Low Current Bipolar Devices?

[solis365]

I am looking to build a 1 nA current source. I have a technique to sense the current in the load and a feedback loop with high gain to control the current, but my trouble is with devices.

What should I look for in the datasheet of a BJT to determine if it will function properly at collector currents of 1 nA?

The best way to tell is with a Gummel plot, but datasheets do not often have those printed. Such a plot would show the beta at very low currents. Is there a certain specification on the datasheet that might give good information about this? I have been looking at Collector Cutoff Current, but I am not sure if this is a good metric for the application.

I have been using SPICE2 models of devices to simulate Gummel plots, but I'm not sure if I trust the simulator or the model (Cadence OrCAD 16 - PSPICE A/D) at such low currents. Also, not all devices on the market have published models, especially matched pairs (which I am using).

 

[dick_freebird]

If you have tabular Gummel data you could turn it into a beta-vs-Ic curve.
This is more instructive, although Gummel data is usually not taken
where you'd operate the device.

1nA collector current is mighty low, especially if you have to
operate at extremes of temperature. And vendors tend to
publish curves, when they do at all, only at room temp.

Emitter current density is the key design variable in respect to
beta vs current. It's going to push you to a minimum sized
device, which helps matching none.

 

[solis365]

The only gummel data I can get is from simulation, where I can measure over any current range I like. Once the devices arrive in the mail, I can take real gummel plots. I have an agilent B1500A parameter analyzer available to use so I can probably take gummel data for pretty low current ranges.

The beta-vs-Ic curve would just be a plot of the ratio of Ic to Ib versus Ic, would it not? You can see pretty easily how beta varies with collector current from just a plain gummel plot, just look at the height difference between the two curves.

The system does not have to operate at temperature extremes. It will only be used in a climate and humidity controlled room, which is pinned at 27C plus or minus just a couple degrees. The devices on the circuit board will be kept away from other components that emit heat, if there are any. Since it is only passing 1nA I dont expect that the devices will heat up at all.

I am stuck with using discrete matched pairs in a TO-78 or TO-71 package, so I have no control over device size. I found a couple that are specified to a Collector Cutoff Current of 0.2 nA. Hopefully this was the right spec to look for and hopefully its good enough. I am controlling with a DAC and the system is designed so that 1nA is the LSB step. Since the DAC is only +/- 1LSB accurate, any matching in the system is going to have to be only good to +/- 1 nA. Make that 100pA, or 10%, and all other error gets buried in the DAC's INL and offset error.