LM358B: Why do we differentiate small signal and large signal when measuring slew rate and settling time?

Hello Yunsik, 

Yes this is indeed a difficult topic, especially with many op amp datasheets treating things slightly differently - but, 

1. The amplifier at whatever external set up it has will have some small signal AC response shape

2. Be that 1st or 2nd order, an applied step will try to follow the ideal response shape as well as possible on the output

3. 2nd order is more interesting, and it will have a smoothly varying instantaneous dV/dT. 

4. if that dV/dT exceeds the device slew rate, the linear response will break off and the output will "slew" at the available slew rate

5. By definition, if the output is in a slew limited transition, the loop is open and you will typically see a single overshoot then recovery to final value - not a linear response at that point. 

6. Normally, that slew limited step will have a much longer settling to final value than a step that stays linear. 

7. Some amplifiers have different rising and falling edge slew rates, not clear which one is being used in the spec tables sometimes, should use the slowest, but not always the case. 

So, Large signal slew limited steps are never used for settling time reporting - although they will certainly have a number, you just can't relate it to the small signal response shape. 

One of the more interesting new relationships I worked out a couple of years ago, is that the peak dV/dT on a 2nd order response shape is approximately 2.85*Vstep*F-3dB. You can use this in 2nd order cases to see if your required step size is going to slew limit - probably more than you ever would want to know, but that is all worked out in these two articles - I was mainly after the required slew rate in active filter stages - and this gives it to you nicely, 

https://www.edn.com/design/analog/4443250/What-is-op-amp-slew-rate-in-a-slew-enhanced-world--Part-1

https://www.edn.com/design/analog/4443345/What-is-op-amp-slew-rate-in-a-slew-enhanced-world--Part-2

Oh and I might add while that simple estimate of what the output signal step would require in a peak dV/dT gives you a minimum for the amplifier being used. Normally I would then guardband that by 2X in picking parts for the application as slew rate does have some variability part to part and over temp. 

Hi Yunsik,

an OPAmp is a very complex circuit. It's the result of decades of electronic development. It's not easy to fully understand an OPAmp. And most datasheets only show simplified schematics of the internal circuit. Another approach is to think of an OPAmp as a black box with a very certain behaviour. As an OPAmp is usually the faster the smaller the output signal is, datasheets usually contain two data about speed, the "Settling Time" for very small signals and the "Slew Rate" for very large signals. This is just convention and is historically grown. Another important speed data is the "Maximum Peak Output Voltage vs Frequency".

Kai

Hi Kai,

The Settling Time specified in an op amps datasheet is usually a large signal (10V step, 2V step, etc.) not a small signal response. See below.

Bandwidth can be used to calculate the small signal settling of an op amp. Rise Time = 0.35/Fc where Fc is the cut off frequency (or bandwidth of an op amp).

Thank you,

Tim Claycomb

Hi Tim,

oh yes, you are right! Sorry!

Kai

Well what Tim was saying did kind of point out how confusing this is. 

1. That spec snip was not from the LM358B data sheet, but the point was there is (on the slower parts) a settling time where the transition is slew limited

2. If your large signal settling spec is slew limited, can't really use the time constant ideas to relate bandwidth to settling

3. The rise time equation is correct for a 10% to 90% single pole, but again not settling time

4. The settling time to a target window for single pole is a number of time contants. 0.01% is 10 time constants,

The LM358 spec is here, 

that is a 0.1% settling to in 4usec on a 2V step (why there is a Cl who knows). can't find that equation right now, but lets say 7 time constants for 0.1% and estimate that as 1/(2pi1.2MHz) = 130nsec. So 7 times that is 928nsec far faster than the reported 4usec. So, it must be slewing . a 2V step  with 0.5V/usec is 4usec, so this totally a slew rate set spec in this case, not related to time constants. And probably wrong as the slew rate takes up the whole 4usec but you still need time from there to get into the 0.1% window. 

In higher speed parts, when settling time with different step sizes were needed, we would slow the input edge rate to stay out of slew limiting to stay within a linear response type settling time. 

So here is the plot we put into the similar FDA THS4551, this is simulated and as we use larger steps we slow the input edge down to stay out of a slew limited edge, that is always a good idea for better settling time if you can do it (this is just saying you need check and reduce if necessary the BW of the incoming signal)

If you can stay linear, the settling times to the 0.01% level are the same with step size, these are obviously not 1st order responses and there is low level ringing that is in the actual time response. 

Thank you very much

those videos really helped me understand the difference.

Thank you for your reply!

I'm glad we were able to help!