To: E614 Collaboration
From: Robert Henderson
Date: 30 Nov 1999
Re: Preliminary cross-talk results


                    Preliminary cross-talk results
                    ==============================


Pulser studies of Preamp
------------------------

  Our preamp board has the following features:

  - Input connector, 0.1" pitch, no intermediate grounds.
  - Input HV resistors, 0.1" pitch.
  - HV capacitors, alternate sides and staggered, 0.2" pitch, buried 
    ground plane in middle of PCB.
  - Fan-in to VTX input, 0.1" pitch to 0.05" pitch.
  - The VTX, either a 16 or 24 channel unit.
  - Output connector, double row 0.1" pitch, ground pairs at ends and 
    middle only.

  Initial pulser setup seemed to show that preamp contributed almost 3%
cross-talk??!! Later studies with more reasonable input pulse shape did NOT
confirm this. Doug Mass did a variety of pulser tests on the preamp, his
results were posted and can be summarized as follows:

Features and their contribution: 

  - Output connector      => Very little contribution.
  - VTX circuit           => About 0.5" cross-talk, also small effect at
                             beginning and end of the dies.
  - Fan-in                => Very little contribution.
  - HV Caps               => Very little contribution.
  - Input HV resistors    => Very little contribution.
  - Input connector       => About 1% contribution.

  Doug modified a 16 channel preamp with a staggered 2 row 0.1" input
connector. He also managed to modifiy the gas-box output for a similar
staggered male connector. Pulser tests indicated this modification virtually
eliminated this contribution.


Chamber Tests of modified preamp with Fe55 source (Ar/Iso 25/75)
----------------------------------------------------------------

  Unfortunately, tests with the modified preamp did NOT show a large
reduction in cross-talk. I compared channels on modified (staggered) preamp
and adjacent standard preamps. I get 2.65% compared to 2.95%, NOT the 1%
reduction I hoped to see. I believe that the pulser tests were misleading,
since the VTX circuit contains a shaper circuit, so choise of input pulse
shape is difficult. Also pulser tests are nominally at 50 ohms input
impedance, while the detector might be 150 ohms or more. For these reasons,
tests with chamber pulses are the most significant.



                       New discovery from chamber
                       ==========================


  After the above tests, I looked more closely at the chamber outputs. I
noticed something that I'd overlooked. The 16 channel preamp has two
8-channel groups coming in, the 24 channels preamp has three such groups. I
looked at the cross-talk on adjacent channels which cross between groups.

*** The observed cross-talk was VERY small!!?? So small it's hard to measure,
    but I estimate a cross talk value of 0.16%. This is a factor of 18
    smaller than the 2.95% typically seen!

  The first qustion people will ask is "How can I see 0.16%?" After all, the
VTX alone contributes 0.5%! (published result). The answer is, that the 0.16%
is the summation of TWO effects:

1) Electronic cross-talk. This produces a SAME polarity pulse (but NOT an
   exact copy, it's frequency dependant).
2) Avalanche image pulse. When an avalanche occurs on a wire, it is a partial
   discharge of a HV capacitance. An image pulse of OPPOSITE polarity is
   induced on all nearby surfaces. i.e. on adjacent wires.

  So, when I see 0.16%, it just means that these two effects (both probably
about 0.8%) are cancelling. This is confirmed, when I look at cross-talk
between adjacent channels on entirely different preamps. Then I see about
-0.8% cross-talk. i.e. small opposite polarity pulse.

  I have asked Mahir to look at our beam test results. Specifically to see if
we see low CT between the 8-channel groups within a preamp. He will report
the results, but indeed there is greatly reduced CT, it had been masked
somewhat by dilution with CT from channels on `opposite side', that is NOT
across group boundary.

  The exciting conclusion of these observations is that we should be able to
achieve far better than 3%, perhaps less than 0.5%. Had I made this
observation earlier, we could have elminated the worry of CT on the preamp
boards! Cleary the CT contributions are earlier, that is inside the chamber.



                               Further tests
                               =============


  I had planned to do puler tests inside the chamber, but the new discovery
changed the plans. When we look inside the chamber, there are several regions
to consider:

1) Wire-to-wire crosstalk between sense wires.

   Calculations indicate this should be insignificant. Since I see NEGATIVE
gross-talk for sense wires across boundary between preamps, this is
confirmed.

2) First part of Lamel traces heading toward readout.

  Since all channels look the same at solder pads and nearby, this is NOT a
region of concern.

3) Output connector of detector, 0.1" pitch.

  Since the CT between 8-channel groups on the origional preamp is observed
at 0.16%, everything to do with the preamp is NOT a concern. This includes
the chamber output connector.


  Elinating these three regions, we are left with:


4) The `grounds' on lamel.

  There are grounds between channels, bigger islands beteen 8-channel groups 
and large islands between preamps. There is also a `ground bus' on the back
side of the lamel near the outer edge, connecting all the grounds.


