1
\$\begingroup\$

I am designing a circuit with a very sensitive analog part. In order to eliminate SMPS noise from a regular power supply, the device has its own non-SMPS power supply followed by an ultra low noise linear regulator (LT3042). However even after this, the device still sees some noise coming from mains socket if a noisy equipment is connected on the adjacent socket.

Usually, simply choosing a socket on a different wall, or moving the equipment power cables apart, significantly reduces noise. I see this as free noise reduction over and above what can be done in the circuit.

I would like to have some mechanism in the device that can inform the user about noise on the line. I don't need exact ripple voltage measurements just something good enough to tell that cable arrangement B is better than cable arrangement A. This would allow the user to try different sockets and choose one that has the least noise.

Alternatively, the device can simply have a noise threshold below which a "Power Good" indication will be provided to the user.

Just FYI, this is for an industrial application and the position of the device or its wiring won't change very often. If during operation the noise level rises above threshold, say due to some other equipment connected nearby, the device can notify the master via RS485.

\$\endgroup\$
1
  • 2
    \$\begingroup\$ I think you'll actually need to figure out a way to make your equipment less sensitive to this noise if you want it to pass emc tests, rather than just band-aiding the issue. Have you identified where the noise is being injected into your circuit (ground plane? measurement path? etc)? \$\endgroup\$ Oct 15 '17 at 6:26
1
\$\begingroup\$

Well duplicated here and on the web are imperfect desires for low noise with no specs on: signal {V,I,Z, BW} or Noise {...}. The answers do not depend so much on choice of PS , rather an understanding that noise ingress into high Z signals can be induced by every mode; conducted , radiated, common mode impedance ratio from source that converts to differential noise.

The answers are everywhere, Baluns, twisted pairs shielding, CM chokes. Pi filters , Cable orientation and proximity , decoupling caps. Each of many solutions depends on specs (a priori)

The expertise and wisdom comes from simple impedance divider relations from source to destinations using CM Impedance., Zcm and differential Impedance Zdm, from unbalanced lines. CM chokes serve to raise Zcm over some limited spectrum to create more balance and greater attenuation. Twisted pairs and shielding serves to lower Zdm(f ) from stray coupling either pF from 240Vac rms E fields or nH from current nearby induced power cords.

You will need many articles or a good book (Henry Ott) on EMC, EMI reduction, to fully comprehend.

\$\endgroup\$
3
  • 1
    \$\begingroup\$ Isn't the OP asking how to detect the relative magnitude of the noise rather than how to reduce the noise? \$\endgroup\$
    – Dan Laks
    Oct 15 '17 at 8:49
  • 1
    \$\begingroup\$ Yes. but this is a rather unskilled way that may not resolve the real problem which ignores the cause of common mode noise. If you can fix the sensor cabling with above solutions then it will be immune to socket A or B. (XY problem) \$\endgroup\$ Oct 15 '17 at 8:54
  • \$\begingroup\$ I agree with Tony: measuring the "noise" is not the right method to solve this. I agree that the root cause is not well understood. I am not even sure that technically this condition is ok to pass the EMC certification tests. I liked "Printed Circuit Board Design Techniques for EMC Compliance: A Handbook for Designers" from Montrose a lot for its practical approach. Generally, circumventing issues this way will make the core issue just hit you harder in the future. Or is it required to just "dumbly" reply to the literal question rather than the real need detected in the question? \$\endgroup\$
    – le_top
    Oct 15 '17 at 10:17
1
\$\begingroup\$

You could, perhaps, arrange to have your device short its own input and measure the noise floor, however that may or may not accurately represent what the noise is like when genuine signal is present, particularly if the signal is coming from outside the instrument. Chances are the external noise on the power is coming in as common-mode noise and changing the input circuitry around much may change the effect of the noise.

\$\endgroup\$
0
\$\begingroup\$

[TITLE] Measuring ripple voltage

[QUESTION] to eliminate SMPS noise from a regular power supply

What you mention in the title and what is elaborated in the body of this question induces the possibility that you are confusing some concepts. The Ripple has a well defined frequency spectrum and could be better measured by a transformer that could amplify it to at least give an estimate of its magnitude. In other hand, the Noise is a phenomenon usually of origin not precisely defined. In any case, it would only be possible to determine whether this approach works by experimentation itself. It may be good in theory, but just simulation or assembly with real components confirm.

\$\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.