On Thu, 5 Feb 2004, Aaron Joseph Maschinot wrote:
>
> high theta hits that hit the (lorentz-angle-shifted) isochrone region but
> reconstruct outside the cell aren't too common (at least i don't see too
> many in the monte carlo). maybe we could play some game with trying to
> calculate the 1/2-way falloff point of the reconstructed position
> histgram and (somewhat arbitrarily) get a first guess at k?
>
I second very much this point. And I want to make 3 points : (!!!)
1) Of course we can argue about cell boundaries and probably intercellular
events are a better definition (at least logically). But we should not
miss on the issue of "size". Maybe we wouldn't quibble as much over 1 mm,
but when X_drift spans a range of 7 cm instead of 7.8 then it is clear
that we are not using the wch data correctly. Maybe this was only the
case
2) The run plan is to reverse the field. It is possible that - for the
same calibration - the momemntum resolution is bad at reverse field and
closer to design values for inbending. But really the comparison is to be
made for calibration files of *equivalent* quality ! In other words, as I
continue to argue, the most important priority is to image back the wires
and the chamber, and force the calibration to be consistent with what we
expect. So let's determine K-factor for inbending and outbending. When we
either find the same factors, or have 2 calibrations files showing similar
quality for X_drift in the 2 BLAST configurations, then we can compare
reconstruction results and possibly conclude about an intrinsic limit to
the reconstruction accuracy. In this sense btw, we don't need any new data
to study the issue, but I won't argue against that.
3) Sure there are caveats. But I hope the caveats won't be strong enough
to explain a 10 % shorter cell lenght ! In a worst case scenario, events
from the edge of the cell are simply lost due to some unexpected intrinsic
inefficiency, our max tdc value corresponds to only 6.5 cm from the wire
and our k-factor are a mockery.'
Well, I think there is some way to control the K-factors other then
"arbitrarily" drawing a cell boundary at 7.8 cm. Infact if the drift
velocity is really (0.95*v_drift) this will show at all drift distances
not just the largest. I suggest the following method: call x1,x2,x3 the
measured drift distance for wires 1,2,3. Use any pair of position xn,xm to
determine the 3 or missing position and adjust the scale factor for the 3
wire.
Once it is repeated for all pairs, or fitted simultaneously, this method
shoud give the "best" or most consistent k-factors relative to geometry
and straight tracks (assume a track is "straight" in a cell). With OOPS
(any other spectrometer really) we used this method to determine the
precise alignement of 3 wire chambers in a detector package. It is our
"resolution tune up"
In this way you force the calibration to be consistent with the data.
It is really ok. Nobody should force a physical interpretation of the
calibration parameters, but that's ok too. It would be interesting to
plot
(x1+x3)/2 vs x2
where x1+x3, after the staggering has been taken into account, gives
the position of the track at plane 2. Data in this plot should cluster
around the line with slope =1, intercept =0. Else we need to fix x2. And
maybe be ready to adjust the slope *and* the offset (a cumulative
correction to x0,t0) which requires looking at both side of the track.
When x2 is "fixed", you move on iteratively until you find a stable
point.
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:30 EST