3
\$\begingroup\$

If I will build this circuit in real life, I will probably see a mV noise with the scope due to high cascaded gain and probably some oscillations.

I use LPC358 which is specified to (Input-referred voltage noise) 178nV/SQRT(HZ). I am trying to figure out how small can be the input signal (now is GND at C3) to be masked by noise generated by circuit itself.

My simulation is clearly wrong, showing a clean output with some nV flicker.

How can I add some noise in LTSpice, based on the fact that any opamp has specified noise, that will emulate real worlds noisy output so I can select from several opamps, gain, etc?

enter image description here

enter image description here

\$\endgroup\$
5
  • 1
    \$\begingroup\$ Have you tried a LTspice noise analysis? Even with this tool, many opamp models are crude. You get output noise optimistically low. \$\endgroup\$
    – glen_geek
    Jul 9, 2017 at 15:45
  • \$\begingroup\$ Yes, I tried, but this shows nothing to me compared to real life that I can measure. I try to estimate for a real prototype. \$\endgroup\$ Jul 9, 2017 at 16:00
  • 2
    \$\begingroup\$ The specifics of your physical measurement can fool you; make sure you are probing immediately next to the ground reference. Ground loops in measurements are surprisingly easy to create. \$\endgroup\$ Jul 9, 2017 at 16:15
  • \$\begingroup\$ Does LTSPICE provide a transient-random-noise mode? \$\endgroup\$ Jul 10, 2017 at 4:24
  • \$\begingroup\$ @analogsystemsrf If you search in the LTspice Yahoo Group for tranoise, you'll find one Robert Macy who made a calibrated transient+noise analysis. You'll also find steps to make it. \$\endgroup\$ Jul 10, 2017 at 5:58

2 Answers 2

3
\$\begingroup\$

It looks like you have 30 kHz bandwidth so take the square root of 30,000 and multiply that by the first opamp's noise voltage then you have the equivalent noise at the input due to the first opamp voltage noise. Multiply that by the total gain and you have, near enough, the output noise as an RMS figure. That should be near enough to get a picture in your mind's eye. I wouldn't use any sim for estimating noise by the way. What I have described is the normal approach.

\$\endgroup\$
1
  • \$\begingroup\$ I accepted this because it's closest to my needs. However, I injected a VSOURCE with equivalent voltage noise (random) at first stage input and have a picture on output amplified noise. Thanks. \$\endgroup\$ Jul 12, 2017 at 7:01
1
\$\begingroup\$

Here is transient noise (+10dB SNR) atop a squarewave signal-source at 200KHz. The "system" is RC low-pass-filter at 159KHz.

enter image description here

With 100 samples per input squarewave, the samplerate is 20MHz, with Nyquist freq being 10MHz.

The tool, Signal Wave Explorer, uses the Nyquist freq at the noise bandwidth. You will notice almost all the noise have been removed, by the 0.159/10 ratio of LPF_cutoff / noise_cutoff.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.