Maurice Garcia-Sciveres Lawrence Berkeley National Lab SiD tracking meeting
Power lines Problem of “long” distance electrical power distribution is not new The novelty is that “long” is getting shorter Wrong units? Voltage is driven by oxide thickness inside the IC: units of length. => Miniaturization inside IC also affects distances outside IC! Eventually all power distribution distances become “long” => can only be solved at IC level
ATLAS pixel detector cable plant Built in the traditional way
Scaling from present pixel detector with same cable plant down to PP2
Options well known, Once accepted that we are dealing with long distance power transmission - Most ideas discussed today can be found in patents filed by Tesla well before 1900.
Existing efforts There is a combined ATLAS R&D proposal: Serial power: - SP demonstrator pixel modules have been produced at Bonn some time ago
- Plan to make realistic serial power stave prototype in the near term
- SCT implementing serial power with external regulators
- More recent but also more prototyping work in SCT at present (RAL, LBNL)
DC-DC - Switched capacitor development at LBL
- 2 version of ASIC submitted 1 month ago
- This should produce a real regulator ready to power SCT stave prototypes and pixel modules but late summer.
- Magnetic “buck” converter
- Work started by Satish Dawn of Yale >1 year ago. Evaluate commercial parts and options for industry partnership
- Parallel effort started at CERN in 2006. Significant manpower. Plan is to design a magnetic regulator controller and build a regulator from the ground up
- Need to evaluate AC magnetic field issues.
Why switched capacitors? Commercial DC-DC down-converters for power applications are all inductive. Why then study switched capacitors for power? - Cannot use ferrites in magnetic field => performance penalty for magnetic converters
- Fringe AC magnetic fields may produce pickup in detectors (must study case-by-case)
- Ceramic capacitor miniaturization makes great advances year after year (air-core inductors cannot be improved).
- Over-voltage safety considerations
Switched capacitor credits IC design: Peter Denes Simulation: Bob Ely, Peter Denes Testing (so far only first prototype): Bob Ely, Seung Ji, Sami Hynynen, M. Garcia-Sciveres
Test configuration used: divide-by-4 stack
Other configurations Many capacitor arrangements are possible with different advantages - Minimum number of capacitors for a given ratio (less than for stack)
- Minimum voltage drop across switches (more than for stack), etc.
Problem has been solved in general: Makowski, D. Maksimovic, "Performance limits of switched capacitor DC-DC converters," IEEE PESC, 1995 Record, pp. 12151221)
First prototype test chip 50V (s-d) 0.35m HV CMOS process Minimum size (adequate for ~100mA) Switch transistors only- no auxiliary circuitry Learned about process simulation, radiation hardness, and bulk isolation Did not work as a useful converter due to bulk isolation problems Results presented at 12th Workshop on Electronics for LHC and Future Experiments, http://ific.uv.es/lecc06/
HV Transistor characteristics after irradiation The most important result is that the drain source resistance has increased by about 10% Measured Rds also exceeded model predictions even before irradiation. => Increased switch size.
Second prototype Same 50V 0.35m HV CMOS process Submitted February 2007 (expected back in May) Sized for 1A output. 4.3 x 4.9 mm Contains auxiliary circuits. All capacitors external
Top level schematic
Simulation results 1
Simulation results 2
Simulation results 3
Serial power vs. DC-DC trade-offs Power: Both increase power at/near the module by a similar amount Mass: Serial power is in principle less massive because regulators can be built into chips. Radiation: SP naturally rad-hard if regulators built into chips. DC-DC not yet rad-hard enough Generality: DC-DC can be an off-the-shelf solution. A given converter can be used in several applications Complexity: With SP the detector basic unit is a super-module. With DC-DC the basic unit is still a module. Control: individual module control is simple with DC-DC, not with SP. With DC-DC module voltages are adjustable- not easily with SP. Cable reduction: Both reduce copper mass by same amount. SP also reduces the cable count bu default. DC-DC allows trade-off between cable count and control granularity.
Regulators can be external or inside readout chips Production pixel readout chips already have internal regulators built-in (but not used). - This shows that the serial power pixel R&D has a long history by now
Why is there no conductive interference (noise) between serial powered modules?
Serial-powered strip sensors
Serial power references
Conclusion Powering has become a hot topic with lots of work going on Expect usable prototype DC-DC converters by year’s end Focus tends to be on powering existing chips, because that’s what people can make tests on Reduction of current at the source (chip) has NOT YET received nearly as much attention within HEP. It should How do we reduce the analog current keeping good performance? How do we reduce the digital current? - An extreme is stacked logic domains- serial power inside the chip. Reference: http://www.bioee.ee.columbia.edu/
- Less aggressive approaches are possible. I/O protocol. Clock distribution. Architecture.
Dostları ilə paylaş: |