0

I am using the gmid method to design a common-source amplifier. The gmid method forgoes the use of the square-law equations for lookup-tables/charts to provide accurate results when designing circuits.

Consider the common-source amplifier shown below that is biased with an ideal current source. It turns out that to meet certain specification, the transistor needs to be biased in moderate inversion. I have found the current needed, the W and L for the transistor. However, the gmid method doesn't tell you what voltages to use. Let's say I want the output voltage to be centered at \$0.5 V_{DD}\$. How do I calculate the DC value of \$V_{in}\$. Since the transistor is operating in moderate inversion, I cannot use the square-law equations/rules. I was thinking of using an op-amp with a feedback, but this seems like a massive overkill. Any ideas how to set \$V_{in}\$?

enter image description here

RussellH
  • 12,496
  • 2
  • 9
  • 34
Leonhard Euler
  • 374
  • 3
  • 10
  • Overkill isn't overkill if it's the only thing that does the job. My advice: ditch the FET and just use an op-amp to achieve what you want or, explain why a FET must be used. – Andy aka Nov 22 '22 at 21:42

1 Answers1

0

You simply can't set Vin in such a way that the FET is biased at half-supply, at least not reliably. The problem is that the gain of a MOSFET with a current-source drain load is so high that even the tiniest variation in input voltage (or threshold voltage!) is going to send the output straight to VDD or ground. Any tiny temperature variation is going to cause the circuit to malfunction.

Such an amplifier circuit only works when it's used inside of a feedback loop. The feedback will then ensure the correct bias automatically. An OpAmp is likely the easiest solution. You could of course design a discrete feedback circuit, too. (It'd be even easier if you just replaced the entire common-source amplifier stage with an opamp.)

Jonathan S.
  • 14,865
  • 28
  • 50
  • Thanks for your reply. If I had used a resistor instead of a current source, would that make it possible to set Vin without a feedback loop? – Leonhard Euler Nov 22 '22 at 22:51
  • Yes, then it'd be much easier to find a proper setpoint. The gain will be lower, though, and you'll also have to use the actual quadratic MOSFET equations. You need to take channel-length modulation into account, as well as the maximum possible variation of the threshold voltage. The output bias voltage will still change quite a lot, though (threshold voltage tolerance multiplied by the amplifier's gain). It'll become infeasible if the stage's gain is greater than 10 or so. (100mV Vth variation would turn into 1V drift at the output). – Jonathan S. Nov 22 '22 at 23:07
  • Look at the datasheet of any Mosfet to see that some are very sensitive and others have low sensitivity even when they have the same part number. The curves in the datasheet are for a "typical" one then all very sensitive and not sensitive ones are different. Your circuit needs some negative feedback to reduce the variations in Mosfet sensitivity. – Audioguru Nov 22 '22 at 23:32
  • @Audioguru OP is designing an amplifier stage on an integrated circuit (hence the W/L parameters), not with discrete transistors. Vth can be controlled better on ICs. – Jonathan S. Nov 22 '22 at 23:53