Project

General

Profile

Zeropoint trends for Nov 1-19

I started the analysis with i-band, so the i-band results are presented first
(with full documentation). The results from the rest of the standard DES filters
(grzY) are then presented, and then finally the results from the DECam u-band.

i-band

  • Form of the Photometric Equation:
  i_std = -2.5log10(counts[ADU]/sec) - a_i - k_i*X - b_i*( (i-z)_std - 0.09) )

  where:
    i_std is the standard, or calibrated, magnitude of the star
    (i-z)_std is the standard, or calibrated, i-z color of the star
    a_i is the photometric zeropoint in i (one for each CCD)
    k_i is the first-order extinction in i
    b_i is the instrumental color term coefficient in i (one for each CCD)
    X is the airmass
    0.09 is a "fiducial" reference color for stars in i-z (based on SDSS experience)

    Note that there is a separate photometric zeropoint (a_i) and 
    instrumental color term coefficient (b_i) for each CCD.
  • Note that the solved values for the a's and k are highly correlated. Therefore,
    it makes sense to calculate an effective zeropoint for at a typical DES airmass
    (say, X=1.3) as "a + k*X" (for a given CCD).
  • Immediately following this bullet point is a plot of the effective zeropoint
    (at X=1.3 airmasses) vs. Night for one of the two central CCDs (CCD 35) relative
    to that effective zeropoint for the first night of DES SV observations (November 1).
    The error bars are calculated as 1.4xRMS of the night. Nights 20121107, 20121108,
    and 20121109 have non-photometric solutions mostly due to cirrus. Note that the
    values on the Y-axis are consistent with 0.00 at slightly larger than the 1-sigma
    level.

  • Immediately following this bullet point is the histogram of the Y-axis of the
    above plot. If we exclude the 3 points from the 3 non-photometric nights,
    the RMS standard deviation of the remaining values is 0.010 mag.

  • Now that we have shown the stability of CCD 35 effective zeropoint over time, now
    we need to show that zeropoints of the other CCDs across the DECam focal plane
    are stable relative to CCD 35 over time. In other words, how stable is the
    photometry across the DECam focal plane over time? Is it consistent over time,
    or is is variable? To do this, we want to compare the effective zeropoint of
    CCD n to CCD 35: (a_n + k*1.3) - (a_35 + k*1.3). We want to compare this
    quantity relative to its value on the first night of DES SV observing (Nov 1)
    -- i.e., relative to (a_n + k*1.3)_Nov1 - (a_35 + k*1.3)_Nov1. Or, altogether:
    [(a_n + k*1.3) - (a_35 + k*1.3)] - [(a_n + k*1.3)_Nov1 - (a_35 + k*1.3)_Nov1].
    Since the value of the first-order extinction, k, is the same for all CCDs
    for a given night, this measure simplifies to (a_n - a_35) - (a_n - a_35)_Nov1.
    In the following figure, we plot these values for all CCD's independently, so
    there are 62 points per night. Note that, aside from a couple bad nights
    (like the non-photometric 20121107 and 20121108) and aside from the outlier
    CCD 61 in nights 20121108 and following, the full spread of points on a given
    night is typically 0.02 mag or less.

  • Now let's plot the above figure in a slightly different way. Here, we plot
    the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov1" vs. night. We exclude
    CCD 61, to avoid this extreme outlier from affecting the results. To give
    a sense of the spread of points, the error bars are plotted using the
    RMS standard deviation (per observation RMS), not the error in the mean
    (i.e., the RMS standard deviation was not divided by root-N).

  • Immediately following this bullet point is the histogram of the Y-axis of the
    above plot. The RMS standard deviation of the distribution is 0.007 mag, and
    this value INCLUDES the 3 non-photometric nights.
    (Be sure to note that the
    values along the X-axis tick marks are multiplied by 10^{-2}.)

g-band

  • Form of the Photometric Equation:
  g_std = -2.5log10(counts[ADU]/sec) - a_g - k_g*X - b_g*( (g-r)_std - 0.53) )

  where:
    g_std is the standard, or calibrated, magnitude of the star
    (g-r)_std is the standard, or calibrated, g-r color of the star
    a_g is the photometric zeropoint in g (one for each CCD)
    k_g is the first-order extinction in g
    b_g is the instrumental color term coefficient in r (one for each CCD)
    X is the airmass
    0.53 is a "fiducial" reference color for stars in g-r (based on SDSS experience)

    Note that there is a separate photometric zeropoint (a_g) and 
    instrumental color term coefficient (b_g) for each CCD.
  • The same plots as for the i-band are presented below for the g-band, and in the
    same order. Nights 20121107, 20121108, and 20121109 are treated as non-photometric,
    just as for i-band.
  • Here is the plot of the effective zeropoint relative to the effective zeropoint
    for Nov 1 vs. Night for CCD 35. Note that the values on the Y-axis are consistent
    with 0.00 at about the 1.5-sigma level, and consistent with each other at slightly
    larger than the 1-sigma level.
    (The discrepancy is due to the Nov 1 effective
    zeropoint being slightly higher than average.) Note that 20121104 residuals have large
    scatter due to a strange g-band PSM solution for a night with otherwise good solutions
    (see, e.g., this residual plot for g-band for this night).

  • Here is the histogram of the Y-axis from the above plot. If we exclude the
    3 points from from the 3 non-photometric nights and the strange g-band result
    from 20121104, the RMS standard deviation of the remaining values is 0.019 mag.

  • Here is the plot of (a_n - a_35) - (a_n - a_35)_Nov1 for all 62 CCDs plotted
    independently. Aside from a couple bad nights and aside from the outlier
    CCD 61 in nights 20121108 and following, the full spread of points on a given
    night is typically 0.03 mag or less (a bit worse than for the other DES filters).

  • Here is the plot of the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov1"
    vs. night (with CCD 61 excluded). As with i-band, the error bars are plotted
    using the RMS standard deviation (per observation RMS), not the error in the mean.

  • Here is the histogram of the Y-axis of the above plot. The RMS standard deviation
    of the distribution is 0.007 mag, and this value INCLUDES the 3 non-photometric nights
    and the strange g-band solution from 20121104.

r-band

  • Form of the Photometric Equation:
  r_std = -2.5log10(counts[ADU]/sec) - a_r - k_r*X - b_r*( (g-r)_std - 0.53) )

  where:
    r_std is the standard, or calibrated, magnitude of the star
    (g-r)_std is the standard, or calibrated, g-r color of the star
    a_r is the photometric zeropoint in r (one for each CCD)
    k_r is the first-order extinction in r
    b_r is the instrumental color term coefficient in r (one for each CCD)
    X is the airmass
    0.53 is a "fiducial" reference color for stars in g-r (based on SDSS experience)

    Note that there is a separate photometric zeropoint (a_r) and 
    instrumental color term coefficient (b_r) for each CCD.
  • The same plots as for the i-band are presented below for the r-band, and in the
    same order. Nights 20121107, 20121108, and 20121109 are treated as non-photometric,
    just as for i-band.
  • Here is the plot of the effective zeropoint relative to the effective zeropoint
    for Nov 1 vs. Night for CCD 35. Note that the values on the Y-axis are consistent
    with 0.00 at slightly larger than the 1-sigma level.

  • Here is the histogram of the Y-axis from the above plot. If we exclude the
    3 points from the 3 non-photometric nights, the RMS standard deviation of the
    remaining values is 0.016 mag.

  • Here is the plot of (a_n - a_35) - (a_n - a_35)_Nov1 for all 62 CCDs plotted
    independently. Aside from a couple bad nights and aside from the outlier
    CCD 61 in nights 20121108 and following, the full spread of points on a given
    night is typically 0.02 mag or less.

  • Here is the plot of the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov1"
    vs. night (with CCD 61 excluded). As with i-band, the error bars are plotted
    using the RMS standard deviation (per observation RMS), not the error in the mean.

  • Here is the histogram of the Y-axis of the above plot. The RMS standard
    deviation of the distribution is 0.005 mag, and this value INCLUDES the 3
    non-photometric nights and the strange g-band solution from 20121104.

z-band

  • Form of the Photometric Equation:
  z_std = -2.5log10(counts[ADU]/sec) - a_z - k_z*X - b_z*( (i-z)_std - 0.09) )

  where:
    z_std is the standard, or calibrated, magnitude of the star
    (i-z)_std is the standard, or calibrated, i-z color of the star
    a_z is the photometric zeropoint in z (one for each CCD)
    k_z is the first-order extinction in z
    b_z is the instrumental color term coefficient in z (one for each CCD)
    X is the airmass
    0.09 is a "fiducial" reference color for stars in i-z (based on SDSS experience)

    Note that there is a separate photometric zeropoint (a_z) and 
    instrumental color term coefficient (b_z) for each CCD.
  • The same plots as for the i-band are presented below for the z-band, and in the
    same order. Nights 20121107, 20121108, and 20121109 are treated as non-photometric,
    just as for i-band.
  • Here is the plot of the effective zeropoint relative to the effective zeropoint
    for Nov 1 vs. Night for CCD 35. Note that the values on the Y-axis are consistent
    with 0.00 at slightly larger than the 1-sigma level (for photometric nights).

  • Here is the histogram of the Y-axis from the above plot. If we exclude the
    3 points from the 3 non-photometric nights, the RMS standard deviation of the
    remaining values is 0.020 mag.

  • Here is the plot of (a_n - a_35) - (a_n - a_35)_Nov1 for all 62 CCDs plotted
    independently. Aside from a couple bad nights and aside from the outlier
    CCD 61 in nights 20121108 and following, the full spread of points on a given
    night is typically slightly higher than 0.02 mag.

  • Here is the plot of the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov1"
    vs. night (with CCD 61 excluded). As with i-band, the error bars are plotted
    using the RMS standard deviation (per observation RMS), not the error in the mean.

