A Monitoring Algorithm for the TABs

updated: 10-Mar-03

Overview & Goals

• Online monitoring information is collected only for events associated with an L1 Accept. The main goal of the L1Cal online monitoring scheme is to provide some level of access to L1Cal data taken on those beam crossings that do not come from L1 triggers. To do this we require some mechanism to store a "history" of data taken on bunch crossings before the one corresponding to the L1 Accept.
• Two pieces of information are actually monitored for each Trigger Tower.
  1. <Et> calculated over n BC's
  2. rms(Et) calculated over n BC's
• In the Run IIa system this is accomplished by storing the EM and HD Et's for 8 BC's preceding the current one and transmitting all this information to the TFW on a CollectStatus request.
• In Run IIb it may be more efficient to calculate the mean and rms (or Sum(Et) and Sum(Et²) over an arbitrary number of BC's in the TABs and then transmit only these numbers to the TFW.

Algorithm

Running Sum(Et) Calculation

1. Parallelize serial input Et from ADFs ==> et_in[7..0]
2. Delay et_in to keep in step with Et² calculation
3. et_in to n-deep delay register
4. Subtract oldest Et from Sum(Et) register:
   sum_reg[31..0] - et_in_delay_n[7..0] ==> sum_sub[31..0]
5. Update Sum(Et) register
   sum_sub[31..0] - et_in_delay_2[7..0] ==> sum_reg[31..0]

Running Sum(Et²) Calculation

1. Parallelize serial input Et from ADFs ==> et_in[7..0]
2. Multiply:
   et_in[7..0] x et_in[7..0] ==> et2_in[15..0]
3. et2_in to n-deep delay register
4. Subtract oldest Et² from Sum(Et²) register:
   sum2_reg[31..0] - et2_in_delay_n[7..0] ==> sum2_sub[31..0]
5. Update Sum(Et²) register
   sum2_sub[31..0] - et2_in_delay_2[7..0] ==> sum2_reg[31..0]

Running Average & RMS Calculation (if necessary)

1. Note: this assumes n = 2^x
2. Average: Sum(Et) / n:
   ave = sum_reg[31..x]
   output ave[15..0]
3. Average²: [Sum(Et)/n]²
   ave2[31..0] = sum_reg[15+x..x] x sum_reg[15+x..x]
4. RMS: Sum(Et²)/n - <E>²
   rms = sum2_reg[31..x] - ave2[31..0]
   output rms[15..0]

Notes

1. Types of Monitoring Requests
   • Synchronous with L1 Accept (online monitoring):
     This requires that the running sums be synchronized in a similar way to L2/L3 readout. When an L1 Accept for event i arrives at the L1Cal the running sums registers should contain information for:
     BCi, BCi-1, BCi-2, ..., BCi-n
   • VME-initiated collect monitoring requests:
     These can come at any time.
2. Reading Out Running Sums Registers
   Delay registers should be cleared after information is transmitted to readout buffers.
3. Truncation (for n = 8 BC's)
   
   • <Et> = 8 ADC counts; RMS = 1-2 counts
     ==> Sum(Et) ~ 64
     ==> Sum(Et²) ~ 512
   • Should be safe in truncating numbers to 16-bits for output