Ivan PogrebnyakSupervisors: Paramonov, Drake, ProudfootATLAS, MSU, ArgonneTile week upgrade sessionJune 10, 2015

TileCal requirements for ELMB upgrade

TileCal low voltage distribution system

2

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015

200V DC from USA15

LVPS

+10V +10V +10V +10V

+10V +10V +10V +10V

ELMB MotherboardELMB

Mini-drawermain board

POL Regs

Daughter Board

Mini-drawermain board

POL Regs

Daughter Board

Mini-drawermain board

POL Regs

Daughter Board

Mini-drawermain board

POL Regs

Daughter Board

https://indico.cern.ch/event/394814/

ELMB monitored LVPS valuesin current implementation

3

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015

Currents Voltages Temperatures

Brick input ×8 Brick input ×8 Water ×2

Brick output ×8 Brick output ×8 Brick ×2 ×8

Brick sense line ×5

Total: 55 values per LVPS

https://indico.cern.ch/event/394814/

The main issue: Identifying point of failure

• We'd like to know what type of failure caused a trip− 200 VDC from USA15 to LVPS− LVPS internal brick failure− Main board, i.e. brick load, e.g. short circuit

• Not having to do repairs by trial and error will save a lot of money, time, and effort

• LVPS is much easier to access then the rest of the drawer− If only the power supply failed, it can be replaced without

opening the detector

4

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015 https://indico.cern.ch/event/394814/

LVPS trip monitoring concept

• Failure of 200 VDC input from USA15 can be easily identified by simply measuring LVPS input voltage

• No discriminating information about failure type can be obtained from measuring LVPS output voltage

• Distinguishing between brick internal failure and a trip caused by the load requires analysis of transient behavior of output current− A short circuit in the load will result in a sharp increase of

output current− Other trip types will exhibit decreasing current transient without

the initial rise• The monitoring device (ELMB) must take multiple

measurements during the current transient

5

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015 https://indico.cern.ch/event/394814/

LVPS trip due to load short circuit

• A test of transients was done with a +5V MB brick at Argonne• Current transients exhibit a characteristic rise to overcurrent value• Voltage simply decays with the time constant determined by brick

output capacitance

6

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015

Current during load short circuit Voltage during a trip

https://indico.cern.ch/event/394814/

Current ELMB version limitations

• At Argonne, we considered many options for ELMB upgrade• The current ELMB cannot be feasibly improved

− No pin compatible faster ADC− Any solution with replacing ADC on the current board will be

costly and will reduce reliability− Several ELMB component are no longer produced

• A new ELMB version must be designed

7

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015

Component Limitation

CS5523 ADC Nominal maximum digitization rate of 600 Hz

32S12C 32.768 kHz ADC crystal Limits the ADC maximum digitization rate to 101 Hz

HCPL0731 optocoupler betweenthe analog and digital stages

Limits the ADC maximum digitization rate to 30 Hz

https://indico.cern.ch/event/394814/

New ELMB requirements

• Fast sampling rate: 10 kSPS to 1 MSPS• In situ failure identification (will require on-board memory)• Push notifications to DCS, alarms on failure (trip)• Retain independent power• 12 bit ADC will be sufficient

• Radiation tolerance and resistance to SEU should not be a problem with modern small feature size components

• If the new ELMB will be possible to produce before Phase-II upgrade, CANbus and Mother Board compatibility will be required

8

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015 https://indico.cern.ch/event/394814/

New ELMB design

• ELMB++ group already exists, but no development started yet, need sub-detector input

• References: Stefan Schlenker (https://indico.cern.ch/event/388112/)

• TileCal is currently the only sub-detector actively involved in defining requirements for ELMB++

• A lot of room for improvement using modern technology• FPGA (instead of microprocessor) based solution is promising• g-2 experiment is designing a monitoring system based on

ADS8363, a 16-bit 1MSPS 8 parallel channel ADC− Approximately the same size as ELMB− Radiation tolerant

9

Ivan Pogrebnyak. Tile week upgrade session. June 10, 2015 https://indico.cern.ch/event/394814/