OpAmp: What The Difference Small and Large Signal Frequency Response?

[Mechatrommer]

in opamp datasheet i always see graph for "small signal frequency response" and "large signal frequency response" (example pictured below), i know this must be a simple thing, but googling give me offtopic/advanced answer. whats the difference? and what is defined as "small" and what is meant by "large"? is it input signal? or the output signal (Vo in the graph)? i know i like laymans terms, but this one not

[w2aew]

In general, small signals refer to signals where the peak-peak voltage of the signals are a few hundred millivolts or less.  Large signals typically mean very large peak-peak signals, maybe even near rail-rail or at least approaching maximum available swing.  The factors that affect the bandwidth are different in most circuits under these two conditions.  For small signals, the circuits are generally operating in their linear region, and thus the frequency response is due to gain-bandwidth products, node impedances, etc.  For large signals, many times there are active devices that are turning on and off, or even going from cutoff to saturation.  Thus, the frequency response will be limited by things like slew rate, recovery times, etc. 

I don't know if any "hard and fast" rules though governing what is considered large or small signals.

[saturation]

Its always been fairly vague when used for non-transistor based terms, but in reference to op amps, small signals are limited by the gain bandwidth product and large signals are limited by slew rate.

For example, if Vin was 7vpp its Vout at G=1, the op amp frequency response would be less than if Vin were 1Vpp instead, at the same gain, G=1.   If Vin were 0.1Vpp, and you needed it at 1Vpp, meaning a G=10, it frequnecy response would be less if Vin were 1Vpp.  

The bottom line is to balance these effects for the maximum frequency response you need.  There is a sweet spot for every op amp based on those parameters.

[tecman]

One other explanation.  The difference between small signal and large signal is usually a difference that occurs then the voltage out reaches a non-linearity known as slew rate limiting.  When you reach a slew rate limit, the output voltage will not change any faster, no matter what the input is.  The output is generally limited by the available current in the driver and the total capacitance of the circuit.  If you compare input to output, below the slew rate limitation, an input sine wave will result in a sine wave output.  When the limit is reached, the slope with the greatest rate of change will start to rise at a linear rate of dV/dT.  Sine waves will get linear slopes, more like a triangle wave, with a rounded top and bottom.  Most high performance op-amps will give a slew rate in volts/uS.

paul

[Mechatrommer]

i know this must be a simple thing,

no this is not a simple thing. but i guess i got the picture now. its vague, its slew rate limited, near saturation and the like. thanx guys.

The bottom line is to balance these effects for the maximum frequency response you need.  There is a sweet spot for every op amp based on those parameters.

any simple rule to find this sweet spot? from sim or datasheet? before putting it in circuit? or do we have to try and error in circuit?

or should i rely on datasheet info? from my limited experience, when i calculated my real circuit behaviour using simple GBW math, its not what is stated in the datasheet. so

[Zero999]

The maximum output voltage depends on the frequency.

Slew rate = 2pi*Vpk

Rearrange the formula, to determine the maximum frequency or voltage for the slew rate listed on the datasheet: this is the large signal bandwidth. There's also often a graph on the datasheet showing the large signal bandwidth vs the frequency and output voltage.
http://en.wikipedia.org/wiki/Slew_rate
http://hyperphysics.phy-astr.gsu.edu/hbase/electronic/a741p3.html

For the small signal bandwidth, look at the graph of gain in dB vs frequency on the datasheet. You can also calculate the bandwidth because it's normally linear: after the first corner frequency, the gain will normally drop by a factor of 10, if the frequency goes up by a factor of 10.

[Bored@Work]

Slew rate = 2pi*Vpk

You didn't figure that not even the measurement units add up in your formula? The slew rate of a sinusoidal signal (i.e. v = Vpk * sin(2 * pi * f * t + phi)) is SR = 2 * pi * f * Vpk.

[Zero999]

Yes, you're right, I missed out the frequency.

[Ronnie]

You can also calculate the bandwidth because it's normally linear: after the first corner frequency, the gain will normally drop by a factor of 10, if the frequency goes up by a factor of 10.

This is assuming the op amp is fully compensated (aka unity gain stable); if the op amp is de-compensated it will be a different story....