Difference between revisions of "CDC prototype more on timing"
Line 5: | Line 5: | ||
#Search forward for sample x where adc value > high threshold n<sub>1</sub> sigma | #Search forward for sample x where adc value > high threshold n<sub>1</sub> sigma | ||
#Step back p samples to sample x-p, take adc value of sample x-p to be local pedestal value | #Step back p samples to sample x-p, take adc value of sample x-p to be local pedestal value | ||
− | #Subtract local pedestal value from a number (10+p) of samples starting at sample x-(p+5) to the LE algo for | + | #Subtract local pedestal value from a number (10+p) of samples starting at sample x-(p+5) to the LE algo for desampling |
− | #Search through | + | #Search through desampled data, start with last sample (highest adc value) search backwards until adc value < low threshold n<sub>2</sub> sigma |
#Calculate time where interpolated unsampled data cross n<sub>2</sub> sigma, and add to time of sample x-p, this is the estimated drift time. | #Calculate time where interpolated unsampled data cross n<sub>2</sub> sigma, and add to time of sample x-p, this is the estimated drift time. | ||
− | + | Below are plots of desampled adc values (z) using different values for p (local pedestal lead time ahead of first/high threshold crossing) for high threshold of 5 sigma. x-axis=0 corresponds to sample x-p-4.2, x-xaxis=5 corresponds to 1+x-p-4.2, etc. | |
− | 5 | + | 5 desampled values per 8ns sample, all events for ch17 (central straw) included (no tracking) |
{| border="0" cellpadding="2" | {| border="0" cellpadding="2" | ||
|width="450pt"| | |width="450pt"| |
Revision as of 16:26, 24 October 2011
Current analysis code procedure:
- Calculate s.d. of pedestal for first 100 samples, 100 events, save for later use (sigma)
For each event...
- Calculate mean pedestal over 100 samples ending 10 samples before the trigger time (every 4th of these samples also works)
- Search forward for sample x where adc value > high threshold n1 sigma
- Step back p samples to sample x-p, take adc value of sample x-p to be local pedestal value
- Subtract local pedestal value from a number (10+p) of samples starting at sample x-(p+5) to the LE algo for desampling
- Search through desampled data, start with last sample (highest adc value) search backwards until adc value < low threshold n2 sigma
- Calculate time where interpolated unsampled data cross n2 sigma, and add to time of sample x-p, this is the estimated drift time.
Below are plots of desampled adc values (z) using different values for p (local pedestal lead time ahead of first/high threshold crossing) for high threshold of 5 sigma. x-axis=0 corresponds to sample x-p-4.2, x-xaxis=5 corresponds to 1+x-p-4.2, etc. 5 desampled values per 8ns sample, all events for ch17 (central straw) included (no tracking)
High threshold 4 sigma
High threshold 6 sigma
Looks like p=3 should work best.
Previous best resolutions for 50/50 gas mix using single threshold of 4sigma:
Resolutions for ch20 ignoring hits before tzero, using 50/50 mix in Garfield and wire diameter of 17.8um instead of 20um, 2100V (left) and 2050V (right).
Current best resolutions for 50/50 gas mix and 17.8um wire, 2100V, using high threshold 5sigma followed by low threshold 1sigma, p=3
Ignoring hits before tzero (tzero = fitted value + dt, dt=6ns)
Resolution of ch20 (central straw, 17) with early hits moved to t0 + 1ns (left) and t0 + 2ns (right), only showing events with early hit in ch20
Moving early hits to tzero+1ns
Below: 2100V with high threshold 5sigma, low threshold 1sigma (left) high threshold 5sigma, low threshold 2sigma (center) and 2050V with high threshold 5sigma, low threshold 1sigma (right)