CMS Endcap Muon DDU: DMB-DDU Data Format
Second Prototype (and First-Updated) Version

DMB-to-DDU Transmission (for pre-July 2005 DMB firmware)

This document describes the format structure for the Event Record arriving at the CMS Endcap Muon system Device Dependent Unit (EMU DDU). The data originates from the Data Acquisition Motherboard (DMB) in response to a Level 1 Accept (L1A) from the trigger system. Although the event size depends on the conditions within the cathode strip chamber (CSC), the event structure is constrained to a set of fixed format rules. Throughout this document it is to be assumed that bits are ordered with the LSB (bit 0) on the right.

Note 1: The current format for the 16-bit DDU input has been designed to easily interface with a 64-bit readout system through S-LINK64 (see DDU Output Format). This requires that all data sent to the DDU must have an integer multiple of 4 16-bit words for each event. Furthermore, the highest bit of each 16-bit word from a DMB is reserved as a special flag for `DDU Code' words; since we are constrained by a 16-bit fiber-optic transmission system, only the lowest 15 bits are available for physics data.

Note 2: The 16-bit DDU Code words are always sent in groups of 4 to make a 64-bit DDU word. The lowest 16-bit portion of a 64-bit DDU word is the first word received by the DDU, and is designated as word "a". The second 16-bit portion received by the DDU is word "b", the third is word "c", and the last (and highest) 16-bit part of a 64-bit DDU word is designated as word "d". The DMB-DDU word labels can be distinguished from the HEX codes in this page which use the capital letters A-F, sometimes preceded by the "0x" designator as used in C programming.

DMB-DDU Control Codes

> bit15 set HIGH indicates a special DDU Code word < When the most significant bit of the word is high, the next three most significant bits indicate the DDU Code word:

Pattern of Bits 15 14 13 12	DDU Code	Definition and uses for lowest 12-bits
1000	8	DMB Lone Word: L1A received, but no matching LCT on CSC, and no useful data is present.
1001	9	DMB First Header Word: More data follows for this event (also Header Word 1; see below). this word is removed by zero-suppression logic in 1st Prototype DDUs (all big red PCBs)
1010	A	DMB Generic Header Word: DMB Event information (there is now 1 of these, Header Word 2; see below).
1011	B	Error/Interrupt : Not fully implemented yet. (3 bit error type + last 9 bits have debug data; see below).
1100	C	FMM/Status Word: (not used yet)
1101	D	Trigger Trailer Word: Last word from the TMB or ALCT for this event (generated by TMB and ALCT).
1110	E	DMB Last Trailer Word: Marks the LAST word from the DMB.
1111	F	DMB First Trailer Word: Indicates End of DMB; the NEXT word is the LAST word from the DMB.

Overview of the DMB-DDU Event Record

In an EMU DMB there are 2 types of events: those that contain a local charge track segment (an "LCT") corresponding to the L1A and therefore have interesting CSC data, and those that do not have interesting data (the DMB "No Data" events). This information is encoded in the first word of the DMB record as DDU Code word 9 or 8, respectively. In the case of matching LCT.and.L1A, additional data words follow that include the Trigger Motherboard (TMB) data, Anode and Cathode LCT data, Cathode Front End Board (CFEB) data, as well as Header and Trailer words containing DDU control information. Error conditions are described below.

For LCTxL1A match on CSC:

2 Header Words (all are `DDU Codes')

Header Word	Highest 4 bits	DDU Code	Lowest 12 bits [11:0]
1a	1001	9*	DMB_L1A[11:0] recent change, 9/24/04
1b	1001	9*	DMB_L1A[23:12] recent change, 9/24/04
1c	1001	9*	TMB_DAV(1) + ALCT_DAV(1) + CFEB_ACTIVE(5:1) + CFEB_DAV(5:1) recent change, 9/24/04
1d	1001	9*	DMB_BXN[11:0]
2a	1010	A	TMB_DAV(1) + ALCT_DAV(1) + Active-DAV Mismatch(1) + TMB_DAV(1) + ALCT_DAV(1) + TMB_DAV(1) + ALCT_DAV(1) + CFEB_DAV(5:1)
2b	1010	A	DMB_CRATE(8) + DMB_ID(4)
2c	1010	A	CFEB_MOVLP(5:1) + DMB_BXN[6:0]
2d	1010	A	DMB-CFEB-Sync[3:0] + DMB_L1A[7:0]