  • Here is the histogram of the Y-axis of the above plot. The RMS standard
    deviation of the distribution is 0.007 mag, and this value INCLUDES the 3
    non-photometric nights and the strange g-band solution from 20121104.

Y-band

  • Form of the Photometric Equation:
  Y_std = -2.5log10(counts[ADU]/sec) - a_Y - k_Y*X - b_z*( (z-Y)_std - 0.05) )

  where:
    Y_std is the standard, or calibrated, magnitude of the star
    (z-Y)_std is the standard, or calibrated, z-Y color of the star
    a_Y is the photometric zeropoint in Y (one for each CCD)
    k_Y is the first-order extinction in Y
    b_Y is the instrumental color term coefficient in Y (one for each CCD)
    X is the airmass
    0.05 is a "fiducial" reference color for stars in z-Y (a guess based on SDSS and UKIDSS)

    Note that there is a separate photometric zeropoint (a_Y) and 
    instrumental color term coefficient (b_Y) for each CCD.
  • The same plots as for the i-band are presented below for the z-band, and in the
    same order. Nights 20121107, 20121108, and 20121109 are treated as non-photometric,
    just as for i-band.
  • Here is the plot of the effective zeropoint relative to the effective zeropoint
    for Nov 1 vs. Night for CCD 35. Note that the values on the Y-axis are consistent
    with 0.00 at the 1-sigma level (for photometric nights).

  • Here is the histogram of the Y-axis from the above plot. If we exclude the
    3 points from the 3 non-photometric nights, the RMS standard deviation of the
    remaining values is 0.018 mag.

  • Here is the plot of (a_n - a_35) - (a_n - a_35)_Nov1 for all 62 CCDs plotted
    independently. Aside from a couple bad nights and aside from the outlier
    CCD 61 in nights 20121108 and following, the full spread of points on a given
    night is typically slightly higher than 0.02 mag.

  • Here is the plot of the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov1"
    vs. night (with CCD 61 excluded). As with i-band, the error bars are plotted
    using the RMS standard deviation (per observation RMS), not the error in the mean.

  • Here is the histogram of the Y-axis of the above plot. The RMS standard
    deviation of the distribution is 0.011 mag, and this value INCLUDES the 3
    non-photometric nights and the strange g-band solution from 20121104.

u-band

The u-band results tend to be quite a bit worse than for the other filters.
Perhaps this is due to the sparseness of available standard stars per CCD
compared with the other filters. The analysis technique used here is too
noisy to give strong results for u-band. Maybe using purely relative
photometry of the standard star fields rather than running things through
PSM would give stronger results.

  • Form of the Photometric Equation:
  u_std = -2.5log10(counts[ADU]/sec) - a_u - k_u*X - b_u*( (u-g)_std - 1.39) )

  where:
    u_std is the standard, or calibrated, magnitude of the star
    (u-g)_std is the standard, or calibrated, u-g color of the star
    a_u is the photometric zeropoint in u (one for each CCD)
    k_u is the first-order extinction in u
    b_u is the instrumental color term coefficient in u (one for each CCD)
    X is the airmass
    1.39 is a "fiducial" reference color for stars in u-g (based on SDSS experience)

    Note that there is a separate photometric zeropoint (a_u) and 
    instrumental color term coefficient (b_u) for each CCD.
  • NOTE: There weren't any standard star observations in u-band on the first
    night of DES SV (Nov1) -- or at least any in the first (DES) half of the night
    that were good -- so here we use the second night of DES SV (Nov 2) as the
    reference date.
  • The same plots as for the i-band are presented below for the u-band, and in the
    same order. Nights 20121107, 20121108, and 20121109 are treated as non-photometric,
    just as for i-band. (Note that there did not seem to be any standard star u-band
    observations for 20121107 in any case.
    )
  • Here is the plot of the effective zeropoint relative to the effective zeropoint
    for Nov 2 vs. Night for CCD 35. Note that the values on the Y-axis are consistent
    with 0.00 at slightly above the 1-sigma level (for photometric nights).

  • Here is the histogram of the Y-axis from the above plot. If we exclude the
    non-photometric nights, the RMS standard deviation of the remaining values is
    0.032 mag.

  • Here is the plot of (a_n - a_35) - (a_n - a_35)_Nov2 for all 62 CCDs plotted
    independently. The spread is clearly worse than for the other filters, possibly
    due to the relative sparseness of the u-band standard star coverage.

  • Here is the plot of the nightly mean of "(a_n - a_35) - (a_n - a_35)_Nov2"
    vs. night (with CCD 61 excluded). As with i-band, the error bars are plotted
    using the RMS standard deviation (per observation RMS), not the error in the mean.

  • Here is the histogram of the Y-axis of the above plot. The RMS standard
    deviation of the distribution is 0.020 mag, and this value INCLUDES the
    non-photometric nights (but EXCLUDES the night 20121103 which has a huge
    outlier for CCD 43)