Post History

The problem with cable emissions from HS data and SMPS noise is very common even with UTP and ribbon cable with adjacent grounds on differential signals. 

I have seen this frequently on HDD testing at EMC sites, whereas the HDD emits nothing of interest.  The problem is due to the imbalance of the cable single-ended to free-space relative to the drivers and receiver, while the differential impedance error is not a radiation factor. Thus the cable becomes an effective antenna per unit length. The best UTP twisted pairs or ribbon cable might be around 60 dB for imbalance to ground, even with adjacent gnd lines.

 Remember "ground" is defined as your 0V reference, wherever that is, and in this case for EMI the detector antenna and Rx ground. 

Unlike a laser-trimmed INA with 120 dB CMRR, the physical tolerances of insulation thickness and conductor thickness with skin depth are impossible to control to that equivalent of 1 ppm.

The choices are always;

1. improve the CMRR balance, if possible.
2. suppress the CM radiation with a single or double shield or CAT-5 STP cable.
3. Raise the CM impedance sufficiently high with a balun or CM choke rated for at least 2 decades of f in the region of concern. This raises the impedance ratio to the shunt capacitance of the CM emissions with the shield capacitance in STP cable "ground". The shield termination important and as done in high speed Ethernet it is RC coupled with 75 ohms + 4.7 nF (?) to 0Vdc.  
4. For non-DC baseband data signals consider a Mag-jack PHY transformer which has the CM choke.
5. Even SMT CM chokes for HDMI signals might be insufficient alone ( observed in ASUS designs.) improved with the shielding CM balun to add capacitance and bypass CM signals t local gnd.

If you have ever looked EEG or EKG designs you will notice they do not rely on the INA CMRR of 120 dB to reject line hum or other noise because of the imbalance of the cables and high CM noise of the subject. An additional 30 to 60dB CM rejection is done with filters.

6. The CM choke solution is not limited to data but also the power interface cables.


simple Lab tests
---

Using a shorted loop scope probe is a good way to start sniffing for CM emission noise in the near-field noise. Then attach that to a spectrum analyzer and compare with the noise on your computer cables.  You can also use a short piece of wire to scope probe or BMC T on a coax cable for an antenna for sniffing emissions.  Of course, there are fancy probes, if you want to get calibrated E/H fields at a standard distance.