Because of the holidays, TI E2E™ design support forum responses will be delayed from Dec. 25 through Jan. 2. Thank you for your patience.

Similar topics
Recommended
Bruce Trump

I’m not referring to op amps in your parts bin. Those should be in anti-static bags or conductive foam. What about the one on your circuit board—the unused op amp in a quad or dual package. Hummm??

A recent question on our forums spurred me to address this subject but in the process, I ran across a great article by my colleague, Todd Toporski. He did an excellent job of covering the important issues and reasons. Check it out, here. I’ll summarize and add some additional thoughts.

It’s best to connect the op amp in a real op amp circuit with feedback; a unity-gain buffer is an obvious choice since no additional components are required. Then connect the input to a voltage within its linear input and output range. Connections (or open-circuit inputs) that can potentially overload the input, output, or leave the amplifier in an indeterminate noisy state are undesirable.

A suggestion on circuit board layout: Position any unused op amps for possible future modifications. You may find a use for a spare op amp in a redesign or future product spin. Think ahead. Make the connections to the spare op amp on top and bottom circuit board layers where minor surgery can easily test your changes. You might even provide layout positions for feedback components with traces to tie-off nodes that can be easily cut.

Another possibility… you could avoid all these issues completely by selecting an amplifier type that has single, dual and quad versions—OPA322 is one example. This can allow an optimized circuit board layout with no orphans while using op amps with the same specs and behavior.

A word of comfort to those who may not have used a preferred method to tie off an unused amplifier: You are unlikely to greatly disturb the working op amp(s) in the same package. While you may be drawing some extra current in the unused amp, your system is unlikely to crash and burn. Most modern op amps have independent biasing circuitry, unperturbed by overloads in other channels on the same chip. If your circuits are working, relax and follow best practices in your next design.

Other readers (and I) would appreciate reading your additional tricks and ideas in comments below.

Thanks for reading,

Bruce       email:  thesignal@list.ti.com (Email for direct communications. Comments for all, below.)

  New...   Table of Contents for all The Signal blogs, organized by topic.

 

Anonymous

  • Christian--  Interesting question. An op amp that is not unity-gain stable should not be connected in unity gain, even if it is not being used to pass a signal. The oscillations that would likely result could easily couple to other amplifiers. You could, of course, use two additional resistors to set an acceptable gain for that amplifier and connect the non-inverting input to a node within its common-mode range. If adding these resistors is undesirable, it would be wise to check with an expert on our E2E forums. For any specific amplifier, there may be other acceptable options. The OPA4022 is supported on the high speed amplifiers forum and an expert on this device can provide advice. -- Bruce

  • I wonder whether the unity gain approach still applies to parts like THS4022 which are not unity gain stable. If not, what would you recommend in this case?

  • Dox--  Rarely (I'd say never, but there might be some exception) would an op amp be designed to turn off internal bias circuitry when an amplifier is forced to saturate. This would likely prevent or greatly delay recovery in real signal overload conditions. Far more likely is the case described where the current increases as it "strains" to accomplish the signal condition that has been applied. Best approach--keep an unused op amp in linear operating range.

  • While I always use the "unity gain follower" approach, I sometimes wonder if it would make more sense to force the input and/or output stage to saturate in a particular direction to turn off internal current mirrors and shut down the opamp.  Any advice on this approach?

    If this makes sense, shouldn't the approach be explicitly described in the datasheet?

  • If I have the space and the extra cost doesn't cause the marketing dept to have a fit, I will usually put a couple of resistors in - one from output to one of the inputs and then resistors from each input to (as mentioned) a node that is in the I/O range. I'll use resistor values that are already used someplace else (so number of part types being used doesn't increase). Setup like this, it makes it much easier to 'mod' the board in production if needed.