Voltage dip on MOSFET switchable current shunt

current measurementltspicemosfetsimulationtransistors

I have the following circuit built up in LTSpice which will be used to switch in different sense resistors for current measurements.

Branches
Switch Mechanism

The mAEn signal switches the M1 transistor on at t=10ms which allow current to flow through the 30m Ohm resistor in branch 2. M3 in branch 1 is turned off the entire time.

Here is a plot of the voltage on Vdev.

Output 1

The dip goes below 3.2V for about 10ns.

Now, if I remove Branch 1 like so.

Circuit 2

I simulate this circuit, there is no voltage dip on the Vdev net as can be seen below.

Output 2

Just cutting the line on the branch 1 path gives the same result (ie the following circuit also produces the same voltage as above with no dropout).

Circuit 3

Why is it that turning on the transistor with less line impedance (branch 2) the voltage drops. This is like having a pipe with a small amount of water flowing, then opening a valve to another pipe with much higher water pressure and the pressure in the original water pipe dropping when you do it? Can somebody explain what is going on here?

What modifications can be made to the circuit to remove the transient voltage drop but have multiple branches?

Best Answer

It's possible that it's just a simulation singularity/bug. I have done a lot of SPICE modelling and simulations and they are quite frustrating indeed. What happens if you put a simple 10uF capacitor from VDev to ground?

the original poster followed this, and fixed their issue.

Related Solutions

Electronic – Defining a nonlinear capacitor in Ngspice

As you say, the behavioral source in ngspice can only generate a voltage or current as a function of other node voltages or branch currents in your circuit. Nonetheless, you can use this, with some extra elements, to produce a nonlinear capacitor like you seem to need:

enter image description here

Here I used a linear CCCS and a linear capacitor to do the integration of the incoming current and track the charge variable. Then a nonlinear VCVS provides the capacitive behavior of your nonlinear capacitor. The gigohm resistor is there because SPICE requires every node to have a dc path to ground in order to obtain a solution; it will not affect the circuit solution significantly.

There's a similar example, using ngSpice syntax on page 89 in the ngSpice manual here.

Alternate solution

Also, be aware, ngSpice offers a behavioral capacitor model, described on page 71 in the manual I linked. The syntax is

CXXXXXXX n+ n C = ’expression’ <tc1 = value > < tc2 = value >

Here you should understand that the capacitance defined is the differential capacitance

\$C \equiv \frac{\mathrm{d}Q}{\mathrm{d}V}\$.

To put your equations in an appropriate form you need to re-work them a little bit. You have

\$V = K Q^2\$.

If you turn this around you have

\$Q = \sqrt{\frac{V}{K}}\$.

From which you can get

\$ \frac{\mathrm{d}Q}{\mathrm{d}V} = \frac{1}{2\sqrt{KV}}\$

Which you can easily implement in the ngSpice expression syntax.

What’s the best MOSFET for switching small currents

"Why was everyone saying this would be so difficult? Am i missing anything here?"

Does it bother you at all that the BSS84 used in your model has a leakage spec that is about 2X the 50nA you are trying to switch?

Single MOSFETs are designed for low Rds for die size, not so much for switching extremely low currents (although tcrosley found one that will almost do what you want). To switch low currents like this, one would use a low leakage analog switch. For example something like DG2012 might work.

Remember that all models are wrong, and it is up to you to figure out if it is right enough to be useful.

Related Topic