Teledyne LeCroy Eye Doctor II - Manual (Page 43)

Operator's Manual
As noted previously, this value should not be changed.
The DFE taps are optimized together with this value; changing it changes the clock skew into DFE and
degrades the result. Instead, to shift the eye diagram horizontally, use the Deskew control on the SDA II
Clock Recovery dialog.
Conclusion for case 3: beyond a certain level of degradation, it becomes very difficult to compensate with
only receive equalization. Eye Doctor II can compensate by using both a CTLE and DFE, or both a CTLE
and FFE, but real silicon implementations are unlikely to do either. In such cases it is necessary to use TX
equalization (meaning de-emphasis or pre-emphasis). Eye Doctor II can simulate TX equalization (by add-
ing it to a waveform that doesn't have it) using controls under the Emphasis tab.
FFE and DFE Training Algorithm
EyeDoctor II includes a training algorithm that determines the optimal tap values to use in the FFE and
DFE filters. The method used is the Levenberg–Marquardt algorithm. The variable that is minimized is the
range of the high and low voltages near the center of the eye, as shown in the figure below. Minimizing
this quantity maximizes the eye opening.
You can configure certain arguments that are part of the training algorithm. Both FFE and DFE include
the following controls, which can be set independently:
Number of Taps: The training algorithm will create a filter with the number of taps selected.
Auto find levels: Checking this will have the algorithm find the upper and lower levels of the equalized
eye as well as the decision level to use. When selected, the voltage levels of the minimized locations of
high and low states will tend to match the outmost voltage levels of the unequalized eye.
Upper Level: Voltage to place the upper level of the equalized eye.
Lower Level: Voltage to place the lower level of the equalized eye.
Precursor taps(FFE only): The training algorithm create a filter that includes the specified number of pre-
cursor taps.
Max UIs for train (DFE only): Maximum number of unit intervals to use for training. We suggest leaving
this value at 1000. Higher values make the training slower, and much lower values may reduce the train-
ing accuracy.
922961 Rev A 41