Procedure to Take and Analyze Allan Variance Data 20 Mar 01 In OBST do BACKEND AOSA AOSB AOSC FBS (or just the one(s) you'll be testing). Normally, one would use the AOS noise source(s) for this test. Turn these on manually with the button(s) on the AOS rack in the computer room (next to the control room). The button for turning the internal noise source on is red and labeled "Noise On"; press it and it should light up. To turn the noise source off after the Allan Variance test, press the green button labeled "Noise Off". Alternatively, one might occasionally be asked to use the Hayward external test source in the chilled rack. In this case type @TEL$OBS:EXT_ALLAN in the OBST window on Frankenheim. This should set the IF switches to connect the Chilled Rack Noise Diode ("Test Source") to the backends and then PAUSEs. When finished using the external noise source for Allan Variance tests, type C in order to return the switches to their normal state. Note that whereas the Hayward test source provides the same noise to all spectrometers, it does so via the IF converter on the receiver level. Consequently, it is susceptible to temperature variations in both the computer room and the receiver room. Only very rarely will one be asked to use the IF passband of a receiver as the noise source for an Allan Variance test. In that case, one would choose a receiver (RX command in OBST), tune it as if you were about to start normal observations, and make sure that BE_FE is properly configured to have the backends looking at that receiver. When the Hayward test source is used, one may need to remove some attenuation in order to provide a decent noise level. On the SIS345 PC (HOST1) in the left receiver room, make a note of the attenuation in IF-A and IF-B (so they can be restored later) and set both to 0 dB. Also note that turning on the noise source on an AOS rack disables the input of the Hayward test source to that AOS. Consequently it is possible to use the Hayward noise source for some AOS's and the AOS internal noise source for others. Now you are ready to start taking data. In an SMTSYS window on Kronen, set your default directory: SET DEFAULT KRAW:[SMT.AOS] To collect AOS data to be used for Allan variances, type @AOSTEST and answer the questions (which AOS, how many samples, output file name, start time if any). This will submit a batch job which will start collecting and writing the data to directory KRAW:[SMT.AOS]. (Careful! If you ask for the maximum number of samples--100000--the test will go on for 55.5 hours!) Do this for each AOS you wish to collect Allan Variance data from. To collect filter bank data for Allan variances, type @FBSTEST and answer the questions; note that only the Hayward test source can be used for filter bank Allan Variance data. For the output from the AOS's the file name convention is .AOS where the is the month, day, and year and is A, B, or C. For the filter banks the file name convention is .FB. Normally, data will be taken every two seconds and it will take several hours to get enough data for an Allan Variance calculation. AOSTEST will normally type a comment that you may need to RUN a program BE1, BE2, or BE3. Normally, this will be necessary. After submitting the batch job with @AOSTEST, type RUN KDATA:[DMUDERS.AOSTEST]BE (where =1,2, or 3) for each AOS to be tested. When you RUN it, if the AOS is NOT idle, you will get the complaint MICRO IS NOT READY TO ACCEPT COMMANDS ! IGNORE? (N OR ) If you get this, strike . Normally, the next (first?) question asked will be ENTER BANK TO BE USED > Choose 1. Next a "MAIN MENUE" will be printed. Select COMMAND EXECUTION by typing a 6. The next prompt will be ENTER COMMAND > Type START. (This must be in upper case otherwise you'll get an UNKNOWN COMMAN complaint.) After a few seconds strike . Each time you do this, the 4 MICRO numbers will be updated. The MICRO I-FRAM should increment every 2 seconds or so. To exit the program, type 0 a couple of times. Do this for each of the AOS's from which you are taking Allan Variance data. While the Allan variance data is being taken, one should not do any other observing as the backends will not be connected to the receiver and commands from BEORGA to the backend(s) being tested will interfere with the tests. If you need to stop the test early, in an SMYSYS window find the batch jobs for taking Allan Variance data with SHOW QUEUE/BATCH These batch jobs will have job names like RASC_BATCH; filter bank batch jobs will have name FBS_BATCH. To delete one of these from the queue type KILL where is the entry number (first column from the show queue command). If the Hayward test source was used, when the tests are over, restore the IF-A and IF-B attenuators in the left receiver room to their original values. Then type C on OBST to reset the IF switches. To analyze the Allan Variance data, ftp the raw binary data files (file names .AOS and .FB) from KRAW:[SMT.AOS] to /home/smtop/aostests on Jever. The analysis program wants a text file with two columns (the counts of two AOS channels roughly 100 channels apart). In addition to that there is some commentary blurb at the beginning of the file and also a line with the (negative) dark current counts. To make such a file from the Vax binary file, you have to use the "al" program. The syntax is: al filename channel1 channel2 integ.time > outfile The naming scheme for the output file is .aos_ where is intnoise, extnoise, or sisxxx to indicate if the noise source was the internal noise diode, external (Hayward) noise diode, or SIS reciver xxx. For example, if the input Vax binary file transferred from Kronen were named something like 031301.aosc, then the program might work something like this al 031301.aosc 800 900 2 > 031301_800_900.aosc_sis230 The file 031301_800_900.aosc_sis230 could then be used for the Allan plot: allan 031301_800_900.aosc_sis230 When the plot first comes up, the total power counts of the first channel (800 in our example) will be plotted. To get the total power counts of the second channel (900 in our example), type 2 in the display window. These individual total power counts may drift over time but they should be the same for all channels (i.e. the bandpass just scales up or down but does not change shape). More meaningful is the Allan Variance of the difference of the two channels; this can be displayed by typing 0 in the display window. (0 is the most meaningful plot, as this cancels out most of the noise source variations and is closer to how real observations are reduced.)