Results from AFEB reliability test.

The CSC AFEB boards are the 16-channel anode preamplifier/shaper/discriminator boards used in the CMS Endcap Muon CSC Chambers (see details about the anode front end electronics in the Review, see also CMS anode front-end project at CMU and US CMS).

Since Nov. 20, 2000 we have had the 99 boards (with CMP16F chips on them) heated in an oven at the temperature of 110 C during 5.4 months. The boards are powered and the discriminator thresholds are minimum (default at power on) to keep the boards oscillating. The goal of this test is to check the reliability of the anode front-end boards. According military standard the time of work of the board heated by 20 C degrees above the operation temperature is equivalent to about the doubled time of work at this temperature. Therefore one week of work at 110 C degrees corresponds to 16 weeks of operation at 30 C degrees, and, correspondingly, 5.4 months of heating - to the 7 years of operation time.

The board ID numbers are 91, 97-176, 178-179 and 235-250. We were checking the boards periodically on the ASIC test stand at Fermilab looking for dead channels and measuring the threshold, noise, gain, discriminator offset, time resolution and the slewing time. The measurements have been taken in the runs with one week interval between them (Runs 1 - 4), two weeks (Runs 5 - 6), one month (Run 7) and 2.5 months (Run 8). Run 0 has data taken before the beginning of the test. It is used for the reference. Note that the board #130 has had channel #5 dead since Run 0.

There were no permanently dead channels found in the process of about 3 months long heating of almost 1600 channels. However, in all but Run 1 runs one board was found with only one channel having no output signal in the threshold curves measurements:

        -----------------------------
        Run    Board ID    Ch #       
        -----------------------------
            0        100           9
            1       none
            2        250           1
            3        148          16
            4        114          16
            5        128          16
            6        142          16
            7        122          16
        -----------------------------

The boards above have different and even IDs and have one and the same not working channel number (#16) in the last 5 runs. There are two slots on the ASIC stand to measure two boards simultaneously and the even boards have been placed almost always in the right slot, the odd boards - in the left slot (or vice versa). The time measurements for the same channels were screwed up as well. All this suggests a bad contact in the cable but why it happens only once in the 49 measurements in each run - a question to the authors.

Another feature of the threshold measurements - a small "inefficiency" of channels on the plateau of the threshold curve where we expect to get always 400 counts corresponding to the 400 input test pulses. The fraction of such channels is less than 2% for about 1600 channels, measured in runs 0-7 (see page 1 of Fig.1 (*.ps), Fig.1 (*.gif) ; to look at all pages of picture in *.gif format click anywhere in the opened "ImageMagick" window, go to "File" in the opened "Commands" menu and click "Next"). The number of boards with such observations vs board ID and the run # is presented on page 2 of Fig.1 (*.ps), Fig.1 (*.gif) . It was minimum in runs 2,6 thanks to the better grounding. There is a clear excess of such events in Run 5 and for boards ## 110,123,138 and 149 (see page 3 of Fig.1 (*.ps), Fig.1 (*.gif)). The "inefficiency" is due to the way the DAQ was set up. The multi hit TDC have been readout in the single hit mode. For the oscillations it was the last detected hit which has come outside the presetting window used to count the output pulses.

The parameters obtained from the data in run 0 for each channel in each board were used as a reference. We calculated the differences between the reference parameters and their values measured in runs 1 - 7. The results have been averaged over 16 channels for each board in each run. We also looked at the maximum (as the most positive) and minimum (as the most negative) values of these differences in each board (see example for Board 101 (*.ps), Board 101 (*.gif) ). Two boards, 201 and 202, were kept outside of oven and were measured almost each time (in Run 0 and then starting from Run 4) together with the tested boards to watch the stability of the measurement conditions, see Board 201 (*.ps), (*.gif) and Board 202 (*.ps), (*.gif) . To monitor the variations versus heating time the results were plotted for 99 boards board by board:

- Fig.2a (*.ps), Fig.2a (*.gif) for changes in the threshold,

- Fig.2b (*.ps), Fig.2b (*.gif) for changes in the noise,

- Fig.2c (*.ps), Fig.2c (*.gif) for variations in the gain,

- Fig.2d (*.ps), Fig.2d (*.gif) for discriminator offset changes,

- Fig.2e (*.ps), Fig.2e (*.gif) for changes in the calculated chip internal Cinj.

The threshold, noise, gain and offset distributions for all channels together from reference Run 0 and Run 7 (after 3 months of heating) are here:

- Fig.2f (*.ps), Fig.2f (*.gif) for threshold and noise,

- Fig.2g (*.ps), Fig.2g (*.gif) for gain and discriminator offset.

The boards were practically stable during 3 months of heating at 110 C degrees. The deviations of the thresholds are less than +-2 fC. The noise remains almost the same. There are some noticeable deviations of parameters of a few boards in some runs mainly due to operator's errors. For example, data for board 126 were mismeasured in Run 0. Boards 136-145 in Run 7 have been powered up at the wrong settings of the low voltage power supply. Boards 161-162 have stable thresholds in Runs 1-7 and they are all different from Run 0.

The threshold, noise, gain and offset distributions for all channels together from reference Run 0 and Run 8 (after 5.4 months of heating) are here:

- Fig.3a (*.ps), Fig.3a (*.gif) for threshold and noise,

- Fig.3b (*.ps), Fig.3b (*.gif) for gain and discriminator offset.

Note for Run 8 the shift in the offset and gain resulting in the shift of the threshold by 3 fC. The monitoring boards 201 and 202 show the same effect. The data in Run 8 were taken on the new test stand having different hardware. In particular a new pulse generator was used with the pulse rising time of 10 nsec instead of 40 nsec as it was in runs 0-7. Indeed, using old generator on the new stand we observed the shift of the threshold of 2 fC. Therefore the data from Run 8 also confirm the stability of parameters of the boards after 5.4 months of heating. No new dead channels were found.

For reminding, we define the slewing time as the difference between max. and min. mean time of the AFEBs output signal in the interval of the input charge 50 < Qin < 600 fC. Data were taken at Ud = 1400 mV and attenuations of 6 dB and 16 dB. Two attenuators were needed in order to take into account the contribution of the pulser slewing time (see example in Fig.4 (*.ps), Fig.4 (*.gif) ).

- Fig.4a (*.ps), Fig.4a (*.gif) present the mean time and the slewing time distributions in reference Run 0 and Run 7 (after 3 months of heating). The small peaks at 110 ns and 135 ns in mean time distributions are due to mismeasured boards 126 in Run 0 and 136-145 in Run 7. Zero slewing time data in Run 0 are due to board 126 as well. We do not present data from Run 8 because these data don't have correction for the generator contribution to the slewing time.

The time resolution was measured as RMS of the peak in the time distribution at given Qin.

- Fig.5a (*.ps), Fig.5a (*.gif) give the RMS distribution for all boards at Qin = 50,100 and 150 fC in Run 0 and Run 7.

- Fig.6a (*.ps), Fig.6a (*.gif) the same in Run 0 and Run 8. The time resolution in Run 8 turned out to be slightly better than in Runs 0-7 because of use the new generator with the test pulse rise time of 10 ns.

After 5.4 months of heating at 110 C no dead channels were found in all 99 boards (1584 channels total). The parameters of boards were stable.