GCT meeting – by phone - Friday 16 December
Present:Bob, Geoff, David C, Jim, Dave N
Apologies:Greg H, Steve Q, Rob, Magnus, Greg I, Ben, Azmat
1.Minutes
2Actions from previous meetings
Actions from the last minutes were reviewed and those completed removed from list.
Location of CMs. There are two in CERN, two at RAL. The “missing” fourth one is stated to be boxed up in the lab at RAL.
3Integration progress, and plans for the coming week.
3.1 CERN status (Jim)
Checked TPM output data by looping back signals to TPM input. The TPM is not configured to do this without dedicated firmware for correct timing so it is not possible to fully verify that data are error free. There appears to be an error on the parallel side (single-bit) which will be investigated.
Some work has been done on DTPM integration, providing synch signals. This still needs more work.
Backplane cables have arrived from RAL. There should be 31 in total in CERN which, with others to be sent from CERN, should be enough to configure 3 TPMs, populated as in jet-type TPMs.
Plans for next week
Investigate TPM operation with Dave N. DTPM integration.
3.2 TPM and firmware status (Dave N)
Completed RCT firmware, and handed on for implementation. Following visit to CERN next week will release firmware for January test.
Plans for the coming week:
System-ACE debugging. In CERN to work with Jim on TPM investigations.
3.3 IM status (David C)
RCT emulator:
Firmware to drive clock and control signals to slave boards has proved trickier than anticipated. Expect to get help from Dave N today, or David C will complete. This is believed to be the only remaining obstacle to comple of emulation of an RCT crate, now needed for IM and IM-TPM tests.
It was established this week, by comparison with the IM module in CERN, that the second IMv3 does not behave identically to the first one, i.e. there is something else wrong with it which needs investigation. If the second IMv3 is not available, delivery of the IMv4’s is essential to complete the hardware for the January test.
The daughterboard can be further tested once the RCT emulator is working. It is expected to be compatible with IMv4.
IMv4 production:
The job did not start in Cemgraft as hoped on 14 December. Kit checking has begun and assembly start is now scheduled for 21 December, with delivery on 4 January. Cemgraft will not shut for as long as usual over Christmas.
Plans for next week
Still aim for RCT emulators to be fully commissioned. Daughterboard should be fully tested in Bristol. Aim for a completely working test stand as soon as possible.
3.4 TPMv3 design (Greg)
Discussions held with Azmat and also with Chris Day. The main topic was the set of comments from Christoph Schwick on the layout of the fast LVDS serial lines at the front of the board. These were circulated by e-mail and incorporated into the latest version of Azmat’s master list of TPM changes. We also went over some other points on the TPM re-design.
LVDS serial lines:
Looking at the design, we realised that the 4X connector is “upside down”. The schematic assumes that LVDS signal pair number 1 is at the top of the connector, but it’s actually at the bottom. The same mistake appears to have been made with the 1X connectors.
The signal connections need to be reversed on the schematic, for all seven connectors (four 4X and three 1X). Once this has been done, it will be possible to route all the serial lines either on the top layer or the first inner layer, as suggested by Christoph.
We discussed with Chris Day some of the other suggestions from Christoph. He agreed that the following were possible: changing the termination to improve common mode rejection; moving some decoupling caps to the component side of the board to improve high frequency performance; going to wider track widths and spacing to improve attenuation. The discrete components on the top of the board were likely to get rather crowded. We agreed that the priority was to move the caps on the analogue supply pins AVDD/AGND to the top, and the decoupling for the digital circuitry could remain on the bottom.
We do not feel there is much to be gained from the suggestion of moving some serdes chips to the back of the board. This was supposed to allow the use of shorter lines, but in fact the distances to some of the devices in the first “row” are as long as those to the second.
Other points:
On the schedule, Chris said that the earliest available time in the Drawing Office was late Feb/early March. Their clear preference is to have a complete updated schematic, and list of required layout changes, before they start work.
Azmat has started to look at alternative suppliers of quarter-brick DC-DC converters. One company, Galaxy Power, states that its supplies have “very low common-mode noise”. Azmat has contacted them to try to get more details of this claim and some sample devices.
We have talked through some ideas on the design of the daughtercard for the power supply, and its connection to the motherboard. More design work is needed here.
The list of required changes is being re-structured to separate schematic from layout changes.
Comment from Dave N: consider mounting serdes chips onto daughterboards.
4 Integration and expected progress
4.1Definition of January test (Greg)
A revised definition of the January test milestone has been agreed with Wesley.
