Using constant current source as voltage regulator

[Phisci]

Newbie, so looking for some clairity. If I take a JFET with GS shorted to ground via R and drain to V+, this should give me a constant current (Idss) through the resistor.

Could one use this constant current to create a voltage reference?

 

[Aiace]

drive the current to a resistor. If you need a voltage reference with an output resistence of 1 ohm, just drive the current to a 1 ohm resistor. The voltage you'll got is R*I....it's quite simple.

 

[powerbob]

Yes the Jfet will create a voltage reference. The other factor is how stable you need it to be.

http://www.vishay.com/docs/70596/70596.pdf

But I think most of the time people use a precision shunt voltage reference, kind of a zener diode that is temperature compensated.

LT1004 - Micropower Voltage Reference - Linear Technology

“The LT1004 Micropower Voltage Reference is a 2-terminal bandgap reference diode designed to provide high accuracy and excellent temperature characteristics at very low operating currents.”

+- 4mV initial accuracy LT1004-1.2

 

[Phisci]

OK, thanks. I am quite keen to build a discrete low noise regulator and thoiught that perhaps a cascoded JFET set-up as a constant current source woul be a start?

 

[jan.didden]

A voltage reference mimics an ideal battery, that is, zero output impedance. That means that when the load on the reference varies, the reference does NOT vary (because the 'ideal battery' gives out the same voltage whatever the load variation).

Your current source into a resistor gives a reference voltage that depends on the resistor value (assuming for the moment ideal current source). The load circuit that 'hangs' off your reference also has a 'resistance' so right there that messes up your reference voltage. Further more, if that load circuit's resistance or impedance varies with signal, your reference varies with signal.
So no, your idea for a reference is not a good one.

jan didden

 

[zigzagflux]

IMO, best solution is a CCS feeding a true voltage reference. Best of both worlds.

 

[Phisci]

That's a great answer Janneman and really helps my understanding.

 

[jan.didden]

You're welcome!

 

[CBS240]

I do this often and it works great. But like Jan said, any variance in current pulling off of the voltage reference will vary the reference. So I like to use it in this manner...

Cascode the low Gm J-fet with the high Gm J-fet to get a more acurate Idss. Use a J-fet gate as the load on the voltage reference to steer the differential. (the N-Jfet and the P-Jfet act like a diff amp in this arrangement) If much higher output voltage is needed, you can always add another cascode to the J201. Way better than using noisy Zeners for a voltage reference IMO.

 

[Phisci]

Awesome, I will built your circuit very soon! How do you determine output voltage?

 

[jan.didden]

I do this often and it works great. But like Jan said, any variance in current pulling off of the voltage reference will vary the reference. So I like to use it in this manner...

Cascode the low Gm J-fet with the high Gm J-fet to get a more acurate Idss. Use a J-fet gate as the load on the voltage reference to steer the differential. (the N-Jfet and the P-Jfet act like a diff amp in this arrangement) If much higher output voltage is needed, you can always add another cascode to the J201. Way better than using noisy Zeners for a voltage reference IMO.

Nice. Any measurements for us to compare to a purpose-designed voltage reference driven by a CS? I would be interested in Zout over freq and nosie levels.

Note to Phisci: this is also a reference build from a current into a resistor, but the difference with your idea is that here there is a negative feedback loop that modifies the current source 'strength' to compensate for reference changes with load changes (and a few other things), and thereby stabilizes the ref voltage.

jan