Transcript mm_considerations

1
Problems and Challenges in the
mm/submm
2
• Effect of atmosphere on data:
• Tsys
• Mean Refraction
• Phase fluctuations
• Correction techniques
• Other facility considerations
• UV-coverage / Dynamic range
• Mosaicing
Constituents of Atmospheric Opacity
• Due to the troposphere (lowest
layer of atmosphere): h < 10 km
Column Density as a Function of Altitude
• Temperature  with  altitude:
clouds & convection can be
significant
• Dry Constituents of the
troposphere:, O2, O3, CO2, Ne,
He, Ar, Kr, CH4, N2, H2
• H2O: abundance is highly variable
but is < 1% in mass, mostly in the
form of water vapor
• “Hydrosols” (i.e. water droplets in
the form of clouds and fog) also
add a considerable contribution
when present
Stratosphere
Troposphere
3
Optical Depth as a Function of Frequency
4
• At 1.3cm most opacity
comes from H2O vapor
• At 7mm biggest
contribution from dry
constituents
total
optical
depth
optical
depth due to
H2O vapor
optical depth
due to dry air
22 GHz
43 GHz
100 GHz
1.3cm
7mm
3mm
K band
Q band
MUSTANG
ALMA
• At 3mm both
components are
significant
• “hydrosols” i.e. water
droplets (not shown)
can also add
significantly to the
opacity
5
Opacity as a Function of PWV (PWV=Precipitable Water Vapor)
Sensitivity: System noise temperature
In addition to receiver noise, at millimeter wavelengths the atmosphere
has a significant brightness temperature (Tsky):
For a perfect
antenna,
ignoring
spillover and
efficiencies
Tnoise ≈ Trx + Tsky
where Tsky =Tatm (1 – e-) + Tbge-
so Tnoise ≈ Trx +Tatm(1-e-)
Receiver
temperature
Tatm = temperature
of the atmosphere
≈ 300 K
Tbg = 3 K cosmic
background
Emission from
atmosphere
Before entering atmosphere the source signal S= Tsource
After attenuation by atmosphere the signal becomes S=Tsource e-
Consider the signal-to-noise ratio:
S / N = (Tsource e-) / Tnoise = Tsource / (Tnoise e)
Tsys = Tnoise e ≈ Tatm(e -1) + Trxe
The system sensitivity drops rapidly (exponentially) as opacity increases
6
Atmospheric opacity, continued
Typical optical depth for 230 GHz observing at the CSO:
at zenith 225 = 0.15 = 3 mm PWV, at elevation = 30o  225 = 0.3
Tsys*(DSB) = e(Tatm(1-e-) + Trx)= 1.35(77 + 75) ~ 200 K
assuming Tatm = 300 K
 Atmosphere adds considerably to Tsys and since the opacity can change
rapidly, Tsys must be measured often
Many MM/Submm receivers are double sideband (ALMA Bands 9 and
10 for example) , thus the effective Tsys for spectral lines (which are
inherently single sideband) is doubled
Tsys*(SSB) = 2 Tsys (DSB) ~ 400 K
7
Interferometric MM Measurement of Tsys
• How do we measure Tsys = Tatm(e -1) + Trxe without constantly measuring Trx and
the opacity?
• The “chopper wheel” method: putting an ambient temperature load (Tload) in front of
the receiver and measuring the resulting power compared to power when
observing sky Tatm (Penzias & Burrus 1973).
Load in
Load out
Vin =G Tin = G [Trx + Tload]
Vout = G Tout = G [Trx + Tatm(1-e-) + Tbge- + Tsourcee- ]
assume Tatm ≈ Tload
Comparing
in and out
Vin – Vout
Vout
=
Tload
Tsys
Tsys = Tload * Tout / (Tin – Tout)
Power is really observed but is  T in the R-J limit
• IF Tatm ≈ Tload, and Tsys is measured often, changes in
mean atmospheric absorption are corrected. ALMA will
have a two temperature load system which allows
independent measure of Trx
SMA calibration load
swings in and out of beam
8
Example SMA 345 GHz Tsys Measurements
Tsys(4)
Tsys(8)
Tsys(1)
Good
Medium
Poor
Elevation
Note sharp rise in Tsys at low elevations
For calibration and imaging
VisibilityWeight 
1
Tsys (i)Tsys ( j)
9
SMA Example of Correcting for Tsys and
conversion to a Jy Scale
10
Tsys
S = So * [Tsys(1) * Tsys(2)]0.5 * 130 Jy/K * 5 x 10-6 Jy
SMA gain
for 6m dish
and 75%
efficiency
Raw data
Correlator
unit
conversion
factor
Corrected data
Mean Effect of Atmosphere on Phase
• Since the refractive index of the atmosphere ≠1, an electromagnetic wave
propagating through it will experience a phase change (i.e. Snell’s law)
• The phase change is related to the refractive index of the air, n, and the
distance traveled, D, by
fe = (2p/l) ´ n ´ D
For water vapor n  w
DTatm
so
fe  12.6p ´ w
l
w=precipitable water vapor (PWV) column
Tatm = Temperature of atmosphere
for Tatm = 270 K
This refraction causes:
- Pointing off-sets, Δθ ≈ 2.5x10-4 x tan(i) (radians)
@ elevation 45o typical offset~1’
- Delay (time of arrival) off-sets
 These “mean” errors are generally removed by the online system
11
Atmospheric phase fluctuations
12
• Variations in the amount of precipitable water vapor (PWV) cause phase
fluctuations, which are worse at shorter wavelengths (higher frequencies),
and result in
– Low coherence (loss of sensitivity)
– Radio “seeing”, typically 1 at 1 mm
– Anomalous pointing offsets
– Anomalous delay offsets
You can observe in apparently excellent
submm weather and still have terrible
“seeing” i.e. phase stability
Patches of air with different water vapor
content (and hence index of refraction)
affect the incoming wave front differently.
