[evlatests] WIDAR wobbles vs. time, contd.

[Barry Clark]

The fact that you see a weak signal at 9.6 Hz is interesting.  This is
the noise diode modulation, which is, it says here, a simple additive
noise that should not show up in correlations at all.  I can think of
three mechanisms for having it do so.  1). Radiation from one antenna
to another.  This seems unlikely to me. 2).  Some non-linearity in
the system.  My offhand candidate for the non-linearity would be
unequally spaced levels in the A/D.  From that source, 0.1% sounds
plausible, 1% does not.  3.  Introduction of a short spike which
actually correlates.  We have baselines of order a microsecond here,
so to correlate, the spike would have to last that long, ringing away
at 5 GHz.  Seems implausible - the required Q is too large.

Vivek Dhawan wrote:
> At this morning's meeting it was though worthwhile to dig a bit more.
> 
> For my Rupee's worth (~2 cents) this is not the cause of the imaging
> troubles, I think it averages out. But that's just a guess - it would
> be nice to squash it.
> 
> -----------------------
> 
> Data is C band, July 09 UT 21-22:30, 3C273 (~30Jy say).
> 
> a. The wobbles are non-closing (not antenna based), details below.
> 
> b. There are a variety of periods, a few Hz, quite a bit faster then the
>    fringe rate.
> 
> c. The periods do change over an hour, but I see no way to relate them to
>    the corresponding fringe rate changes.
> 
> d. There is no differential in the wobble across the band = it is _not_
>    a delay change = there is no delay jitter associated with the wobbles.
> Plot 1: red= phase wobbles, averaged over band. blue= phase jump at
>    delay change, differential between low/high half of band. time span =
>    10 sec, 40ms steps.
> 
> e. The noise-diode switching at 9.6Hz is seen at ~few % in the autocor
>    and is at about 0.1-1% of the cross-correlation amplitude on most
>    baselines. Assuming 30K Tsys, 3K noise injection, that is ~30Jy, so
>    the f_shift is suppressing the noise diode by a factor of >100 (1000
>    on most baselines), else it would be comparable to 3C273.
> 
> f. The 9.6Hz is strongest on all baselines to ant 19, which is an offender
>    at other wobble frequencies. Otherwise, I see no relation between the
>    strength of the 9.6 and the other modulations. i.e. no clear sign of
>    cross talk between antennas. I dont see how to rule out beat notes.
> 
> g. I see no way with astronomical or even radiated cross-talk, to make
>    the amplitude and phase signatures we see. Not sure of this argument,
>    but I think all these have to be a small vector rotating at a smooth
>    and Earth-related rate, perturbing a large signal vector. This gives
>    equal and orthogonal real/imaginary wobbles, not what we see.
> 
> Plot 2. = Example of modulation = power spectrum of visibility
>    variation. baselines are 2-18, 2-19, 18-19 (top left, Right, bottom)
>    Peak is off the top at ~300 units.
> 
> h. Details of which baselines have what: (most have 9.6Hz at 0.2% or so).
>    I think this says that the antennas with even slot IDs are worse -
>    right Michael?
> 
> 1-*   clean
> 2-18 ~1.9Hz 1% of cross corr peak on 3C273
> 2-19  3.7Hz 5%
> 2-28 ~1Hz   2%
> 3-*   clean
> 9-*   5Hz   weak ~0.2%
> 9-18  6.5Hz 1%
> 18-19 6.5Hz 2%
> 19-*  9.6Hz 1-2% noise diode
> 19-24 2Hz   2%
> 19-28 3Hz   8%
> 23-*  clean (23-3, 23-19, 23-25 have 9.6Hz 1%)
> 
> ----------------
> 
> Vivek Dhawan wrote:
>> Michael inflicted upon me 80 minutes of data taken at 40msec
>> integrations, 20GB, to look for changes in wobbles vs time, etc.
>>
>> 1. load data and index it in short scans - it makes subsequent
>>    tasks start quicker.
>>
>> 2. FRING on 10sec of data, zero rates, then apply to all data.
>>
>> 3. BPASS on 1min, apply to all.
>>
>> 4. Average up across band, plot phase/amp vs time.
>>
>> 5. Wobbles seen mainly between ant 2, 18, 19, 28; regardless of
>>    whether the delay and BPass are applied or not, so their
>>    details are not crucial.
>>
>> 6. FRPLT (= Fringe rate plot, Fourier transform of phase vs. time)
>>    of 1min of data was done every 10 min to see changes. The period
>>    of the wobbles does change in come cases by a factor 2 over 80min.
>>
>> 7. I need to relate the wobble period to the natural fringe rate,
>>    not yet done.  Anyone have a handy program?
>>
>> 8. I did find a more curious thing: The amplitude wobbles may be
>>    different from the phase wobbles on the same baseline! e.g.
>>    Amp 8%, phase 5deg (all pk-pk) on ant 19-28
>>    amp 1%, phase 8deg on 2-19.
>>
>>    Even more curious, the period of the amp wobbles (0.4 sec) is
>>    half that of the phase wobbles on the same baseline (19-28).
>>    In most of the cases seen, phase and amp have the same period.
>>
>> Heard enough? I have - see y'all tomorrow
>>
>> Vivek.
>>
>>
>>
> 
> ------------------------------------------------------------------------
> 
> This body part will be downloaded on demand.