Transcript home.strw.leidenuniv.nl

The Nova Fringe Tracker (NFT):
a second generation cophasing facility
for up to six telescopes at the VLTI
Jeffrey A. Meisner, Sterrewacht Leiden
Walter J. Jaffe, Sterrewacht Leiden
Rudolf S. Le Poole, Sterrewacht Leiden & TNO Science & Industry
NFT design (shown for 6 telescopes): simplicity!
Incoming beams
from M1 (pick-off)
M5 mirrors
PRS
Pinhole
Mask
M4
mirrors
Local Switchyard
Polarization
Reversers
Monolithic prism block of
polarizing beam splitters
Concept for the NFT proposed to
ESO in 2010 by the (former) NOVA
consortium which included:
Spectral prism
Wollaston prism
Camera lens
Detector
Back end of NFT,
Not sensitive to OPD,
No visibility loss possible due
to alignment etc.
Outline of this talk:





Fringe-tracking requirements at the VLTI and the Phase-A
studies towards a second-generation fringe-tracking facility,
ESO's (non-)decision, and subsequent (in-)action
Overview of the NFT, its technological choices, and
performance features
Strategies for the fringe-tracking control loop and simulation
results based on the proposed NFT design.
The particular problem of “photometric crosstalk,” and the
solution proposed using the Polarization-Based Collimated
Beam Combiner topology: how the NFT works.
The road forward toward a 2nd generation fringe tracker at the
VLTI (hopefully....)
So how did we get here?
A timeline:
2009: ESO calls for Phase-A studies to propose a concept
for a Second Generation Fringe Tracker for the VLTI
Requirements:
 Accept beams from 4 or 6 telescopes (not just 2) either from
science target itself or from off-axis reference star
 Measure phase for control of VLTI delay lines (fringe locking) for
long coherent exposures & phase referenced imaging
 Sensitive to group delay for dispersion control and fringe-jump
detection
 Tolerant of wavefront and photometric fluctuations
 Tolerant of different (possibly small) visibilities on some baselines
 Possibility of combining AT (1.8 meter telescope) with UT (8 meter
telescope), 20x brighter!
 Rapid update rate possible (up to 2 KHz) with best possible
limiting sensitivity (of course!)
Two well developed proposals submitted in response to the call
by ESO:
• POPS (Planar Optics Phase Sensor) submitted by LAOG
• NFT (NOVA Fringe Tracker) submitted by the NOVA
consortium
Concepts presented in May 2010 for ESO's consideration
Meeting at ESO in December 2011 to discuss and compare the
concepts, with inconclusive results.
It was generally agreed that:

Theoretical sensitivity of NFT significantly higher.
Technology used in POPS proposal already had been tested
and verified in on-sky interferometry, unlike that of the NFT.

The road forward: no agreement at all.
Differences:
NFT
POPS
Bulk optics interferometric
implementation, using a novel beamcombination topology

Only partial spatial filtering using
pinholes, adaptable to conditions


Ultra-wide-band (1.2 – 2.4 micron)
Direct combination of 4 baselines,
other two inferred

Nominally measures single
interference phase in order to generate
delay-line correction

Integrated optics beam-combination
using a fiber-fed IO device

Full spatial filtering by virtue of using
optical waveguide components


IO chip built either for K or H band
Direct combination of each of the 6
telescope pairs

Measures both quadrature phases of
interference

Similarities:

Both use pair-wise interference.
Both based on static measurement of interference phase (like PRIMA) rather than
phase stepping/scanning (like FINITO).

Both use low-resolution spectral dispersion to detect group-delay offset while cophasing (again, like PRIMA!).

Various concerns raised by ESO, and answered by additions to
the NFT design or the NFT analysis:
Questioning of only detecting one quadrature phase, with no
OPD modulation (whereas POPS detects both quadrature
phases, like PRIMA).

Questioning of not obtaining photometric information (POPS
can obtain the photometry of the individual telescope beams
indirectly).

Questioning of how the fringe OPD sensor would drive a
feedback loop to control the delay lines, as the NFT outputs did
not seem compatible with the current fringe tracker paradigms.

Questioning the “risk” of employing system technologies that
had never been tested on stellar light.

The NFT has carefully addressed the first three concerns, and
the fourth one can only be addressed (to the satisfaction of the
skeptics) by doing it!
Factors driving the design of the NFT:
Problem: a fringe tracker is needed to cophase 4 (or in the future 6)
VLTI telescopes in order that the science instrument can coherently
integrate (thus obtaining a good SNR on dim targets) and also to
obtain visibility phase either with respect to an off-axis reference star
or with respect to the target itself at the fringe-tracker's wavelength.
Therefore the ability to perform interferometry on a faint target is
dependent on the fringe-tracker's ability to lock onto its phase or the
phase of a reference star found nearby.
Thus the three most important performance goals driving the design of
the NFT are:
• Sensitivity
• Sensitivity
• Sensitivity
Achievement of highest limiting sensitivity (dimmest star on
which reasonably reliable tracking can be performed) through:

High throughput:
- Use of free space optical paths rather than integrated optics
- No need for spatial filtering (fiber injection losses) but still use pinholes for spatial mode
reduction in order to maximize performance (as demonstrated by simulations)
- Minimal number of non-cemented optical surfaces (AR coated)
Resulting throughput calculated: 85%, guaranteed: 80%

Minimize wavefront degradation before beam-overlap:
- 3 reflecting and 3 transmitting surfaces including VLTI pickoff and PBS surface
Result: calculated V=.977 at 2 m
Minimal spectral resolution (but variable) of 4 or 5 spectral channels over 1.2 – 2.4
m, just in order to gauge dispersion and fringe-jumps

In original configuration, use of only one of the two quadrature phases for
interference! Full quadrature detection wastes half of the light measuring the real
part of visibility which supplies no first-order information regarding OPD offset.

