Cmos inverter vs opamp noise simulation

[dschwartz]

Hello masters!!
I have been playing with LTspice lately, making multiple feedback filters, and comparing the behaviour of cmos inverters versus opamps.
I made two identical circuits, ordinary MFB low pass filters with high Q, and made a noise analysis..
To my surprise, the inverter version showed significantly less noise than the opamp simulation (4049 vs tl072)..like 300nv/√hz versus 500nV/√Hz
Are cmos quieter than opamps when not being pushed for gain or is that the cmos models don't include the noise parameters?..

[Rob Strand]

To my surprise, the inverter version showed significantly less noise than the opamp simulation (4049 vs tl072)..like 300nv/√hz versus 500nV/√Hz
Are cmos quieter than opamps when not being pushed for gain or is that the cmos models don't include the noise parameters?..

My trust in the noise simulation for a 4049 spice model would be extremely low.  The reasons:
- Mosfet noise isn't so easy to model.
- For digital chips like the 4049 I'd only have a small amount of faith in the model being accurate for analog.  Something like obscure like noise would hardly get on the radar for the model.

That's all IMHO.  I'm happy to be corrected!

BTW 300nV/rtHz and 500nV/rtHz seems high for an opamp.  Is that the noise at the peak of the high-Q response?

[dschwartz]

Yea..the thing is that in my experience when using inverters linearly with low gains, they are pretty quiet until you really go high gain, like the red llama (inverter with no input resistance, that's noisy) i had the hope that their limitations were the cause of the low noise in simulations.

And yes, the 300 and 500uV noise are on the resonant peaks. The rest is 40uv or so..pretty good for a filter, considering the TL072 has a 20uV noise density. Maybe the open loop gain has something to do?

[Rob Strand]

Yea..the thing is that in my experience when using inverters linearly with low gains, they are pretty quiet until you really go high gain, like the red llama (inverter with no input resistance, that's noisy) i had the hope that their limitations were the cause of the low noise in simulations.

You have to be a little careful interpreting noise.   There's input referred noise (apparent noise injected into the input) and output noise (the actual noise coming out).   An opamp will have a certain noise voltage and that is equivalent to an input referred noise.   If the opamp has a gain of 10 the output noise will be 10 times the input noise.  If you have a gain of 1000 the output noise will be 1000 times the input noise.  If you listen to the output noise the gain of 1000 case is obviously noisier but the opamp is still contributing the same noise.

And yes, the 300 and 500uV noise are on the resonant peaks. The rest is 40uv or so..pretty good for a filter, considering the TL072 has a 20uV noise density. Maybe the open loop gain has something to do?

The best thing to do is wire up a gain of 1 buffer and see what noise you get out if that.  That will give you an idea of the noise of your opamp model - it could be wrong.   A source of additional noise is the resistors.  If you have large resistors it will increase the overall noise.   Try 1k or 10k.