Fully differential amplifier, how to bias input?

[thinkfat]

I'm working on a zero-crossing detector with a fully differential amplifier and a comparator. The input signal is from a double-balanced diode mixer. The differential output of the mixer is floating, but I need to bias it so that it doesn't drift out of the common mode voltage range of the opamp.

I'm using a LTC6362 FDA. It has a "Vocm" output, can I use that to bias the input signal path?

See attached schematic.

[TimFox]

Are you sure the IF output of your DBM is floating?  Since the IF output is good down to DC, it is commonly connected between the center taps of the LO and RF transformer secondaries, and often one end is grounded (to the case and low ends of the two primaries).

[thinkfat]

The mixer is a MiniCircuits ADE-1+. As far as I can measure, the IF output is not internally grounded. It hasn't got a shield or metal case either.

PS: I'm actually not sure if I really need to bias the input. The output of the opamp will be centered around VOCM anyway and there is a DC feedback path between outputs and inputs. Therefore, as the differential input voltage will be zero and VOCM = VICM, it should "just work", right?

[TimFox]

The MCL data sheet does not make clear if the three “Gnd” pins are internally connected, but does show the recommended layout with a tight ground pattern to all three.  It also does not make clear which “Gnd” belongs to IF.  My experience with similar circuits dates back to the good old days of split supplies for analog circuits, where the grounded IF common was acceptable.  For high-frequency reasons, perhaps you need a large capacitor to ground on the low IF pin?

[exmadscientist]

Fully differential amplifiers are fun little guys, and very useful. They're designed to drive differential ADC inputs, but they're plenty good for many other things.

The VOCM pin has no direct connection to or effect on the input. You use it to set the average value of the output: it's the Voltage at the Output, measured as Common-Mode. When driving an ADC, you need to drive VOCM to match the ADC's VICM. Many ADCs provide a VICM for you, and the ones that don't just expect that you will be creating it anyway. (Typically VICM is VREF/2 or something along those lines.) As well, most FDAs have weak biasing resistors to default VOCM to midsupply if you don't override them.

VOCM only affects the input through your feedback networks. Because both sides of the FDA are operating in the inverting mode, distortion is low, input impedance is low, and the FDA's actual inputs are often sitting near mid-supply and move little with applied signals. This gives FDAs a lot of common-mode range and makes them very good at level shifting... a useful thing when driving an ADC.

This also means you usually don't have to worry very much about input common-mode range, given reasonable input and output supply rails. The actual calculation is not hard but presents many opportunities for sign mistakes, so I usually just simulate it. Don't expect the amplifier's input nodes to move around much, just like they won't for the common inverting op-amp circuit.

[thinkfat]

Thanks, that was just the confirmation I needed. The amplifier is part of a DMTD system I'm trying to make. It is the first stage in a zero-crossing detector, the input is a 1Hz - 10Hz (heavily distorted) sine wave the represents the offset between two oscillators. I've attached the full (almost) schematic of one channel. The second channel is identical. I'm hoping that the jitter on the output will represent just the two oscillators and not much else