[sdiy] basic inverting op amp question 101

[Mike Beauchamp]

On 12/30/19 11:38 AM, Next Expanse wrote:
> (On top of what has been said about current v noise and bias current...)
> 
> The output stage of op amps is generally class AB, and so greater 
> loading puts that A part of the section in a smaller part of the signal, 
> and therefore might cause more distortion. TL07x’s in particular are 
> known to get much better results for lower output currents. Doug Self 
> published somewhere (that I can’t find again) that in tests, a 22k 
> feedback gave the least THD + noise for the TL07x opamps. Unfortunately, 
> (IIRC) he didn’t say much about the current draw of the following stage, 
> whether this was for unity gain or something else, etc., so I’m not sure 
> how useful that number is. At any rate, it’s a decent ballpark figure. 
> This is part of why the TL07x often comes out worse in tests than people 
> experience it in real world circuits; most other audio opamps can do at 
> least 10k if not 1K with no degradation, and I’ve yet to see a 
> comparative test of the TL07x that gives it optimum loading.
> 
> Second thing (and I’m still chewing on this one so corrections are very 
> much appreciated), distortion aside, there’s actually an optimum 
> feedback impedance for the lowest noise characteristics for an op-amp, 
> and that’s equal to the voltage noise divided by the current noise. This 
> doesn’t take into account the Johnson noise of the resistor itself. 
> Basically, you can think about this by imagining a constant voltage at 
> the inverting input, tied via a resistor to another constant voltage. 
> There’s then some current flowing into the inverting node, which has to 
> be equal to the current flowing out of it, across the feedback resistor, 
> to the output of the op amp. That current plus the feedback resistance 
> determines the output voltage. Now think about voltage noise on the 
> inverting node. Tiny changes in voltage make tiny changes in current in 
> accord with V/R=I. So the higher the resistance, the less effect voltage 
> noise has on the output. Conversely, if we think about current noise, 
> this is going to require voltage changes on the output in accord with I 
> R = V, so lower resistance is going to give less effect to the current 
> noise. The optimum is at R = Vn/In. For ne5532, that’s about 6.8k, and 
> you’ll see this value in a lot of circuits. However, because this 
> doesn’t take Johnson noise into account, the value’s going to skew a bit 
> lower, depending on total noise of the opamp, gain, etc.
> 
> The good news is that rarely do you have to worry about any of this all 
> that much if you aren’t designing a mic preamp. Just don’t use anything 
> too low. Often people just use 10k for everything that isn’t a TL07x, 
> and 50k or so for those. Actually, in the synth world, people often just 
> use 100k for everything and they seem to be happy, but I wouldn’t go 
> quite that far, personally.
> 
> -Evan

Great post, thanks Evan!

Mike