Atmospheric phase fluctuations, continued…
log (RMS Phase Variations)
Phase noise as function of baseline length
• “Root phase structure function”
(Butler & Desai 1999)
• RMS phase fluctuations grow as a
function of increasing baseline
length until break when baseline
length ≈ thickness of turbulent layer
Break
• The position of the break and the
maximum noise are weather and
wavelength dependent
log (Baseline Length)
RMS phase of fluctuations given by Kolmogorov turbulence theory
frms = K ba / l [deg]
b = baseline length (km)
a = 1/3 to 5/6
l= wavelength (mm)
K = constant (~100 for ALMA, 300 for VLA)
13
Residual Phase and Decorrelation
14
Q-band (7mm) VLA C-config. data from “good” day
An average phase has been removed from
Coherence = (vector average/true
absolute flux calibrator 3C286
visibility amplitude) = V/ V0
Where, V = V0eif
The effect of phase noise, frms,
on the measured visibility
amplitude :
Short baselines
V = V0 ´ eif = V0 ´ e-f2rms/2
(Gaussian phase fluctuations)
Example: if frms = 1 radian (~60
deg), coherence = V = 0.60V0
(minutes)
Long baselines
Residual phase on long baselines have larger
amplitude, than short baselines
For these data, the residual
rms phase (5-20 degrees)
from applying an average
phase solution produces a
7% error in the flux scale
VLA observations of the calibrator 2007+404
at 22 GHz (13 mm) with a resolution of 0.1 (Max baseline 30 km)
one-minute snapshots at t = 0 and t = 59 min
with 30min self-cal applied
Position offsets
due to large scale
structures that are
correlated 
phase gradient
across array
self-cal with t = 30min:
15
Sidelobe pattern
shows signature
of antenna
based phase
errors  small
scale variations
that are
uncorrelated
self-cal with t = 30sec:
All data:
Reduction in peak
flux (decorrelation)
and smearing due
to phase
fluctuations over
60 min
No sign of phase
fluctuations with
timescale ~ 30 s
 Uncorrelated phase variations degrades and decorrelates image
 Correlated phase offsets = position shift
 Phase fluctuations severe at mm/submm wavelengths, correction
methods are needed
• Self-calibration: OK for bright sources that can be detected in a few seconds.
• Fast switching: used at the EVLA for high frequencies and will be used at
ALMA. Choose fast switching cycle time, tcyc, short enough to reduce frms to
an acceptable level. Calibrate in the normal way.
• Phase transfer: simultaneously observe low and high frequencies, and
transfer scaled phase solutions from low to high frequency. Can be tricky,
requires well characterized system due to differing electronics at the
frequencies of interest.
• Paired array calibration: divide array into two separate arrays, one for
observing the source, and another for observing a nearby calibrator.
