I was curious about the analog to digital converter stages, and how it would fare in the radiation.
Turns out they're a bit lucky I guess. As noted in the paper[1]:
Transient errors affecting one or a few data samples due to SEUs can be tolerated and corrected by offline data analysis since the LAr pulse shape is analytically known.
This was used in their design:
The architecture incorporates a 9-subrange (3.2 bits) MDAC before a 12-bit, two-stage SAR, with the MDAC used to relax the dynamic range requirement for the SAR.
The MDAC and SAR are memory-less and their states are determined by the 40-MHz precision sampling clock and an intermediate 80-MHz internal clock signal. Therefore, any radiation-induced upset will only affect an individual sample.
The other responder suggested that the coordination is handled differently.
If you know the periodicity of a signal you can violate the Shannon Coding Limit. It's just that most signals you have to have a high enough sampling rate to detect the period, which gets you back to Shannon.
The LHC signal has out of band data that you can use to establish exactly when to take a sample, and the samples happen exactly 40 million times per second. So once you tune the time offset, the period is precise.
tux3•2d ago
That's actually a design parameter! The particles circulate in the ring in a series of bunches, and they're spaced apart so that bunches collide at 40MHz
So the particles aren't colliding continously, they're injected and collided bunch by bunch
hinkley•2d ago
So the clock skew is high resolution but the clock rate is low. Makes sense.
SiempreViernes•2d ago
Pretty sure they measured the pulse and fitted a function to describe it themselves, its not like the bought the scintillator on Alibaba and it happened to come with a spec sheet listing the analytical form.
magicalhippo•2d ago
Seems you're right[1]:
LAr cells pulse shapes can be predicted from calibration
pulses by several models such as Response Transformation Method (RTM) and First Principles Method (FPM). Both model requires that the drift time be measured.
magicalhippo•2d ago
Turns out they're a bit lucky I guess. As noted in the paper[1]:
Transient errors affecting one or a few data samples due to SEUs can be tolerated and corrected by offline data analysis since the LAr pulse shape is analytically known.
This was used in their design:
The architecture incorporates a 9-subrange (3.2 bits) MDAC before a 12-bit, two-stage SAR, with the MDAC used to relax the dynamic range requirement for the SAR.
The MDAC and SAR are memory-less and their states are determined by the 40-MHz precision sampling clock and an intermediate 80-MHz internal clock signal. Therefore, any radiation-induced upset will only affect an individual sample.
[1]: https://ieeexplore.ieee.org/document/11017335
hinkley•2d ago
brcmthrowaway•2d ago
hinkley•2d ago
If you know the periodicity of a signal you can violate the Shannon Coding Limit. It's just that most signals you have to have a high enough sampling rate to detect the period, which gets you back to Shannon.
The LHC signal has out of band data that you can use to establish exactly when to take a sample, and the samples happen exactly 40 million times per second. So once you tune the time offset, the period is precise.
tux3•2d ago
So the particles aren't colliding continously, they're injected and collided bunch by bunch
hinkley•2d ago
SiempreViernes•2d ago
magicalhippo•2d ago
LAr cells pulse shapes can be predicted from calibration pulses by several models such as Response Transformation Method (RTM) and First Principles Method (FPM). Both model requires that the drift time be measured.
[1]: https://inspirehep.net/files/cb39822f5ead85e84d7a1221bcd68fa...