Discussion:  Figure 1 shows a plot of the current measured at both ends of the injection clamp defined in IEEE Std 802.3™-2002 Annex 40B which is shown in Figure 2. The clamp and its connection to a test board are shown. Although the Fischer F-65 current probe is shown positioned next to the test board, in fact it was placed as close as possible to the right end of the clamp to generate the data in Figure 1.

Looking at the top trace of Figure 1, one can see three well defined resonances due to the injection clamp at about 310 MHz, 620 MHz, and 930 MHz. The resonances were in the clamp, not in the test setup. Note that up to 250 MHz, the upper frequency used for 1000Base-T Ethernet testing, the injected current supplied to the test board is well behaved. The downward slope is likely due to the inductance of the 20 cm cable between the clamp and the test board. For 10GBase-T Ethernet testing, the clamp is being used up to 1000 MHz. Taking into consideration the data shown in Figure 1, this is probably not a good idea.



Figure 3 shows ferrite cores on the 10GBase-T Ethernet cable on the opposite side of the clamp. IEEE Std 802.3™-2002 Annex 40B calls for just two ferrite cores here. I do not believe one can achieve enough isolation between the uncontrolled common mode world over a bandwidth all the way to one GHz with just two cores. Lab experience bears this out as touching the cable to the left of only two cores, in one test, significantly changed the current plot of Figure 1.

Figure 3 shows a solution, 11 ferrite cores comprised out of a mixture of low and high frequency cores arranged in a non-repeating pattern. When the current probe was positioned shown in Figure 3, the bottom trace of Figure 1 results. The lower trace is near the noise floor of the analyzer at many frequencies. The ratio between the two traces is 20+ dB, enough to insure adequate isolation of the uncontrolled common mode environment past the ferrites, so that common mode variations do not significantly affect the current delivered to the test board or EUT. The 11 ferrites used here seem to be the minimum needed unless much better ferrite cores can be found. For the case of just two high frequency ferrites, the two plots of Figure 1 do not show much separation, especially below 500 MHz.

An additional problem may exist as well at the high end of the frequency range used for conducted immunity testing of 10GBase-T Ethernet ports. The problem is caused by the clamp which ends abruptly and may capacitively couple more to one wire of a twisted pair than the other wire of the same pair. This seemed to be happening during this test. When the clamp was moved about 1/2 cm, comparable a twist length of the pairs in the cable (the twist lengths are all slightly different to minimize crosstalk), a difference was seen on some pairs in the resulting differential mode interference. This effect needs to be more thoroughly investigated.