Electronic – Amplify 0 to 3.3V input to +/-3.3V

You are ill-advised in using the lower 50 mV and the upper 50 mV of an ADC's range unless the details in the data sheet assure you that the zero offset, gain error and reference errors allow part of these regions to be used. It is my experience that you will never find an ADC that realistically specifies 0 to Vref as the full input range even if it says so on page 1 of the data sheet.

Gain error: -

Offset error: -

So, estimate the usable range of the ADC and rework your formulas taking into account op-amp offset errors. I would recommend using LTSpice (free from LTI) to double check your values.

Alternatively just use the range as it i.e. 256 mV to 449 mV - this as a percentage of 3.3 volts (Vref) is 5.8% or, for a 12 bit ADC, 239 bits of resolution. This is 1.26 LSBs per degree and if you want more resolution (not accuracy of course) then take multiple samples and average. Noise will cause the values to dither and you get process gain by this action that gives you a resolution significantly greater than that implied by 12 bits.

If you are unhappy with that just use an amplifier to raise the 449 mV to about 3.25 volts (a gain of 7.238). Your new map will be: -

• 256 mV ==> 1.853 V
• 449 mV ==> 3.250 V

That will use 1.397 volts of the 3V3 range of the ADC or 9.13 LSBs per degree C.

Don't get hung up on maximizing the range and incurring extra errors due to resistor tolerances etc. Think simple and think accuracy/drift.

To put it simply, you cannot have your cake and eat it too with this type of circuit. You need to separate the DC constant-current sink from the AC pulse generator.

1) I have built many circuits like you have but stayed with a DC stable op-amp like a TL072. The secret is to insert a 1 K ohm resistor in series with the mosfet gate. Insert a 10 K resistor in series with the feedback from R sense. Insert a 100 nF capacitor from the op-amp output to its (-) input.

2) Now you have a ultra-stable current sink. Use a potentiometer to adjust Vin, with Vin being sourced by a stable voltage ref or regulator. To span many decades of range use a rotary switch where R sense is and use 0.1% resistors if possible.

3) You can buy ultra-stable .025% resistors made by Caddock from Digi-key or Mouser but they get expensive. Use 1.00 ohm, 10.00 ohms, etc, on up to 1 K ohm. With this circuit the voltage at R sense will match Vin to within .1% or better. For protection insert a 15 volt zener across R sense so when the contacts are open the servo loop is not broken.

4) Now to inject an AC signal do it by using a 100 nF capacitor in series with a 10 ohm resistor and connect it to the mosfet gate. The resistor is to prevent parasitic oscillation of the mosfet. Use a pulse generator to inject signals less than +/- 15 volts p-p into the capacitor.

5) The op-amp will keep the constant current stable but slow wide pulses may cause a 50 uS correction time in the servo-loop.

6) Option 2 is to add an NPN such as a 2N2222 to the gate of the mosfet to strobe it off very fast (collector goes to the gate, emitter to ground, base to signal generator). Minimum of 100 KHZ or the op-amp will become unstable.

7) Personally I prefer the capacitor injection method as it keeps both the AC and DC servo-loop DC isolated and the ability to keep a constant current (when on) is much better. Remember it is best if the op-amp is a stable voltage feed-back type, then inject the pulses. Place a DVM in series with your load if possible to make sure your current is exact without any pulses.

8) Use an oscilloscope if possible to measure the drain pin of the mosfet. Look for signs of saturation by your pulse generator or instability from the current sink being over driven.

9) The following circuit will give you 3 decades of range if a 3 pole rotary switch is used. Amps, milliamps, and microamps, plus your modulation scheme.

^{simulate this circuit – Schematic created using CircuitLab}