def dofit(xs, slambdas, alpha, sinbeta, gamma, delta, band, y):
xs = np.array(xs)
ok = np.isfinite(slambdas)
lsf = Fit.do_fit_wavelengths(xs[ok], lines[ok], alpha, sinbeta,
gamma, delta, band, y, error=slambdas[ok])
(order, pixel_y, alpha, sinbeta, gamma, delta) = lsf.params
print "order alpha sinbeta gamma delta"
print "%1.0i %5.7f %3.5f %3.2e %5.2f" % (order, alpha, sinbeta,
gamma, delta)
ll = Fit.wavelength_model(lsf.params, pix)
if DRAW:
pl.figure(3)
pl.clf()
pl.plot(ll, spec)
pl.title("Pixel %4.4f" % pos)
for lam in lines:
pl.axvline(lam, color='r')
pl.draw()
return [np.abs((
Fit.wavelength_model(lsf.params, xs[ok]) - lines[ok]))*1e4,
lsf.params]
def fit_line_with_sigclip(xs, data, i=0):
ps = Fit.do_fit_edge(xs, data)
pf = lambda x: Fit.fit_bar_edge(ps, x)
residual = np.abs(pf(xs) - data)
sd = np.std(residual)
ok = np.where(residual < 2.5 * sd)[0]
if len(ok) == 0:
return [lambda x: NaN, []]
ps = Fit.do_fit_edge(xs[ok], data[ok])
pf = lambda x: Fit.fit_bar_edge(ps, x)
return [pf, ok]
def fit_spec(data, pos, alpha, sinbeta, gamma, delta, band):
global DRAW
ar_h_lines = np.array([1.465435, 1.474317, 1.505062, 1.517684, 1.530607,
1.533334, 1.540685, 1.590403, 1.599386, 1.618444, 1.644107,
1.674465, 1.744967, 1.791961])
Ar_K_lines = np.array([1.982291, 1.997118, 2.032256, 2.057443,
2.062186, 2.099184, 2.133871, 2.15409, 2.204558, 2.208321,
2.313952, 2.385154])
lines = ar_h_lines
bmap = {"Y": 6, "J": 5, "H": 4, "K": 3}
order = bmap[band]
d = 1e3/110.5
pix = np.arange(2048)
ll = Fit.wavelength_model([order, alpha, sinbeta, gamma, delta], pix)
spec = np.median(data[pos-3:pos+3, :], axis=0)
[xs, sxs, sigmas] = find_peaks(ll, spec, lines)
slambdas = d/order * np.array(sxs)
[deltas, params] = dofit(xs, slambdas, lines, alpha, sinbeta, gamma,
delta, band)
return (params[1], params[2], params[3], params[4],
np.median(deltas), np.std(deltas), pixel, sigmas)
def dofit(xs, slambdas, lines, alpha, sinbeta, gamma, delta, band):
xs = np.array(xs)
ok = np.isfinite(slambdas)
lsf = Fit.do_fit_wavelengths(xs[ok], lines[ok], alpha, sinbeta,
gamma, delta, band, error=slambdas[ok])
(order, alpha, sinbeta, gamma, delta) = lsf.params
print "order alpha sinbeta gamma delta"
print "%1.0i %5.7f %3.5f %3.2e %5.2f" % (order, alpha, sinbeta,
gamma, delta)
pix = np.arange(2048.)
ll = Fit.wavelength_model(lsf.params, pix)
return [np.abs((
Fit.wavelength_model(lsf.params, xs[ok]) - lines[ok]))*1e4,
lsf.params]
def fit_edge_poly(xposs, xposs_missing, yposs, order):
'''
fit_edge_poly fits a polynomial to the measured slit edges.
This polynomial is used to extract spectra.
fit_edge_poly computes a parabola, and fills in missing data with a
parabola
input-
xposs, yposs [N]: The x and y positions of the slit edge [pix]
order: the polynomial order
'''
# First fit low order polynomial to fill in missing data
fun = np.poly1d(Fit.polyfit_clip(xposs, yposs, 2))
xposs = np.append(xposs, xposs_missing)
yposs = np.append(yposs, fun(xposs_missing))
# Remove any fits that deviate wildly from the 2nd order polynomial
ok = np.abs(yposs - fun(xposs)) < 1
if not ok.any():
error("Flat is not well illuminated? Cannot find edges")
raise Exception("Flat is not well illuminated? Cannot find edges")
# Now refit to user requested order
fun = np.poly1d(Fit.polyfit_clip(xposs[ok], yposs[ok], order))
yposs_ok = yposs[ok]
res = fun(xposs[ok]) - yposs[ok]
sd = np.std(res)
ok = np.abs(res) < 2*sd
# Check to see if the slit edge funciton is sane,
# if it's not, then we fix it.
pix = np.arange(2048)
V = fun(pix)
if np.abs(V.max() - V.min()) > 10:
info ("Forcing a horizontal slit edge")
print("Forcing a horizontal slit edge")
tmp = yposs_ok[ok]
fun = np.poly1d(np.median(tmp))
#fun = np.poly1d(np.median(yposs[ok]))
return (fun, res, sd, ok)
def find_peaks(ll, spec, lines):
xs = []
sxs = []
sigmas = []
pix = np.arange(2048.)
