[Vetroc_daq_sbs] 16-channel tdc 'dropout' at high-rate

[Benjamin Raydo]

Hi Evan,

The current VETROC firmware can handle an average of 3.9MHz of hit rate per channel. Basically channels are grouped in sets 16 and they must be serialized into a stream that can only handle 125MHz of edges per second (so 125MHz / 16 channels / 2 edges per pulse gives 3.9MHz). It would be wise to not run too close to that rate though (a truly random, but average, rate of 3.9MHz on would be okay).

If too many hits/per second are going into the TDC event builder server 16 channels you will basically cause that group of 16 to loose data - likely a significant amount (if not all).

This is just a firmware limitation and I think it would be not a big problem to support a much higher rate so let me know if the experiment really need to support a higher rate. Another important aspect is to consider the trigger rate and readout window size as this will tell you how much data per second needs to be readout on the VME - too high rates in trigger or occupancy can certainly make for an impossible VME readout solution!

Ben

----- Original Message -----
From: "R. Evan McClellan" <randallm at jlab.org>
To: "Benjamin Raydo" <braydo at jlab.org>
Cc: "Vetroc_daq_sbs" <vetroc_daq_sbs at jlab.org>
Sent: Monday, May 15, 2017 3:40:31 PM
Subject: 16-channel tdc 'dropout' at high-rate

Hey Ben,

I'm seeing a loss of TDC hits from the VETROC, in 16-channel blocks, when I run the PMTs at very high-rate (~1 to ~10 MHz).

I replaced one PMT signal with a pulser signal, and checked that channel as I took many runs, slowly increasing the average PMT rate. At some 'threshold' average rate, the number of hits suddenly plummets, and drops to zero. The channels seem to drop out in independent groups of 16 (0-15, 16-31, etc.).

I see the same loss (at the same average rate-per-channel) whether all inputs are connected, or only a single NINO is connected (16 channels). Therefore, I guess the loss must come somewhere in processing that is done separately for each 16-channel group.

I have checked that the NINO output shows no losses up through 10MHz per channel.

I also checked that it's not simply a decoder problem. The hits are missing in the raw data.



Perhaps I have asked you this before, but what is the expected rate capability of the VETROC TDC firmware? And, perhaps more importantly, what is the expected failure mode when the rate is too high?

I will be available this week (except for meetings Tuesday afternoon). I can meet you at ESB if it would be helpful.

(It's possible that these rates are already above the experimental requirements, but I want to make sure to understand the mechanism for the loss of hits.)

Cheers,
Evan



R. Evan McClellan, PhD
Hall A Postdoctoral Fellow
Jefferson Lab