From sragland@coolstars.cfa.harvard.edu Wed Jun 18 08:59:33 2003 -0400 Date: Wed, 18 Jun 2003 08:59:32 -0400 (EDT) From: Sam Ragland To: John Monnier cc: pschuller@cfa.harvard.edu, Jean Philippe Berger , carleton@cfa.harvard.edu, epedretti@cfa.harvard.edu, mlacasse@cfa.harvard.edu, rafael@ipac.caltech.edu, schloerb@fcrao1.astro.umass.edu, Wes Traub Subject: Re: Origin of readout dependent closure phase offset Hi John, I noticed this 'readout dependent instrumental closure phase' about a couple of months back and had a discussion with Wes earlier. We were speculating various possibilities including 'change of delay between pixels'. I was arguing against this possibility. Let us assume (for simplicity) that the ADC sampling interval is 10 microseconds, and consider only three PICNIC pixels (0,2,4) out of six outputs. Case 1: 1 loop & 2 reads ------- ------------------------------------------------------- Pixel 0 pixel 2 pixel 4 ------------------------------------------------------- Sampling 0,10 40,50 80,90 time (microsec) (2 reads) ------------------------------------------------------- Mean samp. 5 45 85 time (microsec) -------------------------------------------------------- Delay between pixels: 45 - 5 = 40 microseconds Case 2: 2 loop & 2 reads ------- ------------------------------------------------------------ Pixel 0 pixel 2 pixel 4 ------------------------------------------------------------ Sampling 0,10 40,50 80,90 (loop 1) time (microsec) 120,130 160,170 200,210 (loop 2) ------------------------------------------------------------ Mean samp. 65 105 145 time (microsec) ------------------------------------------------------------ Delay between pixels: 105 - 65 = 40 microseconds I get the same delay irrespective of # of loops & reads. Am I missing something? Sam ---------------------------------------------------------- Sam Ragland Harvard-Smithsonian Center for Astrophysics (MS20) 60 Garden Street Cambridge MA 02138 Tel: (617) 495-7158 (W) (617) 389-7461 (R) Fax: (617) 496-0121/ 617 495-7467 E-mail: sragland@cfa.harvard.edu; sam_ragland@yahoo.com ----------------------------------------------------------- On Tue, 17 Jun 2003, John Monnier wrote: > > This may be obvious to some, but I thougth I would send this to the wider > audience. (Sam: can yo put this on the webpage). > --- > > Statement of problem: > > Measure clsoure phase depends on readout pattern. > Example: > > readout Cphase (degs) > 1r7l 141 > 2r7l 160 > 3r7l 167 > 4r7l 169 > > Explanation: > With only 1 loop, there is ~ t_int/6 (sample/6) delay between pixels (or > sample/3 between beam combiners). For the 'fast' fringe which only has 4 > samples per fringe, then this amounts to 1/12 fringe, or ~30degrees. For > 2 loops x reads, then this effect is half has bad.. so the difference > between 1l xreads and 2l x reads is ~ 15 degress, in agreement (ballpark) > with observations. This also means that the difference gets progressively > less.. this is waht is observed by me: > > W/o taking into account the other fringes (8pixels/fringe), we would > predict differences similar to the seen above (within a few degrees). This > also says the TRUE instrumental closure phase is the one that corresponds > to a readout with many loops (where time differences becomes negligible) > (~170 degrees in the above example -- I note that there is a possible 180 > degree ambiguity in the above at the current level of analysis). This may > be important when deriving a color calibration. > > Note: This effect would not happen if the piezos were really STEPPING > between pixels, which is how the ADC is programmed. This readout dependent > closure phase suggests that the actual motion of the piezo ramps are > smooth during the whole scan, the steps are so small and sharp that the PI > servo can not hope to correct this (there is low frequency filter, both > electronically and probably mechanically that makes the ramp smooth and > linear. Not surprising I guess. > > Comments are welcome. > ------------------------------------------------- > John D. Monnier, Assistant Professor of Astronomy > University of Michigan 941 Dennison Building > 500 Church Street Ann Arbor, MI 48109-1090 > monnier@umich.edu 734-763-5822 (FAX 734-763-6317) > From sragland@coolstars.cfa.harvard.edu Wed Jun 18 08:59:33 