Abstract
Over the last couple months, we have become aware of uncertainties in the silver epoxy connection to back plane of the sensors on modules that were built at University of California, Santa Barabara. We first discovered that there was a problem during the production of 28 TOB qualification modules for the new type 20 hybrids in December 2004. During the last 2 months we continued to see the problem in limited production with both ST and HPK sensors. We started doing various tests over the last month (January 2005) in an effort to pin point the cause and come up with potential solutions. We found that the connection between the silver epoxy and the sensor back plane surface is where the connection inconsistency is always found. Initial tests have shown that light abrasion (sanding) of the back plane surface eliminates this problem. The purpose of this note is to present what we have found out so far and with the help of the tracker community come up with viable and easily implemented solutions.
- Becoming aware of the silver epoxy bias connection problem
In December 2004, during the assembly of the 28 TOB modules with the type 20 qualification hybrids and pre-qualification ST sensors, we found one module with an open connection between the sensor and the frame during ACRS testing. It is hard to measure the resistance to the back plane of each sensor once they are bonded. If one of the sensors has a good connection the other sensor has a similar resistance going through the wire bonds or through the bias connection if the sensor is exposed to light. We found the resistance to the sensor with the bad bias connection would vary when you covered and uncovered the sensors from the room lights. If we did the same test with the sensor with a good connection it did not change. We used the spare pad on the back of the kapton circuit to make a new connection which worked fine. After this problem was observed we started checking the resistance to each sensor right after the module’s final survey the morning after it is built. In the rest of this batch of 28 modules we found 2 other sensors with higher than 40k ohm resistance and repaired them. There were others between 11.3K and 40K. The resistance in the HVcircuit for TOB modules should be close to 11.31K ohms and this is checked before we build each module. Values higher than 11.31K to the sensor back plane are added resistance from the silver epoxy connection. We did not spend extra in at this point trying to analysis what was happening since there was urgency to get these modules built and tested. Also they were not modules that would ever be used in the Tracker. They were made with pre-qualification ST sensors and though that may have been a factor. They all ARCS and LT tested fine. What we observed in this build had made us very aware of a possible problem with the silver epoxy bias connection.
- 51 modules built 6-1-05 to 1-2-05
During January, we proceeded with building HPK TOB modules with type 19 hybrids at a reduced rate and monitoring the resistance of the glue connection. Any module where the conductance looked suspicious (even slightly over the 11.31K ohms) we set aside and monitored. See Table 1. Modules with normal resistance are not shown.
We talked to Tra-Con who makes the Tra-Duct 2902 silver epoxy that we are using after we built the 1st 6 modules. They thought it might be a mixing problem. We then implemented strict mixing rules to make sure it was done properly and consistently. We checked the left over epoxy from each batch and started putting epoxy dots between glass slides with a ~100micron gap from each batch. The epoxy always conducted fine everywhere except in the module.
Next we increased the amount of silver epoxy we were using to the point where it was just squeezing out from under the side of the sensor.
PICTURE OF EPOXY BEAD
As can be seen in Table 1, this did not improve the bias connection, but it allowed us to check the resistance from the squeezed out epoxy bead to the sensor and also to the HV bias wire on the frame. When we checked these resistances on all the following modules with an exposed bead of epoxy (approx. a 1/3 of the sensors), we found very consistent results. The resistance from the bead to the HV bias wire was always 11.3K ohms. The silver epoxy to frame connection looked fine. The resistance from the bead to the sensor back plane was a few ohms on good connections and was high on all the high resistance connections. The high resistance without exception was in the connection from the exposed silver epoxy bead to the sensor back plane.
We switched to building with ST sensors and saw no difference.
We suspected that the problem may be a layer of oxidation on the sensor back plane, but it was also suggested that the problem could be that we were letting the epoxy sit on the frame too long (10 – 15minutes) before placing the sensors. This also could lead to a good connection to the frame and a poor connection to the sensor. We tried doing some tests with old ST sensor test pieces to confirm this but the results were inconclusive. You can see in Table 1 were we changed our procedures so that the silver epoxy was put on immediately before sensor placement. The results were the same as before.
The last thing we tried was to switch to the Epoxy Technology EE129-4 silver epoxy that is used by the TEC community. The Tra-Duct 2902 is used at both UCSB and FNAL. The results in Table 1 for the EE129-4 was not different than the Tra-Duct 2902.
