[evlatests] 3-bit calibration

[Vivek Dhawan]

A summary, as preamble to tomorrow's discussion.

There is good news:

1. 8-bit digitizers look good in the calibration tests, just as they do
    in the SNR tests.

2. 3-bit digitizers: The characteristics are perfectly repeatable over
    +_3 dB T_sys (total power) variation, if and only if the T304 atten-
    uator and equalizer are held constant. This is true of all working
    units.

The other news:

1. There is diversity in properties across various units, and only faint
    resemblance to theoretical 3-bit devices.

2. While most of the digitizers do have a stable characteristic, there
    are several that change despite 'nominally identical' setup conditions.

    The reason still eludes me. Plots will be shown and ideas solicited.

3. Frequency structure

    Variation within a baseband - The family of 16 subbands from a sampler
    baseband are somewhat alike, but each really needs its own polynomial.

    Variation between RF bands (e.g. K,Q tested so far) - many samplers
    appear to change little between bands, which is great. But some do show
    a change. Attempts to relate this to variations in the (2GHz) baseband
    appear fruitless.

    Along the same lines, another variable is which part of the passband
    is used in the T304 IF converter. This has not yet been explored
    systematically (only incidentally, as when changing bands K/Q).

4. All these tests were with bit levels 1,2,3. We have not thought about
    the other choices yet.


I have one more thing to try: Setup a series of tipping curves, and
manually vary the equalizer and/or attenuators on a few units, to try
and relate these to changes in sampler characteristic.




Rick Perley wrote:

> 
>     2) PDif Compression with the 3-bit system. 

>     10AM, Room 317, 09 February.