50Hz hum in photodiode amplifier

Hello Erik,

The obvious question is "Does the hum disappear if you cover the sensor". This can be a trick question, since many materials can pass infrared wavelengths - including black electrical tape and your fingers IF there is a large infrared source near by (Incandescent or Halogen bulbs). The sensor could be "seeing" this. I will assume you have done this and the hum is not due to incedent light.

Is the hum steady? Or does it change? Does it change as you move the board around in X/Y/Z directions? If the sensor is covered, and the hum changes as you move it around, then the "hum" is probably inductively coupled. Long ground or signal traces can be responsible for this. Are their any power transformers near by?

Does the hum change as you bring your hand close to the board? If so, then it is most likely electrostatically (capacitively) coupled. One trick is to use the floating output of a signal generator cable, running at a few KHz (or a frequency within the BW of the amplifier) and move it around the board to find the "sensitive" spot. The "probe" can be just a few inches of (insualted) wire on the end of the BNC cable (or a clip lead) to act as a little antenna. Watch for the generators signal on the output as you move the "probe" around a few mm's above the circuit.

Most likely the "hum" is coming in through the power supply or a ground trace (classic "ground hum"). Remember you are dealing with nA's to pA' so any extraneous ground currents will easily show up.

Make sure the diode return path ("Gnd or common") is not shared with ANY external or shielding grounds. Use seporate grounds for shields.

You mentioned a guard - is this a buffered guard? Drivien by a low impedance? Is the guard also covered with a ground plane? A guard is not a ground plane, and a guard should be treated as a signal path. The guard should run between the input signal and ground, with the ground providing protection from the outside world. The guard is usally at a higher impedance than ground, so any "modulation" (noise) applied to the guard may also modualte the input. So do not use the guard as the shield - use ground as the shield.

As to your guard question - the best results are when the input is surrounded by the guard. But you also want to minimise stray capacitence to the input - so there is a tradeoff.

Burying the input in the lower layers can increase the capacitence, and it also can cause problems with triboelectric charges if the board is flexed, stressed or tapped.

Assuming you have the top of the board shielded with an enclosure, place the input traces on the top of the board and place a guard trace next to the input and a wide guard trace on the layer below (at least 3x the width of the input trace). The input trace should "see" the guard in all directions. The guard should only be below the critical input node connections (photodiode, input trace, input pin, feedback elements, etc). Making the guard too big, or as a plane under "everyting", can run the risk of allowing noise form other parts of the circuit to be injected into the guard.

My guess is that the "hum" is getting in through the power supply or bias lines. EIther through the diode biasing circuit or any reference or other biasing line.

Are you "biasing" the diode with an external voltage? Where is that voltage coming from? Normally this is applied through a resistor to the "other" end of the diode. Is there a large filter capacitor at the bias end of the diode? This capacitor should be there for two reasons: 1) Provide filtering for DC voltage and 2): Kill AC current and voltage noise due to the resistor Johnson noise.

Are you using a resistor-divider off the supplies to create a bias volatage (ie: Vs/2). A simple two-resistor Vs/2 divider will take any AC noise on the supply, attenuate it by 6dB, and inject it right into the reference or bias line. The reference node should have as large of capacitor as possible on this node to filter and keep the pole as low as possible.

Another nasty way "hum" can make it into a signal is through the amplifer itself. The amplifer has ESD protection diodes from the input to the power supplies. These diodes have a small parallel capacitence associated with them. This creates small capacitors (<1pF) from the power supply line to the input pin. Any "Hum" or "Noise" can be capacitively coupled directly into the input. Normally this is not an issue with low impedance circuits - but when in the multi-Megohms input impedance territory, this becomes significant. This path renders the amplifers native PSRR completey useless since it comes into the input directly. Keep the supplies CLEAN!

You want to make sure the power supplies sourcing the circuit are as clean as possible. Do not share supplies with other high-current cirucits. The amplifier should have it's own local, dedicated, low noise regualtor.

If you could provide the schematic, and possibly a photo of the layout, then we may be able to help further.

Regards,