At this point we decided to stop module production and do more tests to try to reproduce the problem on HPK test strip pieces we had received. We also decided to wire bond 4 of the modules with high resistance connections to see if the difference could be see in modules testing.
- Testing results - Tony
- Silver epoxy tests on sensor and aluminum test strips
In January when we continued to see bias connection problems we started to put together various test pieces to reproduce what we were seeing in the modules. We first tried aluminum to aluminum test pieces. With polished aluminum we got very good (less than 10 ohms) connections between the pieces. We placed ~100 thick tape on one of the test pieces so that when they went together we would end up with the same glue gap as the sensor does when it is placed on the frame. The problem with these initial tests was we always got good connections whether the aluminum was untreated, cleaned with alcohol or citranox, or sanded. Next we tried the aluminum pieces on sensor test strips with different types of preparation and at different time delays. Again we saw very little variation in the resistance of the connection.
At this point we decided it was necessary to try to duplicate the gantry in the placement of the sensor test strips on the aluminum test pieces. In all the tests up to this point we were placing one test piece on top of the other by hand. This is impossible to do with out moving it around as you place it. We had a small vacuum chuck built and attached to a vertical stage. This allowed us to place the sensor test strips from HPK down on the aluminum test piece in much the same way the gantry places the sensors.
Vacuum chuck and vertical stage Sensor test strip with aluminum test pieces
The first 2 test strips (10 epoxy dots) we assembled in this manor with parts that had no special preparation showed high resistance in most of the connections. See Table 2. A couple of the connection were in the mega and kilo ohms range much as we had seen in the modules.
Next we did 2 control sensor test strips (no preparation), 2 test strips with ethanol cleaning and 2 test strips with light sanding with 600 grit sand paper. The control and ethanol cleaned strips were pretty much the same. No real high resistances (2 dots in the hundreds of ohms, the rest in the 10’s of ohms). The sanded test strips all had resistances under 10 ohms. The sanding had made a clear improvement.
The most recent test we did is shown in Table 3. The difference between the 2 control strips and the sanded strips is very dramatic. Again all the dots on the sanded sensor had under 10 ohms thru the epoxy to the sensor surface.
- Conclusions – additional information
The measurements thru the epoxy bead on sensors with high bias resistance and the reduction in resistant when we sand the back of the senor suggest that there is some sort of oxidation layer on the aluminum back plane of the sensor. This layer does not always cause increased resistance in the bias connection, but seems to do it on a consistent basis on both the ST and HPK sensors.
In table 1, all the sensors that had high resistance were repaired by applying Tra-Duct 2902 to the extra bias pad. The resistance of all the repaired connections were measured and listed in the last column on the right for each module. As can be seen, none of them had high resistance even though they were making connections on sensors that had already demonstrated a high resistance first connection. The only difference was that the second connection was made manually using a curved tool (looks like a dental tool) that could apply the epoxy between the kapton and the sensor surface thru the hole in the center of the extra pad. Below if a picture of the tools used and a repaired pad.
The minor amount of rubbing the applicator tool on the surface of the sensor during application seems to be enough to break through the oxidation layer. We will continue to experiment with less aggressive methods of breaking thru this oxide layer on the sensor back plane than sanding. We are also consulting with HPK on what might be the best method to achieve a good low resistance connection to the sensor back plane.
As can be seen in the data in Table 1, if the connection has a high resistance it can be unstable. The connections with no extra resistance did not vary while many of the high resistance connection would jump up and down in resistance form day to day. This would seem to indicate an unreliable connection.
Even with a good initial connection to the sensor back plane it may be wise to have a redundant connection with the extra pad on the TOB modules. On the TEC modules 2 dots of epoxy are already applied to each sensor, so this redundancy is already there. It may be possible to actually make both of these connections on the gantry for the TOB modules. This will have to be investigated.
It is obvious from what we have learned that checking of this connection before bonding on all modules should be a standard procedure. One observation in working with the 2 different epoxies is that the EE-129-4 took ~48 hours for the connection to stabilize where the 2902 didn’t seem to change much after ~ 24 hours. There were not consistent differences in the conductivity of the 2 epoxies as far as we could see.
Just this week FNAL built 6 modules and then measured the resistance the next day. They found 5 of the 12 sensor bias connections had higher than normal (11.3K) resistance. The values of the high resistance connections were: 22.0, 22.5, 48.8, 169.2 and 178.8K ohms.