Massimo, I am not familiar specifically with LVDS, but I have experience with very high speed data transfer using differential signaling (Fibre Channel-like circuitry - 531 MHz fundamental, 1.2 volts on each side). In my experience, EMI is greatly influenced by imbalance between the two differential lines. A difference between the two signal paths produces an imbalance called delay skew. Delay skew is an imbalance that creates a net current down the differential pair. This net current is a common-mode current on the pair that must return on the inside of the shield of the cable. This makes shielding of the cable (and connector) important.
If you don't shield the cable, the common-mode current will still return, but will find whatever path it can - probably on other wires in your cable. This produces an uncontrolled situation in which you, as an EMI engineer, are at the mercy of fortune. An impedance discontinuity will cause a reflection of the differential signal, which may, or may not influence EMI. EMI is caused by the common-mode signal. Since your differential signaling structure supports both Odd Mode propagation (differential) and Even Mode propagation (common-mode), you must look at how the impedance discontinuity appears to the common-mode. If you get a significant reflection of the common-mode current, you may increase the apparent radiated EMI. So, how should one terminate a differential signaling transmission line? The safe way is to provide a termination impedance that properly matches both even and odd modes. You must characterize the even and odd mode characteristic impedances of your cable and then provide a matching network (a Pi, or a T) that will match these two impedances. The simple differential termination of a resistor between the two differential lines can match the Odd Mode, but presents an infinite impedance to the Even Mode (maximum reflection). Does this always make a practical difference in terms of radiated EMI? No! It depends a lot on the details of your system. Since the differential signal is creating EMI, rather than a clock on the board, this means that the data pattern of the signal is also an important factor in determining whether you have an EMI issue. Repetitive signals are more troublesome than complex data patterns. A discussion of effects of skew and data patterns may be found in: Hoeft, L.O., et. al., "Spectral Analysis of Common Mode Currents on Fibre Channel Shields Due to Skew Imbalance of Differential Signals Operating at 1.0625Gb/s'" Record of the 1998 IEEE International Symposium on Electromagnetic Compatibility, Denver, Vol. 2, pp. 823-827. I hope this is helpful. Jim Dr. Jim Knighten e-mail: [email protected] Senior Consulting Engineer NCR 17095 Via del Campo San Diego, CA 92127 http://www.ncr.com Tel: 619-485-2537 Fax: 619-485-3788 ---------- From: Massimo Polignano Sent: Wednesday, October 14, 1998 12:40 AM To: [email protected] Subject: LVDS technology We are about using LVDS (Low Voltage Differential Signaling) technology to transfer high speed video data to a XGA display 1 m from the driver board. Differential signaling is very good to improve EMC characteristics of the link, but in my opinion is also very critical in actual layout: any lack of symmetry can jeopardize its performances. Has anybody esperience with these circuits? How much EMI behaviour depends on equal signal paths? Is it advisable to shield cables? What does a discontinuity on the transmission line (change in characteristic impedence) produce? Thanks in advance for any comment Massimo ------------------------------------------------------- ESAOTE S.p.A. Massimo Polignano Research & Product Development Regulatory Affairs Via di Caciolle,15 tel:+39.55.4229402 I- 50127 Florence fax:+39.55.4223305 e-mail: [email protected] --------- This message is coming from the emc-pstc discussion list. To cancel your subscription, send mail to [email protected] with the single line: "unsubscribe emc-pstc" (without the quotes). For help, send mail to [email protected], [email protected], [email protected], or [email protected] (the list administrators). --------- This message is coming from the emc-pstc discussion list. To cancel your subscription, send mail to [email protected] with the single line: "unsubscribe emc-pstc" (without the quotes). For help, send mail to [email protected], [email protected], [email protected], or [email protected] (the list administrators).