*9-code words are removed by zero-suppression logic in 1st Prototype DDUs (all big red PCBs)

ALCT Data (if present*) - See Trigger Data Format, below
TMB Data (if present*) - See Trigger Data Format, below
CFEB Data (CFEB 1 to 5, if present*) - See CFEB Data Format, below
2 Trailer Words (both are `DDU Codes')

Trailer Word	Highest 4 bits	DDU Code	Lowest 12 bits [11:0]
1a	1111	F	Repeat Header 1c recent change, 4/20/05
1b	1111	F	CFEB_MOVLP(5:1) + TMB_HALF(1) + OVLP_HALF(1) + CFEB_HALF(5:1)
1c	1111	F	DMB_L1PIPE(8) + DMB_BXN[3:0]
1d	1111	F	Repeat Header 2b
2a	1110	E	TMB_FULL(1) + CFEB_FULL(5:1) + TMB_MT(1) + CFEB_MT(5:1)
2b	1110	E	ALCT_TIMEOUT(1) + CFEB_ENDTIMEOUT(5:1) + TMB_TIMEOUT(1) + CFEB_STARTTIMEOUT(5:1)
2c	1110	E	Repeat Trailer 2b
2d	1110	E	Repeat Trailer 2b

The First Trailer is a flag for the DDU indicating the DMB is nearly done. The Second Trailer is actually the Last Word from the DMB.

For NO LCTxL1A match on CSC (i.e. DMB "No Data" event):

Single Word Event (1 header only), also known as a "DMB Lone Word": (see `DDU Codes')

Lone Word	Highest 4 bits	DDU Code	Lowest 12 bits [11:0]
1a	1000	8	DMB_L1A[11:0] recent change, 4/20/05
1b	1000	8	DMB_L1A[23:12] recent change, 4/20/05
1c	1000	8	0000.0000.0000 recent change, 4/20/05
1d	1000	8	DMB_BXN[11:0]

* "if present" means that an LCTxL1A match was found on that unit

DMB-DDU Header/Trailer bit definitions (some bits get repeated for bit error mitigation)

TMB_DAV: non-zero indicates TMB data exists for this event
ALCT_DAV: non-zero indicates ALCT data exists for this event
CFEB_ACTIVE: indicates which CFEBs should have sent data to DMB
CFEB_DAV: indicates which CFEBs have sent data to DMB
DMB_L1A: L1A Event Number, count from DMB
DMB_BXN: Bunch Crossing Number, count from DMB
DMB_CRATE: peripheral crate ID from DMB
DMB_ID: board number of DMB
DMB-CFEB-Sync: bits used for DMB-CFEB synchronization check
CFEB_MOVLP: the time sample for this event has multiple overlaps with samples from previous events
TMB_HALF: zero indicates that the TMB FIFO on the DMB is half-full
OVLP_HALF: zero indicates that the Overlap FIFO on the DMB is half-full
CFEB_HALF: zero indicates that the CFEB FIFO on the DMB is half-full
DMB_L1PIPE: number of L1A Events backed-up in the DMB
TMB_FULL: zero indicates that the TMB FIFO on the DMB is full
CFEB_FULL: zero indicates that the CFEB FIFO on the DMB is full
TMB_MT: zero indicates that the TMB FIFO on the DMB is empty
CFEB_MT: zero indicates that the CFEB FIFO on the DMB is empty
ALCT_TIMEOUT: indicates that the start and/or end of ALCT data was not detected within the time-out period
CFEB_ENDTIMEOUT: indicates that the end of CFEB data was not detected within the time-out period
TMB_TIMEOUT: indicates that the start and/or end of TMB data was not detected within the time-out period
CFEB_STARTTIMEOUT: indicates that the start of CFEB data was not detected within the time-out period.

Trigger Data Format

For a TMB or ALCT Data Word (bit15 low)

See TMB and ALCT documentation.
The last Trigger word is a DDU Code word containing the word count and CRC-22 (see Trigger Trailer below).

