EPD Trigger is very similar to BBC, except EPD uses hit count instead of ADC sum for trigger.
Trigger layers:

1. 14 QTC boards with TAC, 108 TAC tiles for each east and west.
2. Layer zero: EP001(east) & EP002(west) DSMs.
3. Layer one: EP101 DSM.
4. Layer 2: Vertex DSM
5. TCU bits EPD-E, EPD-W, EPD-TAC; corresponding to number of good hits above hit count threshold and in TAC window.

Useful links:

1. Trigger Algorithm document (Explains the algorithms in QTs and DSMs for triggering)
2. Cheat sheet of QTs and DSMs
3. EPD QT mapping (please see the excel sheet at the end)
4. EPD online monitoring (Has summary plots, ADC & TAC distributions, bit map checking)

EPD trigger shows some mismatch (few percent to 20%) from QTC to L0 DSM.
Issue is more likely the Algorithm and/or output latch issues and can not be hardware issue (Many thanks to Steve and Jack fixing hardware issues).

Current Tier1 file is set to provide EPD-E, EPD-W and EPD-TAC
To test the trigger capabilities of EPD following runs are taken only with triggers: EPD-E + EPD-W + EPD-TAC
19075034 (short run)
19075035 (long run)
19075053 (during the vernier scan, very long run)

Following thresholds are used in above runs:

1. 300 < TAC < 4000
2.  16 < ADC
3. All the QTC boards are enabled to pass
4. A TAC difference is between 3895 and 4295 applied (corresponding to ~ 30 cm).
5. Hit count is required to have greater than 0 for both east and west.

Following are some QA plots from run 19075035

EPD TAC difference East vs. West	(a)Earliest TAC East, (b) Earliest TAC west, (c) TAC diff. Blue histo is calculated from QTs and Red is from EP101 So the difference between red and blue tells the bits mis-match!
Hit count from Ep101, Note that we have 108 TAC channles	Top left panel is for BBC only, rest compare BBC with EPD

Applied cut is doing its job on selecting events from ~30 cm cut.

Now the issues:
EP001 and EP002 shows earliest TAC of 0x20, 0x800 and 0x820 for east and west for about 5-20% of the time for all the QTC boards.
Please following bit map check plot, plot compares mis-match from QTC algorithm against bits received by EP001/2. Bits mismatch at 0x20, 0x800 and 0x820 clearly visible.
For 19075034

This gives peak TAC difference centered around 4096 (0x800-0x800 or 0x820=0x820 or 0x20-0x20, plus 4096)
   and 0x20 above and below 4096 for 0x800 and 0x820 combos.
Following plot from low statistic run 19075034

We confirmed that when there is no good hit in QT, no output is generated from QT algorithm or in EP001/2.

Current proposal is to include EPD trigger in production running with limited band with and play with QT algo latches and output latches.
Any thoughts, comments or suggestions?