for lam in lines:
f = 0.9985
roi = (f*lam < ll) & (ll < lam/f)
if not roi.any():
xs.append(0.0)
sxs.append(np.inf)
sigmas.append(0.0)
continue
lsf = Fit.mpfitpeak(pix[roi], spec[roi],
error=np.sqrt(np.abs(spec[roi])))
if lsf.perror is None:
xs.append(0.0)
sxs.append(np.inf)
sigmas.append(0.0)
continue
if lsf.status < 0:
xs.append(0.0)
sxs.append(np.inf)
sigmas.append(0.0)
continue
mnpix = np.min(pix[roi])
mxpix = np.max(pix[roi])
if (mnpix + 4) > lsf.params[1] < (mxpix-4):
xs.append(0.0)
sxs.append(np.inf)
sigmas.append(0.0)
continue
if mnpix < 7:
xs.append(0.0)
sxs.append(np.inf)
sigmas.append(0.0)
continue
if mxpix > 2040:
xs.append(0.)
sxs.append(np.inf)
sigmas.append(0.0)
continue
xs.append(lsf.params[1])
sxs.append(lsf.perror[1])
sigmas.append(lsf.params[2])
return [xs, sxs, sigmas]
def fit_line_with_sigclip(xs, data, i=0):
ps = Fit.do_fit_edge(xs, data)
pf = lambda x: Fit.fit_bar_edge(ps, x)
residual = np.abs(pf(xs) - data)
sd = np.std(residual)
ok = np.where(residual < 2.5 * sd)[0]
if len(ok) == 0:
return [lambda x: NaN, []]
ps = Fit.do_fit_edge(xs[ok], data[ok])
pf = lambda x: Fit.fit_bar_edge(ps, x)
return [pf, ok]
def fit_edge_poly(xposs, xposs_missing, yposs, order):
'''
fit_edge_poly fits a polynomial to the measured slit edges.
This polynomial is used to extract spectra.
fit_edge_poly computes a parabola, and fills in missing data with a
parabola
input-
xposs, yposs [N]: The x and y positions of the slit edge [pix]
order: the polynomial order
'''
# First fit low order polynomial to fill in missing data
fun = np.poly1d(Fit.polyfit_clip(xposs, yposs, 2))
xposs = np.append(xposs, xposs_missing)
yposs = np.append(yposs, fun(xposs_missing))
# Remove any fits that deviate wildly from the 2nd order polynomial
ok = np.abs(yposs - fun(xposs)) < 1
if not ok.any():
error("Flat is not well illuminated? Cannot find edges")
raise Exception("Flat is not well illuminated? Cannot find edges")
# Now refit to user requested order
fun = np.poly1d(Fit.polyfit_clip(xposs[ok], yposs[ok], order))
yposs_ok = yposs[ok]
res = fun(xposs[ok]) - yposs[ok]
sd = np.std(res)
ok = np.abs(res) < 2 * sd
# Check to see if the slit edge funciton is sane,
# if it's not, then we fix it.
pix = np.arange(2048)
V = fun(pix)
if np.abs(V.max() - V.min()) > 10:
info("Forcing a horizontal slit edge")
print("Forcing a horizontal slit edge")
tmp = yposs_ok[ok]
fun = np.poly1d(np.median(tmp))
#fun = np.poly1d(np.median(yposs[ok]))
return (fun, res, sd, ok)
def plot_linefits(y, ff):
global plot_arr
pl.figure(3)
pl.subplot(*plot_arr)
ps = Fit.do_fit(y, Fit.residual_single)[0]
xx = np.arange(len(y))
xxx = np.arange(0,len(y)-1,.1)
fun = Fit.fit_single(ps,xxx)
pl.plot(xx,y, '*')
pl.plot(xxx, fun)
x0 = ff(0)
x1 = ff(0) + ps[4]
pl.plot([x0, x0], [0, Fit.fit_single(ps, x0)])
pl.plot([x1, x1], [0, Fit.fit_single(ps, x1)])
pl.figure(4)
pl.subplot(*plot_arr)
fun = Fit.fit_single(ps,xx)
r = (y - fun)/np.sqrt(np.abs(fun))
pl.plot(xx, r, '*')
pl.ylim([-5,5])
def dofit(xs, sxs, alpha, beta, band):
xs = np.array(xs)
sxs = np.array(sxs)/1000.
lsf = Fit.do_fit_wavelengths(xs, lines, alpha, beta, band, error=sxs)
(order, alpha, beta, gamma, delta) = lsf.params
print "alpha beta gamma delta"
print "%5.7f %3.5f %3.2e %5.2f" % (alpha, beta, gamma, delta)
ll = Fit.wavelength_model(lsf.params, pix)
if False:
pl.figure(3)
pl.clf()
pl.plot(ll, spec)
pl.title("Pixel %4.4f" % pos)
for lam in lines:
pl.axvline(lam, color='r')
pl.draw()
return [np.abs((Fit.wavelength_model(lsf.params, xs) - lines))/lines,
lsf.params]
def do_fits(params, pos, dy):
(alpha, sinbeta, gamma, delta) = params[1:]
thispos = pos - dy
yposs.append(thispos)
spec = np.median(data[thispos-3:thispos+3, :], axis=0)
ll = Fit.wavelength_model(params, pix)
[xs, sxs, sigmas] = find_peaks(ll, spec, lines)
slambdas = d/order * np.array(sxs)
[deltas, params] = dofit(xs, slambdas, lines, alpha, sinbeta,
gamma, delta, band)
results.append(np.array(params))
resdelt.append(deltas)
def fit_spec(data, pos, alpha, beta):
global DRAW
lines = np.array([1.49339, 1.49904, 1.51442, 1.51950, 1.52349, 1.53524,
1.54118, 1.6027, 1.62728, 1.64097, 1.71666])
order = 4
d = 1e3/110.5
pix = np.arange(2048.)
