Photocurrent Bias

[Enthalpy]

Hello everybody!

Some sensors need a huge input impedance: if highly resistive, if capacitive, at low frequencies - reasons vary. Commercial resistors exist up to 22Mohm, uncommonly 100Gohm, and high values integrate badly on a chip; instead, I propose to polarize the amplifier by photocurrents.

The loop can provide a feedback at frequencies lower than the signal, or include the signal frequencies, or set a zero when not sensing the signal - or other uses and their combinations.

Over a resistor, photodiodes have the advantages of a huge impedance up to some 0.2V, especially if used in the photovoltaic mode (=without external bias) which has zero residual current. This reduces the noise, especially where capacitances and frequencies are small.

Over a zero switch, photodiodes advantageously inject no switching charge nor leakage current.

Discrete photodiodes exist with a capacitance <<1pF. Within a chip, photodiodes can be made even smaller, adding very little capacitance to a Mos input, and protecting against static charges. In both cases, an optical attenuator can match their sensitivity to the range of light sources. If >0.2V are needed, the diodes can be in series, more can be connected - but two diodes with a bigger bandgap and small leakage would be preferable.

I didn't try that one, but it can only work.
Marc Schaefer, aka Enthalpy

[Enthalpy]

Photocurrent(s) can make the feedback too, not just the bias of a low-noise amplifier.

From a transimpedance amplifier, this eliminates the resistor and its current noise. The photocurrent feedback only multiplies the source's shot noise by sqrt(2), and this noise is often smaller than the signal.

When the source has essentially no conductance nor background current, as is the case for example with a PIN diode at zero bias, and the amplifying components neither, as MOS practically achieve, the current noise drops to about nothing.

If someone worries about the thermal noise of the huge input resistance: it's shunted by the input capacitance. Increasing the input resistance improves this noise by pushing the corner of the frequency range where the capacitance dominates. Increasing the resistance even reduces its voltage noise shunted by the capacitance, because the thermal current noise shrinks.

In a detector for rare particles for instance, such a stage can await an event for long. If the equivalent voltage noise of the amplifying components is small enough, the circuit adds no false detections.

Marc Schaefer, aka Enthalpy