– Will not remove fluctuations caused by electronic phase noise
– Only works for arrays with large numbers of antennas (e.g., CARMA, EVLA, ALMA)
16
Phase correction methods (continued):
• Radiometry: measure fluctuations in TBatm with a radiometer, use these
to derive changes in water vapor column (w) and convert this into a
phase correction using
183
GHz
fe  12.6p ´ w
l
w=precipitable water
vapor (PWV) column
22 GHz
(Bremer et al. 1997)
Monitor: 22 GHz H2O line (CARMA, VLA)
183 GHz H2O line (CSO-JCMT, SMA, ALMA)
total power (IRAM, BIMA)
17
Testing of ALMA WVR Correction
Two different baselines Jan 4, 2010
Data
WVR
Residual
There are 4 “channels” flanking the peak of the 183 GHz line
•Matching data from opposite sides are averaged
•Data taken every second
•The four channels allow flexibility for avoiding saturation
•Next challenges are to perfect models for relating the WVR data to
the correction for the data to reduce residual
WVR correction will have the largest impact for targets that
cannot be self-calibrated and for baselines > 1 km
18
Test Data from Two weeks Ago
Raw
WVR Corrected
19
20
Absolute gain calibration
There are no non-variable quasars in the
mm/sub-mm for setting the absolute flux
scale
Flux (Jy)
Instead, planets and moons are typically
used: roughly black bodies of known size
and temperature:
ΔSn= 10 Jy
Uranus @ 230 GHz: Sn ~ 37 Jy, θ ~ 4
Callisto @ 230 GHz: Sn ~ 7.2 Jy, θ ~ 1.4
MJD
• Sn is derived from models, and can be
uncertain by ~ 10%
Flux (Jy)
• If the planet is resolved, you need to use
visibility model for each baseline
ΔSn= 35 Jy
• If larger than primary beam it shouldn’t be
used at all
MJD
Antenna requirements
• Pointing: 10 m antenna operating at 350 GHz the primary beam is ~ 20
a 3 error  (Gain) at pointing center = 5%
(Gain) at half power point = 22%
 need pointing accurate to ~1
 ALMA pointing accuracy goal 0.6
• Aperture efficiency, h: Ruze formula gives
h = exp(-[4psrms/l]2)
 for h = 80% at 350 GHz, need a surface accuracy, srms, of 30mm
 ALMA surface accuracy goal of 25 µm
• Baseline determination: phase errors due to errors in the positions of the
telescopes are given by
q = angular separation between
f = 2p ´ b ´ q
source & calibrator, can be
large in mm/sub-mm
l
 to keep f < q need b < l/2p
b = baseline error
e.g., for l = 1.3 mm need b < 0.2 mm
21
Summary
• Atmospheric emission can dominate the system
temperature
– Calibration of Tsys is different from that at cm wavelengths
• Tropospheric water vapor causes significant phase
fluctuations
– Need to calibrate more often than at cm wavelengths
– Phase correction techniques are under development at all
mm/sub-mm observatories around the world
– Observing strategies should include measurements to
quantify the effect of the phase fluctuations
• Instrumentation is more difficult at mm/sub-mm
wavelengths
– Observing strategies must include pointing measurements to
avoid loss of sensitivity
– Need to calibrate instrumental effects on timescales of 10s of
mins, or more often when the temperature is changing rapidly
22
Challenges
 Image sources larger than the primary beam (PB)
• at 1mm a 12m dish has PB~21”
 Mosaic
 Image sources with structure larger than the largest angular scale
• For shortest baseline of 15m (1.25*diameter) ~14” at 1mm
 Add total power from single dish
 Accurate continuum images in presence of copious line emission
and accurate delay calibration (bandpass)
 Spectral line mode all the time
 Sensitive linearly polarized feeds
• Many quasars are linearly polarized
 Full polarization calibration always
Image Quality
M17 VLA 21cm
Image quality depends on:
 U-V coverage
 Density of U-V samples
• Image fidelity is improved when high density
regions are well matched to source brightness
distribution
• U-V coverage isn’t enough
 DYNAMIC RANGE can be more important than
sensitivity
3’ = 1.2 kl
Heterogeneous Arrays
eSMA
8 6m
1 15m (JCMT) + 1 10.4m (CSO)
CARMA
6
10.4m
10
6.1m
8
3.5m
1
10.4m for total power
ALMA
50-60
12
4
12m antennas in main array (two designs)
7m antennas in ACA (Atacama Compact Array)
12m with nutators for total power
Large Nearby Galaxies
SMA ~1.3 mm observations
• Primary beam ~1’
3.0’
• Resolution ~3”
ALMA 1.3mm PB
ALMA 0.85mm PB
CFHT
1.5’
Petitpas et al. 2006, in prep.
Galactic Star Formation
BIMA 46 pointing mosaic
covering 10’ x 15’
CO(1-0) at ~115 GHz
~10” resolution
ALMA 0.85mm PB
Williams et al. (2003)
Mosaicing Considerations
 Each pointing ideally should have similar U-V coverage and hence
synthesized beams – similar S/N is more important
 Nyquist sampling of pointings
 On-the-fly mosaicing can be more efficient at lower frequencies
 Small beams imply many pointings
 At higher frequencies weather conditions can change rapidly
 Push to have very good instantaneous snapshot U-V coverage
 Polarimetry even more demanding for control of systematics due to
rotation of polarization beam on sky
 Accurate primary beam characterization
 Account for heterogeneous array properties
< 3 minutes!
Total Power Considerations
 Adding Single Dish (SD) zero-spacing tricky because
it requires
 Large degree of overlap between SD size and shortest
interferometer baseline in order to accurately cross-calibrate
 Excellent pointing accuracy which is more difficult with
increasing dish size
 *Comparable* sensitivity to interferometric data
 On-the-fly mapping requires rapid (but stable, i.e. short settle
time) telescope movement
 SD Continuum calibration – stable, accurate, large throws
(i.e. nutators)
Model Image
Spitzer GLIMPSE 5.8 mm image
• CASA simulation of ALMA with 50
antennas in the compact
configuration (< 100 m)
• 100 GHz 7 x 7 pointing mosaic
• +/- 2hrs
50 antenna + SD ALMA Clean results
Model
Clean Mosaic
+ 12m SD
+ 24m SD