1) The article says T5 serves as a surge protector. But if the gate is connected to the bottom of RS (RS-SEL in this schematic), wouldn't the
Vgs of the JFET be 1V? Therefore, the FET would never turn on under
normal conditions, which kind of defeats the current source? I'm
obviously missing something.
They're pulling a bit of a trick here — a JFET can also function as an ordinary diode. For example, look at T9 down below. If a positive surge is applied to K2, the gate-channel junction of T5 will be forward biased, which connects it directly to the output of the opamp, a low-impedance point.
The advantage of doing this is that in normal operation, T6 has very little leakage and very little capacitance, so that it doesn't disrupt the low-current settings.
Remember that a JFET is a depletion-mode device. It will conduct current unless the gate is driven more negative than the channel (relative to either the source or drain terminal) by the specified threshold voltage. With a VGS of -1V, the FET still conducts. The drain-source resistance doesn't upset normal operation because of negative feedback, IC6 will raise its output voltage to the level required to force the desired current through RS.
2) According to the article, all of the source current from IC6 travels through R59 as long as the user-selected current is 100uA or
less. I believe I understand that because at 100uA, Vbe of T6 would be
360mV, which is less than Vbe(on) for T6, so T6 would be off. But
wouldn't R59 contribute a very large error in series with R28? The
combined resistance of R59 and R28 is 13.6k, which would result in a
current of 1V/(10k+3.6k) = 73.5uA. That's pretty far from 100uA.
No, because R59 is inside the feedback loop for IC6, the opamp automatically compensates for its effects.
3) For user-selected currents above 100uA, how does R59 and T6 affect the voltage going into the resistors downstream? Wouldn't they
contribute significant voltage drops that mess up the 1V reference
calculations? I can't figure out an intuitive feeling for how the
resistor and transistor work together here.
Again, because of negative feedback, when T6 conducts, IC6 reduces its output voltage to maintain the correct voltage across RS.
The general principle is that components between the output of the opamp and the feedback point don't matter (within certain limits), because the opamp will act to reverse their effects and maintain the desired voltage at the feedback point. It can be tricky sometimes to understand exactly where the "feedback point" is in some circuits. In this case, it is the node labeled TP6
.
A MOSFET always consumes power from the circuit. It has no mechanism to convert energy from some other form to electrical energy.
Therefore, the currents through a MOSFET always flow from a higher potential to a lower one.
This means, for an n-channel FET, if the drain is biased higher than the source, current will flow from drain to source (through the channel). If the source is biased higher than the drain, current will flow from source to drain (through the body diode).
Now, Lets say I have connected Drain to higher than Source and current is flowing through the body diode, But at the same time I want to turn on the MOSFET Source to Drain channel to allow current to flow from the channel from source to drain parallel with body diode.
If you turn on the FET, you may get parallel conduction through the channel and body diode, but both currents will flow from drain to source, because the drain is at a higher potential.
Best Answer
As others have said, it's a JFET, which behaves similarly to a depletion mode MOSFET.
For a voltage across the device that is much bigger than than the cutoff voltage (about 2V) the current is fairly constant. Here is a typical curve from the Onsemi datasheet:
So, at Vgs = 0, the current for voltages of more than a few volts only varies from about 3.5 to 4.0mA (+/- 7% or so). However the actual current is poorly controlled- it might be as low as 3mA or as high as 7mA (parameter IDSS on the datasheet).
Since it needs so much forward voltage it isn't all that useful a circuit compared to alternatives. You can buy more tightly specified devices in a two-lead package- they are called current regulator diodes, but they also require a lot of forward voltage to regulate well and have never been very popular so they are expensive.