“Action: Send data from two RCT crates to two complete IMs to two TPMs to DTPM to GT.
Required Hardware: Two RCT Crates, 2 IMs with daughterboards, 3 TPMs: 2 TPM & DTPM, cable backplane, GCT, GT infrastructure.
Test: Load up 256 crossings of data in RCT, transmit over 12 RCT-GCT cables to 2 IM, transmit over 1.44G LVDS links to 2 TPMs, transmit between FPGAs on TPM, process data on TPM, transmit from 2 TPMs over cable backplane to TPM to GT and to DTPM to readout
Data:Long series of 0 -> one bit 1
Long series of 1 -> one bit 0
Alternating patterns of ffffffff-00000000 (or n times ffffffff – n times 00000000)
Random Data
Other Conditions:
All links should operate at the same time: This means that all IM links are active and all receivers on the TPM are active.
The internal data busses of the TPM should be all active (could be fed with random data or data close to the expected data pattern when running.)
The backplane links should be all active.
Dataflow:
RCTIM1.44 LVDSTPM*Vitesse Cable BPTPM*GT readout
DTPMCMVME or SLINK readout
*transfer data between FPGAs on TPM and process data on TPM”
Comments:
A new feature, with respect to my proposed test definition of the past week or so, is that we are required to implement some sort of algorithm processing and checking. This may be as trivial as adding two inputs together. Work on this is foreseen to start shortly after Christmas.
The transfer of data to the DTPM is a required part of the test. Wesley states that it is his understanding Greg Iles should be available to us to complete the integration of his firmware.
We have concerns about whether the required facilities for checking captured data against expectations can be made to work in time for the end of January, both in the GCT and GT systems. I have made a verbal request for the precise definitions of the input data from the RCT to be finalised by 13th Jan. Discussions on the synchronisation are ongoing, to understand the timing relationship between the BC0 signal and the data patterns in different RCT Receiver cards.
4.2Procurement of cables (from Laurie with comments from Greg)
Backplane cables: I note you have 50, I have asked Molex for progress following my previous emails.
Infiniband cables: No reply from Volex, have asked for progress on my request for quote. Note your email from Leoni.
TCC order: Should be despatched to CERN this week via UPS.
Heat sinks: One further glitch, I had to arrange to put European Thermodynamics Ltd on CCLRC approved suppliers list. This is now done so order should proceed.
Comments
50 backplane cables: Jim now has 31 in CERN. There are 18 in the lab at RAL (now sent to CERN) and one in Bristol.
Leoni, the original supplier of halogen-free Infiniband assemblies, were asked for update on latest request for quote. Promised immediately but no further communication yet. Asked to copy the quote to Laurie. Particular concern about the 1X-1X cables, where we currently have only one with acceptable performance for the GT link.
The TCC order refers to the 15 metre cables for the RCT signals, so there should be no problem having the required twelve of these in time for the test.
4.3 Progress towards January test (Greg)
Targets December 14
Test/verify GT link (Jim+GT)
Under investigation by Jim.
Define strategy for checking data within GCT; define software/firmware components needed for this and assign responsibilities
Done, software document was circulated for Monday's meeting
Define TTC arrangement for tests in discussion with TTC, RCT, GT groups.
In hand: Jim reported on discussions in Trigger integration meeting.
Four TPMs in CERN, check whether any require repair (Dave, Jim)
Done
Targets December 21
BER tests of RCT-IM link with six cables (Jim)
This will not been done before CERN shutdown.
DTPM available and tested (Dave, Jim+GI)
Under way, but stated to be "unlikely, but possible" a couple of days ago.
TPM firmware configuration with full data routing for tests (Greg, Dave)
Routing has been defined and implemented but not yet released.
5Summary and Actions
Carried forward/ongoing:
David C: complete IM test-system and define tests for IM acceptance.
New
All: study the version of TPMv3 changes circulated on 8th December, and comment
Azmat: implement the required changes to the TPMv3 schematic.
Azmat: consider possibility to mount serdes chips on daughterboards in TPMv3
Laurie: try to progress the Infiniband cables, with Volex and Leoni
David C; send plastic clips to configure backplane cable loom
Jim: procure more halogenated hydra cables (for IM-TPM link) in case insufficient supplies of halogen-free cable in time.
Laurie/Greg (?): send additional 18 backplane cables from RAL to CERN
6Next meetings
6 January – by phone
then weekly on Fridays (to be confirmed).
GH: 16 December 2005
1