4
\$\begingroup\$

I have a test instrument that measure spectral energy density versus frequency. The measurement is extremely low noise. The RF input on the test instrument is an SMA-connector where the inner conductor is for the signal and the outer conductor is electrically connected to earth ground (the 3rd prong in the power outlet; as typical for high-performance scopes and other RF test equipment).

I connect an SMA cable between this input and an SMA bulkhead connector that sits in a penetration panel of a shielded enclosure. The cable's outer conductor is electrically connected to the shielded enclosure, which is a solid (welded) metal box that is only grounded via the shielded cable to the test instrument. On the inside of the shielded enclosure, the SMA bulkhead connects via another SMA cable to a device under test (DUT) that produces a signal for the test instrument to measure.

Using the test instrument, I see a lot of low and high frequency spurious noise on the signal.

If I DISCONNECT the DUT, I see the same spurious noise, which leads me to believe the shielded enclosure (which is roughly a 4 foot cube) is acting like an antenna and picking up electromagnetic fields, which I'm sensing as a small voltage via the outer conductor of the SMA cable connected to the test instrument.

Is there any way to remove or reduce this noise (whose frequency could be anywhere from 1 Hz to 18 GHz)?

\$\endgroup\$
  • 1
    \$\begingroup\$ Can you draw a grounding diagram showing all of the faraday cages, electronics cards (and various signal grounds that each may have), and power supplies? That would help this discussion quite a bit. \$\endgroup\$ – kjgregory Jul 3 '14 at 2:14
  • \$\begingroup\$ [Disclaimer: I'm not an RF specialist.] Would it make sense to transformer-couple the RF signal and break the ground loop? Just an $0.02. \$\endgroup\$ – Nick Alexeev Jul 3 '14 at 2:14
  • \$\begingroup\$ I think a transformer could be a part of a solution, yes. \$\endgroup\$ – kjgregory Jul 3 '14 at 2:27
2
\$\begingroup\$

You likely have a ground loop in your setup.

I have a bit of a hard time following the description of your setup, but if you can illustrate your setup in the form of a grounding diagram, I would be willing to get into more detail with giving you specific advice.

The bottom line is that there are two keys to minimizing EMI (electro-magnetic interference) in your setup:

  1. Ensure that your entire setup is enclosed in a continuous faraday cage. This will shield your detecting elements, amplifiers, and wires that may act as antennas from EMI.
  2. Ensure that the faraday cage is connected to your signal ground at only one location. You can experiment with exactly which location works best. The purpose here is to ensure that return currents travel through signal grounds rather than through the faraday cage.

Wikipedia has a good article that may help you understand ground loops here.

It may also serve you well to use differential signaling on any long cables that you have. This would involve using something like the AD8138 for driving analog cables along with twinaxial cables or for digital signals using LVDS, RS-422, or similar signaling method.

One more piece of advice: Make sure you are using a linear power supply as opposed to a switching power supply. Switching supplies, while more efficient, can contribute to noise in your amplifiers through the supply rails.

Good luck!

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

If possible, use a high input impedance follower at the DUT end and use it to drive the shield. Then do not connect the shield at your test instrument end. If your "earth" ground has no current, it can be the reference for both devices. The driven shield will protect against a number of ills. The same technique is used to guard analog inputs on PCBs.

If you can use shielded cable inside a shield, drive the outer shield and use the inner as usual.

\$\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.