Is there any benefit to using opamps / unity gain buffers / instrumentation amplifiers between the signals and ADC ports on a micro?
The reason I ask is I am currently drawing up an ADC breakout for my varsity project board and am not sure what extra components to include on the board.
Several quick thoughts:
On many AVRs the impedance of the source driving the ADC is spec'd to be < 10K ohms. (Edited, JC)
If your sensor, (source, whatever), has a high output impedance then an op amp makes for an easy buffer.
The equivalent input circuit for the AVR ADC's is usually shown in the data sheet. There is a capacitor which is part of a sample and hold circuit. The signal driving the input has to charge up this capacitor.
For a fast sampling rate one has to charge the sample & hold cap very quickly, and do so without changing the signal being measured. Op amps hence make perfect buffers. High input impedance so as not to load the sensor, low output impedance so they can easily drive the ADC nput circuitry.
Additionally, one aften used the input op amps as part of a filter, or dc offset adjustment.
Recall that for much of the theory behind processing digitally sampled circuits the input signal is assumed to be band limited to less than 1/2 the sampling rate. Wiki Nyquist-Shannon sampling theorem, or pick up any signal processing text, for more on this.
Once again, the op amp makes the filter design easy.
Last thought is that if properly designed the output of the op amp can be clamped to what ever you want. This is sometimes helpful, although I suspect most people just attenuate the input to begin with.
The hard core analog guys can discuss the difference in the noise factor for the two circuits...
When I put together my first Xmega testbed I put a few op amps on it for the ADC inputs and the DAC outputs. Easy to do when designing the PCB, much harder to route the signals around afterwards.
Many sure you pick a good rail-to-rail op amp.
JC
Recently I was looking at a real small signal, like 2 counts. A *100 amplifier made it nicely visible to the A/D
The largest known prime number: 282589933-1
In my humble opinion, I'm always right. 
For dealing with weak resistive signals like a PTC/NTC thermistor or a thermocouple, an op amp buffer can be useful.
For amplification, if you need it then you need it.
There is one more tangible thing, although it just might be me. Op amps are cheap, and you can find ones with better ESD protection than what is on your microcontroller. Also you get easily get opamps in DIP packages where as it is not possible to get DIP for many micros. If ESD blows the opamp, I just pull it out and pop in a new one in the socket.
I put a DIP opamp buffer infront of my ADC in one designs, gave me a peace of mind, but the opamp never blew so I never got a chance to reap the rewards. Now I just plant TVS diodes everywhere and forgo the OP amp buffer.
The 10k-ohm number is the recommended maximum for the source driving the input, not the input itself. The input pin has a specified resistance of 100M-ohm. But there is a capacitance to be charged, through up to 100k-ohms (figure 24-8 of the data sheet for ATmega48A/88A/168A/328P). I actually don't know what they mean by " 1..100k " ohms. Maybe it means 1 to 100k or 1k to 100k. But 100k would be worst case.
Anyway, looking into the pin the time constant would be 100k times 14pF or 1.4us. ...
I'm straying off the subject. I think if your source resistance is less than the recommended maximum of 10k-ohms, the op-amp would definitely be optional.
Nick
Sorry Nick, I stated it incorrectly, above, and will edit it.
I knew what I wanted to say, it just didn't come out that way... :oops:
JC
Sometimes source impedance can be brought down with a capacitor on ADC pin.
For example it is not reasonable to measure battery voltage with resistors that make the impedance less than 10kohms, because it would drain batteries too quickly.
Instead, you can use megaohm-sized resistors and put a 33nF cap on ADC pin. Now the impedance is low enough compared to sampling rate and the charge transfer from this external cap to ADC sampling cap is enough to make accurate measurements of battery voltage.
Of course it makes an RC filter so that must be taken into account for quickly varying signals, but for low varying signals this is fine. Also the cap must be let to charge back to original value through the high valued resistors until performing another measurement that gulps charge, so the maximum sampling rate limit must be calculated as well.
Yes an op-amp would solve all these, but sometimes the capacitor is enough.
What the op-amp is really needed for is gain (as already said) but also handy for converting current signals or resistances into voltage signals.
Or the amplifier can be used to convert differential signals to single-ended. Some AVRs actually have a differential amplifier built-in, with programmable gain.
An amplifier can provide lower drive impedance to the ADC if the signal source is really high resistance.
That is one of he few uses for a unity gain buffer for an ADC.
Yes, you can convert differential to single ended. Yes, you can make signal gain. But, these things generally don't go with a unity gain buffer.
Jim
Ideally you would use as few components as possible. Opamps are good to boost the signal to readable levels, but the higher the gain the worse your error will become. Have a look at stuff like gain error and input offset error. If an opamp has an input offset error of 100uV and you have a gain of 100 then that 100uV will become 10mV. Stuff like that is certainly worth considering at the early stages.
As mentioned above, Op Amps can operate as level shifters. This can be valuable in two ways. In the first, if your input range is really small, you are under-utilizing your ADC. You can use an op amp to "stretch" the input to use the full ADC range. In the second, you can convert negative signals to positive signals (which the ADCs in the MCUs require). There's a really good app note (I think it was written by by Ron Macini) from T.I. about how to use Op Amps to take convert any input range to any output range (i.e., -1mV to -1V to 0V to 5V). I think it was called SLOA030 or something like that. If I recall correctly, it was for single supply amps, but the theory holds true for dual supply amps as well.
Wiki Nyquist-Shannon sampling theorem, or pick up any signal processing text, for more on this.
Nice to see someone (an Ohioan, no less!) include the Wolverine on that theorem name, rather than just mentioning Nyquist. There is a very nice bust of Claude outside the EE building in Ann Arbor...
