wiki:CommonATCA

Version 21 (modified by dgastler, 13 months ago) (diff)

--

SVN: http://gauss.bu.edu/svn/common-atca-blade

Reference Material

Consider making mezzanine board 7U high (277.8 mm, see Eurocard info). Depth probably non-standard ~ 200mm as standard depths 160mm and 220mm are not convenient.

P23 Connections (CMS only)

P23 Row Port Pair CMS Function
4 0 Tx LDAQ out 10Gb/s
4 0 Rx LDAQ in 10Gb/s
4 1 Tx Reserved STDIO
4 1 Rx Reserved STDIO
3 2 Tx TCDS2 Out (throttling, TTS+)
3 2 Rx TCDS2 In (timing, TTC+)
3 3 Tx Precision Clock
3 3 Rx LHC Clock

P23 Connections (Base ethernet)

P P23 Row Port Pair Base Channel Function
6 0 tx 2 BI_DA (10/100 TX )
6 0 rx 2 BI_DB (10/100 RX )
6 1 tx 2 BI_DC
6 1 rx 2 BI_DD
5 0 tx 1 BI_DA (10/100 TX )
5 0 rx 1 BI_DB (10/100 RX )
5 1 tx 1 BI_DC
5 1 rx 1 BI_DD

NOTE: "Precision Clock" is reverse direction from standard!

This is from this draft of the document "CMS ATCA Crate Specification and Hub- and Node Board Requirements"

Meetings:

2018-05-31

Thinking about a common blade for ATLAS L0MDT and CMS IT-DTC.
Issues with the current ATLAS design:

  • FMC stacking height at 8.5mm is too high

Issues with the current CMS design (Cornell one):

  • Really don't like the interposer concept with metal plates, springs etc
  • Prefer stand-alone mezzanine card operation

Plus, there are two different blade designs for no obvious reason.

First, looking at low stack height connectors, but not Z-ray. (See Samtec guide) Samtec families with low stack heights:

Family Series Stack Height Notes
Sea-Ray LPAM/LPAF 4mm
Edge Rate EDM6/EDF6 5mm
Q Series QTH/QSH 5mm? 28 Gbps+, pairs, power
QTE/QSE 5mm 25A GND plane
Ultra-Micro EDM6/EDF6 5mm
LSS/LSEM/LSHM 5mm
ST4/SSR 4mm

Below is a figure with dimensions excerpted from the ATCA standard.

http://gauss.bu.edu/svn/atlas-phase-2-muon-upgrade/Meetings/2018-05-31_EDF/figs/stack_heights.png Seems as if we can find a 4mm or 5mm solution.
This gives us about 15.7mm above a mezzanine board for components.

Next, thinking about the logic. Is it really that hard to make a common blade? Here are some thoughts:

  • Build a blade to accommodate 3 mezzanine cards
  • Plan for only "low speed" mezz-to-mezz communications
    (SelectIO or < 10Gbps SERDES)
  • Mezzanine cards would have the optics on board along with DC/DC converters and regulators
  • Split the CMS IT-DTC into 3 parts
  • Eliminate the "blade FPGA" from the ATLAS design and designate one mezz card as "master"
    (this could be a different design).

ATLAS requirements

  • Sector logic