CAEN power supplies The neverending story
description
Transcript of CAEN power supplies The neverending story
CAEN power suppliesThe neverending story
Jennifer PursleyJohns Hopkins University
Silicon Workshop II, May 10-12, 2006University of California, Santa Barbara
J. Pursley - Silicon Workshop II 2
System overview:
Diagram courtesyof J. R. Mumford
J. Pursley - Silicon Workshop II 3
Infrastructure: the SY527Universal Multichannel Power Supply system
CAEN mainframe, has 10 board slots NOT custom-made, also used for plug power supplies Communicates via serial connection (RS232 port or
CAENET coaxial cable) Control power supplies
(settings and on/off) from front panel
NOT radiation hard, but located in the collision hall!
These are what you hockerize (reboot the crate CPU)
J. Pursley - Silicon Workshop II 4
Collision Hall Map: 16 crates Mounted on CH
walls, with fib racks
4 crates in each corner, numbered clockwise (sort of…)
Roughly, 2 PS crates = 1 fib
Even number fib is SVX, odd is ISL/L00Diagram courtesy
of M. Stanitzki
J. Pursley - Silicon Workshop II 5
Crate Naming Conventions
Crate Reset Panel PS Interlocks
Crate 1 SVX NW Top 1 NW Top Top
Crate 2 SVX NW Top 2 NW Top Bot
Crate 3 SVX SW Top 1 SW Top Top
Crate 4 SVX SW Top 2 SW Top Bot
Crate 5 SVX SW Bot 3 SW Bot Top
Crate 6 SVX SW Bot 4 SW Bot Bot
Crate 7 SVX NW Bot 3 NW Bot Top
Crate 8 SVX NW Bot 4 NW Bot Bot
West Side
J. Pursley - Silicon Workshop II 6
Crate Naming Conventions
Crate Reset Panel PS Interlocks
Crate 9 SVX NE Top 1 NE Top Top
Crate 10 SVX NE Top 2 NE Top Bot
Crate 11 SVX SE Top 1 SE Top Top
Crate 12 SVX SE Top 2 SE Top Bot
Crate 13 SVX SE Bot 3 SE Bot Top
Crate 14 SVX SE Bot 4 SE Bot Bot
Crate 15 SVX NE Bot 3 NE Bot Top
Crate 16 SVX NE Bot 4 NE Bot Bot
East Side
J. Pursley - Silicon Workshop II 7
The Workhorse: SVX Modules (A509) Occupies 1 slot in SY527 1 board powers 1 wedge
5 Bias channels 5 Low voltage (AVDD & DVDD) 2 portcard (2V & 5V DOIMs) Total: 18 channels
73 supplies in CH (72 for SVX wedges, 1 for a L00 wedge)
VMax = 250V, IMax = 5 mA Cable pinouts designed for SVX
Layer 0 1 2 3 4
VMax (V)
170 170 60 140 60
J. Pursley - Silicon Workshop II 8
Next up: ISL Modules (A510) Occupies 2 slots in SY527 1 board powers 1 wedge
10 Bias channels 5 Low voltage (AVDD & DVDD) 2 portcard (2V & 5V DOIMs) Total: 23 channels
30 supplies in CH VMax = 250V, IMax = 5 mA Bias Adapter, double LV cablesLayer 00,01 10,11 20,21 30,31 40,41
VMax FWD
60 140 60 140 60
VMax CNTL
140 140 140 140 140
J. Pursley - Silicon Workshop II 9
Finally: L00 Modules (A509H) Occupies 2 slots in SY527 1 board powers 1 wedge
4 + 1 Bias channels 4 Low voltage (AVDD & DVDD) 2 portcard (2V & 5V DOIMs) Total: 15 channels
11 supplies in CH VMax = 500V, IMax = 36/23 mA Sense, LV, and Bias adapers! Extra feature: crowbar on bias line
Layer 0 1 2 3a,3b
VMax (V) 90 90 90 160
J. Pursley - Silicon Workshop II 10
L00 Crowbars
Protect Si from PS failure 2 mA fuse on each bias line Now frequently blown in beam incidents
(eg kicker prefires or nasty quenches) Blown fuse = no bias on sensor Most L00 sensors draw measurable current now; check
for blown crowbars by biasing L00, look for 0 current
Automatic crowbar tester!
No light = blown fuse
J. Pursley - Silicon Workshop II 11
Junction Cards Same junction card used for SVX, ISL, and L00
Extra LV and Bias connectors to accommodate ISL Cables in CH run from PS racks down into the bore
And across the COT face… This is the closest we can get to the silicon!
Diagram courtesyof J. R. Mumford
J. Pursley - Silicon Workshop II 12
Common Failure Modes – SY527 Crate CPU gets in a funny state
Some symptoms: “Crate xx lost communication” (or you see anything turn
blue in IMON) All supplies in one crate spontaneously turn off Garbled readback of voltage/current of a ladder (could
lead to trigger inhibit) Solution: hockerize!
Fan failures: frequent in plug crates, but none of ours (yet…)
1 damaged backplane (not fun!) Remember: must turn crate off for ~10 mins
before removing power supplies, or could blow a PS fuse
J. Pursley - Silicon Workshop II 13
Common Failure Modes – PS Overcurrent trips
Maybe the current is going high – try raising limit, watch plots Also a common PS failure, usually fixable at FCC
Overvoltage/Undercurrent trips Approx. the same thing – if the voltage is set above the limit,
the supply clamps it down before tripping Usually a PS failure
Undervoltage trips Supply can’t get to the set voltage Common PS failure (esp. of ISL supplies), NOT fixable at FCC
Software protection trips Voltage/current doesn’t trip the hardware Software limits stricter than hardware, but requires the value
stay above the limit for several mins before tripping Usually denotes a readback problem, check values in IMON
J. Pursley - Silicon Workshop II 14
Less Common Failure Modes Transistor regulating the Bias voltage blows
Symptoms: ladder voltage ramps up to VMax (250V or 500V) and doesn’t trip
Can’t be turned off, have to cut power to the whole crate This is the failure mode crowbars were introduced to prevent!
Large current offsets Erratic or oscillatory voltages/currents (seen on DVDD
and Bias) Crate doesn’t recognize power supply
Not fully connected to backplane, X28HC256 prom is garbled… None of the other supplies in the crate will work if there’s one
in there the crate can’t recognize! Supply won’t work in one particular slot/crate, but is
fine in another Usually require expert diagnosis!
J. Pursley - Silicon Workshop II 15
Summary Love ‘em or hate ‘em, CAENs are here to stay Increasing frequency of failures may be due
to radiation exposure or aging 9 PS swaps in 2005; 4 in first 3 months of 2006
Do our best to: Work w/ FCC to minimize downtime from common
failure modes (eg hockerization) Hassle CAEN to make more spare crates and PS
(and to fix the broken ones faster!) Get creative (such as, put a PS with failure on one
ladder in for a wedge where that ladder is out of the HWDB for other reasons!)
Backup Slides
J. Pursley - Silicon Workshop II 17
Loadbox testing Must test a new
PS before hooking it up to the detector
Do this by cabling it to a junction card with a loadbox attached
Loadbox uses constant and variable resistances to mimic a silicon wedge