Op Amps for Everyone

Beware: The guide on how to calculate noise of op-amp amplifiers is full of errors, which resulted in great confusion as I tried to learn this topic and my calculations disagreed depending on which instructions I followed.

First big mistake: They combine the noise of the resistors with the output noise of the op-amp:

Instead, they should be combining the noise of the resistors with the input noise of the op-amp, and then multiplying this by the (noise) gain to get the output noise.

Second big mistake: They sum the noise of parallel resistors together. Instead, they should be calculating the equivalent resistance seen by the source, and then using that. For instance, they combine a 10 MΩ resistor and a 100 kΩ resistor as if they were in show more series (10.1 MΩ → 57.1 μV), but they should be combining them as if they are in parallel (99 kΩ → 5.6 μV).


And no, that equation doesn't match the text describing it, either. There are a number of other miscalculation/typo errors in this section, making it useless for learning how to calculate noise.

There are other fundamental errors, too: In another part of the book, they claim that "unity gain stable" means that an inverting amplifier with gain less than unity is unstable, so you have to attenuate first and then amplify:

This is the most often misdesigned circuit of all the cases. Many inexperienced designers create an unstable stage by attempting to extrapolate an inverting gain stage to the attenuation case by making RG greater than RF. The easiest fix for this problem is to use a voltage divider followed by a unity gain buffer, as described in Section 5.2. If inverting gain is absolutely required, then a similar solution can be implemented by adding a voltage divider to the input of an inverting gain stage.

This is wrong, but it's still being echoed around the internet, confusing n00bs. There is nothing unstable about making Rg greater than Rf. The truth is that "unity gain stable" applies to the noise gain, not the inverting signal gain. Noise gain of this circuit can never drop below unity. Their proposed solution increases both parts count and noise for no reason:



(And that equation is wrong, too. It should be m = (Rf * Ratten)/(Rin*(Rin 2*Ratten)).)

So yeah, don't trust anything you read in this book until you've double-checked it. show less