ll = alpha * d/order * (np.sin(18./(250e3) * pix) + np.radians(beta))
spec = np.median(data[pos-3:pos+3, :], axis=0)
if DRAW:
pl.figure(2)
pl.clf()
pl.figure(3)
pl.clf()
pl.figure(1)
pl.clf()
pl.plot(ll, spec, '-+')
for lam in lines:
pl.axvline(lam, color='r')
pl.title("Pixel pos: %i"% pos)
pl.figure(2)
pl.clf()
xs = []
sxs = []
pks = []
rats = []
for lam in lines:
roi = ((lam-.002) < ll) & (ll < (lam+0.002))
if not roi.any():
xs.append(0.0)
sxs.append(np.inf)
continue
lsf = Fit.mpfitpeak(pix[roi], spec[roi],
error=np.sqrt(np.abs(spec[roi])))
if lsf.perror is None:
xs.append(0.0)
sxs.append(np.inf)
continue
if lsf.status < 0:
xs.append(0.0)
sxs.append(np.inf)
continue
#if DRAW:
if False:
pl.plot(pix[roi], spec[roi])
pl.plot(pix[roi], spec[roi], 'o')
pl.plot(pix[roi], Fit.gaussian(lsf.params, pix[roi]), 'r--')
xs.append(lsf.params[1])
sxs.append(lsf.perror[1])
if False:
pl.axvline(lsf.params[1])
if DRAW:
pl.draw()
def dofit(xs, sxs, alpha, beta, band):
xs = np.array(xs)
sxs = np.array(sxs)/1000.
lsf = Fit.do_fit_wavelengths(xs, lines, alpha, beta, band, error=sxs)
(order, alpha, beta, gamma, delta) = lsf.params
print "alpha beta gamma delta"
print "%5.7f %3.5f %3.2e %5.2f" % (alpha, beta, gamma, delta)
ll = Fit.wavelength_model(lsf.params, pix)
if False:
pl.figure(3)
pl.clf()
pl.plot(ll, spec)
pl.title("Pixel %4.4f" % pos)
for lam in lines:
pl.axvline(lam, color='r')
pl.draw()
return [np.abs((Fit.wavelength_model(lsf.params, xs) - lines))/lines,
lsf.params]
[delta, params] = dofit(xs, sxs, alpha, beta, 'H')
return (params[1], params[2], np.median(delta), np.std(delta), pixel)
def fit_spec(data, pos, alpha, sinbeta, gamma, delta, band):
global DRAW
ar_h_lines = np.array([1.465435, 1.474317, 1.505062, 1.517684, 1.530607,
1.533334, 1.540685, 1.590403, 1.599386, 1.618444, 1.644107,
1.674465, 1.744967, 1.791961])
Ar_K_lines = np.array([1.982291, 1.997118, 2.032256, 2.057443,
2.062186, 2.099184, 2.133871, 2.15409, 2.204558, 2.208321,
2.313952, 2.385154])
Ne_J_lines = np.array([1.149125, 1.16719, 1.17227, 1.194655, 1.202994,
1.211564, 1.214306, 1.234677, 1.240622, 1.244273, 1.249108,
1.27369, 1.280624, 1.296021, 1.301182, 1.321761, 1.327627,
1.331685, 1.337077, 1.341026, 1.350788, 1.362638])
Ne_J_lines = np.array([1.149125, 1.16719, 1.117227])
AR_Y_lines = np.array([0.966043, 0.978718, 1.005481, 1.04809,
1.067649, 1.07039, 1.088394, 1.110819])
# NeAr H (1961, 1949)
lines = np.array([1.465435, 1.474317, 1.505062, 1.517684, 1.530607,
1.533334, 1.540685, 1.590403, 1.599386, 1.618444, 1.644107,
1.674465, 1.744967, 1.791961, \
1.493386, 1.499041, 1.523487, 1.5352384, 1.5411803, 1.5608478,
1.6027147, 1.6272797, 1.6409737, 1.6479254, 1.6793378,
1.7166622])
lines = np.array([ 1.540685, 1.5411803])
lines.sort()
bmap = {"Y": 6, "J": 5, "H": 4, "K": 3}
order = bmap[band]
d = 1e3/110.5
pix = np.arange(2048.)
ll = Fit.wavelength_model([order, pos, alpha, sinbeta, gamma, delta], pix)
spec = np.median(data[pos-1:pos+1, :], axis=0)
if DRAW:
pl.figure(2)
pl.clf()
pl.figure(3)
pl.clf()
pl.figure(1)
pl.clf()
pl.plot(ll, spec, '-+')
for lam in lines:
pl.axvline(lam, color='r')
pl.title("Pixel pos: %i"% pos)
pl.figure(2)
pl.clf()
xs = []
sxs = []
pks = []
rats = []
for lam in lines:
f = 0.9985
roi = (f*lam < ll) & (ll < lam/f)
if not roi.any():
xs.append(0.0)
sxs.append(np.inf)
continue
#lsf = Fit.mpfitpeak(pix[roi], spec[roi],
# error=np.sqrt(np.abs(spec[roi])))
lsf = Fit.mpfitpeak(pix[roi], spec[roi],
error=np.sqrt(np.abs(spec[roi])))
print lsf.params[2], lsf.params[1]
print "LSF"
pl.figure(6)
pl.clf()
pl.scatter(pix[roi], spec[roi])
pl.plot(pix[roi], Fit.gaussian(lsf.params, pix[roi]))
lsf = Fit.mpfitpeaks(pix[roi], spec[roi], 2,
error=np.sqrt(np.abs(spec[roi])))
print lsf.params[2], lsf.params[6], lsf.params[8]
print "LSF"
pl.figure(5)
pl.clf()
pl.scatter(pix[roi], spec[roi])
pl.plot(pix[roi], Fit.multi_gaussian(lsf.params, pix[roi]))
if lsf.perror is None:
xs.append(0.0)
sxs.append(np.inf)
continue
if lsf.status < 0:
xs.append(0.0)
sxs.append(np.inf)
continue
mnpix = np.min(pix[roi])
mxpix = np.max(pix[roi])
if (mnpix + 4) > lsf.params[1] < (mxpix-4):
xs.append(0.0)
sxs.append(np.inf)
continue
if mnpix < 7:
xs.append(0.0)
sxs.append(np.inf)
continue
if mxpix > 2040:
xs.append(0.)
