I would appreciate if anyone could hint at fallacy in my thought process in regard to picking the appropriate op-amp based on certain but few specifications. I would also appreciate if you could check the calculations just to make sure I didn’t leave out any pi (3.14) in my calculations.

Say I have to pick an op-amp for the system, wherein the op-amp only buffers the signal. This input signal has a moderate common-mode range so op-amp doesn’t need to work rail-to-rail on its input. Also the op-amp shouldn’t affect the system’s bandwidth considerably.

Task 1: Say the input signal is 500 KHz, and the system prior to the op-amp has a bandwidth of 750KHz. Task is to make sure that with the op-amp the overall bandwidth is at least 650 KHz.

Task 2: Make sure that slewing is not going to be an issue.

First thought: Picking an op-amp with a unity gain closed-loop bandwidth of 2*750KHz  (1.5MHz) should still roughly give an overall system bandwidth of 670KHz. This is based on RMS roll-up of 750KHz and 1.5MHz, i.e. 1/sqrt(1/750K^2+1/1.5M^2)

Second thought: The above calculation implies that the Opamp should have a minimum gain-bandwidth product (Ft) of 1.5MHz

Third thought: Say the slew rate (SR) specification of the op-amp matches quite close to 2*Vt*Ft (around 75KV/sec at room temperature, where Vt=KT/q=25mV), as observed in typical BJT design, so the signal’s maximum input rate of change needs to be lower than the SR of the op-amp.

For a sinusoidal signal input to the op-amp, dV/dt|max=Am *2*pi*fmax. So as long as dV/dt|max<SR then the op-amp should not-slew. Likewise a similar dV/dt|max treatment for a step input. Hence if the maximum input sinusoid signal frequency could be say 650KHz, then the maximum input level can be 115.4mVpeak (=2Vt*2*pi*Ft/(2*pi*fmax))

Fourth thought: If input signal shall be bigger than the above calculated 115.4mVpeak, and if the maximum input frequency could not be lowered, then the only option is to pick another op-amp with higher unity-gain bandwidth (Ft).

Thank you very much for reading and thanks in advance for replying.

subgold
First of all thanks for taking the time to reply.

About BW: As I mention "unity-gain buffer" of the op-amp, I meant op-amp with feedback acting as a buffer. The intended circuit is supposed to be linear (as linear as the specs). Just to make sure that this buffer doesn't become the bottleneck, its bandwidth is chosen to be 1.5MHz. So I just rolled up the bandwidths of the two components (first component not spcified but has a BW of 750KHz, second component is the buffer of 1.5MHz)  in a RMS fashion to get an approximate answer.

About slewing: Slewing is dependent upon rate of change of input. Since sinusoid will have max dV/dt at zero crossing, the op-amp could slew momentarily and as a result cause non-linearity; not a desired property in linear systems, of course.

kataria0 wrote on Mar 26^th, 2009, 10:16am:



of course your opamp is in feedback loop. The feedback i meant was the feedback from opamp to any other previous blocks. if there is no such feedback, then the overall system bandwidth is just roughly equal to the slowest block (error tolerance is not considered for this moment). i think that is also the opinion of rajasekhar.

kataria0 wrote on Mar 26^th, 2009, 10:16am:



first, slew rate is the changing rate of the output, not input. second, the buffer opamp should be designed in such a way that its output swing can always sustain the signal amplitude, and the opamp is always in linear region. even momentarily entering nonlinear region is not accepted, otherwise large distortion will be introduced and the buffer is poorly designed.

i understand what you are worrying about, but that is not slew-rate. if your opamp is correctly designed (e.g. the output swing issue i mentioned), then the problem you are concerning is still within the scope of small-signal property, which is guaranteed by BW.