X-Stream Operator’s Manual
WM-OM-E Rev I 431
keyword:
PLL_INTEG_GAIN
arguments:
<value 0 to 1e-10>
required?:
no
Notes:
The value is multiplied by a code difference from the expected level times slope. The result is to
produce a period correction as a fraction of bit cell time. The DDA uses 4e-11 for PR4, EPR4 and
E2PR4. If this keyword is not seen its value remains unchanged. If it has never been set since power
on the default is 4e-11. See notes for PLL_PROP_GAIN above.
possible error
messages:
"PLL_INTEG_GAIN must be >= 0 and <= 1e-10"
keyword:
PLL_PROP_GAIN
arguments:
<value 0 to 1e-8>
required?:
no
Notes:
The value is multiplied by a code difference from the expected level times slope normalized to
codes/bit cell near the sample. The result is to produce a phase correction as a fraction of a bit cell
time. The phase error tends to have a significant random component so this value should be kept low.
The DDA uses 5e-9 for PR4, 2e-9 for EPR4 and 1e-9 for E2PR4. If this keyword is not seen, its value
remains unchanged. If it has never been set since power on, the default is 1e-9.
Notes about the PLL loop:
We steer our software PLLs phase (proportional) and period (integral) with coefficients
phase_steering_gain and spacing_steering_gain. They are used like this:
spacing correction = phase error * spacing_steering_gain
phase correction = phase error * phase_steering_gain
where:
phase error = distance from sample to level * slope * scope samples per bit cell
Notes:
1. Distance is normalized (-7200 = -1 level, +7200 = +1 level)
2. Slope is (normalized codes/scope sample) around the channel sample position
3. Multiplying by samples per bit cell normalizes for scope sample rate: slope
(codes/scope sample) is cut in half when sample rate doubles
codes/scope sample * scope sample/bit cell = codes/bit cell
The gain terms must be small enough so that the correction is mostly uncompleted within the latency
(STEERING_LATENCY + FIR delay) of the control loop. Otherwise the loop will be unstable.
The spacing and phase corrections are in fractions of a bit cell.
possible error
messages:
"PLL_PROP_GAIN must be >= 0 and <= 1e-8"