sxs.append(np.inf)
continue
#if DRAW:
if DRAW:
pl.plot(pix[roi], spec[roi])
pl.plot(pix[roi], spec[roi], 'o')
pl.plot(pix[roi], Fit.gaussian(lsf.params, pix[roi]), 'r--')
xval = lsf.params[1]
xvals = lsf.perror[1]
xs.append(xval)
sxs.append(xvals)
if DRAW:
if np.isfinite(xvals):
pl.axvline(xval)
if DRAW:
pl.draw()
def dofit(xs, slambdas, alpha, sinbeta, gamma, delta, band, y):
xs = np.array(xs)
ok = np.isfinite(slambdas)
lsf = Fit.do_fit_wavelengths(xs[ok], lines[ok], alpha, sinbeta,
gamma, delta, band, y, error=slambdas[ok])
(order, pixel_y, alpha, sinbeta, gamma, delta) = lsf.params
print "order alpha sinbeta gamma delta"
print "%1.0i %5.7f %3.5f %3.2e %5.2f" % (order, alpha, sinbeta,
gamma, delta)
ll = Fit.wavelength_model(lsf.params, pix)
if DRAW:
pl.figure(3)
pl.clf()
pl.plot(ll, spec)
pl.title("Pixel %4.4f" % pos)
for lam in lines:
pl.axvline(lam, color='r')
pl.draw()
return [np.abs((
Fit.wavelength_model(lsf.params, xs[ok]) - lines[ok]))*1e4,
lsf.params]
slambdas = d/order * np.array(sxs)
[deltas, params] = dofit(xs, slambdas, alpha, sinbeta, gamma, delta,
band)
if 0.4 < np.median(deltas) < 0.8:
prev = np.median(deltas)
[deltas, params] = dofit(xs, slambdas, params[1], params[2],
params[3], params[4], band)
print "Previous MAD: %f is now %f" % (prev, np.median(deltas))
return (params[1], params[2], params[3], params[4],
np.median(deltas), np.std(deltas), pixel)
print "-------------==========================------------------"
print "Finding Slit Edges for %s starting at %4.0i" % (ssl[target]["Target_Name"], y)
tock = time.clock()
yposs = []
widths = []
xposs = []
x = 936
for i in range(-49, 50):
delt = 1900/100.
xp = x+delt*i
v = data[y-roi_width:y+roi_width, xp-2:xp+2]
v = np.median(v, axis=1)
ff = Fit.do_fit(v, residual_fun=Fit.residual_disjoint_pair)
if (0 < ff[4] < 4):
xposs.append(x+delt*i)
xs = np.arange(len(v))
yposs.append(ff[0][1] - roi_width)
widths.append(ff[0][5])
else:
print "Skipping: %i" % (x+delt*i)
(xposs, yposs, widths) = map(np.array, (xposs, yposs, widths))
fun = np.poly1d(Fit.polyfit_clip(xposs, yposs, 3))
wfun = np.poly1d(Fit.polyfit_clip(xposs, widths, 3))
res = fun(xposs) - yposs
sd = np.std(res)
pars = [1500, 20, 100000, 10000]
pars.extend(np.zeros(len(pixs)))
pars = np.array(pars)
parinfo = []
for par in pars:
parinfo.append({"fixed": 0, "value": par})
pl.ion()
y = Wavelength.beta_model(pars, pixs)
parinfo[0]["fixed"] = 1
parinfo[1]["fixed"] = 1
parinfo[2]["fixed"] = 1
parinfo[3]["fixed"] = 1
merit_fun = Fit.mpfit_residuals(Wavelength.beta_model)
lsf = Fit.mpfit_do(merit_fun, pixs, all_betas, parinfo)
parinfo = []
for param in lsf.params:
parinfo.append({"fixed": 0, "value": param})
lsf = Fit.mpfit_do(merit_fun, pixs, all_betas, parinfo)
print lsf
pl.figure(2)
pl.clf()
pl.plot(all_pixs, all_betas, '.')
y = Wavelength.beta_model(lsf.params, pixs)
d = np.abs(np.diff(y))
rois = np.where(d>0.01)[0]
prev = 0
def find_edge_pair(data, y, roi_width, edgeThreshold=450):
'''
find_edge_pair finds the edge of a slit pair in a flat
data[2048x2048]: a well illuminated flat field [DN]
y: guess of slit edge position [pix]
Keywords:
edgeThreshold: the pixel value below which we should ignore using
to calculate edges.
