Ahh - Cool! (Or to mock politicians and other fakers - "
I'm glad you asked me that!".)
To me it's implicit that there is no loading of the OEM circuit - otherwise use a voltage follower etc (an op-amp type buffer circuit that mimics input without loading).
In this case with its 1.8k resistors etc, I'd want at least 10x the resistance, preferably 100x. Hence the 1.8k equivalent being >18k or pref 180k. Hence maybe 100k.
On the other side, we can't have too high an impedance/resistance for the PIC/uPC etc. But since these are usually quite high - eg, requiring mere nA else mA - that usually isn't a problem.
Before I mention
skewing, re resistor tolerances...
Yes - resistors and all components have a tolerance. Now while therefore you might use 1% resistors if 1-2% accuracy is required, it may not matter....
VIZ - those tolerances represent the max difference between its marked or intended manufacture value and its actual value. Those tolerances do not indicate the natural variation of a component (due to temperature or arbitrary drift etc).
IE - an
actual 1.005 meg resistor is the same whether it be a 1% or 5% or 10% tolerance.
The point being that using two 5% tolerance resistors might give an output that deviates up to 10% from your ideal calculated values, but that deviation remains fixed - eg 4% higher or 2% lower. And that is just a matter of recalibration.
Hence trimpots etc in analog circuits, and for digital - a reprogrammed REF value or a programmable
scaler to modify whatever REF values or maps are used.
In some cases calibration isn't necessary - eg, in this case we may need to detect a ~60% swing (4V versus 7V) or a 20% swing (that 3V difference between 4V & 7V relative to a supply of up to 15V).
Hence 10% tolerance resistors might do (not that they exist anymore), but 5% should be fine. 1% or 2% would be overkill (but not as expensive as they used to be??).
[ Incidentally, in ye olde days one could almost guarantee that in a batch of 5% tolerance resistors you would never find any within 1% or 2% because that had been removed to supply as 1% & 2% resistors respectively. I'm not sure about manufacturing these days - I'd suspect all are manufactured as 1% and merely arbitrarily marked as 1% & 5% etc. Just as all 1N400x diodes might be 1N4007s. ]
So component
tolerance is more a question of whether you can calibrate the final product (trimpot or a software/firmware variable).
Back to
skew (for want of a better or correct term)...
Except where the aforementioned loading of the OEM circuit effects OEM operation, any skew caused by parallel (linear) components is similar to the component tolerance issue.
EG - if our resistance is so high that the uPC's input affects the voltage divider's output, it may not matter - we simply recalibrate. (Alas constant current inputs are more complex but that only matters if wanting a
continuous output like a temp or volt display etc instead of a simple REF point or two.)
How'z that sound? Yep, I've been there & done that...
Hence MY 'glad you asked that' because I do know the answer(s) - unlike most politicians and others that use that phrase (IMO it's a dead giveaway!).
Of course I'm even gladder if you see a flaw in my arguments or come up with something I haven't considered.
BTW - I too often thought of zeners etc as a linear drop-down device but came to the conclusion it's only good as a drop down of
[level - eg, a 46V to 47V swing to 1-2V etc, aka an expanded scale.
But if that
level changes, or the swing is too high eg 46-54V is 8V which is too high for a 5V or 3.3V ADC etc.
Or if the zener drop is too variable with temperature, or the zener is too unreliable or costly.
And zenors have their tolerance too - and vary with current.
So given the cheapness and simplicity of a voltage divider; the greater selection of resistors; and that it's the same circuit/PCB configuration but to which filter caps are easily added...
Hence my
standard 2 resistor (+ cap + 2 diode) PCB input design. Any zenor drops would be external to that - eg, maybe to drop 450VDC CDI supplies to reasonable voltages.