2003 -0400 Date: Wed, 18 Jun 2003 08:59:32 -0400 (EDT) From: Sam Ragland To: John Monnier cc: pschuller@cfa.harvard.edu, Jean Philippe Berger , carleton@cfa.harvard.edu, epedretti@cfa.harvard.edu, mlacasse@cfa.harvard.edu, rafael@ipac.caltech.edu, schloerb@fcrao1.astro.umass.edu, Wes Traub Subject: Re: Origin of readout dependent closure phase offset In-Reply-To: Message-ID: References: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Status: R X-Status: X-Keywords: Hi John, I noticed this 'readout dependent instrumental closure phase' about a couple of months back and had a discussion with Wes earlier. We were speculating various possibilities including 'change of delay between pixels'. I was arguing against this possibility. Let us assume (for simplicity) that the ADC sampling interval is 10 microseconds, and consider only three PICNIC pixels (0,2,4) out of six outputs. Case 1: 1 loop & 2 reads ------- ------------------------------------------------------- Pixel 0 pixel 2 pixel 4 ------------------------------------------------------- Sampling 0,10 40,50 80,90 time (microsec) (2 reads) ------------------------------------------------------- Mean samp. 5 45 85 time (microsec) -------------------------------------------------------- Delay between pixels: 45 - 5 = 40 microseconds Case 2: 2 loop & 2 reads ------- ------------------------------------------------------------ Pixel 0 pixel 2 pixel 4 ------------------------------------------------------------ Sampling 0,10 40,50 80,90 (loop 1) time (microsec) 120,130 160,170 200,210 (loop 2) ------------------------------------------------------------ Mean samp. 65 105 145 time (microsec) ------------------------------------------------------------ Delay between pixels: 105 - 65 = 40 microseconds I get the same delay irrespective of # of loops & reads. Am I missing something? Sam ---------------------------------------------------------- Sam Ragland Harvard-Smithsonian Center for Astrophysics (MS20) 60 Garden Street Cambridge MA 02138 Tel: (617) 495-7158 (W) (617) 389-7461 (R) Fax: (617) 496-0121/ 617 495-7467 E-mail: sragland@cfa.harvard.edu; sam_ragland@yahoo.com ----------------------------------------------------------- On Tue, 17 Jun 2003, John Monnier wrote: > > This may be obvious to some, but I thougth I would send this to the wider > audience. (Sam: can yo put this on the webpage). > --- > > Statement of problem: > > Measure clsoure phase depends on readout pattern. > Example: > > readout Cphase (degs) > 1r7l 141 > 2r7l 160 > 3r7l 167 > 4r7l 169 > > Explanation: > With only 1 loop, there is ~ t_int/6 (sample/6) delay between pixels (or > sample/3 between beam combiners). For the 'fast' fringe which only has 4 > samples per fringe, then this amounts to 1/12 fringe, or ~30degrees. For > 2 loops x reads, then this effect is half has bad.. so the difference > between 1l xreads and 2l x reads is ~ 15 degress, in agreement (ballpark) > with observations. This also means that the difference gets progressively > less.. this is waht is observed by me: > > W/o taking into account the other fringes (8pixels/fringe), we would > predict differences similar to the seen above (within a few degrees). This > also says the TRUE instrumental closure phase is the one that corresponds > to a readout with many loops (where time differences becomes negligible) > (~170 degrees in the above example -- I note that there is a possible 180 > degree ambiguity in the above at the current level of analysis). This may > be important when deriving a color calibration. > > Note: This effect would not happen if the piezos were really STEPPING > between pixels, which is how the ADC is programmed. This readout dependent > closure phase suggests that the actual motion of the piezo ramps are > smooth during the whole scan, the steps are so small and sharp that the PI > servo can not hope to correct this (there is low frequency filter, both > electronically and probably mechanically that makes the ramp smooth and > linear. Not surprising I guess. > > Comments are welcome. > ------------------------------------------------- > John D. Monnier, Assistant Professor of Astronomy > University of Michigan 941 Dennison Building > 500 Church Street Ann Arbor, MI 48109-1090 > monnier@umich.edu 734-763-5822 (FAX 734-763-6317) >