To build the perfect amplifier that we all want - we need to know how transistor Ft varies with Vce. So we can optimise.
Bob Cordell's book (p22) says Ft droop at lower Vce at hi current, with a "Vce ... as little as 5V ... current ... several amperes" in an example.
The On semiconductor datasheet for the MJL4281A (fig.9) has the reverse, with better Ft at Vce=5 than Vce=10 for the entire current range shown (.1A to 10A)
A net search didn't show much accessible research. I even overworked my brain with the original Gummel_Poon paper in the Bell Systems Journal.
I suspect there's several different effects that interact in a complicated way. Any pointers to more information or an explanation?
Or it could just be a mistake in the datasheet.
Edit: Which is what it is.
Bob Cordell's book (p22) says Ft droop at lower Vce at hi current, with a "Vce ... as little as 5V ... current ... several amperes" in an example.
The On semiconductor datasheet for the MJL4281A (fig.9) has the reverse, with better Ft at Vce=5 than Vce=10 for the entire current range shown (.1A to 10A)
A net search didn't show much accessible research. I even overworked my brain with the original Gummel_Poon paper in the Bell Systems Journal.
I suspect there's several different effects that interact in a complicated way. Any pointers to more information or an explanation?Or it could just be a mistake in the datasheet.
Edit: Which is what it is.
Last edited:
Hi Dave,
some 10 years later, I had a look at the data sheet, and the very same thing struck me. I, too, would have expected the opposite, as parasitic capacitance goes down with increasing CE voltage.
Unfortunately, other manufacturers don't show fT as a function of VCE, and it even seems to be a one off for On Semi.
It could be thermal, but I think these were pulsed tests. So probably a mistake.
some 10 years later, I had a look at the data sheet, and the very same thing struck me. I, too, would have expected the opposite, as parasitic capacitance goes down with increasing CE voltage.
Unfortunately, other manufacturers don't show fT as a function of VCE, and it even seems to be a one off for On Semi.
It could be thermal, but I think these were pulsed tests. So probably a mistake.
Exactly, well put.
But it turned out to be a simple mistake in the datasheet, so we can all run our cascodes as normal.
Best wishes
David
The answer is simple and complicated at the same time... if beta at DC is large, then it can be shown that the BJT unity gain frequency (or the gain-bandwidth product, same thing) can be approximated as Wt=gm/(Cbe+Cbc).
From here, things get messy; obviously, gm depends on the collector current (increases), but so does increase Cbe, so it is not straightforward to conclude that Wt goes up or down with the collector current, but it is now rather intuitive there will be a possible maximum in the Wt(Ic), which is often observed in the data sheets.
If we keep the Ic constant, then gm and Cbe are, in a first approximation, constant. Therefore, Wt increases with decreasing Cbc, that is, increases with Vcb or Vce. So Bob is correct, at least in this small signal approximation.
At large signal, it is even more messy, since large charge injection and recombination can be important. A non linear charge controlled model would apply, but it does not change significantly the trend: Wt increases with Vce, although the rate can be quite different (in general, smaller) from the small signal approximation model.
From here, things get messy; obviously, gm depends on the collector current (increases), but so does increase Cbe, so it is not straightforward to conclude that Wt goes up or down with the collector current, but it is now rather intuitive there will be a possible maximum in the Wt(Ic), which is often observed in the data sheets.
If we keep the Ic constant, then gm and Cbe are, in a first approximation, constant. Therefore, Wt increases with decreasing Cbc, that is, increases with Vcb or Vce. So Bob is correct, at least in this small signal approximation.
At large signal, it is even more messy, since large charge injection and recombination can be important. A non linear charge controlled model would apply, but it does not change significantly the trend: Wt increases with Vce, although the rate can be quite different (in general, smaller) from the small signal approximation model.