Accelerating the measurement of low ber
By william b. gardner
There are many possible causes of bit errors in digital lightwave systems: thermal noise, shot noise, laser intensity noise, crosstalk, jitter, and others. The bit error ratio (ber) is the fraction of bits that are erroneous. Working Party 4 of the International Telecommunication Union`s (itu`s) Study Group 15 has published a number of Recommendations governing digital systems (for a recent summary, see Lightwave, February 1998, p. 44). The itu`s acceptable ber level for lightwave systems has dropped from 10-10 to 10-12. This level of performance is so good that it`s actually difficult to verify. At a 2.5-Gbit/sec bit rate, a 10-12 ber means that an error occurs only once every 400 sec on average. The accumulation of 10 errors (a bare minimum for determining the ber) would then require over an hour of measurement. Even longer times would be necessary at lower bit rates.
One solution to this problem is to accelerate the error rate in a controlled way, so that a few hundred errors occur in a matter of minutes rather than hours. The ber for an unperturbed system can then be determined by extrapolation. Ways of doing this have been documented in Optical Fiber System Test Procedure (ofstp)-8 by the Telecommunication Industry Association (tia). ofstp-8 has been balloted by the tia Joint Subcommittee on Singlemode Systems FO-2.1/6.6, chaired by Allen Cherin of Lucent Technologies (cherin@ lucent.com). ofstp-8`s editor is Rick Neumann of Vixel Corp. (rneumann@ denver.vixel.com), with assistance from Felix Kapron of Bell Communications Research ([email protected]). The same test method is at the committee draft for comments stage in the International Electrotechnical Commission`s (iec`s) Subcommittee 86C Working Group 1 on Fiber Optic Systems and Subsystems, also chaired by Allen Cherin of Lucent Technologies.
ofstp-8 offers two methods for accelerating the measurement of ber:
Sinusoidal interference method: A sinusoidal signal is added to the digital signal at a point before the receiver`s decision circuit, to increase the measured ber. The sinusoid may be injected optically or, if there is access to the receiver circuitry, electrically. ber measurements in the range of 10-5 to 10-10 are performed for several amplitudes of the sinusoid, so that short measurement times suffice. On a semi-log plot, the ber is then extrapolated back to zero amplitude, where the ber measurement would be too time-consuming to be routinely performed.
Threshold modification method: In this method, a controllable DC offset (either optical or electrical) is used to vary the receiver`s decision threshold level. In a manner similar to the previous method, faster measurements of larger ber values are performed at several threshold biases, and the unperturbed ber is then determined by extrapolating the data to zero bias. Since an electrical offset can be either positive or negative, a plot of ber versus bias in this case gives not only the ber of the system as presently adjusted, but also the minimum ber that could be achieved by optimizing the decision threshold.
In both approaches, mathematical formulas for the ber are provided to facilitate the graphical extrapolation procedure. Experimental results are also presented which confirm that the methods described above agree with each other, and with the ber measurement performed in the conventional (unaccelerated) manner. o