Learn Three Ways You Can Create Bode Plots - Workbench Wednesdays 64

At the end of the video, before I rushed The Intern to the ER, I showed this picture:

I didn't show in the video how closely the LM358-based circuit matched between simulation and measured. (The -3dB point was off by a few kilohertz--but close enough.)

So any thoughts as to why the TL081-based circuit has such a different response in simulation versus being measured?

For reference, here are the LTSpice files I used.

wbw64 - lm358 and tl081 LTspice sims.7z

Here's a possible theory you might test.

The TL081 has a pair of offset pins which you appear to have left unconnected. As well as using them to adjust the offset, which you wouldn't need to worry about for ac-coupled signals with a lowish gain, one of them could also be used to adapt the compensation for better bandwidth (see the part of the TI datasheet that refers to feed-forward compensation - there's a graph and a schematic, but it's not all that clear on a quick, casual read that they relate together or what the circuit does). On your breadboard, with all the parasitic capacitances between traces, it might just be that you're doing your own adaption of the compensation. A simple way to find out is to bend both those legs so that they're pointing straight up and have minimal capacitance to anything else and see if the Bode plot changes.

I have been able to change the simulation to match the measured result.

Interesting idea on the test with offset pins. I'm curious what will happen, so I'll give it a try.

I think BE just had bad luck with the intern/murphy. The TL081 spice model doesn't include any capacitance on its input, let alone the parasitic capacitance on a breadboard.

Green: No cf, Red: Cf=4pF

On the LM358 the output and inverting input pins are adjacent so you get the breadboard row-row capacitance a helpful bonus by introducing a zero to help counteract the pole created in the feedback network from capacitance to ground. 

If the peaking problem disappears with smaller feedback resistors say 1kOhm, then the gain was likely peaking due to capacitance on the inverting input node (reducing the closed loop phase margin).

Gosh. It's almost like there was a reason I made sure to attach the TL081 model I used.   (Not sure if you used that one or another. I think I got that one from TI.)

I found just 5 pF got a similar result. (I had a Cf cap at one point too.) Enough to realize what was going on.

I knew the prolific members here would likely catch this one. But it did throw me for a few minutes.

And by "me," of course, I mean The Intern.

I tried your circuit with a TL072 (don't have a 071) and got a substantially different results, the gain peaked maybe 5-10%.

So I was intrigued to dig a little deeper. I had the bend the output pin up and solder the one leg of feedback resistor to demonstrate the effect.

Again, great episode!

Cool! My plan is to solder a duplicate circuit to a perfboard (with generous solder) and do the sweep/plot again when I talk about how to automate the measurement in python (for those that don’t have an FRA option on their scope.)

But your way sounds more fun. 

That sounds wonderful! Being able to automate your measurements is a game changer for sure!

To clarify the reason I didn't see the issue nearly as pronounced as you had was because the breadboard was doing me a favor! 

I got this capacitor for free:

That sounds wonderful! Being able to automate your measurements is a game changer for sure!

To clarify the reason I didn't see the issue nearly as pronounced as you had was because the breadboard was doing me a favor! 

I got this capacitor for free: