[Halld-pid] f1TDC timing resolution

[Alexander Ostrovidov]

Beni,

As a follow up, I tried to repeat your F1 TDC resolution study in our
setup. I also plugged 2 outputs of the TI board into F1TDC. The 
resolution I measured is 0.866*56.6ps=49.0ps (see the attached plot).
While it is not that far off from your value of 41.7ps (based
on your plot with 56ps bin width), it is still 15% worse. At this point,
I don't know if  the newer F1TDC works 15% better than the older
one which we have, or if I should blame our VME crate with its 
problematic -12V for that.

The good news is that directly measured 49ps is pretty close to
55ps resolution which I got with a pulser-simulated 3-bar method.
This means that, apart from F1TDC, there is very little contribution
to the resolution (i.e., about sqrt[55ps^2-49ps^2]=25ps only) either 
from a discriminator jitter (for a fixed-amplitude signal), or from the
assumptions of the method itself (in which the 6-signal and 4-signal
distributions are used to extract the resolution of the remaining 
2-signal distribution).

Sasha 
 

On Monday, February 21, 2011, Beni Zihlmann wrote:
> Hi Sasha,
> attached are two fits to the same data with different bin sizes.
> one very fine with 25ps/channel and one with 56ps/channel
> the actual bin size of the F1TDC. As you can see the fit results
> do change a little but not drastically. The chi2/ndf is bad in
> both cases because it is not really a Gaussian distribution
> but closer to a square.
> 
> cheers,
> Beni
> 
> > Hi Beni,
> > 
> > Thanks for making these measurements! I have a quick question
> > about your timing resolution plot. Why did you choose a bin size of
> > 50ps for this histogram? After all, TDC measurements (and their
> > differences) are integers, and it would be natural to histogram them
> > in these units, with a bin size of 1 TDC count or, at least, with the
> > proportional 56.6ps bin size. With your bin size choice, a bin with
> > about 3000 counts is followed by a bin with 0 counts, followed by
> > a bin with 5000 counts. This is "statistically unnatural" and is a
> > pure artifact of a non-aligned binning. Fitting a very narrow
> > (literally, a couple of bins) histogram to a gaus might be very
> > sensitive to how the binning is done. Seeing chi2 / ndf = 4892 / 9
> > makes me wonder how stable the obtained 36.7ps resolution value
> > is. Would it be possible to plot your data with the "natural" bin size
> > of one TDC count? Thanks.
> > 
> > Sasha
> > 
> > On Monday, February 21, 2011, Beni Zihlmann wrote:
> >> Hi All,
> >> to follow up on the discussion about the timing resolution of our
> >> F1TDC I looked at the data I took to test the Hamamatsu H10534 PMT.
> >> Attached is a plot that shows the timing difference between the
> >> trigger signal going to the TI board and the trigger signal output
> >> from the TI board. The Gaussian fit gives a width of 36.7ps while
> >> the position is 31.85 ns. This means that the TI response time to
> >> the triggers was 31.85ns with a timing resolution of 36.7ps. This
> >> resolution includes the timing jitter
> >> of the TI board as well as the intrinsic timing resolution of the
> >> F1TDC. In other words the intrinsic timing resolution of the F1TDC
> >> is expected to be better than 36.7ps.
> >> 
> >> On a side note: looking at the open PMT and its base the active base
> >> circuitry
> >> 
> >>                            is based on 3 transistors.
> >> 
> >> cheers,
> >> Beni
> > 
> > _______________________________________________
> > Halld-pid mailing list
> > Halld-pid at jlab.org
> > https://mailman.jlab.org/mailman/listinfo/halld-pid

-------------- next part --------------
A non-text attachment was scrubbed...
Name: f1tdc_resolution.pdf
Type: application/pdf
Size: 14110 bytes
Desc: not available
URL: <https://mailman.jlab.org/pipermail/halld-pid/attachments/20110223/72c766c7/attachment-0002.pdf>