Electrical tester

The quality of every MSGC substrate, over 300 substrates were produced for HERA-B, had to be checked right after the production. A maximum number of 50 broken anodes was accepted and no single short between an anode and a cathode was tolerated.

Figure 4.5: Principal of the electrical substrate tester.

Checking every substrate with our semi-automated microscope (see Chapter4.2) was not

4.3 Electrical tester 49

feasible because it takes about two working days to control one single substrate. There-fore we developed a testing apparatus for a fast and reliable quality control. Checking a substrate with this setup takes about 2 minutes, including the analysis, graphical dis-playing, and producing the protocols. For a more detailed description see [48].

(a)cover opened (b)cover closed, during measurement

Figure 4.6: Picture of the electrical substrate tester.

The principal of the electrical substrate tester is depicted in Figure4.5. High voltage of typical 300 V is applied to the cathodes. The current flowing through the high resistivity substrate coating is measured and analyzed with a LabView application running on a PC.

Making quickly a temporary, and reliable connection to 768 anode pads with a pitch of 270µm is a challenging task. The solution choosen is to map the anode pad layout to the electronic board and to interconnect the two pad structures with gold plated wires incor-porate in a silicone rubber piece (see Figure4.7). The current signals of the anodes are 256-fold (16×16) multiplexed and fed into a analog to digital converter¹⁰for measuring.

4.3.1 Measurement procedure

The control, readout, and analysis program was completely programmed as a Lab-VIEW [47] application. Its purpose is to:

10National Instruments LabPC+

pad on MSGC substrate silicone rubber pad on electronic board

wire

(a) (b)

Figure 4.7: Interconnection between the MSGC substrate pad and the readout electronic

• initialize the tester hardware.

• assign each tested wafer a consecutive number.

• perform the actual measurement.

• analyze the data.

– find the anode- and cathode-breaks.

– find the shorts.

– report a misaligned wafer in the tester.

• display the raw data and the analysis results.

• save the raw data in a human and machine readable form as a short summary and as an extended report.

• read back the raw data of previous measurement, displaying and reanalyzing it.

For the duration of the measurement the high voltage is switched on. Its actual value is read back just before and after the measurement of a whole plate via an ADC. The currents flowing through the anodes are measured consecutively in groups of three, the n-th, the n+256-th, and the n+512-th are measured together. To suppress the 50 Hz noise, each signal is sampled with a frequency of 400 Hz 8 times and averaged afterwards.

Because the varying surface resistivity and the different length of anodes one can not use predefined values as threshold to discriminate good from broken anodes. Therefore in a first step the expected current of each anode, which is the normal expected current that would flow through this specific anode has to be determined (the blue line in Figure4.8).

4.3 Electrical tester 51

and / or cathode empty

pads

Figure 4.8: Typical substrate control measurements done with the electrical tester.

top: good substrate, middle: substrate with all detectable defects, bottom: substrate with produc-tion residuums

This is done in an iterative way averaging over the seven neighboring anodes (-3 — +3), excluding runaways. In each iteration only values in a window around the mean value, determined in the previous iteration, are included in calculating the new local average.

The window is decreased with each iteration to exclude more and more runaways. Ac-tually most runaways are the broken anodes and shorts we are looking for. As starting value for the first iteration we use the overall mean of all measured currents.

Due to the special shape of the substrate not every anode has the same length. The mea-sured current is to the first order proportional to the anode length, thus the signals vary, too (see Figure4.8top, “beam pipe clipping”). Especially in the region, where the lengths change strongly, the algorithm above described fails. Hence before building the local average, every signal is divided by its geometrical anode length taken from a table. Af-terwards the resulting local mean value is again multiplied by the anode length to get the correct expected signal.

From the comparison to these reference values for every single anode, we are ready to check for different kinds of defects:

• A signal below 95% of the expected current indicates a broken anode. This ratio provides an estimation of how much of the anode is still intact, and the position of the anode break.

• Two adjacent, broken anodes are considered as a potentially broken cathode.

• A tester channel connected to two anodes sees a doubled signal. Hence a signal which is 1.2 times higher than the expected one points to a misaligned wafer in the tester.

• An anode–cathode short leads to high current in the affected anode. Additionally the 15 neighboring anodes belonging to the same cathode group show almost no signal due to the voltage drop over the protection resistor. Thus a signal of over 99%

of the measurement range, combined with an expunged signal in the neighboring anodes clearly indicates an anode–cathode short.

Every measured wafer gets a new, increased number. Even if a wafer is measured a second time after a repair, it gets a new number. This ensures that none of the wafers get the same number. The raw and the analyzed data is saved as ASCII-data to a file. The file format is described in AppendixB.