Moves along the edge of a slit image
- At each location along the slit edge, determines
the position of the demarcations between two slits
Outputs:
xposs []: Array of x positions along the slit edge [pix]
yposs []: The fitted y positions of the "top" edge of the slit [pix]
widths []: The fitted delta from the top edge of the bottom [pix]
scatters []: The amount of light between slits
The procedure is as follows
1: starting from a guess spatial position (parameter y), march
along the spectral direction in some chunk of pixels
2: At each spectral location, construct a cross cut across the
spatial direction; select_roi is used for this.
3: Fit a two-sided error function Fit.residual_disjoint_pair
on the vertical cross cut derived in step 2.
4: If the fit fails, store it in the missing list
- else if the top fit is good, store the top values in top vector
- else if the bottom fit is good, store the bottom values in bottom
vector.
5: In the vertical cross-cut, there is a minimum value. This minimum
value is stored as a measure of scattered light.
Another procedure is used to fit polynomials to these fitted values.
'''
def select_roi(data, roi_width):
v = data[y-roi_width:y+roi_width, xp-2:xp+2]
v = np.median(v, axis=1) # Axis = 1 is spatial direction
return v
xposs_top = []
yposs_top = []
xposs_top_missing = []
xposs_bot = []
yposs_bot = []
xposs_bot_missing = []
yposs_bot_scatters = []
#1
rng = np.linspace(10, 2040, 50).astype(np.int)
for i in rng:
xp = i
#2
v = select_roi(data, roi_width)
xs = np.arange(len(v))
# Modified from 450 as the hard coded threshold to one that
# can be controlled by a keyword
if (np.median(v) < edgeThreshold):
xposs_top_missing.append(xp)
xposs_bot_missing.append(xp)
continue
#3
ff = Fit.do_fit(v, residual_fun=Fit.residual_disjoint_pair)
fit_ok = 0 < ff[4] < 4
if fit_ok:
(sigma, offset, mult1, mult2, add, width) = ff[0]
xposs_top.append(xp)
yposs_top.append(y - roi_width + offset + width)
xposs_bot.append(xp)
yposs_bot.append(y - roi_width + offset)
between = offset + width/2
if 0 < between < len(v)-1:
start = np.max([0, between-2])
stop = np.min([len(v),between+2])
yposs_bot_scatters.append(np.min(v[start:stop])) # 5
if False:
pl.figure(2)
pl.clf()
tmppix = np.arange(y-roi_width, y+roi_width)
tmpx = np.arange(len(v))
pl.axvline(y - roi_width + offset + width, color='red')
pl.axvline(y - roi_width + offset, color='red')
pl.scatter(tmppix, v)
pl.plot(tmppix, Fit.fit_disjoint_pair(ff[0], tmpx))
pl.axhline(yposs_bot_scatters[-1])
pl.draw()
else:
yposs_bot_scatters.append(np.nan)
else:
xposs_bot_missing.append(xp)
xposs_top_missing.append(xp)
info("Skipping wavelength pixel): %i" % (xp))
return map(np.array, (xposs_bot, xposs_bot_missing, yposs_bot, xposs_top,
xposs_top_missing, yposs_top, yposs_bot_scatters))
[[xslice, yslice], extent] = make_slice(pos, 6, 25)
if extent[0] == extent[1]:
cnt += 1
continue
if extent[2] == extent[3]:
cnt += 1
continue
cnt += 1
fits = []
xs = np.arange(-10, 10)
for i in xs:
tofit = data[pos[1] - i, xslice]
y = median_tails(tofit)
ps = Fit.do_fit(y, Fit.residual_pair)
fits.append(ps[0])
fits = np.array(fits)
m = [np.mean(fits[:, i]) for i in range(5)]
s = [np.std(fits[:, i]) for i in range(5)]
means.append(m)
sds.append(s)
[ff, ok] = fit_line_with_sigclip(xs, fits[:, 1])
pl.figure(5)
pl.subplot(7, 7, cntfit)
pl.plot(xs, fits[:, 1] - ff(xs), '*-')
pl.plot(xs[ok], fits[ok, 1] - ff(xs[ok]), 'or-')
[[xslice, yslice],extent] = make_slice(pos, 6,25)
if extent[0] == extent[1]:
cnt += 1
continue
if extent[2] == extent[3]:
cnt += 1
continue
cnt+=1
fits = []
xs = np.arange(-10,10)
for i in xs:
tofit = data[pos[1]-i, xslice]
y = median_tails(tofit)
ps = Fit.do_fit(y, Fit.residual_pair)
fits.append(ps[0])
fits = np.array(fits)
m = [np.mean(fits[:,i]) for i in range(5)]
s = [np.std(fits[:,i]) for i in range(5)]
means.append(m)
sds.append(s)
[ff, ok] = fit_line_with_sigclip(xs, fits[:,1])
pl.figure(5)
pl.subplot(7,7,cntfit)
pl.plot(xs, fits[:,1] - ff(xs), '*-')
pl.plot(xs[ok], fits[ok,1] - ff(xs[ok]), 'or-')
def go(fname):
global plot_arr
print fname
(header, data) = IO.readfits(fname)
bs = CSU.Barset()
bs.set_header(header)
bars = []
means = []
sds = []
deltas = []
deltas_mm = []
poss = []
poss_mm = []
poss_y = []
request = []
qs = []
bars = range(1, 93)
fit_fun = Fit.residual_single
for bar in bars:
pos = bs.get_bar_pix(bar)
if bar % 8 == 0:
print "%2.0i: (%7.2f, %7.2f)" % (bar, pos[0], pos[1])
if is_odd(bar):
if (bs.pos[bar] - bs.pos[bar-1]) < 2.7:
fit_fun = Fit.residual_pair
else:
fit_fun = Fit.residual_single
width = 19
