follow-up on preamp/termination discussion

[Gerard Visser <gvisser@indiana.edu>]

Hi Mitch,
	I thought perhaps I should explain further, my thoughts on the 
termination. If you have any comments please do write back...
	The rationale for double-termination is as follows:
	1. It should almost completely eliminate reflections from prior pulses. 
With a line terminated only at the end, completely mismatched at the 
source, we may expect to have reflections probably of order of 5% of the 
pulse amplitude following the main pulse (after 3*100ft/c=305ns). With 
double termination these are reduced to 0.25%. [FYI Elton reminds us 
rates are 600kHz on inner channels of the detector.]
	2. The cost of doing this is just a small increase in the noise level 
due to the cable receiver amplifier (which now has to have 2x the gain). 
However, the preamp stage should hopefully be the dominant electronic 
noise source so the cable receiver's noise is negligible. In practice 
the receiver circuit I have prototyped has a noise of about 13 
nV/sqrt(Hz) so 92 uV in 50 MHz. This should be invisible in comparison 
to the preamp noise, given the full scale signal swing on the order of 
500 mV.
	3. Besides the reflection issue, I am pushing the idea that on one end 
of the line there be a high common-mode impedance to ground. The point 
is, ground voltage differences will otherwise cause significant 
common-mode currents to flow through the signal cables, and inevitably 
then also through the ground lines between the preamp boards and the 
chamber (which is the ground reference point for the preamp boards). The 
impedance of the gound connections there, translates this into a voltage 
source in series with the chamber capacitance, putting charge into the 
preamp inputs. This worries me (at least) - although Fernando states 
that in the CLAS setup they do not maintain a high common-mode impedance 
termination on either end, and don't have any spurious ground noise 
pickup problems. We did have such problems on AGS E896 drift chamber, 
but I don't know really that my high common-mode impedance plan would 
have mitigated them... I'm just being paranoid, not to have such problems.
	Anyway, since the preamp has a very limited voltage compliance on its 
outputs, the common-mode termination has to be set at the preamp board. 
Without using transformers that means a differential-mode termination as 
well. So, we are lead to use a double termination, by this 
consideration. [Ok, we could imagine no termination at the far end of 
the line, only high impedance... Rather nonstandard but possibly worth 
considering.]

	Specifically, assuming a 100 Ohm cable, I suggest that the preamp 
boards have a split termination (50 Ohms each) to 1.25V (and will need 
an opamp circuit to provide this 1.25V reference without a lot of wasted 
quiescent power). The ADC boards will have a 100 Ohm differential 
termination.
	The ASIC test boards should have a split termination (50 Ohms each) to 
1.25V, or a Thevinin equivalent using 4x 100 Ohm resistors. It should 
also have a differential termination using two equal resistors in 
series, with the midpoint accessible for probing (to observe the 
common-mode output which ideally is zero). At some point, we might want 
to study the ASIC pulse response to some standard input pulse, as a 
function of the common-mode termination voltage. To that end, it would 
be better if there was a 1.25V termination supply (that could be 
adjusted), common to all channels of course, rather than the Thevinin 
equivalent.
	[But I acknowledge that a hand-wired quick test setup probably won't 
have all that on all channels.]
	The preamp board test fixture probably uses the real cable (full 
length) and has at the end either the real ADC board or a test 
termination board and scope connection. That termination board ideally 
should be able to look at the common-mode voltage, i.e., should use two 
equal resistors in series for the differential termination.
	Thanks,

	Gerard