def __init__(self, msids, start, stop):
super(MSIDset, self).__init__()
self.tstart = DateTime(start).secs
self.tstop = DateTime(stop).secs
self.datestart = DateTime(self.tstart).date
self.datestop = DateTime(self.tstop).date
slot_datas = {}
slots = set(slot_for_msid(confirm_msid(msid)) for msid in msids)
for slot in slots:
# get the 8x8 data
tstop = self.tstop + 33.0 # Major frame of padding
slot_data = aca_l0.get_slot_data(
self.tstart, tstop, slot,
imgsize=[8], columns=ACA_DTYPE_NAMES)
# Find samples where the time stamp changes by a value other than 4.1 secs
# (which is the value for 8x8 readouts). In that case there must have been a
# break in L0 decom, typically due to a change to 4x4 or 6x6 data.
# t[0] = 1.0
# t[1] = 5.1 <= This record could be bad, as indicated by the gap afterward
# t[2, 3] = 17.4, 21.5
# To form the time diffs first add `tstop` to the end so that if 8x8 data
# does not extend through `tstop` then the last record gets chopped.
dt = np.diff(np.concatenate([slot_data['TIME'], [tstop]]))
bad = np.abs(dt - 4.1) > 1e-3
slot_data[bad] = ma.masked
# Chop off the padding
i_stop = np.searchsorted(slot_data['TIME'], self.tstop, side='right')
slot_data = slot_data[:i_stop]
# explicitly unmask useful columns
slot_data['TIME'].mask = ma.nomask
slot_data['IMGSIZE'].mask = ma.nomask
slot_data['FILENAME'].mask = ma.nomask
slot_datas[slot] = slot_data
# make a shared time ndarray that is the union of the time sets in the
# slots. The ACA L0 telemetry has the same timestamps across slots,
# so the only differences here are caused by different times in
# non-TRAK across the slots (usually SRCH differences at the beginning
# of the observation)
shared_time = np.unique(np.concatenate([
slot_datas[slot]['TIME'].data for slot in slots]))
for msid in msids:
hdr3_msid = confirm_msid(msid)
slot = slot_for_msid(hdr3_msid)
full_data = ma.zeros(len(shared_time),
dtype=slot_datas[slot].dtype)
full_data.mask = ma.masked
fd_idx = search_both_sorted(shared_time,
slot_datas[slot]['TIME'])
full_data[fd_idx] = slot_datas[slot]
# make a data dictionary to feed to the MSID constructor
slot_data = {'vals': HDR3_DEF[hdr3_msid]['value'](full_data),
'desc': HDR3_DEF[hdr3_msid]['desc'],
'longdesc': HDR3_DEF[hdr3_msid]['longdesc'],
'times': shared_time,
'hdr3_msid': hdr3_msid}
self[msid] = MSID(msid, start, stop, slot_data)
def __init__(self, msids, start, stop):
super(MSIDset, self).__init__()
self.tstart = DateTime(start).secs
self.tstop = DateTime(stop).secs
self.datestart = DateTime(self.tstart).date
self.datestop = DateTime(self.tstop).date
slot_datas = {}
slots = set(slot_for_msid(confirm_msid(msid)) for msid in msids)
for slot in slots:
# get the 8x8 data
tstop = self.tstop + 33.0 # Major frame of padding
slot_data = aca_l0.get_slot_data(
self.tstart, tstop, slot,
imgsize=[8], columns=ACA_DTYPE_NAMES)
