT is much more predictable than Vgs.

The PN junction of JFET and BJT function totally different, one is reverse biased and BJT has two where one is forward biased and one is reverse. There is no similarity between the two. JFET is much close to MOSFET.

I mean scale as in shrinking the geometry and size of the device, like making MOSFETs smaller so more will fit in a given area. Everything I have read has said that MOSFETs scale better but does not say why.

And I understand that a JFET operates under the same mechanism as MOSFET, but its composed of pn junctions rather than an isolated gate, and it can act as a diode when Vgs is forward biased on the gate, so I meant that way that it is similar to a BJT.

 

[yungman]

Yes, you can make MOSFET much smaller because it is really a rectangle block with the gate laying on top for depletion FET or a simple NPN block with the gate on top of the P for enhancement mode ( something like that!) The voltage cause the inversion layer to conduct. The structure is a lot simpler than BJT and you can make it a lot smaller. I am not an expert in semiconductor, but I believe the enhancement mode MOSFET has PN junction, but it is a lot easier to fab than the BJT.

I am not sure about JFET why it cannot be made smaller.

 

[es1]

I can't think of any theoretical reasons why a BJT cannot be made as small as a MOSFET. A MOSFET does takes less fabrication steps which I guess makes it simpler to make smaller.

My only guess is no one has put as much thought into shrinking BJTs because BJTs are not as good for digital due to their lower input impedance.

I think probably what they mean is the gain reduces more quickly with feature size for a BJT than a MOSFET. This makes a smaller BJT less useful than a small MOSFET.

 

[yungman]

I actually read about the reason, it is the simpler in fab. It takes more steps to fab a BJT. BJT are fab vertically with lightly doped collector first, then heavy dope on top to form the base. Then the emitter in the middle of the base tub. Where MOSFET is pretty much lateral particular the depletion mode where all you need is the N or P body and an insulated gate on top in the middle. The more step you need in fab, the bigger the size of the component because you really cannot absolutely control the boundary of the doping.

Even for enhancement mode MOSFET, the NPN are lateral instead of on top of each other like BJT. The gate just deposit on top of the P so you can apply voltage to create the inversion layer on top of the P.

I think the one most important thing that MOSFET is better for digital circuit because BJT will be working in forward forward mode and has reverse recovery time. Even if you use Schottky diode ( like the AS family), there is still significant propagation delay compare to MOSFET. So if you use MOSFET in the internal of the digital IC, they run really fast. You can see those processor are running in over 2GHz.

Only reason why the discrete CMOS logic don't seem to be that fast is because the internal MOSFET that can run so fast has no drive capability. Before driving out to an external output pin, multiple buffers are needed with progressive larger size transistors. When the MOSFET get bigger, the input gate capacitance become a gating factor and slow the circuit down. that is the reason why the processor can run in over 2GHz inside, but any external I/O can only run is a few hundred MHz at best.

 

[jsgruszynski]

BJT's don't have a nice scaling formula like MOSFETs. This is due to a number of reasons but the biggest is how BJTs work vs. how MOSFETs work.

BJT performance is primarily defined by base-emitter junction width which doesn't scale much with planar dimensions. The bandwidth is primarily defined by how well you control diffusion or ion implant rather than photolithography. In contrast, MOSFETs improve bandwidth as gate lengths and oxide thickness drops.

Added to this, there isn't a direct planar (W, L, Area) relationship to power consumption with BJTs. In contrast there is for MOS.

The net results is that BJTs haven't budged much in performance (dropping size reduces some capacitance but also reduces drive) while MOSFETs have scaled in speed and power with photolithographic scaling. This is also part of why you can now build CMOS op amps that rival BJT op amps in certain performance corners.

 

[DragonPetter]

BJT's don't have a nice scaling formula like MOSFETs. This is due to a number of reasons but the biggest is how BJTs work vs. how MOSFETs work.

BJT performance is primarily defined by base-emitter junction width which doesn't scale much with planar dimensions. The bandwidth is primarily defined by how well you control diffusion or ion implant rather than photolithography. In contrast, MOSFETs improve bandwidth as gate lengths and oxide thickness drops.

Added to this, there isn't a direct planar (W, L, Area) relationship to power consumption with BJTs. In contrast there is for MOS.

The net results is that BJTs haven't budged much in performance (dropping size reduces some capacitance but also reduces drive) while MOSFETs have scaled in speed and power with photolithographic scaling. This is also part of why you can now build CMOS op amps that rival BJT op amps in certain performance corners.

Thanks so much, that was the kind of information I was looking for.

 

[yungman]

MOSFET has no limit in speed, problem is the drive needed to overcome the input capacitance. You can get higher and higher speed when yo shrink the size that lower the capacitance, but then you loss drive. When you interface to the outside world, you need multi progressively larger transistors to get the drive up. That slow down the speed when interface to the external world. BJT have saturation problem regardless of size. Even with schotky diode is not going to help that much.