Work on TCIM Glinks

Our DCIMs once had terrible problems holding Glink lock with the DCMs. The not-so-obvious cause of all or part of this problem was the clock signal from the TCIM which was distributed to the DCIM over the backplane of the VME crate they share. Here is a sketch of the MVD electronics chain, which may help to understand this discussion.

As a result of some modifications made in August, 2001 this problem has been substantially improved, but some problems remain. In particular, a few Glinks still do not lock and we have trouble getting Glink lock for the top output channel in most DCIM modules. A description of the August 2001 changes is here. This page has a sketch of the relevant parts of the TCIM schematic and some scope shots of the clock signals before the August 2001 improvements. Unfortunately, these scope shots were made before I recognized the grounding problem and I do not know which ground reference I used when I made them. There is also a document showing scope shots of the clock signals on the DCIM board -- also made before the fixes described above and some other fixes -- so it is out of date.

I am not sure how much remains to be done. There is a jumper on the side of each DCIM which is used to make a 40 MHz clock out of the 10 MHz clock it gets from the TCIM over the backplane. These jumpers were individually adjusted to maximize the Glink reliability on each DCIM board. Look about half-way down this page on the sketch where something is labeled "this sets the delay from input to output -- doesn't really matter but needs to be set somewhere (fraction of 1 beam clock)". This sketch shows the location of the jumper -- but the comment that it does not matter is wrong. The jumpers were set prior to the modifications to the TCIM (it was during the run and there was no time to readjust them). It is possible that a careful readjustment of all of the jumpers on the DCIMs will reduce or even fix the problem, but I doubt it. It is also possible that adding more copper to the connections between the two ground planes on the TCIM will improve the Glink reliability. I am fairly certain that this will help. Otherwise, further investigation will be needed. Finally, I should add the opinion of Glenn Young -- he thinks the problem may be due to noise from the "+5 Volt" (really about 6.5 V) power supply.

We have also noted (again before the August 2001 changes) that the shape of the clock signal on the VME backplane changes as you move away from the TCIM in the crate. Because of this, we now configure the crates with the TCIM in the middle. In general, we have noticed that the DCIMs work better as they are moved farther from the TCIM.

Because of the uncertainties, and because it is a system-level problem (depends on TCIM, DCIM, their numbers and locations in the crate, the power supply to the crate, and the termination of the backplane) the time needed for this project is not obvious. It is a job for a skilled electronics person or at least a persistent and lucky amateur. I have guestimated the time at 1 month.