BEE2 comments

back to main ATA page

from Gerry Harp      11/19/2007
----------------------------------------------------------------------------------
Hi Ed

I have some more info for you from Billy Barott. My question are in 
asterisks, his answers below. I was wrong about the IBOB's performing a 
first level channlization on the data. There is a coarse sample delay in 
the BEE2 followed by a FIR filter.
I'm also attaching a figure. It reads from the bottom to the top.
Gerry Gerry- responses below. ===================================== 1. What processing is done on the IBOB? E.g. does it perform the coarse delay? ===================================== The ibob "just" digitizes at 838, and down convertes to i & q at 104 mega samples per second. Coarse delay is in the BEE2. ===================================== 2. How long is the FIR filter you apply to the data for delays? I guess there is a separate filter for each channel? ===================================== Fine/fractional delay is implemented with the FIR filter, with one filter for each sky signal. It is currently a complex filter with 8 coefficients. I have selected a complex filter in the hopes that it will give us greater flexibility in nulling experiments than a purely real filter. ===================================== 3. This is longer. How do you calibrate? For the 4x correlator, there are 6 cross correlations. Do you find the largest eigenvalue of this correlation matrix to determine the calibration? ===================================== First, we're only using one reference antenna, so for the 4x correlator there are only 3 cross correlations. I apply a simple slope/intercept estimation of the cross correlation spectra to determine the delay and phase offset between the test antenna and the reference antenna, and apply this data as a correction to the test antenna. ===================================== 4. After calibrating the 4 antennas, do you correlate the sub-beams at the next level (and the next) to get their calibration? Or what is the algorithm? ===================================== Conceptual correction: Each low-level FPGA passes a time-domain copy of the reference antenna to its output, in addition to the summed beam. So each summing-level FPGA receives three sub-beams and three reference antennas. The summing level FPGA selects its own reference antenna from these three, and calculates two useful cross correlations. The results of these correlations are applied to all antennas on the "test" FPGAs. The new reference antenna and summed beam are passed to the output to go to the next level. You might find a diagram helpful. I'll send you a draft of one that I've been working on to explain this process. Billy Ed Thelen wrote: > To Garry Harp - SETI > > I am a retired engineer still fascinated with techie things > and ATA is a techie heaven. > > I've been to ATA twice, and learned some about > the physicals and electronics - > making web pages to aid my never good memory. > > http://www.ed-thelen.org/ATA/HatCreekATA.html > > > I've bumped into a barrier about the beam former. > > ---------------------------------- > > The web page > http://www.ed-thelen.org/ATA/HatCreekInfoPlease.html#DataPath > follows the data path. > > I think I've got it reasonably well through the mixers, > and likely into a BEE2 platform, > where it disappears into flexibility - or chaos ;-)) > What goes on where in the BEE2s is > controlled by FPGA setup and PowerPC programming. > > (An FPGA based variable FIFO does not seem a stress if > the clock rate can be maybe 1/10 of the 800 MHz sampling rate. > > Unfortunately, I do not understand Digital Down Converters - > http://en.wikipedia.org/wiki/Digital_Down_Converter > > But, if I understood Prof Geoff Bower of Berkeley correctly, > > http://www.ed-thelen.org/ATA/HatCreekATA-StrugglingOn.html#BerkeleyOct22 > there is an analog output of the fine resolution phase shifting. > > I do not understand how a BEE2 would: > - get delay resolution less than one clock, ? ratioed between > digital filter taps? > - generate an analog output of the fine resolution as stated in some > reference? > > Your assistance is requested - > I offer a dinner at a restaurant of your choice > while receiving this presumed 1 hour of tutoring. > (No $700 wine please ;-) > or some other suitable thankyou - (assuming this ATA information is > not confidential) > > > Thanks much > Ed Thelen > > > P.S. Is there some publically available document/web site > about the goings on in the ATA beam forming BEE2s ? > I think I've read all the papers available on the web > from SETI and Berkeley RAL and some others around the > world that GOOGLE hit. > You are a busy guy! > > I didn't spot a sidereal clock, nor obvious beam > forming/tracking control station at ATA. > I presume the time critical delay calculations and updates > are handled at ATA - maybe another BEE2? > > Sorry, rambling - > > P.P.S. I leave compensating for variable data path delays > to my betters. > This is one heck of a big tough field!! > Makes system software seem as child's play.