Op-Amp Selection for TIA

[gigavolt]

Hey folks, I am trying to design a transimpedance amplifier. My goal is to output 15 V at 100 nA while being able to see 10 pA signals and have at least a 10 kHz bandwidth. I was able to follow some design guides by TI and determined that I should have a GBWP of 2 MHz. From the noise requirement, I also got that I should have a bias current decently less than my target resolution, say 1 pA.

Putting these parameters into digikey, I can find some pretty cheap op-amps that meet my requirements. For instance, digikey recommends the TL081 (https://www.ti.com/lit/ds/symlink/tl084.pdf) or TLV271 (https://www.ti.com/lit/ds/slos351e/slos351e.pdf). However, when I look at what people use in similar applications, the parts are like 10x the cost (such as the OPA124 or OPA627) and so I feel like I am missing out on something obvious.

Could someone help me understand what the real differences between these parts are and if I might need to go to something more expensive?

[Kleinstein]

The TL081 has quite some noise, especially 1/f noise and also drift with the temperature.  It depends on the application of the low frequency part is relevant.  How important the voltage noise of the OP-amp is depends on the source impedance. The resistor in the feedback is usually large, but there is additional impedance from the input.
There is no need to get the input bias current below the planed for resolution, but it still helps to get a pretty low bias. By itself the bias would not be a problem, but it is usually quite temperature dependent and the temperature dependence is what can be limiting.

The OPA627 is a bit on the expensive side and rarely needed.  If noise matters (relatively low source impedance / capacitoace) I would consider the OPA141 or rather similar OPA1641.
With a high source impedance one could consider lower bias, but higher voltage noise types like OPA377 or LMP7701 or similar - but these often only come as low voltage types an would thus need an added circuit to get 15 V out.

[gigavolt]

Thanks! That's really helpful since I'm still kind of new to this type of electronics. So I guess the idea is that the biases/offsets aren't a big deal since they can be compensated for. It's really the drift and noise performance that separates the parts?

One confusion I had was actually around noise. For the TIA, it looks like voltage noise will just bias the output of the amplifier. So even if I have like 1 mV of noise from the op-amp itself, it will still only show up as 1 mV of noise on the output which is pretty good for me on top of the +/- 15 V range. The input will be high impedance (also like 100 M-ohm scale) by the way. So, I guess it's really current noise that I should be worried about. Am I understanding that correctly?

Also, you mentioned considering lower bias current for a high impedance source. Is that because of the high temperature dependance that you talked about?

Edit: also I'm looking over more parts now and some of them specify an input bias current "over temperature" and it can be on the scale of 1 nA. Is that input bias only spec'ed for like a narrow temp range?

[Kleinstein]

If the source is high impedance, the voltage source is indeed relative unimportant. The feedback resistor (if 100 M range) alone has quite some noise, way more than the low noise OP amps.

A lower bias OP-amps will have less current noise and also less drift in the bias. There can be some cheaper types than the OPA141, like  OPA197 or OPA172 or TL072H.
The TLV271  is a low const version of the TLC271 and as such only good for a 0 / 16 V supply, so not suitable for +-15 V.
To just resolve 10 pA there is no need to have < 1 pA bias

It could help to get away with only +-15 V supply and thus maybe +-13-14 V as the actual output range.

[gigavolt]

Ok, I keep learning more and more about this problem. It looks like the biggest factors in performance for me are going to be the current noise and the drift of the input bias with temperature.

For the dirft, I read in an Analog Devices application note that for JFET op-amps, the input bias current doubles roughly every 10 C in temperature change. I think that's going to be the limiting factor here. If I assume the op amp is running at 30 C above ambient under power (just a number I pulled from a spec sheet), then the bias current will be a factor of 8 larger than the room temp value. If I assume like a 10C drift over time after powering up, then the estimated current drift will be i_bias * (8*sqrt(2) - 8/sqrt(2)) = 5.6 * i_bias. So I need to keep that value below 10 pA and get an op amp with a room temp bias current of like 1-2 pA or something.

[Terry Bites]

A (modern) CMOS input opamp will be advantageous here. eg LMP7717.
The price is right too.
The bias currents and current noise are low.

From your post it seems you want a 5 decade span. Without range switching you'll be down in the dust at the 10pA end.
This article has some good tips on gian setting.
www.analog.com/en/analog-dialogue/articles/programmable-gain-transimpedance-amplifiers.html

TI and AD have photodiode amplifer deisgn wizards. The help you design TIAs. You don't use a photodiode model if you don't want to. Just plug in your own input parameters.
Of course their flogging their products here but you'll get a good idea of how opamp selection is done.

[gigavolt]

Cool, my problem I think is that I was having a hard time finding parts that can reach my 15 V supply rails. All of the really low input bias current parts seem to sacrifice range in terms of the output voltage.

I found something interesting in the AD design wizard you recommended. It yelled at me that my gain was too high when I tried to ask it to make a design with 150 M-ohm gain. Is there a reason that's too much for a single stage?

Edit: I did find a couple of op-amps that fit my requirements. The AD8627 seems to be OK on noise and drift and the wizard recommended the AD795 which is slightly under my bandwidth constraint, but would work if I did two stages.