At 09:01 AM 8/25/2004 -0400, Electronic Log Book wrote:
>Operator: zwart
>
>I added CCR Coulomb meter to the EPICS displays in the counting bay. B McA
>has made three distinct variables, total integrated DCCT, DCCT available
>to exp and DCCT accepted by experiment. The red needle (DCCT on tape) is
>the most important for luminosity optimization.
>
>The numbers are:
>
>Total DCCT: 360 C/hr
>DCCT available to expt: 300 C/hr
>DCCT accepted by expermient: 230 C/hr
>
> >From scalers looks like computer live is ~88%, but the ratio of 230/300
> gives only 76%. The other 12% comes from target flipping? This is an
> "effective" dead time which should be included in the I_min calculation
> for ring refill. We should have EPICS calculate the optimum I_min (at
> least make a suggestion) for a given I_max.
The red needle only counts charge when we are live and taking
data. It is inhibited by the same signal that inhibits our B scalers (an
OR of run not in progress, target flipping, electron spin flip, HV not at
operate). The difference is that the red needle looks back for an hour,
while the scaler display looks back for 1 second. When the target flips,
our BDCCT/DCCT ratio goes to 0 for a second or two (something on that
order, anyhow). The target flips every 5(?) minutes, so that is a small
contribution to subtract from Townsend's instantaneous 88%, maybe 1% or so.
The live time is driven by the event rate, and so the live time is
lower at the top of a fill. I think we are only at about 80% right now at
a fill of 130mA, and it will be more like 75% if we fill to 140mA.
The last contribution to live time that the red needle catches but
you don't think about when looking at the 1-second scalers is the time
between runs. That's typically a minute, every 30 minutes, so that's 3%
dead right there. This puts the overall efficiency at something less than
80%, consistent with the ratio of the red and green needles.
I do have a spreadsheet that shows the optimum dump current for a
given I_max, beam lifetime and current-dependent experiment live time. It
shows that (for deadtimes we were seeing back in the early summer) the
optimum dump current is about 63% of the fill current. This is basically
what Aaron's original sheet showed. The only difference is that for Aaron,
the average current turns out to be equal to the dump current; with
experiment dead time included, the average current to tape is LESS than the
dump current.
The maximum in I_avg as a function of dump current is pretty
broad; you lose about 0.5% in I_avg if you change the dump current to 57%
or 67% of I_max. My spreadsheet doesn't exactly mimic reality, since it
assumes the beam lifetime is independent of current. And if the current
goes too high, we will hit a wall in our event rate (the Ethernet
bandwidth), which I haven't put into the spreadsheet.
I think we can safely say that I_dump=0.63*I_fill, more or less
(and 0.60 is fine, easier to do in your head). To maximize the red needle,
we need to have a beam tune that minimizes the WC scalers (which are
correlated with the 2nd-level trigger rate, which drives our live time),
and we need to start a new run as soon as the old has finished. If our
event rate drops (because the beam tune is cleaner), then our live fraction
will addtionally improve just because runs last longer.
Karen
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:31 EST