Trigger Trailer (bit15 high: DDU Code 0xD)

Recall that "DDU Special Code" words come in groups of 4, so there are 4 "D" words (let's call them Da, Db, Dc, Dd)
These indicate the last word from the TMB or ALCT for this event. Lowest 12 bits:

bit11: high for TMB, low for ALCT (but not for word "Dc")
other bits:

Da[10:0]=lowest 11-bits of CRC-22
Db[10:0]=highest 11-bits of CRC-22
Dc[11:0]= hex "E0F" for TMB, hex "E0D" for ALCT
Dd[10:0]=11-bit Trigger word count (# of 16-bit frames)

CFEB Data Format (for pre-July 2005 CFEB firmware)

> bit15 set LOW for all data words< For each of the 4 or 5 CFEBs, one of three things can happen after an L1Accept:

Empty: Zero words are delivered, as this CFEB has no data.
Error: Some words are delivered, including one or more DDU Code words indicating multiple FIFO overlaps or possible error condition.
Normal: A series of single-time-sample data packets with 100 words each (8 or 16 time samples in series form a full event).

For the 3rd case (normal data transmission), each CFEB data stream is composed of 96 data words per time sample (for the 12-bit digitized cathode strip data), plus 4 additional words at the end of each time sample (a CRC word, 2 status words and 1 dummy, making the word count an integer multiple of 4). Therefore, the full word count per CFEB (CF_WC) depends on the number of time samples (N_ts) digitized:

CF_WC = {[(16 channels) * (6 layers)] + 1 CRC + 2 CFEB_INFO + 1 dummy} * (N_ts)

N_ts (# time samples) is set in the hardware or through slow control, typically set to 16 for early LHC running (perhaps eventually reduced 8) yielding CF_WC(typical)=1600.

Each of the 96 data words has the following format:

CFEB Data Bit Bit Definition

15 Always LOW for data words. HIGH for DDU Code word (e.g. DMB Trailer or Error case 2 above).

14 Overlapped sample flag (normally HIGH; set LOW when two separate LCTs share a time sample).

13 Serialized CFEB-SCA controller data (trigger time and capacitor block number; see below).

12 Out of range flag from CFEB ADC.

11-0 12-bit Digitized CFEB ADC data.

CFEB Data Bit	Bit Definition
15	Always LOW for data words. HIGH for DDU Code word (e.g. DMB Trailer or Error case 2 above).
14	Overlapped sample flag (normally HIGH; set LOW when two separate LCTs share a time sample).
13	Serialized CFEB-SCA controller data (trigger time and capacitor block number; see below).
12	Out of range flag from CFEB ADC.
11-0	12-bit Digitized CFEB ADC data.

The 15-bit CRC word (the 97th word in each time sample) is calculated for the set of 96 data words in a time sample. It is created using a CRC-15 algorithm (description, code, generator) with 1-bit error correct, multiple-bit error detect. Words 98 and 99 contain CFEB Status information:

Word 98: L1PIPE_MT(1) + LCTPIPE_MT(1) + L1PIPE_FULL(1) + LCTPIPE_FULL(1) + LCTPIPE_CNT(4) + NF_SCA(4)
Word 99: BUSY(1) + SCA_FULL(1) + CFEB_PUSH(1) + LCT_POP(1) + L1PIPE_CNT(8)

L1PIPE and LCTPIPE refer to CFEB-SCA Controller internal pipeline status; NF_SCA is the number of free SCA blocks (12 max); BUSY indicates that the system is digitizing data; CFEB_PUSH indicates that data is being sent to the DMB; LCT_POP indicates that a used SCA block is now available. Word 100 is a dummy: 0x7FFF.

CFEB Data Stream
The ordering of the words in the data stream from a single CFEB is described by the following nested loops:

do (N_ts time samples){
do (Gray code loop over 16 CSC Strips; S=0,1,3,2,6,7,5,4,12,13,15,14,10,11,9,8){
do (loop over 6 CSC Layers; L=3,1,5,6,4,2){
}
}
CRC word
CFEB Info word 98
CFEB Info word 99
Dummy word (0x7FFF)
}

Should an error occur, the CF_WC may deviate from this scheme (case 2 above). Whenever that happens, a DDU Code word will be sent indicating an error condition (see below) that modifies the word count.

