How to make repeatable Op-amp amplifier circuit

Question

I did a simple op-amp based 100X gain single ended circuit. I tried to order 100 IC's and manufactured them. But found that out of 100 only 60 devices have same and expected baseline. What could be the reason for this variation and how do i improve my yield? (PS: I use TLV9061) All resistors used are SMD type 0603 1% tolerance resistors. Also sensor that i would be using is a capacitive load of 220pF, will this op-amp be good for that. Test case: If the vsensor is not connected, the output of the op amp should be 0.33 as an ideal condition. I considered OK range as 0.3 to 0.4 . Only 60 of them were marked OK. Rest all are out of range, may be 0.2V, 3.0V, 2.08V etc

Answer 1

Look at section 8.10 in the data sheet. The part you have chosen to use has an input offset voltage of \$\pm 0.3\mathrm{mV}\$ typical, \$\pm 2\mathrm{mV}\$ worst case. Your output offset voltage in an amplifier like this is your non-inverting gain times the input offset voltage -- so you're looking at between \$\pm 30\mathrm{mV}\$ and \$\pm 200\mathrm{mV}\$ of offset.

I think that you're lucky to have a yield of 60%, although the picture may be more grim if you did your measurements over any sort of a temperature range.

On the bright side, with input bias currents in the pA and gain-setting resistors in the k\$\Omega\$, you don't need to worry about offsets from the input current + your mismatched resistors, or from input current bias.

You need to build some offset correction into your circuit, and run the numbers (or do a lot of testing) to verify that the drift over temperature will be acceptable, or you need to choose a part that has significantly better voltage offsets without too much input bias current or input bias current offset.

Answer 2

What could be the reason for this variation

When operating at high gains in DC-coupled situations, input offset voltage can be a problem. The worst input offset voltage given in table 8.10 the datasheet is 2mV, which when multiplied by your gain of 100 means a potential output offset of 200mV, this would explain an output voltage in the range of 0.13V to 0.53V, it doesn't explain why you were getting voltages in the volts though.

However that is not the whole story.

Firstly lets re-read table 8.10 in the datasheet carefully and in particular pay attention to what it doesn't tell us. In particular at the head of the table we find the statement (bold mine).

For V_S (Total Supply Voltage) = (V+) – (V–) = 1.8 V to 5.5 Vat T_A = 25°C, R_L = 10 kΩ connected to V_S / 2, V_CM = V_S / 2, and V_OUT = V_S / 2 (unless otherwise noted)

In other words the parameters in that table only apply when you are in the middle of the amplifiers common mode voltage range. Anywhere else and it may be worse, how much worse? they don't guarantee that!

However they do provide a "typical performance characteristics" graph "figure 4" which implies that this amplifier maintains it's offset performance at the negative rail (but not the positive rail).

So unfortunately I don't have a good explanation for the behavior you are seeing, only several possibilities.

1. The parts you have are not the parts you think you have. Either due to a mistake or outright counterfeiting. Have you checked the markings on the devices against the expected part markings in the "packaging" section of the datasheet?
2. You got unlucky and your parts offset performance at low common mode voltages is much worse than "typical".
3. The lower supply voltage significantly effects the performance of the part (I think this is unlikely)

---

how do i improve my yield?

First ask yourself if you really need to DC couple, if you can get away with AC coupling then offsets (both voltages and currents) become far less of an issue.

Secondly read data sheets very carefully with an eye to what they are not telling you.

Thirdly always buy your parts from reputable sources.

Fourthly try to avoid operating near the rails unless you really have to. Even if the amplifier is advertised as rail-rail.

But ultimately there is always a risk that unknown unknowns that did not show up with the first couple of prototypes will bite you when you try to scale up manufacturing.

Answer 3

Transistors are fickle things. They have a specification but also a min/max specification. The min/max is what gets you. Op-amps are made out of transistors so they inherit the same qualities.

The min/max is in a Gaussian distribution. Most will be around the norm but there will be outliers. I think you were lucky to get even a 60% yield.

There are 3 things you can do.

1. Select very accurate components. Resistors that are 0.1% or better and an Op-amp who's offset voltage is within the parameters. This way even if the offset is off a bit it will still be within tolerance after the 100x gain. (depending on the bandwidth requirements a chopper amp might work)
2. Get an Op-amp that has an offset trimmer pin. Manually trim the offset with a potentiometer.
3. Add trimmer potentiometers to the gain and offset circuits and manually trim it that way.

Option 1 is expensive. Option 2 can adjust the offset but the gain may be askew. It also requires someone to trim the pot. Option 3 can fix issues with both the offset and gain but requires someone to trim 2 pots.