Best pass element, regulation loop for a very low noise 10mA-700mA CC DC supply?

[evb149]

Hi all,

I've got an interest in creating a very low current noise constant and operationally stable DC current type of supply passing somewhere between 10mA and 700mA of current through a ground referenced load which drops ~3V; the supply source voltage might be in the 3.3V-9V DC range selected or pre-regulated as may be optimum for the circuit's function.

I can easily of course get a low noise stable 1.2xV voltage reference and use op-amp like feedback relative to the reference voltage compared to that across a current sense resistor to create a closed loop constant current, though due to the current needed by the load I'd have to select an appropriate "power" pass element (NPN BJT or NMOS FET or so on) to be controlled by the regulator loop's low current output to generate a higher current output for the load.

What are good choices for the "power" pass element wrt. its noise contributions?  Often the noise characteristics of BJTs are specified for small signal RF devices, though one tends to find less data for medium power devices that would have an Ic maximum of say 1A or more and which could stay relatively cool
at the dissipation levels involved depending on the collector current and VCE in regulation.

I've seen some application notes suggesting various model BJTs for reference buffering pass elements though none that suggest why the particular devices were chosen over others that are similar generally.

[T3sl4co1l]

How precise and stable are you talking?  What's the line and load variation (PSRR and output resistance, respectively)?  How low is "low noise"?  What bandwidth -- it's only going to be a good current source at DC, how much capacitance/inductance/complicated reactance can you tolerate, and at what frequencies?

FWIW, TL431 is in the 70uV RMS noise range (I forget if that's 0.1-10Hz or 10-10kHz or what), which out of 2.5V suggests a noisiness of 0.0028%.  You'd of course get worse performance using a traditional current amplifier approach, due to amplifier noise, resistor noise, and environmental limitations (PSRR of the amp itself, bandwidth).  There are better references, and there are worse; this typical case would correspond to a "not trying very hard" ballpark of 19.6uA RMS current noise.

BJTs are typically preferred as they have ~10 times the transconductance of a MOSFET (which is, in turn, maybe 10 times the transconductance of its vacuum tube predecessor).  Either this gives strong amplification, or low impedances (voltage feedback), or high impedances (current feedback).  The intrinsic current mode performance has some limitations (Early effect sets an intrinsic output resistance that can only be improved with feedback; hFE limits collector-to-emitter accuracy, and therefore where you can place the shunt resistor), but it's all around quite handy.

Tim

[T3sl4co1l]

Ah, but those are constant voltage outputs... you need to mirror the output to get a current.

Tim

[T3sl4co1l]

But it can only do that with respect to its common terminal, which means it has to be floating on top of the load.  Which can still work fine and all, but..

An option to improve stability might be a voltage follower sensing the output voltage, adding a constant offset, and using that to supply the ref; so the ref's supply is bootstrapped to its GND + 5V or whatever.

Also, did you have any expectations of output voltage range and compliance?  Positive or negative voltages?

Tim