Post Go back to editing

Instrumentation Amplifier Gain resistor(s)

Hi. I am looking for Instrumentation Amplifiers for an application.  At this stage, we have decided a certain gain value for our Instrumentation Amplifier. The datasheet of the product mention the very precise value of gain resistors to attain certain gain value.

We haven't ordered the products yet as I am confused about whether we will be able to acquire the gain resistors ourselves (from some other manufactures) or they will come with the Instrumentation Amplifier package we will order?  Because these gain resistor values for certain gain sets are very precise which are difficult to get from manufactures. I am attaching the screenshot of AD8237 gain resistors from datasheet as reference  https://www.analog.com/media/en/technical-documentation/data-sheets/AD8237.pdf

Parents
  • Selecting the resistors to match your application is your job, and I doubt very much that the precise gain you claim you need is really a requirement. You'll be probably digitizing your signal and thus will have to do a system calibration anyway, and there are more error sources than just this gain, and the exact gains will be factored away during calibration. If you're clever, this calibration may be internal to the system without much cost impact.

    Selecting precise gain values using real parts is so much harder than writing nice-looking numbers in a table in a datasheet, that many designs typically avoid placing too strict gain limits. A nominal gain plus-minus a few percent is perfectly normal in most cases.

    You'd only need a very accurate analog gain if you're providing analog output that itself has to be related to some absolute units. Say that you're designing a transimpedance amplifier with a transimpedance of 1V/mA. The input terminals measure current, the output terminals provide a voltage. In such a device, the overall input-output relationship (transimpedance) has to be probably controlled better than "a few percent either way is OK"- then you'd either use a manually adjusted trimmer that a technician on the production line would adjust using a calibration fixture, or you could use a digital potentiometer with built-in non-volatile storage to adjust this transimpedance (the equivalent of a gain) - again by connecting the device to a calibration fixture that will "zero-in" the transimpedance and then save it in the digital potentiometer.

    You could also use an in-amp that has built-in nonvolatile storage for offset and gain adjustments. One part family that comes to mind here is AD8555/6/7.

    In other words: I boldly claim that you're facing a so-called XY problem: you think the problem is X, but it's really a part of a larger issue called Y. In your application, where does the signal come from and where does it end will determine the allowable tolerance on gain. Without sharing at least some of this information, there's little anyone can do to effectively help you, I'm afraid. Also, as a rule of thumb, typically relaxing the gain at the cost of e.g. requiring an extra effective bit of ADC resolution is cheaper than precision resistors. Make the ADC have twice the resolution (i.e. one more binary bit) and you can use 5% gain-setting metal film resistor and things will be just fine, whereas the starting point was "infinitely accurate and infinitely expensive" gain resistors - where you'd have "perfectly" matched the in-amp gain to the ADC input range. In practice such designs are unmanufacturable, and indicate enough of a novice approach that other error sources are likely to have been overlooked too, making the whole "let's get our gain perfect" issue completely moot.

Reply
  • Selecting the resistors to match your application is your job, and I doubt very much that the precise gain you claim you need is really a requirement. You'll be probably digitizing your signal and thus will have to do a system calibration anyway, and there are more error sources than just this gain, and the exact gains will be factored away during calibration. If you're clever, this calibration may be internal to the system without much cost impact.

    Selecting precise gain values using real parts is so much harder than writing nice-looking numbers in a table in a datasheet, that many designs typically avoid placing too strict gain limits. A nominal gain plus-minus a few percent is perfectly normal in most cases.

    You'd only need a very accurate analog gain if you're providing analog output that itself has to be related to some absolute units. Say that you're designing a transimpedance amplifier with a transimpedance of 1V/mA. The input terminals measure current, the output terminals provide a voltage. In such a device, the overall input-output relationship (transimpedance) has to be probably controlled better than "a few percent either way is OK"- then you'd either use a manually adjusted trimmer that a technician on the production line would adjust using a calibration fixture, or you could use a digital potentiometer with built-in non-volatile storage to adjust this transimpedance (the equivalent of a gain) - again by connecting the device to a calibration fixture that will "zero-in" the transimpedance and then save it in the digital potentiometer.

    You could also use an in-amp that has built-in nonvolatile storage for offset and gain adjustments. One part family that comes to mind here is AD8555/6/7.

    In other words: I boldly claim that you're facing a so-called XY problem: you think the problem is X, but it's really a part of a larger issue called Y. In your application, where does the signal come from and where does it end will determine the allowable tolerance on gain. Without sharing at least some of this information, there's little anyone can do to effectively help you, I'm afraid. Also, as a rule of thumb, typically relaxing the gain at the cost of e.g. requiring an extra effective bit of ADC resolution is cheaper than precision resistors. Make the ADC have twice the resolution (i.e. one more binary bit) and you can use 5% gain-setting metal film resistor and things will be just fine, whereas the starting point was "infinitely accurate and infinitely expensive" gain resistors - where you'd have "perfectly" matched the in-amp gain to the ADC input range. In practice such designs are unmanufacturable, and indicate enough of a novice approach that other error sources are likely to have been overlooked too, making the whole "let's get our gain perfect" issue completely moot.

Children
No Data