Quote:
Originally Posted by Argus25
However, my main point was that if the signal current in the load is generated by the lower transistor, and passes unchanged through to the load, then overall, the voltage gain of the cascode circuit, is attributable to the lower transistor, not the upper one, so I simply disagreed that the lower transistor contributed no voltage gain.
|
Agree! Overall voltage gain of the cascode is gm (lower) x load resistor (upper). The gm of the upper device does not feature in the calculation.
More accurately, the above formula should be multiplied by:
(hfe (upper) -1)/ (hfe (upper)) to account for base current of upper device. But as that's so nearly 1, we can forget that - Argus25 you made that point.
Where gm of the upper device DOES come into play, as we have seen, is determining the magnitude of the (small) voltage swing at the midpoint. And this small voltage is what we have been debating. Its significance is that it has quite a big impact on Miller fed-back capacitance. I've just been playing with paper and pencil the last couple of days, but David RW has obviously 'been there, done that.'
Quote:
Originally Posted by Argus25
For the transistor case, Horowitz and Hill (pg 103) describe the upper transistor (they label as Q2) being interposed in the collector circuit of the lower transistor (Q1) to prevent Q1's collector from swinging (thereby eliminating the Miller effect) and in their words (which I believe): "passing the collector current through to the load resistor unchanged". So I remarked "pretty well rock solid"
|
H & H is a super book, full of great engineering approximations and tips which work well in practice. But one does do well to challenge things, because that leads to greater insight. Miller effect in a cascode is vastly reduced, but not entirely eliminated. Where I see 'pretty well rock solid' my thought then is, 'how "pretty well" exactly?' and a bit of analysis has shown that it's 'pretty well compared to output, but not compared to input', being equal - and equal whether valves OR transistors.
If the middle node could be TOTALLY rock solid, then Cin = Cgk + Cga (capacitances of lower device) but with the middle node moving as we have found, Cin = Cgk + 2Cga. If Cga = Cgk = 1.5pF (typical values for triodes), then you get Cin = 3pF with the invalid assumption that the middle node is rock solid, and 4.5pF if you do it properly. That's a 50% increase!! And the same with transistors.