0

My objective is to realize a current splitter which could split current from a wire A in the range of pA, nA equally to wire B,C.

enter image description here

I was initially sure what this type of circuit would be called, but google redirected me to Current Conveyors page - https://en.wikipedia.org/wiki/Current_conveyor. (please correct if I am wrong here).

A CC-1 circuit is shown below:

enter image description here

I could correlate (X=A,Y=B,Z=C). However, I am not sure if it would support low currents (pA,nA).

  • Voltage Range: (0-3.3V)
  • Frequency Range: 1-100 kHz.
  • Loads: Current sensing amplifiers, each on B and C. (Assume Input is a gate capacitor of MOS and hence high impedance)
  • Reason on why I am doing this: To compare results of 2 current sensing amplifiers.

Any hint/guide on above will be helpful.

sundar
  • 436
  • 1
  • 6
  • 19
  • 2
    More details would be helpful. Voltage range? Frequency of operation? What are the loads for the two currents? And why do you want to do this? – Elliot Alderson Mar 03 '19 at 21:24
  • @ElliotAlderson Thanks. I have added the 4 points in the question. – sundar Mar 03 '19 at 21:35
  • Is it not possible to just connect the inputs of the two current sense amps in series with each other (It might not, but it depends on exactly what type of current sense amplifier you're using)? – The Photon Mar 04 '19 at 17:13
  • @ThePhoton If its voltages, yes. However, current will definitely split in this case. Also, I fear that it will split non-linear depending on the impedances at two sides. – sundar Mar 04 '19 at 17:40
  • Why not simulate it? http://tinyurl.com/y2hxe4ko Do you know gm and Vt ? – Tony Stewart EE75 Mar 04 '19 at 20:58

1 Answers1

1

Here is a cheap approach, but you need access to a foundary with long-channel FETs. The purpose of using a foundary is the access to layout of matched (interdigitation of alternative strips) FETa and FETb, for matching during implanting and, later, thermal matching during operation.

schematic

simulate this circuit – Schematic created using CircuitLab

How fast can a current splitter function? Or, how much charge is needed?

Assume the FETs are 1U by 1,000U each. Area of 2 FETs is 2,000 (micron)^2 which is about 4 picoFarads, in 0.6 micron process (140 Angstrom oxide). To change the voltage by 0.005 volts requires

Q = C * V

dQ = C * dV [assuming C is constant, and some FETS do vary near Vt]

dQ(coulombs) = 4pF * 0.005 volts = 0.02 picoCoulombs

At 1pA, this allows about 1pA/0.02pC or maybe 50 cycle-per-second operation.

========================

So what to do? buy tiny matched same-die JFETs, tweak the gate up or down by a few milliVolts on one of the FETS, so the current splitting is accurate.

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46
  • 1
    By the way, one picoAmp is 1e-12 amps, or about 10Million electrons. Thus 100,000Hz has only 100 electrons to play with. – analogsystemsrf Mar 03 '19 at 22:30
  • Thanks. Just one verification - Do you feel any nonlinearity that might occur during current copying? (For example, A= 1 pA input; B=0.9 pA, C=0.9 pA outputs; with ideal result being B=1pA, C=1pA) – sundar Mar 03 '19 at 22:50
  • If the DRAINS of he 2 FETS are at the same voltage, the current should be evenly split. – analogsystemsrf Mar 04 '19 at 04:49
  • Is it correct to say that the FETs need to be very well matched for this, and preferably on the same die? That might be worth including in the answer. – Hearth Mar 04 '19 at 14:01
  • @analogsystemsrf I went through research papers and it seems that copying currents (nA,pA) is possible only till kHz ranges. Is it true? I can also post it as a different question. – sundar Mar 04 '19 at 16:42
  • 1
    @ Sundar The FETs have non-zero channel-to-gate capacitance. Lets assume the FETs are very-long-channel (Width 1 micron, Length 1,000 micron) in 0.6 micron-minimum-gate-size process. I memorized 440 square microns of gate will produce 1pF of capacitance in 0.6U process. We have about 2pF. Times 2, for the dual FETs. How many electrons needed to cause 5milliVolts deltaVoltage in 4picoFarad? – analogsystemsrf Mar 04 '19 at 17:13
  • Any reason for a down-vote? – analogsystemsrf Mar 04 '19 at 17:24
  • @analogsystemsrf So, you mean that this can work till 50 Hz? – sundar Mar 04 '19 at 17:55
  • @ Sundar Using the assumptions I gave you, yes. – analogsystemsrf Mar 05 '19 at 02:43