[[xslice, yslice],extent,ystart] = make_slice(pos,0,width,30)
if not is_in_bounds(extent):
fits = [0,0,0,0,0]
[ff,ok] = [np.poly1d(0,0), []]
means.append(fits)
sds.append(fits)
drop_this = True
else:
drop_this = False
fits = []
ys = np.arange(-10,10, dtype=np.float32)
for i in ys:
tofit = data[ystart-i, xslice]
y = median_tails(tofit)
ps = Fit.do_fit(y, fit_fun)
fits.append(ps[0])
fits = np.array(fits)
fits[:,1] += 1
# fit to the ridgeline
[ff, ok] = fit_line_with_sigclip(ys, fits[:,1])
m = [np.mean(fits[:,i]) for i in range(5)]
s = [np.std(fits[:,i]) for i in range(5)]
means.append(m)
sds.append(s)
slit_center_offset = pos[1] - ystart
fc = ff(slit_center_offset)
slit_center_pos = np.float(extent[0] + fc )
if drop_this:
poss.append(NaN)
poss_y.append(NaN)
poss_mm.append(NaN)
else:
poss.append(slit_center_pos)
poss_y.append(ystart)
poss_mm.append(CSU.csu_pix_to_mm_poly(slit_center_pos, ystart)[0])
delta = np.float(slit_center_pos - pos[0])
if drop_this:
deltas.append(NaN)
deltas_mm.append(NaN)
else:
deltas.append(delta)
b = CSU.csu_pix_to_mm_poly(slit_center_pos + delta, ystart)[0]
deltas_mm.append(b - poss_mm[-1])
q = np.float(np.degrees(np.tan(ff(1)-ff(0))))
if drop_this: qs.append(NaN)
qs.append(q)
means = np.array(means)
f = lambda x: np.array(x).ravel()
sds = f(sds)
deltas = f(deltas)
poss = f(poss)
poss_y = f(poss_y)
poss_mm = f(poss_mm)
deltas_mm = f(deltas_mm)
qs = f(qs)
bars = f(bars)
fout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".sav"
print "saving"
tosav = {"bars": bars, "request": bs.pos, "deltas_mm": deltas_mm, "poss": poss, "poss_mm": poss_mm, "deltas": deltas, "means": means, "qs": qs}
scipy.io.savemat(fout, tosav)
save_region(bs, "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".reg")
print "saved"
regout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".meas.reg"
pairs = np.array([poss,poss_y]).transpose()
s = CSU.to_ds9_region(pairs, dash=0, color="blue", label=False)
try:
f = open(regout, "w")
f.writelines(s)
f.close()
except:
print "Couldn't write: %s" % regout
return [tosav, bs]
def go(fname):
global plot_arr
print fname
(header, data) = IO.readfits(fname)
bs = CSU.Barset()
bs.set_header(header)
cntfit = 1
for i in range(1,8):
continue
pl.figure(i)
pl.clf()
first_time = True
bars = []
means = []
sds = []
deltas = []
deltas_mm = []
poss = []
poss_mm = []
poss_y = []
request = []
qs = []
bars = range(1, 93)
for bar in bars:
pos = bs.get_bar_pix(bar)
if bar % 8 == 0:
print "%2.0i: (%7.2f, %7.2f)" % (bar, pos[0], pos[1])
width = 19
[[xslice, yslice],extent,ystart] = make_slice(pos,0,width,30)
if not is_in_bounds(extent):
fits = [0,0,0,0,0]
[ff,ok] = [np.poly1d(0,0), []]
means.append(fits)
sds.append(fits)
drop_this = True
else:
drop_this = False
fits = []
ys = np.arange(-10,10, dtype=np.float32)
for i in ys:
tofit = data[ystart-i, xslice]
y = median_tails(tofit)
ps = Fit.do_fit(y, Fit.residual_single)
fits.append(ps[0])
fits = np.array(fits)
fits[:,1] += 1
# fit to the ridgeline
[ff, ok] = fit_line_with_sigclip(ys, fits[:,1])
m = [np.mean(fits[:,i]) for i in range(5)]
s = [np.std(fits[:,i]) for i in range(5)]
means.append(m)
sds.append(s)
#if bar == 44: start_shell()
#if bar == 43: start_shell()
# End of measurement logic, PLOTS
plot_arr = [12, 8, cntfit]
cntfit += 1
if False:
plot_stamps(data[yslice, xslice], ps[0])
y0 = median_tails(data[ystart, xslice])
plot_linefits(y0, ff)
plot_ridgeline(fits, ok, ys, ff, bar)
plot_widths(fits, ok, ys)
# End of plots, BOOKEEPING
slit_center_offset = pos[1] - ystart
fc = ff(slit_center_offset)
slit_center_pos = np.float(extent[0] + fc )
if drop_this:
poss.append(NaN)
poss_y.append(NaN)
poss_mm.append(NaN)
else:
poss.append(slit_center_pos)
poss_y.append(ystart)
poss_mm.append(CSU.csu_pix_to_mm_poly(slit_center_pos, ystart)[0])
delta = np.float(slit_center_pos - pos[0])
if drop_this:
deltas.append(NaN)
deltas_mm.append(NaN)
else:
deltas.append(delta)
b = CSU.csu_pix_to_mm_poly(slit_center_pos + delta, ystart)[0]
deltas_mm.append(b - poss_mm[-1])
q = np.float(np.degrees(np.tan(ff(1)-ff(0))))
if drop_this: qs.append(NaN)
qs.append(q)
if False: #is_in_bounds(extent):
pl.figure(2)
pl.text(pos[0], pos[1], 'b%2.0i: w=%3.2f p=%5.2f q=%3.2f d=%1.3f' % (bar, np.mean(fits[:,4]), extent[0]+ff(0), q, delta), fontsize=11, family='monospace', horizontalalignment='center')
means = np.array(means)
f = lambda x: np.array(x).ravel()
sds = f(sds)
deltas = f(deltas)
poss = f(poss)
poss_y = f(poss_y)
poss_mm = f(poss_mm)
deltas_mm = f(deltas_mm)
qs = f(qs)
bars = f(bars)
pl.draw()
if False:
[pf, ok] = fit_line_with_sigclip(bars, deltas)