5) The end regions of the lamel fan-ins.
   
  This is a likely region where the traces are about 2 mm apart instead of
the 4 mm at the wire pads. At the very edge of the lamel, there are wider
pads for the ribbon cable conections, here the gaps are even smaller,


6) The ribbon cables.

  Another likey region. The traces are 2 mm pitch and at the ribbon ends the
pads are wider and the gaps less. The ribbon cables terminate in small PCBs
and the female connector. Inside the female connector the metal-to-metal
distance is quite small.


7) The gas-box service PCB.

   This has 2 mm pitch male connectors, then the traces fan-out to 2.54 mm
(0.1") for the preamp input connectors. Somewhat less likely region.


Test 1
------

  I was particularly suspicious of the `ground bus' on the back of the lamel.
The lamel PCB is very thin, perhaps it was coupling channels, primarily those
within the 8-channel group.

  I couldn't just peel off this bus, since the grounds between channels in
the 8-channel groups would then be floating. Instead I installed the wire
plane we had produced with NO GROUNDS!

  I looked at CT on the B2 preamp (24 channel, unmodified, no stagger). I see
a CT value of about 2.8%, little different from the 2.95% mentioned above,
within the measurement error. 


Conclusion  It seems that the CT is slightly better without the HV `grounds',
----------  or at least no worse. Either they are doing nothing in the way of
            shielding or maybe the `bus' is having a negative effect that
            balances off any benefit of grounds between traces.

  This is more significant than it seems. If we can eliminate the `grounds'
or greatly reduce them, we have much more space at the readout arc. We could
certainly go to 2.54 mm pitch at the ribbons instead of 2.0 mm. Perhaps even
4 mm might be possible, but I'm uncertain without doing a complete layout.
Any increase in pitch at the end of the lamel fan-in and ribbon would be very
significant, since the small gaps would increase far more than a factor of
2.54/2.0=1.27, the small gaps might increase by a factor of say THREE.



Test 2
------

  I removed one ribbon cable opposite modified preamp (staggered). I
substituted a bare ribbon cable, i.e. no small PCB, no female connector. I
soldered this ribbon to the edge of lamel and the other end to the male 2 mm
connector on the gas-box service board.

  The cross-talk on the un-modified ribbon was still about 2.65% (staggered
connector). On the modified ribbon the CT dropped to about 2.23%.


Conclusion  There was a significant CT reduction by elimating the female 2 mm
----------  connector and it's PCB. This tends to confirm that regions 4)
            thru 6) are indeed the probelem areas.


Test 3
------

  I opened the chamber again and made the following modification. I put a
layer of kapton tape over the last 2/3 of the lamel fan-in and the
lamel-ribbon solder joints, to insulate them. Then I put down a layer of
copper tape over the Kapton. This copper tape covered the last 2/3 of the
lamel fan-in, the lamel-ribbon solder joints and about 2/3 of the ribbon
length. I soldered this copper tape to `ground'.

  The objective of this modification was to partially `shield' the channels
from each other. Of course, it is incomplete, since the other side of the
lamel is glued to the glass and could not be similarly shielded. Of course, I
expected more capacitance to ground, so might expect a PH reduction and
perhaps a rise-time increase.

  The result of this test was a further reduction of the CT on this group,
from 2.23% to 1.50%. The PH and RT appear unchanged.


Conclusion  There was a further significant CT reduction by partially 
----------  shielding regions 4) thru 6). This test again indicates these are
            the problem areas.

         

                             Future Test
                             ===========



  I think I will `butcher' the 16-channel region of the no-cgrounds lamel and
gas box. The idea would be to `dremmel off' this section of the lamel and
substitude a piece made here at TRIUMF. This section would be a simple PCB of
the same material, but with no plated-thru holes at edge. Similarly, I would
expose that arc of the gas-box PCB and modify it the match the new lamel
piece. Of course, we would not have the appropriate ribbon to join them, but
if we can use 4 mm pitch, I could `modify' existing spare ribbons by peeling
off every second one. They could be usable as-is if we had `grounds' between
adjacent signals.

  This test will take awhile, at least a week of work. There is little chance
this could be done by the dec 10/11 meeting at TRIUMF.



                              CONCLUSIONS
                              ===========


(1) The lamel fan-ins should be redesigned to seperate the channels as much
    as possible. 

(2) Depending on how that layout looks, we might drop the HV `grounds'
    entirely, or at least minimize them.

(3) We might want to think about using a multi-layer lamel to `bury' the
    signal traces between HV grounds. If we can significantly seperate the
    signal traces, this might not be necessary. I hope the latter is true.

(4) We should be able to get cross-talk down to less than 0.5%.

Created for the The Center for Subatomic Research E614 Project Projects Page.

Created by The CoCoBoard.