# Find samples where the time stamp changes by a value other than 4.1 secs
# (which is the value for 8x8 readouts). In that case there must have been a
# break in L0 decom, typically due to a change to 4x4 or 6x6 data.
# t[0] = 1.0
# t[1] = 5.1 <= This record could be bad, as indicated by the gap afterward
# t[2, 3] = 17.4, 21.5
# To form the time diffs first add `tstop` to the end so that if 8x8 data
# does not extend through `tstop` then the last record gets chopped.
dt = np.diff(np.concatenate([slot_data['TIME'], [tstop]]))
bad = np.abs(dt - 4.1) > 1e-3
slot_data[bad] = ma.masked
# Chop off the padding
i_stop = np.searchsorted(slot_data['TIME'], self.tstop, side='right')
slot_data = slot_data[:i_stop]
# explicitly unmask useful columns
slot_data['TIME'].mask = ma.nomask
slot_data['IMGSIZE'].mask = ma.nomask
slot_data['FILENAME'].mask = ma.nomask
slot_datas[slot] = slot_data
# make a shared time ndarray that is the union of the time sets in the
# slots. The ACA L0 telemetry has the same timestamps across slots,
# so the only differences here are caused by different times in
# non-TRAK across the slots (usually SRCH differences at the beginning
# of the observation)
shared_time = np.unique(np.concatenate([
slot_datas[slot]['TIME'].data for slot in slots]))
for msid in msids:
hdr3_msid = confirm_msid(msid)
slot = slot_for_msid(hdr3_msid)
full_data = ma.zeros(len(shared_time),
dtype=slot_datas[slot].dtype)
full_data.mask = ma.masked
fd_idx = search_both_sorted(shared_time,
slot_datas[slot]['TIME'])
full_data[fd_idx] = slot_datas[slot]
# make a data dictionary to feed to the MSID constructor
slot_data = {'vals': HDR3_DEF[hdr3_msid]['value'](full_data),
'desc': HDR3_DEF[hdr3_msid]['desc'],
'longdesc': HDR3_DEF[hdr3_msid]['longdesc'],
'times': shared_time,
'hdr3_msid': hdr3_msid}
self[msid] = MSID(msid, start, stop, slot_data)
def get_slot_image_data(start, stop, slot):
slot_data = aca_l0.get_slot_data(
start,
stop,
imgsize=[4, 6, 8],
slot=slot,
columns=['TIME', 'BGDAVG', 'IMGFUNC1', 'QUALITY', 'IMGSIZE'])
return Table(slot_data)
def plot_bgd(e, edir):
"""
Make a plot of BGDAVG over an event and save to `edir`.
Presently plots over range from 100 seconds before high threshold crossing to
300 seconds after high threshold crossing.
:param e: dictionary with times of background event
:param edir: directory for plots
"""
slots = [int(s) for s in e['slots_for_sum'].split(',')]
plt.figure(figsize=(4, 4.5))
max_bgd = 0
for slot in slots:
l0_telem = aca_l0.get_slot_data(e['cross_time'] - 100,
e['cross_time'] + 300,
imgsize=[4, 6, 8],
slot=slot)
if len(l0_telem['TIME']) == 0:
raise ValueError
ok = (l0_telem['QUALITY'] == 0) & (l0_telem['IMGFUNC1'] == 1)
if np.count_nonzero(ok) > 0:
plot_cxctime(l0_telem['TIME'][ok],
l0_telem['BGDAVG'][ok],
'.',
label=f'slot {slot}')
max_bgd = max([max_bgd, np.max(l0_telem['BGDAVG'][ok])])
plt.title("Hi BGD obsid {}\n start {}".format(
e['obsid'],
DateTime(e['event_tstart']).date),
fontsize='small')
plt.legend(numpoints=1, fontsize='x-small')
plt.tight_layout()
plt.margins(.05)
plt.ylim([-20, 1100])
plt.grid()
filename = "bgdavg_{}.png".format(e['event_datestart'])
plt.savefig(os.path.join(edir, filename))
plt.close()
return filename, max_bgd
def test_get_slot_data_8x8():
"""
Do a validation test of get_l0_images:
- Get 20 mins of image data for slot 6 of obsid 8008 (very nice clean stars)
- Do first moment centroids in row and col
- Compare to aspect pipeline FM centroids for same slot data
This is a deep test that all the signs are right. If not then everything
breaks badly because the star image doesn't move in sync with row0, col0.
