<html>
<head>
</head>
<body text="#000000" bgcolor="#FFFFFF">
Hovanes,<br>
<ul>
<li>
<pre wrap="">How are the error bars calculated? Are these the spec uncertainties on
the ADCs and amplifiers propagated into this signal ratio?
</pre>
</li>
</ul>
<br>
This is a profile histogram using hbook. The error bars are simply
the spread in the values observed within a bin divided by sqrt(n).
The step size in this scan was 0.002 inches, so there are many
samples within a single bin. These error bars show the uncertainty
in the mean output at that displacement, calculated the usual way.
In case you would like to see it, I have created a new profile
histogram of the same data showing each individual step, this time
no error bar, just the data points. See plot below. That's a lot
of points! <br>
<br>
<ul>
<li>
<pre wrap="">What is the origin of the vertical shift of the curve, and apparently
a scale change as well? The limits are not -1 and +1, which means that the
zero-crossing does not indicate the beam position. I assume this can be calibrated out. But then if these signals are not well calibrated, can the 60Hz, 120Hz etc oscillation components that you see in the FFT be interpreted as fast beam current fluctuations resulting in the asymmetry variations
rather than fast position motion?
</pre>
</li>
</ul>
<br>
I have not subtracted any DC offsets that are present in these
preamplifiers. The zero points should be calibrated. I wanted to
show you very raw results, to show that the output is very simple to
interpret and does not require much processing. Of course, I could
have centered it vertically.<br>
<br>
<ul>
<li>
<pre wrap="">Is this the finest step scan that was done?
</pre>
</li>
</ul>
<br>
The step size was 0.002 inches. That was the smallest step size we
took. It is only 50 microns, which seems very small to me. It is
smaller than the size of the beam motion that we see, maybe the
smallest step that the translation stage could do, not sure.<br>
<br>
<ul>
<li>
<pre wrap="">My understanding of the sensitivity of this device to the beam
position was the uncertainty on the location of the "zero-crossing" of a linear fit in the central region (inversely related to the uncertainty of the gain in the central region). Do you have an estimate for that number from a simple fit of this curve in the central region?</pre>
</li>
</ul>
<br>
Yes, that is what I reported. The slope is only part of the story.
It converts an uncertainty in the difference/sum (dimensionless) to
an uncertainty in position (cm). The rest of the story is the
uncertainty on the difference/sum itself. Extracting that from real
data requires one to understand what part of the spread in the
difference/sum comes from real beam motion and what part comes from
detector noise. That separation can only be done with a FFT and the
knowledge that the detector plus preamp response curve is flat up to
the bandwidth cutoff. That was the point of showing the FFT plot.<br>
<br>
-Richard Jones<br>
<br>
<br>
<br>
On 10/31/2011 3:47 PM, Hovanes Egiyan wrote:
<blockquote type="cite" cite="mid:4EAEFB64.6080302@gmail.com">
<pre wrap="">Hi Richard,
thanks for the plot. Indeed, this is the plot I asked about. I have a
couple questions about it:
o How are the error bars calculated? Are these the spec uncertainties on
the ADCs and amplifiers propagated into this signal ratio?
o What is the origin of the vertical shift of the curve, and apparently
a scale
change as well? The limits are not -1 and +1, which means that the
zero-crossing does not
indicate the beam position. I assume this can be calibrated out. But then
if these signals are not well calibrated, can the 60Hz, 120Hz etc
oscillation components that you see
in the FFT be interpreted as fast beam current fluctuations resulting in
the asymmetry variations
rather than fast position motion?
o Is this the finest step scan that was done?
o My understanding of the sensitivity of this device to the beam
position was the
uncertainty on the location of the "zero-crossing" of a linear fit in
the central region
(inversely related to the uncertainty of the gain in the central region).
Do you have an estimate for that number from a simple fit of this curve
in the
central region?
Hovanes.
On 10/31/2011 02:58 PM, Richard Jones wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Hovanes and all,
I have attached a plot of what I think you wanted to see: the
difference over the sum of currents in the two inner wedges as a
function of displacement of the translation stage carrying the active
collimator across the beam. Notice that in the central linear region
of high sensitivity, the error bars are much larger than outside this
region, where the sensitivity to position decreases. The error bars
in the low slope regions on either side of the plot are a good
indication of the intrinsic resolution of the device. The gain of the
central linear region is about 1.8/cm.
In this plot you can see slow drifts in the beam position as bumps and
wiggles in the otherwise smooth transition from the region of -1 to
+1. I estimate the intrinsic resolution at 15nA electron beam current
on a 10^-4 radiation length target to be less than 50 microns at 600
Hz bandwidth.
-Richard Jones
</pre>
</blockquote>
<pre wrap="">
</pre>
</blockquote>
<br>
</body>
</html>