Test Bench Noise at UCSB

• After hooking up the new DAT to the testbeam module for the first time, the noise was at least one order of magnitude higher than what had been seen with the old DAT. The power and signal cable were separate and coiled up on the floor. This was the same configuration used with the old DAT. The cable shield was not connected. Our experience with the old DAT showed that the shield configuration made little or no difference.



• The shield was connected to earth ground. It is not clear from my notebook where this was done, but I think it was done at the HDI-link card. We saw no real improvement.



• We pulled the decoupling capacitors between System 5V and D5 local on the HDI-link card. This improved the situation. The noise was now a factor of 4-5 higher than with the old DAT.



• We put the System 5V for the DAQ-link and HDI-link cards on different supplies. No improvement.



• Disconnected the shield. No improvement.



• Pulled two capacitors on the matching card. My notes unfortunately are not clear as to which ones they were (Max Wilder from UCSC should know). I am sure that one side of these capacitors was connected to Analog 2. Anyway, after removing these capacitors the situation got noticeably worse.



• We switched back to the old DAT. The capacitors on the matching card were still missing. The noise level was about the same as we had previously measured with the old DAT. Therefore, the absence of these capacitors affected the new DAT, but not the old DAT.



• At this point the capacitors on the matching card were reattached, and we switched to the 3x3 module, since the noise levels on the two modules were comparable, and since we were afraid to damage the 7x7 module in the chaos that was quickly taking over all activities in the lab (!).



• We then played with the old DAT to see if we could find anything that made a real difference to the noise level in the hope of learning something about the new DAT. The noise levels were independent of the cable shield configuration (floating, tied at one end to earth ground, tied at both ends, etc. etc.). The metal dark box containing the detector, which had been previously been accidentally connected to earth ground through the case of one of the power supplies was isolated and left floating. Again, this made no difference with the old DAT. In all following measurements, this box was left floating.



• We connected the new DAT to the 3x3 module. In doing this, we made an attempt to have a cleaner connection to our bench top power supply system. Pairs of power and return wires between the HDI-link card and the power supplies were twisted, and kept away from power cords. Also, the power and signal cables going from the HDI-link card to the matching card were twisted together. The noise level was reduced. It was now within a factor of order 2 of the noise levels observed with the old DAT. It is not entirely clear from my notes what the shield configuration was, but I suspect that it was left floating.



• The shields on the power and signal cables were connected together at the matching card end. On the HDI-link card, the shield was connected to system ground (SVT common, earth ground). No real improvement.



• A section of the signal and power cables, which were twisted together, was tightly coiled up about three times, and two toroids were put around it. The noise was reduced by 15-20%.



• We reattached the decoupling capacitors between System 5V and D5 local on the HDI-link card, which had been removed previously. The situation improved considerably, and the noise was now comparable to the noise with the old DAT, but perhaps still of order 20% worse. Why putting back these capacitors made things better, while initially the removal of these capacitors had been found to be very useful, is not understood.



• We verified that the results were the same on both channels of the HDI-link card.



• Since the noise with HDI-link decoupling capacitors was lower than without these capacitors, we tried to increase the values of these capacitors. (The actual values of the capacitors that were added is not in my notebook, sorry about that). However, in this configuration the noise increased by about 30%. (Why ?)



• We put back the original capacitors, and verified that our results were reproducible.



• We made some changes to the bypassing capacitors on the PC matching card. This modification may have improved the noise by 10% or so, although it is hard to tell given the fact that we were only looking at about 50 channels, and small changes were hard to measure. At this point, the noise levels with the new DAT were comparable to the noise levels with the old DAT. A drawing from Max Wilder detailing these modifications to the matching card is shown in Figure 1.


Figure 1: Modifications to the matching card.


• The toroids were removed, and the noise increased by 15-20%. This is in agreement with the previous measurement. So, it definitely looks like the toroids help, and they were put back in the system.



• At this point, we switched back to the 7x7 test beam module, and did a scan on every channel. In Figure 2 and 3 I show the noise distribution with the old and new DAT. (Click here for details of the measurement procedure, and the conversion between Threshold DAC counts and SI units).
  
  
  
  Figure 2: Noise in Threshold DAC counts for all channels on the n- (phi-) and p- (z-) sides of the detector as measured with the new DAT. (Postscript version).
  
  
  
  
  
  Figure 3: Noise in Threshold DAC counts for all channels on the n- (phi-) and p- (z-) sides of the detector as measured on a typical threshold scan with the old DAT. (Postscript version).
  
  The noise with new DAT seems to be still about 10% worse than with the old DAT. It should be mentioned that the data was taken on two different days, so it is possible that the difference is due to different conditions in the lab. In the past we have seen day-to-day variations in our noise measurements at the 20% level. However, more recently, our measurements have been very reproducible.



• Finally, we shorted analog and digital grounds with copper tape on the HDI link card. The grounds were already connected at the power supply. This was an attempt to get a better connection downstream of the rat's nest of wires that brought the power from the power supplies to the HDI-link card. No effect was seen.

• The old DAT seems to be more robust.



• Power and signal cables should be twisted together.



• Toroids help.



• The noise is sensitive to the decoupling capacitors between System 5V and D5 local on the HDI-link card. This is something that people at the test beam should play with.



• The details of the shield configuration did not seem to make a great deal of difference in our test setup.



• The modifications to the matching card detailed in Figure 1 may help.

Return to UCSB test-beam page