"""
start = '2007:002:06:00:00'
stop = '2007:002:06:20:00'
slot_data = aca_l0.get_slot_data(start, stop, slot=6, centered_8x8=True)
files = asp_l1.get_files(8008, content=['ACACENT'])
acen = Table.read(files[0])
# Pick FM centroids for slot 6
ok = (acen['alg'] == 1) & (acen['slot'] == 6)
acen = acen[ok]
# Row and col centroids
times = slot_data['TIME']
# Easy way to do FM centroids with mgrid
rw, cw = np.mgrid[0:8, 0:8]
img_raw = slot_data['IMGRAW'] # np.round(slot_data['IMGRAW']).astype(int)
norm = np.sum(img_raw, axis=(1, 2))
rcs = np.sum(img_raw * rw, axis=(1, 2)) / norm + slot_data['IMGROW0'] - 1
ccs = np.sum(img_raw * cw, axis=(1, 2)) / norm + slot_data['IMGCOL0'] - 1
rcen = interpolate(acen['cent_i'], acen['time'], times)
ccen = interpolate(acen['cent_j'], acen['time'], times)
assert np.all(np.abs(rcen - rcs) < 0.05)
assert np.all(np.abs(ccen - ccs) < 0.05)
def make_images(start, stop, outdir='out', max_images=200):
"""
Make pixel plots of the a time range and write out to a directory.
:param start: start of time range (Chandra.Time compatible)
:param stop: end of time range (Chandra.Time compatible)
:param outdir: output directory
:param max_images: stop making image files if more more than this number
have been made
"""
start = DateTime(start)
stop = DateTime(stop)
slotdata = {}
for slot in range(8):
slotdata[slot] = Table(
aca_l0.get_slot_data(start.secs - 20, stop.secs + 20,
slot=slot).data)
slotdata[slot]['SLOT'] = slot
# Get a list of all the unique times in the set
times = np.unique(
np.concatenate([slotdata[slot]['TIME'].data for slot in range(8)]))
times = times[(times >= (start.secs - 4.5)) & (times <= (stop.secs + 4.5))]
SIZE = 96
rows = []
for time in times:
row = {'rowsecs': time, 'rowdate': DateTime(time).date, 'slots': []}
slot_imgs = []
for slot in range(8):
# Find last complete row at or before this time
last_idx = np.flatnonzero(slotdata[slot]['TIME'] <= time)[-1]
dat = slotdata[slot][last_idx]
imgraw = dat['IMGRAW'].reshape(8, 8)
sz = dat['IMGSIZE']
if sz < 8:
imgraw = imgraw[:sz, :sz]
imgraw = np.clip(imgraw, a_min=10, a_max=None)
if np.all(imgraw == 10) or dat['IMGFUNC1'] != 1:
pixvals = np.zeros((sz, sz))
np.fill_diagonal(pixvals, 255)
else:
# Log scale the images but use a fixed low end at 4
logimg = np.log(imgraw)
pixvals = 255 * (logimg - 4) / (np.max(logimg) - 4)
# Scale the image because the browser isn't great at it
pixvals = np.kron(pixvals, np.ones(
(int(SIZE / sz), int(SIZE / sz))))
# Set a border for stale data
img = (255 * np.ones(
(108, 108)) if dat['TIME'] < row['rowsecs'] else np.zeros(
(108, 108)))
img[6:102, 6:102] = pixvals
slot_imgs.append(img)
# bgcolor = 'gray' if dat['TIME'] < row['rowsecs'] else 'white'
row['slots'].append({
'stale': (dat['TIME'] < row['rowsecs']),
'slot': slot,
'time': dat['TIME'],
'rowsecs': row['rowsecs'],
'bgdavg': dat['BGDAVG'],
'imgfunc1': dat['IMGFUNC1'],
})
im = Image.fromarray(np.hstack(slot_imgs)).convert('RGB')
imgfilename = "piximg_{}.png".format(row['rowsecs'])
im.save(os.path.join(outdir, imgfilename), "PNG")
row['img'] = imgfilename
rows.append(row)
if len(rows) > max_images:
break
return rows