Speaker
Description
Summary
I. POWER DISTRIBUTION SCHEME FOR THE TILE CALORIMETER
The amount of power required by the front-end electronics of the Tile Calorimeter
of the ATLAS experiment imposes the presence of switched power supplies made of
custom radiation tolerant DC/DC converters. The power supplies are located inside
the detector near the front-end electronics.
The noise of the DC/DC converters deteriorates the performance of the detector: it
must be filtered out.
II. EVALUATION OF NOISE PERFORMANCE
The high frequency noise seen by the fast readout electronics is estimated from the
pedestals data acquired with the complete data acquisition chain. The noise of the
system is estimated by a selection of parameters computed over several runs and on
all the channels:
• RMS of the pedestals distribution.
• Gaussian property of the pedestals distribution.
The low frequency noise is evaluated from data sampled by the current integrators
ADC.
On the other hand, the noise performance of the power supply is measured on each
low voltage output as common mode current amplitude and as voltage ripple for the
differential mode component.
III. NOISE PROPERTIES OF THE CONVERTERS
The DC/DC converters emit large common mode currents at the switching frequency and
its harmonics, exceeding the limits in use. Beyond the switching frequency, the CM
current amplitude is below the usual limits.
The common mode current path is modelled and the amplitude is measured. From this,
appropriate filters can be applied.
IV. FRONT-END SUSCEPTIBILITY
The measurements made show that the fast readout electronics is sensitive to high
frequency common mode currents (5MHz to 100 MHz). The DC/DC converters noise was
already below the usual limits, but further filtering is required due to the high
susceptibility of this part of the electronics.
Similar measurements on the slow readout electronics show that this part of the
electronics is very sensitive to low frequency ripple (below few kHz). The
operating point of the converters must be tuned to minimize the ripple.
V. FILTERING METHODS
Several methods allow reducing and filtering the common mode currents. Among those,
the following devices were tested: common mode chokes on the primary and secondary
side of the converters, ferrites on the low voltage outputs, decoupling capacitors
between each input and output pins to the case. The chokes show the best
performance, especially when placed on each output; however, they are too bulky to
fit in the tight space of the low voltage box. As an alternative, ferrites and
decoupling capacitors were successfully tested.
The low frequency ripple is minimized by setting the DC/DC converter operating
point around mid-capacity. This adjusted at the feedback components of the
converters.
VI. CONCLUSION
The noise properties of the DC/DC converters were compared with the susceptibility
of the fast and slow readout electronics. From this a noise coupling model was
established, and different filtering solutions were exercised. Among those, the use
of ferrites and common mode decoupling capacitors appeared to be most suitable
method that allows operating the front end electronics within the detector noise
specifications.
