Your conclusion about the design not having a ground plane is very accurate !
Before looking to fix this problem look at the closed loop the RF-current is going to take. The current is injected in Common Mode on one cable and drained in a controlled impedance level by another cable. This means the you have to create a Low Impedance path (< 0.1 Ohm or so) from all input cable's wires to the corresponding output cable's wires. This effectively leads the RF current around your electronics and blocks the RF current from reaching any non-linear device: conducting diodes, transistors, rusty pipes, FETS, etc. where it WILL be demodulated. A/ All input wires (ground + signal too) should connect to ground for frequencies involved (150 KHz - 230 Mhz).(filter) B/ Some Common Mode serial impedance (in all signal wires + ground) L or R may help to improve the ground connection in A) C/ All output wires should connect to the same ground as the filters are connected to in A) The problem is two-fold now: 1) How to create the low impedance path between the 2 cable connection points ! 2) How to avoid the signal being attenuated by the ground connection. Ad 1) Basically only a large flat conductor (metal sheet) may create a physical large and still low impedance ground bonding between input and output (if remote). A PCB trace performs badly, a GROUND layer may perform better. (remember forever that the RF-impedance is inverse linear dependent of the conductors circumference) Note that you may have to create a separate ground plane, the circuit's GND or AGND may not be sufficient low impedance anyway. Putting input and output cables close together on the PCB improves this, but creates other problems such as unwanted couplings, or insufficient insulation in case to PSTN or other telecom connections. You might however, connect their shielding together at the point of entry. Ad 2) Carefully design the filter time constants, or in case the signal frequencies overlap the test frequencies, use a Common Mode Coil ( minicircuits transformers may perform well as 2 wire CM coil for PSTN purposes) to attenuate the Common mode current while leaving the signal current unattenuated. If you still have any questions, I will be happy to assist by phone ( just mail me privately) or you may ask me to resolve the problem on a commercial basis. An existing design will however need substantial modifications. Regards, Gert Gremmen, (Ing) Ce-test, qualified testing ================================== Web presence http://www.cetest.nl CE-shop http://www.cetest.nl/ce_shop.htm /-/ Compliance testing is our core business /-/ ================================== -----Original Message----- From: [email protected] [mailto:[email protected]]On Behalf Of David Gelfand Sent: Thursday, May 24, 2001 8:27 PM To: [email protected] Subject: EN 55024 Annex A.1 Help! When we inject as per IEC 61000-4-6, the 1 kHz tone is demodulated and is VERY loud in the telephone earpiece! This happens no matter which i/o cable we inject, even the power cord. The conducted RF appears to re-radiate inside the chassis and is being picked up by the line card. We use an AMD/Legerity Am79R70 chip, it is very sensitive at the RSN pin. Has anyone experienced similar problems? Has anyone passed this test with this chip, or with another? Thanks in advance, David. David Gelfand Regulatory Approvals Memotec Communications Inc.
<<attachment: Gert Gremmen.vcf>>
