Other Parts Discussed in Thread: THS4541
I'm looking for an amp to drive an ADC which follows a mux operating at 10MHz. The ADC will be oversampled to 16-bit resolution. Given the ADC's 20ns S/H acquisition time and allowing 10ns for the mux to switch and settle, that leaves only 70ns to settle to 16-bit accuracy (.00153%). The ADC will be operating at 3V but the op-amp rail could be increased to 5V if necessary. From an app note on multiplexing that I found, the BW would need to be 3.75 x the 5MHz Nyquist rate = 18.75MHz and the slew rate required is > 3V/70ns = 43V/us.
Naturally the application is very cost sensitive; whilst the inputs are all low frequency AC signals (20 to 3kHz) with a source impedance < 500 ohms, there are a lot of channels so it would be too expensive to provide the gain prior to the muxes and would have the problem that the gain drift errors over time and temperature would differ for every channel. This would require the use of expensive precision resistors as the equipment can't be re-calibrated after installation.
To make things more difficult I'd like the amp to provide at least x20 gain and preferably x60. Is this realistic within a budget of less than $2 or $3 for the op-amp?
One option would be to have two amps providing the gain prior to a final 2:1 mux feeding the ADC (or an amp driving the ADC) which would increase settling time to 170ns but still has the problem of gain variation between the two channels and extra cost, except that the lower demands might allow lower cost devices to be used.
Another would be to spread the gain over two amps in series but that would presumably require a track and hold between them at extra cost, complexity and gain errors? I guess the settling time of two amps in series will be rather less than twice the datasheet value as the second will see the large step after the propagation delay of the first. Is there a rule of thumb of how to calculate this?
If the one amp solution is not feasible what about the two op-amp options?
To get the best performance out of the ADC it needs to be driven differentially, so is single ended to differential conversion possible as well? Obviously it would cost more but again the settling time is the problem.
I'm not sure which op-amp parameters, apart from slew-rate, settling time and noise performance are relevant for this application. In any case the settling times in the datasheets for high speed amps are typically at low gains of x1 or x2. Is there a simple way to calculate how it changes with gain?
DC offset of < 1mV would be preferable to avoid too much loss of dynamic range with a gain of 60.
Any suggestions for suitable amps?
Thanks, Tony H
