[Olympus] rates in TOFs

[Alexander Winnebeck]

Hi Alexander,

maybe let me add some more information about the fastbus readout.
The drift chamber TDCs are zero suppressed, so the data load is depending on the occupancy. During the commissioning we found a typical amount of data of less than 100 bytes per detector. The TDC of the TOFs is also of the same order. The most data is coming from the TOF ADCs, because they are not zero suppressed. The amount of data is 524 bytes/event.
The maximum rates of 1.3 kHz we could observe during the August run is not a real hardware limit, but was a safe setting for the read out. Philipp is now working on the optimization of it. I have no idea how much speed we could gain.
If we want to speed up with additional hardware, it would make sense to put the TOF ADC into a third crate. To find another crate is not a problem, as you wrote yesterday, but the SFI sequencer is the challenge. I'm not aware of any at DESY or other labs I have access to.

Cheers,

Alexander  

On Aug 24, 2011, at 8:42 AM, Alexander Kiselev wrote:

>  Good mornig colleagues,
> 
>>  Installation of say two more crates would increase the effective readout 
>> speed by a factor of almost 2, as we discussed today. This may be pretty 
> 
>  this factor of 2 is way too optimistic I admit. Let's estimate correct 
> numbers useful for further discussion. Sorry if the below calculations
> look trivial.
> 
>  Indeed numbers quoted during the Monday meeting (max readout rate 1300Hz
> at 2% live time, "typical readout fastbus") have nothing to do with the
> real running conditions. Let's assume ~750us average "event cycle (readout
> time + waiting for next trigger)" for simplicity. 2% live time would mean
> that next trigger waiting time was only ~750us*0.02=15us, so the original 
> trigger load was ~67kHz. Clearly we will never run under conditions like 
> this. 
> 
>  I'd say with so large readout times we should never go beyond ~1kHz 
> original trigger load (~1ms next trigger waiting time in the above 
> calculations). Assuming now ~750us average readout time for simplicity, 
> this would correspond to the on-tape (accepted) trigger rate of ~1/1750us 
> or ~570Hz at a ~43% dead time, obviously. 
> 
>  If we manage to decrease the readout time from ~750us to say ~400us, 
> average event cycle will become 1400us for the same 1kHz trigger load, 
> which would mean ~715Hz accepted trigger rate at a ~29% dead time. So 
> gain would be significant, but no way close to a factor of 2.
> 
>  It should also be mentioned that useful (ep-elastic) trigger rate at 
> nominal luminosity is ~15Hz in BLAST acceptance and about the same in 
> 12-degree Lumi acceptance. Even if we increase average luminosity by 
> a factor of 2, trigger budget will (should!) be always dominated by  
> calibration (inclusive) triggers and random coincidences. Under these
> circumstances with a relatively slow readout hardware it does not make 
> sense to run at higher dead times, since this causes 1) smaller 
> *fraction* of ep-elastic events on tape, obviously; 2) in fact smaller 
> *rate* of ep-elastic events on tape, less obvious; 3) worse signal
> to background ratio if random coincidences start dominating. Derivation 
> of the "best figure of merit" running strategy in terms of DAQ rates
> as a function of luminosity, readout time, random coincidence scaling 
> and allowed fraction of calibration triggers is a rather straightforward 
> excersice, which "everybody else can do" :-)
> 
>  Best regards,
>    Alexander.

--
Dr. Alexander Winnebeck
winnebeck at mit.edu

Massachusetts Institute of Technology (MIT)
77 Massachusetts Avenue
Bldg 26 / Rm 441
02139 Cambridge, MA, USA
Tel: +1-617-253-3680

DESY
Bldg 66 / Rm 002
Notkestrasse 85
22603 Hamburg / Germany
Tel: +49-40-8998-6402