Hi folks,
>
>
>
> On Mon, 11 Mar 1996, Chris Walter wrote:
>
> > Chan: 3090000
> > VME WFD Node
> > Channel: 3090000
> > Tics to data: 0
> > Number of buckets: 2172
> > Time Timewd dBits AD0 AD1 AD2 AD3
> > 0 61198 8888 28 28 27 32 -1.4 -1.4 -0.6 -4.6
> > 327690 61196 8888 31 35 33 32 -3.8 -7.0 -5.4 -4.6
> > 327710 61192 8888 32 29 33 35 -4.6 -2.2 -5.4 -7.0
> > 327730 61188 8800 37 34 30 27 -8.7 -6.2 -3.0 -0.6
> >
> > OK now look at the time words. You see the first one goes from 61198
> > to 61196. This isn't really supposed to happen. Time Words are
> > always supposed to be separated by 4. This one is only separated by
> > 2. Now what happens if you use these words to expand the time into an
>
On Tue, 12 Mar 1996, Ash Sanzgiri wrote:
> Isn't this exactly what you would expect if the Stop Enable bit (bit 6)
> of the control register is turned on? When this bit is enabled, the last
> data in the RAM contains the stop time rather than the last sample time:
> so the difference between the last two timewords could be 1 to 4. I think
> this bit is set in the control register, so I am surprised why this kind
> of thing happens so infrequently.
Ash is of course right. I use the same code Chris W does to decode
the WFD buffers and it deccodes the times correctly (it doesn't assume
that the first two times will be separated by greater than 4 ticks). So
I assume that there is some bug in the text output above only, and that
this was just a bad example of some problem Chris has found with this
channel. In the past we have replaced channels that have incorrect
timing like this in the MIDDLE of data.
The reason it doesn't happen that often is because there are usually
usecs between radioactivity pulses. (good thing or zero suppression
would be kind of useless)
> Concerning the missed rollovers, is the Rollover Enable bit (bit 4) in the
> control register being set? Since this forces a sample storage cycle when
> the timeword rolls over, it should provide an additional handle to
> detect a rollover apart from an increment in the timeword.
>
We don't enable the Rollover Bit. Ed tells me that this feature is
buggy. Enabling the Rollover bit does properly force this extra
storage cycle, but it apparently can cause ADC data to repeat between
samples in an unpredictable way. Not having this feature does make us
vulnerable to overly quiet channels. But for now I am assuming that if
channels do frequently miss rollovers (like the ones Chris found) it is
due to some other problem, like an improperly set threshold. (more on
this in the near future).
- Chris O.