hello,
Bjt speed depends on the base width, Mosfet speed depends on Channel lenght. In verical BJT's width defined by diffusion where as Channel lenght defined by Lithography. With diffusion we can finer dimentions compared to lithography, hence BJT is faster than mosfet.

Thanks,
Raj.

agree with Jerry fully. You need to compare apples to apples. What does it mean when someone says that a BJT is faster than a MOSFET? For any realistic comparison, you need to look at BJTs and MOSFETs in the same or similar technology nodes.

1. BJTs as bulk devices will have higher carrier mobility. MOS as surface device will have lower mobility.

2. Intrinsic gm/Id is higher for BJTs than for MOSFETs. SOI MOS will tend to fare better than a bulk MOS.

3. Depending on make, BJTs have some speed limitation due to base charge recombination time. I think that modern SiGe doped HBTs don't suffer from this limitation anymore.

4. On the other hand, MOS dimensions can be shrunk quite aggressively, and the intrinsic drift time may thus be still not be so far from that of a BJT despite lower mobility.

Again, there are too many variables in question for you to make a clear statement. The fact however is that BJTs are HUGE in size and much more expensive to manufacture in MOS process because they come require additional steps.

So nobody would make or use BJTs in modern processes unless they could make them significantly faster than the available MOS devices. The textbook explanation that a BJT is always faster than a MOS is as banal and missing the point as quoting Moore's Law to say that transistor density will keep doubling over time. Someone has to actually work to make those BJTs faster than the best available MOS in each process generation! Until the introduction of commercial SiGe HBTs and vertical scaling of base width a decade or so ago, it was not self-evident that the BJT would survive for long.

Vivek

How do you define 'modern processes'?  No one is making BJTs in a 28nm process, but in >0.13um BJTs are a much better choice if you are designing for high performance and/or low power.  Of course it comes at the expense of higher cost from more complex fabrication.  Also, BJTs offer much better 1/f noise performance and device matching as Kevin described.  If you want to know what determines the speed of a device, have a look at the ft equations for BJT and MOS.  The Gray/Meyer text has a nice section on device operation without getting too heavy into equations.