Follow the instructions from the guide sheet.
In preparation,
go through the regular lab. experiment (i.e. your notebook report)
and make sure you know what you are supposed to do at the test.
At the Lab. test you will
not have to go through the calibration process
(with the Na spectral source).
As soon as you get to the apparatus make sure you see the cross hairs,
and that you are able to focus the spectrometer properly.
Center your spectrometer
on the source to get the maximum brightness.
Adjust the
width of the entrance slit if necessary (as you may recall, a narrow slit
allows you to see only the brightest lines...)
To get the measurements use, as recommended in the guide sheet, the
brightest lines.
If possible, drop by to the lab the week before the test
for a bit of extra training in the identification of those lines.
The practice shows that this is the main challenge: getting the
right correspondence between the lines and
's.
The Graph
Use most of the area of the graph paper.
Hopefully, this will allow you to display the
error bars for at least one of the variables
(independent [ 1/(
-
0) ]
or/and dependent [ y ] )
You expect to get a straight line of best fit, and determine
the slope m right from (and on) the graph.
As already noted, the biggest challenge is to get the right
identification of the lines. Make sure you don't add extra-challenge
by misplacing the points on the graph.
Quote the units throughout.
For the units of y you have
to write "arbitrary units".
For
the units of 1/(
-
0)
you have nm-1 (perhaps with a multiplier,
since you may want to quote numbers like 0.003 nm-1
as (3 × 10-3 nm-1).
The (reading) errors
You are measuring the position of
lines (given by y)
for various
.
As you may remember,
the only reading errors are in y
and
0.
The error in y
adds up from
- a reading error on the vernier's scale:
about 0.02 units (the repeatability error -
read the description in the Lab. Manual, pg. 99, for details)
and, maybe,
- some human error - in case you don't
get the center of the lines correctly :-) .
Of course, this second error can be minimized - just be
careful.
The value of
0
is written on the top of the spectrometer; the error is
typically 0.4 - 2 nm.
For
you are advised to assume no errors at all.
Therefore, the error calculus is greatly simplified.
The error bars for y
are most probably too small to be displayed.
Example:
The brightest lines show up for y in the
interval ( 6 ... 16); hence, the lenght of the axis for
y is 10 units.
Assuming you take 200mm for 10 units of
y,
then 1mm on the graph paper corresponds to
10/200 = 0.05 units of y.
Since the error in y is about
(± 0.02) units, then
the error bar in y
is only ( 1mm × 0.02/0.05 ) = ± 0.4mm which is hardly
displayable.
It seems that you cannot see the error bars for the y...
For the horizontal axis, on which you plot
1/(
-
0), you need to do some
calculus.
For a fast-track error estimate it is probably
a good idea to calculate the error in
the smallest and largest value
of 1/(
-
0)
and then interpolate the errors for the remaining points
(if you have time, calculate them individually - it's not that hard).
In case you are out of time, leave
this error calculation aside.
It might happen that the spread
of your points is so wide that plotting
the reading errors (and hence the error bars) are
of no help or/and significance.
On the other hand, if this wide spread happens to
you, then better check again how you identified the lines
(the
correspondence of
with
y).
Check the plots you have done for
the regular lab. experiment. See how big the error
bars are. You should get the same at the lab. test.
Expectations:
if the error in
0
si 2nm then the error bars should be visible (at least for some
of the 1/(
-
0)
values.
If the error in
0
is 0.4nm or so, then you will not see
the error bars :-)
Error in the slope
The error in m
has to be retrieved using the minimum and maximum slope
fit lines (see Lab. Manual,
pp. 133-140 for details on fitting techniques).
Quote m
with the right units and the right number of significant digits.
Note: using previously recorded data (from a notebook)
it took me about 15 min. to:
- draw the table (properly labeled) and copy the data
- calculate 1/(
-
0)
(a column in the table)
- calculate the error in 1/(
-
0) (a column in the table)
- design the graph (and draw the axes, put the labels, units)
- plot 5 data points
(with no error bars, since they were too small to be displayed)
- draw the best fit line
- retrieve the slope m from the graph
I couldn't get b (see fit formula)
because my horizontal axis
started at 2×
10-3 nm-1, not at 0.
- estimate a reading error for the slope
(I didn't use the
min-max slope lines, since the points were laying very nice on the best
fit line - no room for swinging :-).
Instead, I considered that there is
a 0.5% relative error
in the slope due to reading errors (of the run and rise) from the graph
(see the Common recommendations
page for more details
on how to get the error in the slope).
Moral: you have to train yourself, to make sure you can do
the analysis smoothly.
Conclusion
Write down again you main result
(the value of the slope, with
the right units and also the error, if any).
You may want to comment on the quality of the fit to a straight
line of your data (good/bad/...).
Back to main page
Last modified: July 02, 2003
© Sorin Codoban, 2003