Yes, within ~50mV sounds quite good to me.
If you're trying to amplify very near the supply, well first of all, stop doing that. Second of all, you may still be able to get away with it, if the load pulls towards that rail. Use it as a good old fashioned "single supply" amp, but with your choice of which supply to single out. And with much more load capability than the ~50uA pulldown ye olde LM358 offered! Operating within some mV can be reasonable this way, at least for CMOS types. Mind that bipolar (complementary open collector style output stage) will still have some minimum saturation.
Two common examples I can think of for needing near-rail operation, are:
- Current sense: the input is single-ended, and the output is common-rail. You just need some gain so as to get away with a smaller shunt resistor.
- Limiter circuit: the output is constrained between rails, and maybe you even run the amp from +VREF so its output is precisely (less that annoying VOH/VOL, that is) 0 to REF.
For current sense, maybe the low end doesn't actually matter, you just need to compare with a modest or large threshold/cutoff (e.g. peak current mode SMPS, current sensing/fusing, etc.). In that case: great, who cares!
Or, maybe the low end does matter, and you need to measure near zero say for measurement, metering, etc. purposes. In that case, simply offset the input by bleeding in a little offset voltage. If it's a noninverting amplifier, just run +IN from a divider from +VREF to +IN to shunt. A few mV will do 'ya.
Maybe you even need a slight negative range, say for a power supply with down-programming (bleeder/leakage circuit) so you can measure it accurately over the whole range; or maybe something something battery metering so you can see charge and discharge currents, but the magnitudes aren't symmetrical so you only need a little negative range. Same idea, use an offset or differential measurement, done and done.
For the limiter, three options:
- Redefine the supplies so they are slightly in excess of what's needed (problematic if you don't have a negative supply handy and need even just some 10s of mV below GND!)
- Redefine your signal's range so that the natural saturation levels are just outside of the required range
- Don't use saturation for limiting; you need too much accuracy anyway (i.e. when a ~50mV margin isn't good enough), use a proper active (precision rectifier?) circuit to implement the limit instead.
Also maybe worth mentioning saturation recovery, but most modern types boast of their prompt recovery, supported by waveforms or settling time. Amps like TLV2372 are well behaved, I have no problem using it as a jellybean. Simply check and confirm, and keep shopping if you don't see adequate data regarding this.
Yeah it's a naming convention, but how would you call it otherwise? If you have a better idea?
ARRIO? "Almost rail-to rail"
Now how "almost" are we really taking here...?
Ok. How about this?
For linear regulators we have "Low Drop Out". They don't call them "Zero drop out*" ....."*still need 0.4V".
How about "RRI-LHO" = Rail-Rail-Input, Low-Headroom-Output
Annoyingly, "LDO" has been taken to mean any linear regulator these days. So 7805s are "LDOs" now. Maybe we should adopt the backronym "linear dropping output" or something..?
Regarding op-amps, there are some with missing +V or -V range (RRI but not O, or RRO but not I; or "RI" or "RO" in the sense of, only one rail is met), something to watch out for. I've seen many precision amps of this sort, which is kind of a shame, but also not such a big deal in a lot of applications. What's more annoying is distributors don't track input and output rail compatibility, so you have to flip through a lot of candidates to find the right one.
And again, simply limit signal ranges, don't rely on device ranges if you don't have to. Sometimes this means adding an extra couple volts supply, may be annoying but it's far from intractable these days.
Tim