Muon Tracker Data Stream

D. M. Lee

Los Alamos National Laboratory
(phenix-muon-96-7; submitted: June 28, 1996)

Introduction

The proposed data stream for the high resolution cathode strips includes 5 - 11bit samples for each readout channel. The consequences of this is that the DCM and subsequent Event Builder must be able to handle the data flow as well as uniquely identify each data word. This document is intended to summerize in one place what requirements this places on the FEM, the optical link, the DSP, the Buffer, and the Event Builder.

System Specifications

The following assumptions are made.

Bunch spacing day one:	213.2 ns
Bunch spacing day two:	106.6 ns
Clock ticks for Level 1:	40
Number of high resolution channels:	16000
Number of samples per channel:	5
Number of ADC bits per sample:	11
Optical fiber transmission rate:	1.0 Gbits/sec
DSP throughput:	40 Mhz (20Mhz)
Number of channels per FEM:	192 (144)
Number of channels per DCM:	192 (144)
Total ADC conversion time at 144Mhz:	35.5 sec (no overhead included)
Total max digitization/ transmission time:	40 sec
Total number of words FEM to DCM:	980 (740) ( see below)
Total transmission time:	24.5 (37.0) sec

The choice of DSP throughput rate has not been made so two alternatives are possible, 40 Mhz or 20 Mhz. The main effect is in the front end board size and the number of optical links/DCM. The 5 samples of each pulse will depend on the character of the noise in the system. In the best case the DC level prior to the pulse will be well known by beam off calibrations so it will be a simple pedestal subtraction in the FPGA. The 5 samples will then be at or near the peak of the slow cathode shaped signal. If the DC level prior to the signal suffers from AC fluctuations, i.e. 60 Hz, then one or more of the samples will sample prior to the CSC signal and the rest will be at or near the peak. In any event we expect the specific timing location of the samples will be programmable in the HEAP manager and need not be specified at this time.The format of the data stream to the DCM from the FEM is fixed format/fixed length.

Description of the data format

The data stream from the FEM to the DCM has the following format ( from PN191, On-Line Mtg., ORNL, Oct 1995). All are 20 bit words.

word	description
1	Boundary Word (all ones, either 16 or 18 together with CAV)
2	Detector ID
3	Event Number
4	Module address
5	.
...	.
	.
5n+19	user specified words
5n+20	X-OR of all data
5n+21	Boundary word (all zeros together with CAV)
where n= the number of channels.

The user specified data as well as the above tag words must fit into the available data stream specifications.

Desciption of the user data format

The data format is unique to each subsystem but the CSC data format is similar in concept to the data stream for the EMCAL. The following Data sequence is adequate for transmission to the DCM and includes the label ( 4 bits) and the data (16 bits).

Data Sequence Block from FEM to DCM

word Label (4bits) Data (16bits)

Data sequence for each FEM

1 1 FEM Address

2 2 AMU cell 1st sample

3 3 AMU cell 2nd sample

4 4 AMU cell 3rd sample

5 5 AMU cell 4th sample

6 6 AMU cell 5th sample

Data sequence repeated for each channel

channel 1

7 7 sample 1

8 8 sample 2

9 9 sample 3

10 10 sample 4

11 11 sample 5

channel 2

12 7 sample 1

13 8 sample 2

14 9 sample 3

15 10 sample 4

16 11 sample 5

channel n

7+5*(n-1) 7 sample 1

8+5*(n-1) 8 sample 2

9+5*(n-1) 9 sample 3

10+5*(n-1) 10 sample 4

11+5*(n-1) 11 sample 5

8 user defined Data words from FEM

12 +5*(n-1) 12 User word 1

. .

. .

20 +5*(n-1) 12 User word 8

Data Sequence Block from FEM to DCM
word	Label (4bits)	Data (16bits)
Data sequence for each FEM
1	1	FEM Address
2	2	AMU cell 1st sample
3	3	AMU cell 2nd sample
4	4	AMU cell 3rd sample
5	5	AMU cell 4th sample
6	6	AMU cell 5th sample
Data sequence repeated for each channel
channel 1
7	7	sample 1
8	8	sample 2
9	9	sample 3
10	10	sample 4
11	11	sample 5
channel 2
12	7	sample 1
13	8	sample 2
14	9	sample 3
15	10	sample 4
16	11	sample 5
channel n
7+5*(n-1)	7	sample 1
8+5*(n-1)	8	sample 2
9+5*(n-1)	9	sample 3
10+5*(n-1)	10	sample 4
11+5*(n-1)	11	sample 5
8 user defined Data words from FEM
12 +5*(n-1)	12	User word 1
.		.
.		.
20 +5*(n-1)	12	User word 8

The total number of transmitted user words is 980 for 40 Mhz DSP rate or 740 for 20 Mhz DSP rate.

Zero Suppression and DSP Algorithms

Zero suppression requires the comparison of the stored offset for each AMU cell for each channel with the channel's digitized signals. The algorithm must keep all 5 samples for each channel that is non zero. A non zero channel would have the majority of the samples above some threshold value. Requiring a majority but not all will keep a channel that may have a bad sample. An additional complication is that if one channel is nonzero (meaning a hit) than we want to keep its nearest neighbors. This means that a channel could be eliminated by zero suppression but be kept because its neighbor is above threshold. Much of this depends on where the algorithm threshold is compared to the final software threshold in the analysis. In all cases the raw signal will end up in the final data stream and we do not require any intermediate analysis for peak centroids or the like.

Data Format in the Final Data Stream to the Event Builder

The data format in the final data stream to the event builder will include headers and data words. The header will include event number, subsystem identifier, FEM number and the 8 user words. The data words will include the channel number, the sample number, and the data. The data format will have the structure,

channel(8bits), sample(3bits), and data(11 bits)

for a total of 22 bits. The data format out of the DSP ( from the WEB) is presently an 8 bit label space and 16 bit data space for a total of 24 bits. We will require modification of the label field to be 11-12 bits or a data field with the sample bit appended to equal a 16 bit field.