> Or is this going to become a chicken-or-egg discussion?
Chicken-and-egg is exactly how I see this. My point being
that, in nature, quantities tend to come in pairs ---
chickens and eggs
position and momentum
angle and angular momentum
voltage and amperage
electric and magnetic fields
heat and temperature
intrinsic and extrinsic geometry of space
Steve B. and his wife ---
in which the behavior of each element of the pair
depends on what the other is doing. In fact, there is
a wonderful physical theory, Hamiltonian mechanics,
which is based on such pairs and a mathematical
theory, symplectic geometry, which provides tools
for making such descriptions.
To me, arguing which of the two elements of such a
pair is more important seems about as silly as the
Lilliput egg debate because both are needed to support
each other. Rather, I think it is more fruitful to examine
how they depend upon each other.
> Imagine some physical process(es) where A causes B which
> causes C which causes D, D being some effect whose cause(s)
> we're investigating.
Ah, but what about when the processes are interdependent
so that A causes B and B causes A? What would you then say?
As my list is meant to suggest, such situations are quite
common in physics.
As for this issue of control, returning to your original
question, one can, in principle, control either kind of transistor
either way, but, in practice, not all the possibilities are
feasible, as we can see by considering the actual values
involved --- for simplicity I've picked round values in the
right ballpark.
FET
At pinchoff, the gate of my FET is at 1 volt relative to the source
and there is a charge of 1 picocoulomb stored in the gate.
Therefore, to pinch off my FET, I could either place a charge
of a picocoulomb on the gate, in which case it will adjust its
potential to 1 V or, alternatively, I could connect it to a 1 V
battery, in which case, 1 pC will flow to the gate.
However, 1 pC is an awfully tiny amount of charge, which
is why people doesn't go around trying to control FET's by
trying to control the charge, but rather by controlling the
voltage. The two exceptions I can think of are as follows:
1. When using an FET to measure tiny currents such as leakage
through an insulator. For instance, at a current of a picoampere,
it takes a second to fill up the gate capacitor.
2. When dealing with high frequencies, as pointed out by
Rob Jenkins earlier. For instance, 1 mA, 1 GHz A.C. consists
of around 1 pC of charge flowing back and forth.
The remaining 99% of the time, the charges involved are
way larger than a picocoulomb.
BJT
When the maximum current through the collector of my BJT
is 100 mA, there is 1.00 mA flowing in through the base and the
base is at 0.700 V relative to the emitter. Thus, to make 100 mA
flow, I could either push 1.00 mA through the base, in which case
it would adjust itself to 0.700 V or set its voltage to 0.700 V, in
which case 1.00 mA would flow through the base.
In this case, the quantities involved, 1 mA and 0.7 V, are quite
reasonable, but the rub comes when we consider changes in
them. In real life, nothing is perfect, which is why we have
tolerance bands on resistors. Suppose that our control quantity
is off by 1%. If the base current is 1.01 mA instead, then the
collector current will be 101 mA, which is 1% off. However, if
my base voltage is 0.707 V instead, then the collector current
is 132 mA, which is 32% off. Yeeouch! As if that isn't bad
enough, it only gets worse because the dependence of current
on voltage is exponential. For instance, if we are 2% high at
0.714 V, then the collector current becomes 175 mA. Thus,
in trying to use the base voltage as a handle on the collector
current we have about as much control as one would have
control over the scalpel in trying to perform surgery with oven
mitts on. Thus , pretty much the only time voltage control over
BJT's is used is in digital applications where one only needs to
turn the transistor on or off as opposed to analog applications
where the exact amount matters.
Chicken-and-egg is exactly how I see this. My point being
that, in nature, quantities tend to come in pairs ---
chickens and eggs
position and momentum
angle and angular momentum
voltage and amperage
electric and magnetic fields
heat and temperature
intrinsic and extrinsic geometry of space
Steve B. and his wife ---
in which the behavior of each element of the pair
depends on what the other is doing. In fact, there is
a wonderful physical theory, Hamiltonian mechanics,
which is based on such pairs and a mathematical
theory, symplectic geometry, which provides tools
for making such descriptions.
To me, arguing which of the two elements of such a
pair is more important seems about as silly as the
Lilliput egg debate because both are needed to support
each other. Rather, I think it is more fruitful to examine
how they depend upon each other.
> Imagine some physical process(es) where A causes B which
> causes C which causes D, D being some effect whose cause(s)
> we're investigating.
Ah, but what about when the processes are interdependent
so that A causes B and B causes A? What would you then say?
As my list is meant to suggest, such situations are quite
common in physics.
As for this issue of control, returning to your original
question, one can, in principle, control either kind of transistor
either way, but, in practice, not all the possibilities are
feasible, as we can see by considering the actual values
involved --- for simplicity I've picked round values in the
right ballpark.
FET
At pinchoff, the gate of my FET is at 1 volt relative to the source
and there is a charge of 1 picocoulomb stored in the gate.
Therefore, to pinch off my FET, I could either place a charge
of a picocoulomb on the gate, in which case it will adjust its
potential to 1 V or, alternatively, I could connect it to a 1 V
battery, in which case, 1 pC will flow to the gate.
However, 1 pC is an awfully tiny amount of charge, which
is why people doesn't go around trying to control FET's by
trying to control the charge, but rather by controlling the
voltage. The two exceptions I can think of are as follows:
1. When using an FET to measure tiny currents such as leakage
through an insulator. For instance, at a current of a picoampere,
it takes a second to fill up the gate capacitor.
2. When dealing with high frequencies, as pointed out by
Rob Jenkins earlier. For instance, 1 mA, 1 GHz A.C. consists
of around 1 pC of charge flowing back and forth.
The remaining 99% of the time, the charges involved are
way larger than a picocoulomb.
BJT
When the maximum current through the collector of my BJT
is 100 mA, there is 1.00 mA flowing in through the base and the
base is at 0.700 V relative to the emitter. Thus, to make 100 mA
flow, I could either push 1.00 mA through the base, in which case
it would adjust itself to 0.700 V or set its voltage to 0.700 V, in
which case 1.00 mA would flow through the base.
In this case, the quantities involved, 1 mA and 0.7 V, are quite
reasonable, but the rub comes when we consider changes in
them. In real life, nothing is perfect, which is why we have
tolerance bands on resistors. Suppose that our control quantity
is off by 1%. If the base current is 1.01 mA instead, then the
collector current will be 101 mA, which is 1% off. However, if
my base voltage is 0.707 V instead, then the collector current
is 132 mA, which is 32% off. Yeeouch! As if that isn't bad
enough, it only gets worse because the dependence of current
on voltage is exponential. For instance, if we are 2% high at
0.714 V, then the collector current becomes 175 mA. Thus,
in trying to use the base voltage as a handle on the collector
current we have about as much control as one would have
control over the scalpel in trying to perform surgery with oven
mitts on. Thus , pretty much the only time voltage control over
BJT's is used is in digital applications where one only needs to
turn the transistor on or off as opposed to analog applications
where the exact amount matters.