AMANDA-DAQ:    TDC  Upgrade for Summer 2001/2002

 

 

We suggest to replace all current Camac_TDCs by VME_TDCs of the same type as currently in use:
CAEN-TDC V693 (128/board, price ~13kDM/board ~ 100DM/Channel)   (see here for the board description).

We need to decide on at least the VME_TDC part of the DAQ hardware-upgrade for next season very soon, forced by delivery times.
 
 
 

Status:

Currently, the DAQ reads out all 302 OMs of B10 by Camac_TDCs  LeCroy3377 (plus a few service Channels like triggers).
While being reliable, these boards are  the  slowest part in the AMANDA-Readout (see below and Bartol DAQ-talk).
No readout-performance improvement for the Camac_TCDs can be achieved, apart from branching the Camac-Crates into different DAQ-subsystems A branching of the DAQ (either in AMANDA or GSI version) is however quite unlikely to happen in the upcoming summer 2001/02.
Also,  the TDCs are only accessable in single-word Camac-Readout.
VME_TDCs CAEN_V693 are usable in blocktransfer mode (an operation mode not used in the current DAQ yet).
 

Advantage of a full  VME_TDC setup:

1. Overflow handling:

In case of  a large number of edges arriving  for an event, VME _TDCs keep the first 16 edges in their memory,
rather than the last (like the LECroy CAMAC_TDCs do).
This is , for shower reconstruction, an advantge of the VME_TDCs over the Camac_TDCs.

Note, that for a high noise rate environment, i.e. when the probability for an overflow to occur between two neighbouring triggers is high, VME_TDCs are not  suitable (they memorize the fact that an overflow occured after the last event-readout. But this situation is not desired in AMANDA anyway, i.e. OMs with high noise need to be set to normal mode or switched off.)
 

2. Speed:

In any DAQ-upgrade scenario for this summer, a fast TDC readout will play an essential role in speed improvement.  Recall  the numbers given in  the Bartol DAQ-talk:

- Full_Readout = 1.9 ms
- Only ADC = 0.54 ms
- All_TDC = 1.3 ms, Only CAM_TDC = 0.92 ms, Only VME_TDC = 0.43 ms.
 

The table below summarizes the gain in readout time, that is expected by replacing Camac_TDCs by VME (when blocktransfer is considered, we assume  that VME-blocktransfer speeds up by 5 times - very conservative).
The values reflect the gain in absolute hardware readout time (which is currently ~1.9ms for Majority=24 triggered events) .
Note, that the full deadtime of the system, including additional  delays set up in the NIM-trigger electronics, is larger than the 1.9ms by about  15-25%.

We give numbers for 3 DAQ-scenarios:
(1) no change in DAQ-software, but  replacement  of all Camac_TDC by VME_TDC hardware,
(2) introduce TDC-blocktransfer for all OMs, and
(3) introduce blocktransfer only for those OMs that are now already read out by VME_TDCs.
Of course, there's many more scenarios for the DAQ, but we cover some of the more realistic ones.
 
 
Expected reduction of readout time , in ms  Expected reduction of  readout  time, in %
No-Software-change scenario
(no blocktransfer)		 0.5ms 26%
Only-BlockTransfer scenario
 (all Channels =TDCs - VME)		 1.2ms 63%
Only-BlockTransfer scenario
(only old VME_TDCs)		 0.34ms 18%

- Additonal speedup comes if GSI is used, even without branching (seems up to 50% faster than Amanda-DAQ).
- Additional speedup comes from branching (AMANDA-DAQ or GSI-DAQ).
 
 
 

Actions/Problems:

1. Recabeling.

Needs new recabeling of all the 302 B10  connections to DMADD--VME_TDC.
Suggestion: make extension cables to extend from the CAM_TDCs (Rack3) over to Rack6 (VME).
This avoids 'retouching the major ECL-output section of the DMADD.

Note: Since the Camac_TDC ECL-input cables are very (!) densely arranged now,
some major activities would be needed there anyway.
Estimated time: 4-5 days for 1-2 persons/1 shift.
 

2. T0-calibration.

Parts of the B10 OMs would have to be t0-calibrated after the change. Estimated:
4 OMs per each old TDC, i.e. ~10*4 channels. Surface-T0 are needed for all channels.
 

3. Performance.

Analysis 2000 has so far shown that the TDCs work reliably.
"Higher level problems" (i.e. occuring at low rate) are not excluded, but of low probability.
 

4. Reliability.

This is a real problem: We had of 3 boards operated in 2000 and 2001 each year failing 1 board,
which gives an estimation of ~1/3+-1/3 failure/(board*winter).
4 spares are planned (we have other VME_TDCs as backup).
 
 

Conclusion:

From this estimations it is clear that the effort to change to VME_TDCs (money & work) is reasonable,
if  AMANDA  aims at a readout-improvement of >20% in summer 2001/02.
 
 

Compiled: R.Wischnewski, 01.08.2001.

 Return to previous page