print "** Residual: %4.3f [pix]" % np.std(deltas[ok]-pf(bars[ok]))
evens = range(0,92,2)
odds = range(1,92,2)
pl.plot(bars[evens], deltas[evens],'r*-')
pl.plot(bars[odds], deltas[odds],'b*-')
pl.title("Red: even, blue: odd")
pl.xlabel("Bar #")
pl.ylabel("Delta pixel (Measured - Predicted)")
pl.plot(bars, pf(bars))
fout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".sav"
print "saving"
tosav = {"bars": bars, "request": bs.pos, "deltas_mm": deltas_mm, "poss": poss, "poss_mm": poss_mm, "deltas": deltas, "means": means, "qs": qs}
scipy.io.savemat(fout, tosav)
save_region(bs, "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".reg")
print "saved"
regout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".meas.reg"
pairs = np.array([poss,poss_y]).transpose()
s = CSU.to_ds9_region(pairs, dash=0, color="blue", label=False)
try:
f = open(regout, "w")
f.writelines(s)
f.close()
except:
print "Couldn't write: %s" % regout
return [tosav, bs]
print "Skipping: %i" % (x+delt*i)
return map(np.array, (xposs, yposs, widths))
def fit_edge_poly(xposs, yposs, widths, order):
'''
fit_edge_poly fits a polynomial to the measured slit edges.
This polynomial is used to extract spectra.
input-
xposs, yposs [N]: The x and y positions of the slit edge [pix]
widths [N]: the offset from end of one slit to beginning of another [pix]
order: the polynomial order
'''
fun = np.poly1d(Fit.polyfit_clip(xposs, yposs, order))
wfun = np.poly1d(Fit.polyfit_clip(xposs, widths, order))
res = fun(xposs) - yposs
sd = np.std(res)
ok = np.abs(res) < 2*sd
return (fun, wfun, res, sd, ok)
def data_quality_plots(results):
y = 2028
toc = 0
slits = []
def go(fname):
global plot_arr
print fname
(header, data) = IO.readfits(fname)
bs = CSU.Barset()
bs.set_header(header)
bars = []
means = []
sds = []
deltas = []
deltas_mm = []
poss = []
poss_mm = []
poss_y = []
request = []
qs = []
bars = range(1, 93)
fit_fun = Fit.residual_single
for bar in bars:
pos = bs.get_bar_pix(bar)
if bar % 8 == 0:
print "%2.0i: (%7.2f, %7.2f)" % (bar, pos[0], pos[1])
if is_odd(bar):
if (bs.pos[bar] - bs.pos[bar-1]) < 2.7:
fit_fun = Fit.residual_pair
else:
fit_fun = Fit.residual_single
width = 19
[[xslice, yslice],extent,ystart] = make_slice(pos,0,width,30)
if not is_in_bounds(extent):
fits = [0,0,0,0,0]
[ff,ok] = [np.poly1d(0,0), []]
means.append(fits)
sds.append(fits)
drop_this = True
else:
drop_this = False
fits = []
ys = np.arange(-10,10, dtype=np.float32)
for i in ys:
tofit = data[ystart-i, xslice]
y = median_tails(tofit)
ps = Fit.do_fit(y, fit_fun)
fits.append(ps[0])
fits = np.array(fits)
fits[:,1] += 1
# fit to the ridgeline
[ff, ok] = fit_line_with_sigclip(ys, fits[:,1])
m = [np.mean(fits[:,i]) for i in range(5)]
s = [np.std(fits[:,i]) for i in range(5)]
means.append(m)
sds.append(s)
slit_center_offset = pos[1] - ystart
fc = ff(slit_center_offset)
slit_center_pos = np.float(extent[0] + fc )
if drop_this:
poss.append(NaN)
poss_y.append(NaN)
poss_mm.append(NaN)
else:
poss.append(slit_center_pos)
poss_y.append(ystart)
poss_mm.append(CSU.csu_pix_to_mm_poly(slit_center_pos, ystart)[0])
delta = np.float(slit_center_pos - pos[0])
if drop_this:
deltas.append(NaN)
deltas_mm.append(NaN)
else:
deltas.append(delta)
b = CSU.csu_pix_to_mm_poly(slit_center_pos + delta, ystart)[0]
deltas_mm.append(b - poss_mm[-1])
q = np.float(np.degrees(np.tan(ff(1)-ff(0))))
if drop_this: qs.append(NaN)
qs.append(q)
means = np.array(means)
f = lambda x: np.array(x).ravel()
sds = f(sds)
deltas = f(deltas)
poss = f(poss)
poss_y = f(poss_y)
poss_mm = f(poss_mm)
deltas_mm = f(deltas_mm)
qs = f(qs)
bars = f(bars)
fout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".sav"
print "saving"
tosav = {"bars": bars, "request": bs.pos, "deltas_mm": deltas_mm, "poss": poss, "poss_mm": poss_mm, "deltas": deltas, "means": means, "qs": qs}
scipy.io.savemat(fout, tosav)
save_region(bs, "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".reg")
print "saved"
regout = "/users/npk/dropbox/mosfire/cooldown 9/csu_meas/" + fname.split("/")[-1] + ".meas.reg"
pairs = np.array([poss,poss_y]).transpose()
s = CSU.to_ds9_region(pairs, dash=0, color="blue", label=False)
try:
f = open(regout, "w")
f.writelines(s)
f.close()
except:
print "Couldn't write: %s" % regout
return [tosav, bs]
def find_edge_pair(data, y, roi_width, edgeThreshold=450):
'''
find_edge_pair finds the edge of a slit pair in a flat
data[2048x2048]: a well illuminated flat field [DN]
y: guess of slit edge position [pix]
Keywords:
edgeThreshold: the pixel value below which we should ignore using
to calculate edges.
