Normally I use OPamps, which do show so low offsets voltages, that an offset trimming isn't necessary. The reason is the following: If an OPamp is used which suffers from such high offset voltage that a trimming is neccesary in a certain application, then this offset voltage will at the same time show a relevant drift. Using an OPamp, which shows already very low offset voltage will automaticaly show much lower drift, mostly. A typical example is TL082 and TL052. Latter provides not even much smaller initial offset drift, but also smaller temp drift and long term drift. So, if you need an offset voltage of better than 1.5mV, then it makes much more sense to choose this TL052 (without any trimming) instead of TL082 (plus trimpot).

Today for nearly each application there are lots of OPamps, which don't need any offset voltage trim.


The same as above. If thermal drift is an issue, then you can choose low drift chips today, which don't need any sophisticated temperature compensation. Now you will say 'ok, but these chips are more expensive.' Yes, that's true, but keep in mind, that a sphosticated temperature compensation circuitry will cost much more, especially if you take into consideration the time you need to adjust this temperature compensation...


Yes, this was done in former days, when low offset chips weren't available. But times have changed, fortunately, and it's much better to choose high precision chips instead of this scheme above.
A nice methode to do some offset cancelling is to use a multiinput ADC, which shows very small offset voltage change between the individual inputs. Then, you can apply a fraction of a reference potential (often already provided by the ADC itself) to one input (1/100 of full range for instance), compare the reading with the theoretical value and remove this offset error from all other inputs. This is a very elegant way to improve the offset performance without needing any manual trimming.


Gain linearity can also be improved by choosing a precision OPamp. It's important not to choose too high gain factors when the signal has to be amplified. If high gains are needed, then more than one OPamp should be provided, each one amplifying same portion. For example, if a gain of 100 is needed, then take two OPamps with a gain of 10, each.


In cases where very high gain precision is needed, you should take the precisest resistors you can get, for instance 0.1% tolerance types. Adjusting is done by putting some cermet trimmer (also a digital pot is possible) in series to one of these precise resistors. But choose the value of trimmer to be much smaller than the precision resistor. Remember you must only trim the tolerance. For example: To adjust +-1% put a 100Ohm cermet trimmer in series to 5kOhm precision resistor. This will allow you to adjust the error of ten 0.1% precision resistors being used in a signal chain. The advantage of using this 100Ohm trimmer, instead of a much bigger one, is that any drift or unprecision of this trimmer is only weighted by the factor 100/5000. Means if your trimmer shows a temp drift of 100ppm, then it reduces to only 2ppm in the series combination of 5kOhm and 100Ohm!
There's one remaining error which cannot be trimmed so easily: Long-term drift. To keep this low, do only allow very very low heat dissipation. Less then 10% of maximum power dissipation of resistor should be allowed. Then, a 1% precision resistor will show about less than 0.01% drift per year (differs of course from brand to brand and can be found out by reading actual datasheet). More precise resistors do even show smaller drifts.
But if extreme gain precisison is needed, then only avoiding resistors and other passive components to the absolute minimum number is the remedy.

Kai