Andrew Holme wrote...
They used lots of variations: always with a resistor between PNP collector
and NPN base; often with a large capacitor from PNP collector to ground;
only sometimes with a capacitor from PNP emitter to ground; and sometimes
with a small NPN emitter resistor. None are anything like rail-to-rail
output. Sometimes there's a diode in series with the potential divider -
presumably for temperature compensation. I also found an example where they
used an NPN instead of the PNP, but enough said.
This is from a 500-900 MHz first LO drive circuit:
.------------------------o------- +10V
| |
| .-.
| | | 100-ohm
| | |
.-. '-'
| | |
| | .-------o-------o
'-' | | |
| | | |
| |< --- |
o------| --- 1n C|
| |\ | C| L
.-. | | C|
| | | === |
| | .-. GND |
'-' | | o-------- out
| | | 1k5 |
=== '-' |
GND | |
| |/
in --------------o-------------|
|>
|
|
===
GND
The circuit has three problems, all easily fixed. First,
the bypass cap should go across the sense resistor, etc.,
because the goal is to get a constant base-bias current
to the RF transistor, unaffected by supply fluctuations,
RF currents, etc. Second, we don't want the servo-loop
gain too high at high frequencies, so some PNP emitter
resistance is in order. Third, we'd like to reduce the
wasted voltage drop across the sense resistor, yet avoid
tempco issues with the PNP, so we need a tempco-matching
transistor. Here's the result of these changes.
.. .---------------------+----+------- +Vcc
.. | | |
.. R1 _|_ Rs
.. 250mV --- 200mV
.. | | |
.. v\| .--/\/\--+----+
.. |---, | 50mV |
.. /| | |/v |
.. +-----+----| C|
.. | |\ C| L
.. R2. | C|
.. 5mA | |
.. etc | |
.. | | +-------- out
.. GND | |
.. | |/
.. RF IN------+-----------|
.. |\v
.. |
.. ===
.. GND