Serialized CFEB-SCA Controller Data
Bit 13 of each data word carries the serialized 16-bit CFEB-SCA Controller (SCAC) status word, containing trigger and SCA information. This data word is serialized (LSB first) with one bit in each of the 16 strips read out, yielding the 16-bit word. Due to the innermost-loop over the 6 CSC layers, each bit of a 16-bit SCAC word is actually sent 6 times in a row (once for each layer). Since there are 16 strips read out for each time sample, the complete SCAC word can be reconstructed independently in every time sample. The 16-bit word is defined as follows (highest-to-lowest bit):

TS_FLAG(1) + SCA_FULL(1) + LCT_PHASE(1) + L1A_PHASE(1) + SCA_BLK(4) + TRIG_TIME(8)

TRIG_TIME indicates which of the eight time samples in the 400ns SCA block (lowest bit is the first sample, highest bit the eighth sample) corresponds to the arrival of the LCT; it should be at some fixed phase relative to the peak of the CSC pulse. SCA_BLK is the SCA Capacitor block used for this time sample. L1A_PHASE and LCT_PHASE show the phase of the 50ns CFEB digitization clock at the time the trigger was received (1=clock high, 0=clock low). SCA_FULL indicates lost SCA data due to SCA full condition. The TS_FLAG bit indicates the number of time samples to digitize per event; high=16 time samples, low=8 time samples.

Error Conditions and Codes

Errors which are detected by the DMB will be flagged as such. This may result in a deviation from the anticipated number of words delivered, and it may result in an odd number of words being present. In such a case, the special error word code (DDU Code 0xB) will occupy the upper 4 bits; the remaining twelve bits are described below.

Pattern of Bits 15 14 13 12	Pattern of Bits 11 10 9 (error code)	Code Definition	Lowest 9 bits [8:0]
1011	000	unused	unused
1011	001	CFEB: SCA Capacitors Full	0 + 4-bit Block Number + 4-bit FIFO1 word count
1011	010	CFEB: FPGA FIFO full**	0 + 4-bit FIFO3 word count + 4-bit FIFO1 word count
1011	011	unused	unused
1011	100	unused	unused
1011	100	unused	unused
1011	100	unused	unused
1011	101	DMB: FIFO full*	0 0000 0000
1011	110	DMB: FPGA FIFO full*	0 + 4-bit GFIFO word count + 4-bit LFIFO word count
1011	111	unused	unused

* Note that for these error conditions, the word count can no longer be verified...resynch required!
** The data may be corrupted in this case...resynch required!

Other Endcap Muon DDU pages

DDU Home Page
DDU Output Format (Prototype Version)

Revision Summary:

4 June 1999: Original version (J. Gilmore, gilmore@mps.ohio-state.edu)
29 June 1999: Format update (P. Nylander, nylander@mps.ohio-state.edu)
9 August 1999: Completed the data stream format description (J. Gilmore)
6 September 1999: Modified Header 1 and Header 6 format descriptions (J. Gilmore)
5 September 2001: New format specifications for S-Link64 DDU Readout Scheme (J. Gilmore)
11 December 2001: Finish New Format update for Fast sites (J. Gilmore)
20 August 2002: Publish under main OSU CMS area (J. Gilmore)
24 October 2002: Realign and update all EMU DDU web pages (J. Gilmore)
4 November 2002: Update for new DMB Rev. 5 (Virtex2) data format (J. Gilmore)
9 December 2002: Fixed some bit conflicts in Header 2d of new DMB Rev. 5 data format (J. Gilmore)
16 January 2003: Fix Header 2d definition (J. Gilmore)
1 May 2003: Fix format info for Trigger now that TMB and ALCT have separate data flows (J. Gilmore)
24 Sept 2004: Modify for New DDU, ordering of 9-codes swapped (J. Gilmore)
20 April 2005: Fix typo on Trailer Word 1a definition, update "recent changes" (J. Gilmore)
23 August 2005: Updated to distinguish Prototype & Production versions (J. Gilmore)

CMS Endcap Muon DDU: DMB-DDU Data Format Second Prototype (and First-Updated) Version