Can LM358 be replaced with AD8676

Question

I am trying to build an circuit with an PCB to sample the main AC waveform in an ADC set to 0V to 3.3V range. I noticed that there are some ZMPT101B modules that is based on LM358. I made some modifications to replace the LM358 with AD8676.

I check the datasheet, and so far I haven't noticed any possible issues regarding this substitution. However, I am a little worried if the 3.3V on VCC will be enough.

In the image below, there is the proposed modified circuit.

Could someone please tell me if its design is OK to use?

Best regards, F.Borges

Answer 1

AD8676 isn’t specified below 5V - it may work but there are no guarantees it will behave well. LMV358 might be an option if the performance suits your needs.

As an aside, LM358 have awful crossover distortion as they are essentially class B. This circuit appears to work around that in U1A at least by loading the op amp output to GND so the amp is always sourcing current and never crosses over to sinking current. LMV358 doesn’t have this ‘feature’.

Answer 2

I am somewhat surprised at "I check the datasheet, and so far I haven't noticed any possible issues regarding this substitution." when working with 3.3V supply power and replacing an opamp specified to work with 3V minimum supply voltage with one specified to work with 10V minimum voltage. If that is not enough of a warning flag, the input voltage range for the minimum ±5V supply is specified to be ±3V, namely at least 2V away from each power rail. Even if you wanted to speculate that it will work at lower voltage than specified, it is impossible with 3.3V of supply power to stay at least 2V away from each power rail.

So checking the most basic specs, the obvious answer is "there is no way this will work". It's not just out of spec, it is out of spec in a manner where there isn't even a chance of hoping for it to work in a manner remotely approaching the spec sheet details.

For 3.3V of supply voltage, you need low-voltage opamps, and with the kind of headroom you have available, they will typically need to be rail-to-rail (or beyond) on both input and output.

Answer 3

Putting aside the power supply issue for a moment, you have picked a low-noise opamp, but your schematic is sort of a noise fest.

The AD8676 has 2.8nV/√Hz input-referred voltage noise, and 0.3pA/√Hz current noise. Now you supply your virtual ground to the positive opamp inputs with voltage dividers and no smoothing at all. That means you get all the supply noise, and 5kohm worth of thermal noise which is about 10nV/√Hz.

Admittedly the input transformer (apart from capacitive coupling) turns this into common-mode noise on the first opamp, making it pass through to the output merely buffered instead of amplified.

What else? The negative input of the first opamp sees an additional 9k of impedance good for 17nv/Hz. If you put the volume wiper in the middle, the right opamp gets to see 35k on its negative input, good for about 26nV/√Hz voltage noise of its own, plus about 10nV/√Hz of noise from the opamp current noise, bumping this to 28nV/√Hz (uncorrelated noise power can be added, so for voltages you take the root of the added squares).

With the first input stage amplifying 5-fold (in middle position), this is an additional 6nV/√Hz when referred to the first input of your amp.

All that adds up. Your virtual grounds can just be given a capacitor for smoothing. You can share the same ground between both stages; that way the ground noise from the first stage is passed through the second stage unamplified (thus becoming irrelevant) iff your volume pot also takes this virtual ground rather than the real ground.

And it really should, or you have 1.65V DC across the volume pot which makes for very scratchy volume adjustments.