Operate a current feedback op amp from a single supply

[matthuszagh]

I've got an application that I'd like to use an LMH6702 for, but I'd prefer to operate it from a single supply, rather than a split supply.  What sort of things should I look out for when attempting this? AoE discusses single supply operation (4.6), but not specifically for CFB opamps. Any other good resources to look at? The LMH6702 indicates single supply operation is possible, but provides no detail.

I'm using the amp in a non-inverting configuration. Of course I'll use blocking caps on the input and output. My initial thought was to bias the non-inverting input to mid supply with a voltage divider and then supply the low-impedance common node for the inverting input with a zener or zener equivalent, such as a TL431.

[matthuszagh]

I think something like this should work. U2 is a TL431. AoE x chapters says to watch out for capacitance at the inverting input, but I think with R1 there I think it should be fine, even if I place a capacitor across U2.

My primary concern now is that U2 is pretty noisy. This is designed to buffer a fairly low noise 10 MHz signal - noise floor of about -153 dBm/Hz, or 5nV/rtHz into 50ohm. If the noise from U2 is not picked up equally by both inputs, then I would expect the residual phase noise performance of this to be seriously degraded. I could use a quieter voltage reference, of course, but it would be a bit hard (and probably expensive) to find one that's quiet enough and it seems like the better solution would be to try to get the noise from U2 to couple into both opamp inputs equally so that it's attenuated by common mode rejection. I think this is complicated somewhat by the fact that the non-inverting input is high impedance whereas the inverting input is low impedance. Thoughts?

[mawyatt]

The output of U1 will follow the noise from U2. In other words unless the signal output is also referenced to U2s output (5V), the signal output will simply follow U2 as referenced to ground and you have no Common Mode Rejection when ground referencing the signal output.

A simple thought experiment will show this, with no signal input U1 output is @ U2 voltage (5V). If you change U2 to supply 6V (think of this "change" as "noise"), then U1 will follow to 6V output, likewise if U2 changes to 4V, then U1 follows to 4 volts.....so no rejection of U2 "noise" is present!!

If you must to use this configuration, think you'll need to use a very low noise reference/regulator for the 5V.

Best,

[matthuszagh]

The output of U1 will follow the noise from U2. In other words unless the signal output is also referenced to U2s output (5V), the signal output will simply follow U2 as referenced to ground and you have no Common Mode Rejection when ground referencing the signal output.

A simple thought experiment will show this, with no signal input U1 output is @ U2 voltage (5V). If you change U2 to supply 6V (think of this "change" as "noise"), then U1 will follow to 6V output, likewise if U2 changes to 4V, then U1 follows to 4 volts.....so no rejection of U2 "noise" is present!!

Ah, of course - U2 also biases the output. Nice explanation! I guess that's a limitation of any biasing scheme then. Perhaps a quieter scheme would be to use a very low noise opamp in a unity-gain configuration fed by a resistive divider with sufficiently low equivalent parallel resistance. But, that has no power supply rejection, so I'd need a very clean supply...

[mawyatt]

If you can use a negative supply somehow, then the output is now ground referenced noise-wise. With a negative supply you might be able to eliminate the input/output DC blocking caps if the use permits such.

A negative supply could be generated with a charge pump type (can't remember the chip #) being very careful with the pump frequency and requisite decoupling filtering, and maybe followed with a linear negative regulator.

Best,

[dobsonr741]

Seems like you aim at high bandwidth, thus selected a CFA. Then put ample bypassing caps across the TL431. Sizing/safe range is as https://www.ti.com/lit/an/slva482a/slva482a.pdf.

[magic]

I believe the traditional solution is to generate the reference voltage with resistors and filter it with capacitors to ground.

Then simply AC couple R1 to ground. Feedback resistors are fairly low here, but at 10MHz even a 100nF MLCC has less than 1Ω impedance, hopefully good enough.