Is its possible to use the AD8476 as an attenuator (G = 0.500)? I need to drive a delta-sigma ADC but the range of the ADC is too small so I want to attenuate the input so the full range of my sensor is covered.
I think you can do this by using a resistor at the input. However, the noise gain will be higher and the stability will change because the gain of 1, created by the internal matched resistors.
You may also want to use a dual supply (for the AD8476) and use a lower input voltage that is within the IVR of -Vs+0.05V so that the low IVR of ADC will be reached.
You may also want to use ADI attenuators which you will find here - http://www.analog.com/en/rfif-components/rfif-attentuators-vga-filters/products/index.html
I would like to clarify something unobvious. The internal resistors of the AD8476 are trimmed for the best possible gain accuracy and CMRR. High CMRR requires that the feedback factor at the positive side and the negative side match, therefore the resistor ratios will be the same, but the absolute tolerance of the resistors can vary by ±20%, as noted in the datasheet.
So in the worst-case, if you add two 10k resistors to the input to try to get G=0.5, but the internal feedback on one side is made up of 8k, 8k resistors and the other side is made up of 12k, 12k, then you can end up with CMRR less than 20dB. (See equations in this article: ADC Drivers: Analog Dialogue: Analog Devices)
You can get better performance choosing something like an ADA4940-1 and setting it up for G=0.5 than using the AD8476 this way. Depending on your application, we also have an AD8475, which is a fixed-gain of 0.8 and 0.4, but is higher bandwidth, higher power, and lower noise. Or if your signal is single-ended before the AD8476, perhaps you can attenuate the signal and buffer it with an op-amp before the AD8476.
I'm also attaching an excel spreadsheet I made up that has the basic ADC driver equations, just in case you want to see the effect of resistor imbalance on common-mode and differential signals. And you may want to check out the ever-useful DiffAmpCalc.
I hope this helps.
Retrieving data ...