2
\$\begingroup\$

Assuming that we have a sensitive electronics that is battery powered, which consist of a very sensitive 32 bit adc is reading a voltage from a chemical reaction and you place it inside a chunky 1 inch thick metal shielding (I could not think of a better example). The circuit inside is built to the best ideal configuration. Would connecting the metal shield to the circuit's ground inside affect the result or its best to leave the circuit inside floating to best protected from whatever stray interference.

What if this time the the electronics inside is powered from mains voltage via a wall power adaptor. The metal shielding is still there but this time from the wall power adaptor 2 wires is passed through to the thick metal shielding to give power to the circuit inside. Would connecting the metal shielding to ground now a must?

\$\endgroup\$
7
  • 2
    \$\begingroup\$ I can't speak to general situations. But in a specific case, our best results came from buying dice, learning to wire bond, and building these onto a 3-tiered TE stack for cooling inside a custom (similar to TO-3) package with a sapphire window. The MCU was "mostly" isolated from the analog section and digitized the photon flux for transmission outside the metal package case. The metal case was NOT connected, but isolated from the internal system. We were able to reach photon flocking noise levels as a result, which was great! \$\endgroup\$
    – jonk
    Commented Jul 8, 2020 at 7:46
  • 2
    \$\begingroup\$ I would say make provision and try it either way. I have the opposite experience with @jonk where connecting the electronics to the shield reduced the noise around 10x. This was a silicon photodiode front end with an ADC measuring light within a spectrometry system. \$\endgroup\$
    – regimon
    Commented Jul 8, 2020 at 8:04
  • 1
    \$\begingroup\$ @regimon I can't disagree with you about that. We tried both ways, too! \$\endgroup\$
    – jonk
    Commented Jul 8, 2020 at 8:04
  • \$\begingroup\$ I see if that is how it goes then i will just have to expose the grounds (Digital and Analog) to the standoff holes in my board, to give a bit more flexibility to whatever my case may be which ever gives the best result. \$\endgroup\$
    – DrakeJest
    Commented Jul 8, 2020 at 9:31
  • \$\begingroup\$ Is it a differential measurement using differential balanced amplifiers? \$\endgroup\$
    – Andy aka
    Commented Jul 8, 2020 at 10:12

2 Answers 2

1
\$\begingroup\$

If your interference is mainly electrostatic you should earth the shield .Short low inductance connection is always good .The earth connection quality is much more important at high frequencies . In fact an unearthed electrostatic shield can make things worse than no shield . For magnetic interference earthing is not important .

\$\endgroup\$
-1
\$\begingroup\$

First you should know that metal shielding is for protection against stray magnetic fields(Electromagnetic Interference, which induce eddy currents in wires/conductor). In the first part of your question, there is no need to connect your metal shield to the ground. You can if you want to, but ground loops are generally avoided if you have long wires. In the second scenario, you just have to buy wires that are either co-axial or twisted pair. Both are optimized for minimizing external interference in the signal they are carrying as well as minimizing the magnetic fields they are providing. Also, avoid switching ac to dc power supplies, they are also one of the main causes of electromagnetic interference. But if you do have to work with them. Use a three terminal prong to have a ground/earth reference (it alleviates the problem to some extent) and use a voltage regulator at its output so that ripples are reduced. (Adding external Capacitors and inductors requires alot of experience and fine tuning so voltage regulator is your best bet).

you can also use other components such as an opto-isolator at the ADC output so load and external factors do not affect its output.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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