Moves along the edge of a slit image
- At each location along the slit edge, determines
the position of the demarcations between two slits
Outputs:
xposs []: Array of x positions along the slit edge [pix]
yposs []: The fitted y positions of the "top" edge of the slit [pix]
widths []: The fitted delta from the top edge of the bottom [pix]
scatters []: The amount of light between slits
The procedure is as follows
1: starting from a guess spatial position (parameter y), march
along the spectral direction in some chunk of pixels
2: At each spectral location, construct a cross cut across the
spatial direction; select_roi is used for this.
3: Fit a two-sided error function Fit.residual_disjoint_pair
on the vertical cross cut derived in step 2.
4: If the fit fails, store it in the missing list
- else if the top fit is good, store the top values in top vector
- else if the bottom fit is good, store the bottom values in bottom
vector.
5: In the vertical cross-cut, there is a minimum value. This minimum
value is stored as a measure of scattered light.
Another procedure is used to fit polynomials to these fitted values.
'''
def select_roi(data, roi_width):
v = data[y - roi_width:y + roi_width, xp - 2:xp + 2]
v = np.median(v, axis=1) # Axis = 1 is spatial direction
return v
xposs_top = []
yposs_top = []
xposs_top_missing = []
xposs_bot = []
yposs_bot = []
xposs_bot_missing = []
yposs_bot_scatters = []
#1
rng = np.linspace(10, 2040, 50).astype(np.int)
for i in rng:
xp = i
#2
v = select_roi(data, roi_width)
xs = np.arange(len(v))
# Modified from 450 as the hard coded threshold to one that
# can be controlled by a keyword
if (np.median(v) < edgeThreshold):
xposs_top_missing.append(xp)
xposs_bot_missing.append(xp)
continue
#3
ff = Fit.do_fit(v, residual_fun=Fit.residual_disjoint_pair)
fit_ok = 0 < ff[4] < 4
if fit_ok:
(sigma, offset, mult1, mult2, add, width) = ff[0]
xposs_top.append(xp)
yposs_top.append(y - roi_width + offset + width)
xposs_bot.append(xp)
yposs_bot.append(y - roi_width + offset)
between = offset + width / 2
if 0 < between < len(v) - 1:
start = np.max([0, between - 2])
stop = np.min([len(v), between + 2])
yposs_bot_scatters.append(np.min(v[start:stop])) # 5
if False:
pl.figure(2)
pl.clf()
tmppix = np.arange(y - roi_width, y + roi_width)
tmpx = np.arange(len(v))
pl.axvline(y - roi_width + offset + width, color='red')
pl.axvline(y - roi_width + offset, color='red')
pl.scatter(tmppix, v)
pl.plot(tmppix, Fit.fit_disjoint_pair(ff[0], tmpx))
pl.axhline(yposs_bot_scatters[-1])
pl.draw()
else:
yposs_bot_scatters.append(np.nan)
else:
xposs_bot_missing.append(xp)
xposs_top_missing.append(xp)
info("Skipping wavelength pixel): %i" % (xp))
return map(np.array, (xposs_bot, xposs_bot_missing, yposs_bot, xposs_top,
xposs_top_missing, yposs_top, yposs_bot_scatters))
spec = slitlet[mid-3:mid+3, :]
spec = np.median(spec, axis=0)
continuum = median_filter(spec, 250)
if fiducial_spec.shape == (0,):
fiducial_spec = spec
fiducial_spec /= continuum
fiducial_spec /= fiducial_spec.max()
soc = spec/continuum
soc /= soc.max()
as_per_pix = .18/1.4
shift = -np.median(loc)/as_per_pix
lags = np.arange(shift-300,shift+300,dtype=np.int)
xc = Fit.xcor(soc[800:1300], fiducial_spec[800:1300], lags)
#start_shell()
shift = lags[np.argmax(xc)]
center = width/2 - shift
print center, write_pos
xs = np.arange(len(spec))
pl.plot(xs - shift, spec/continuum)
try:
out[(write_pos - dy):write_pos, center-1024:center+1024] = slitlet
if minl > center-1024: minl = center-1024
if maxl < center+1024: maxl = center+1024
