Buffer amplifier, minimizing PCB footprint

Question

I have a circuit that needs to pass 22 analog signals through a buffer amplifier before feeding them into an ADC (due to the relatively high impedance on the analog signals). I plan to use non-inverting op-amps in unity gain configuration for this. All signals are under 3.3V and low-frequency.

PCB space is at a premium. General-purpose op-amp ICs use 3 pins per op-amp, but since I know I will be in unity-gain configuration, I am hoping there are some specialized ICs that do what I need and need fewer pins per signal.

Q1: Do you know if there are buffer amplifier ICs that use only 2 pins (total — 1 in and 1 out) per amplified signal? For example, maybe there is an op-amp IC that is internally configured for unity-gain?

Q2: Can you suggest search terms I could drop into a parametric search to look for such things? I'd also take suggestions for specific ICs, because I can work my way backwards from that to a parametric search.

For what it's worth, my prototypes will be hand-assembled, and I am comfortable with soldering SMD IC by hand as long as it's a gull-wing package (down to 0.5mm lead pitch), but a no-leads or a BGA package wouldn't be viable at this time.

Answer 1

This task seems to be high-level analog telemetry.

Given 22 "high impedance, low frequency" signals, why not do this:

• a large capacitor on each signal, to remove EMI/RFI; perhaps even R+C LPF; and you may want 2 Clamp Diodes(schottky) across each capacitor, to protect against voltages/spikes/connection_errors to energy outside the range of 0v/+3.3v

• 3 analog multiplexors, 8:1 that feed into a

• 4:1 analog mux, 3 channels for the signals and one channel for Calibration/Zeroing.

• A single FAST unity_gain buffer, into a

• R+C LPF that stores charge for the ADC to sample

• The ADC

Notice the ADC needs to take a sample, then begin conversion. At that point, with the ADC's internal Sample/Hld now in HOLD, the entire analog signal chain can switch to another sensor, changing channels on an 8:1 and perhaps even the 4:1 as needed. Or implement ZERO or FULL_SCALE(VREF) calibration.

There will be capacitive glitches from the gate_to_channel charge movement in the MOSFETS of analog multiplexor ICs. Assume 5 volts and 20 pF, times two because of 8:1 and 4:1 multiplexors may simultaneously switch and inject charge.

If you want 10 millivolt accuracy (you have given no accuracy requirements, so I'll run this analysis with 10 milliVolts), let us compute the injected charge and then compute how large a capacitor must be placed on each sensor output.

We assume the gate_channel capacitance is 20pF.

Q = C * V = 2muxes * 20pF per mux) * 5volts gate_channel voltage step

Notce this is worst case because as one channel is turned on, another channel (perhaps in another mux) will be turned off. Because of lack of guaranteed matching of gate_channel capacitances, we'll just compute worst case.

Q (in coulombs) = 2 * 20pF * 5v = 00 picoCoulombs

Now what size of capacitor to use on EACH OF THE 22 sensor lines?

Using Q = C * V, we re-arrange (use that algebra) for C = Q/V and we have

• C = Q/V = 200 (picoCoulombs) / 1e-2 (volts) = 200e-12 / 1e-2 = 200e(-12+2)

• C_input_line = 200e-10 = 2e-8 = 0.02e-6 = 0.02 uF

Thus you can install 0.022uF on each of the 22 lines, with guarantee of 10 millivolt maximum voltage upset due to analog Mux switching change injection.

Depending on the "high impedance" of your sensors, the charge may well bleed off before your ADC needs to take a sample of that channel.

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

If you want 3X the stabilization time, then cycle thru the 4:1 mux quickly, and over the 8:1 muxes slowly.