[Halld-online] [New Logentry] Performance testing of the FDC F1 TDC crates

[somov at jlab.org]

Logentry Text:
--
We checked performance of the new F1 tdc firmware
provided by Ed using 4 FDC crates. In general, 
the F1 performance was found to be 'robust'  for 
large trigger rates and hit multiplicities.


Setup:
-----

Crates: rocfdc 1,4,13,14
each crate has 12 low-resolution F1 TDC modules
(except for rocfdc 1, where two more tdcs, not used 
by the FDC are installed)

Trigger: TS FP random trigger
         trigger holdoff parameters consistent with fadc 125
   
Hit multiplicity: changed by lowering LV thresholds

Block level = 40, buffer level = 4, readout window = 1.4 usec 

(we also checked performance using different block levels)


Data were read out from TI and TDC modules, but was NOT
transmitted from the crate.

(the maximum transmission rate of a gigabit Ethernet link 
was measured separately using an individual FDC crate and 
a setup of 4 FDC crates and random trigger. The rate is 
about 115 Mbps, see Fig. 1)


Fig 2. Top
----------

Data rate R (Mbyte/sec), read out from TI/TDC modules 
(in rocfdc 13 ) as a function of the trigger rate 
for various hit multiplicities.

( R = Input trigger Rate * Livetime * Size of 40 blocks / 40 ) 

Fig 2. Bottom

Corresponding live times as a function of the trigger rate.


Comments:
---------

There are two factors limiting the performance:
VME data transfer rate and the F1 TDC processing time.

For very large hit multiplicities, the VME transmission
limitations occur at a small trigger rate, when the
processing time is negligible, see blue curve in the
top plot of Fig. 2.

The maximum VME data rate was estimated to be about 
180 Mbyte/sec

At higher rates and modest hit multiplicities, the 
processing time becomes the limiting factor, see 
other curves on the same plot.

The green curve satisfies the GlueX high-rate performance
(expected L1 rate < 100 kHz,  1 Gbit/s Eth.)

Note, the processing time depends strongly on the number 
of hits in the particular F1 TDC module. The hit 
multiplicity corresponding to the green curve (rocfdc 13) 
is dominated by two F1 TDC modules, see Fig. 3. Each 
module has 3 (out of six) 'noisy' chips producing the 
maximum number of hits (64 hits/chip).

The crate performance is expected to be better if hits
are more uniformly distributed over F1 TDC modules in 
the crate.

Note, the new firmware provides better handling of burst
of triggers: data for some triggers can be lost (the 
corresponding flag is reported).

We'll keep on testing the performance.

---

This is a plain text email for clients that cannot display HTML.  The full logentry can be found online at https://logbooks.jlab.org/entry/3445710
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20161211/b85ae513/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: tdc_crate4_data_rate.png
Type: image/png
Size: 416793 bytes
Desc: not available
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20161211/b85ae513/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: f1_performance1.png
Type: image/png
Size: 31808 bytes
Desc: not available
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20161211/b85ae513/attachment-0004.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: occup_example.png
Type: image/png
Size: 10505 bytes
Desc: not available
URL: <https://mailman.jlab.org/pipermail/halld-online/attachments/20161211/b85ae513/attachment-0005.png>