[evlatests] 3-bit tests -- 7 Sept.

[Rick Perley]

    With a little more time this morning, I reviewed the results from 
yesterday. 

    The most mysterious aspect to me is the overall much higher rms 
values for all baselines.  In our tests from last week, the LCP side 
showed noise 'close' to the expected values.  In particular, the 
baseline 15 x 22 in LCP was consistently at about 0.17 Jy - close to the 
expected value of 0.16. 

    But from yesterday's test, this baseline was consistently at 0.34 Jy 
for all power levels!  If we wish to blame the weather, we are asking 
for an additional 30K of noise power -- I didn't think that we can get 
this at a frequency of 5.8 GHz from even a thick thunderstorm.  As it 
turns out, it was not raining at the time of this test -- checking 
wunderground.com for the LWA site (they have a tipping raingauge -- 
very  useful!) KNMSOCOR9, I see it was raining between 1:30 and 2:30 
PM.  But our test was an hour later than that -- 21:04 -- 21:40 IAT = 
15:04 -- 15:40 PM local time. 

    We should repeat this test when the weather is better, and (very 
preferably!) with a parallel 8-bit observation at the same frequency, or 
at least an 8-bit test at the same frequency immediately preceding or 
following...

   

Barry Clark wrote:
> The correlation efficiency is a multiplicative effect, so if the system
> temp is high, it will show up as the same percentage.
>
> I agree, 6dB too little power into the digitizer is like going from
> a three bit to a two bit sampler, so a 10 or 12 percent effect.  But
> 6dB too high power should result in a sigma about equal to the edges
> of the good bins, or about 30% of samples falling in the overflow bins,
> which act like one bit samplers, and their contribution is weighted
> up by nearly a factor of two, so more like a 30% effect.  But we knew
> from the state counts that things get wonky at high power anyhow.  At
> a guess, we are being saved by the zero offset - there are no overflow
> high times overflow low products, so the overflow bins do not, contrary
> to theory, act like one bit digitizers, but get scraped off into the DC
> channel, which we, sensibly, ignore.
>
> Rick Perley wrote:
>   
>>     Ken/Michael arranged to supply the samplers with a range of input 
>> powers, from -38 dBm to -26 dBm in 5 levels (i.e. steps of 3 dB). 
>>
>>     I've reviewed the data.  The power changes are easy to see in the 
>> gain solutions.  But the curious (disturbing?) result is that there is 
>> no effective change in rms noise in the 'blank sky' histograms between 
>> the max and min powers.  The other (disturbing?) effect is that both 
>> polarizations are much noisier than in the test of last week -- 
>> typically .25 to .35 Jy rms, when 0.16 is the expected value.  Two 
>> baselines (22 x 27 and 22 x 28) in RCP have noise over 0.5 Jy. 
>>
>>     Ken notes is was raining at the time, so perhaps the elevated noise 
>> is from excess cloud emission.  We think that, over the range in power 
>> level, the expected effect might be ~10% of the proper noise -- say, .02 
>> Jy, which is difficult to see in the high noise. 
>>
>>     We should try again tomorrow, in hopefully better weather
>>
>>    
>> _______________________________________________
>> evlatests mailing list
>> evlatests at listmgr.cv.nrao.edu
>> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>>     
> _______________________________________________
> evlatests mailing list
> evlatests at listmgr.cv.nrao.edu
> http://listmgr.cv.nrao.edu/mailman/listinfo/evlatests
>