(However doing this over a very wide band requires a more elaborate dispersion compensator
than otherwise required). Partial quadrature detection available with 2011 hardware additions.
Likewise no optical power wasted on photometric monitoring which supplies no
first-order information regarding OPD offset.

-Variable photometric monitoring available with 2011 hardware additions.

Slower detector readout rate, when possible (but ESO requests 1-2 kHz update rates).
Incoming beams
from M1 (pick-off)
M5 x1
PRS
align
pinhole
fixed
Align, coarse
OPD + piezo
fixed
M4 x1
/2 plates
NFT
Upgradability
align
PBS
Spectral prism (optional)
Wollaston prism
Camera lens
Detector
Minimal configuration to verify components
and performance on 1 baseline
Incoming beams
from M1 (pick-off)
M5's
PRS
M4's
Mask
Local Switchyard
Polarization
Reversers
(rotatable)
NFT
Upgradability
End-around
combiner
Spectral prism
Wollaston prism
Camera lens
Detector
Configuration for up to 4 telescopes.....
Incoming beams
from M1 (pick-off)
M5's
PRS
M4's
Mask
Local Switchyard
Polarization
Reversers
NFT
Upgradability
End-around
combiner
Spectral prism
Wollaston prism
Camera lens
Detector
All 6 channels now installed
(unless you want even more!)
Further advantages of the technological choices adopted
by the NFT:
Not acutely affected by input polarization mismatches, no
SNR loss at all due to birefringence between vertical and
horizontal polarizations, since each pairwise beam combination
only uses one or the other.

Beam overlap is performed very early in the optical chain,
after which OPD shifts, vibrations, or poor optical surface
quality has no effect on interferometric visibility or the measured
phase of interference.

A 4 or 6 beam combining topology can be used at full
efficiency with only N telescopes if N is even. Modest SNR loss
if N is odd.

Delay-tracking control loop and simulation of NFT
Control loop not part of NFT proposal per se, but used in NFT
simulations to demonstrate suitability of hardware.
• A basic linear fringe-tracking algorithm proposed, identical to
what the analog electronics in the experimental setup did:
Delay line motion = gain * Im{V} (too simple!).
• A complicated algorithm was developed (but this had not been
a requirement of the original ESO call!) to examine the detector
outputs in 2 or more spectral channels in order to detect fringe
jumps and order correction jumps. Also generates a judgement
of “verified tracking intervals” needed in order for a science
instrument to (properly) employ the fringe-tracker's data to the
fullest. This was successfully simulated.
• Specifically in response to ESO, a method was developed for
the detection of loss of tracking (actually several parallel
indicators have been identified!), and successfully simulated.
Using a frame time of 5ms, tracking is successful at a K band photon flux of only
70,000 per second with a typical tracking rate over 80%. Periods when the tracker was
locked onto a side fringe (three are shown) are detected by the tracking algorithm after
typically 1/2 second, leading to a correction and invalidation of the intervals where it
may not have been correctly tracking (“validated” intervals are marked in green). Well
less than 1% of “validated” intervals are found to actually have been off of the intended
fringe peak, causing an apparent visibility reduction in the science instrument of less
than 1%. OPD is measured in femtoseconds. The left and right plots represent two
different detector noise realizations when tracking the same atmospheric OPD and
incident wavefronts.
1%
criterion
Tracking statistics predicted for 1ms detector frame rate, tracking on H & K band light
The problem of “photometric crosstalk” due to a beamcombiner lacking photometric symmetry. This is a key issue
addressed by the particular topology of the NFT.
An ideal 2 x 2 on-axis beam combiner:
Ideal 50/50
Beamsplitter
EA
E1 Detector
EB
I1
E2
I2
Detector
+
-
|V| cos()
The problem of “photometric crosstalk” due to a beamcombiner lacking photometric symmetry. This is a key issue
addressed by the particular topology of the NFT.
An actual beam combiner:
Ideal 50/50
Beamsplitter
EA
Photometric
crosstalk
E1 Detector
EB
I1
E2
I2
Detector
+
-
|V| cos()
+  (IA – IB)
where:
= ½ (R - T)
Visibility estimator I1 – I2 inevitably includes a
photometric crosstalk term =  (IA – IB)
Naive approach to prevent photometric crosstalk:
Force  =0 by making the beamsplitter's T = R = ½
But:
 It will not generally be constant over wavelength
 It will almost always be different between the polarizations
 Moreover this is difficult to do in the first place!
For instance, the VINCI (fiber) beam combiner, regularly adjusted
for maximum fringe contrast, had a very unpredictable  (almost
never near zero!) as plotted over 3 years:

+.6
.4
.2
0
-.2
-.4
-.6
Cancelling photometric crosstalk term using
photometric pick-offs:
Split off a fraction m of the incoming optical powers, and add
them in the correct proportions to the interferometric outputs.
A
Drawbacks:
m IA
I1
m IB
I2
+
-
• Photometric monitoring
beams rob power from the
interferometric channels
• Added photometric
corrections m IA and m IB
contain their own detector
noise, added to resulting
visibility determination
• Want cancellation in both
polarizations and at all
wavelengths: generally
impossible!
Combating photometric crosstalk by chopping the phase
(modulation of the OPD): This is the most common solution!
For instance: ABCD phase stepping, or scanning of fringe packet
 Then the visibility appears as an AC signal on the photodetector.
 Just ignore the DC (and low frequencies) then.
 So measure
|E1+ E2 exp(j )|2 +  (IA – IB) as a function of (t), and the
photometric crosstalk term can be eliminated (if it stays constant!).
OPD
Modulator
A
B
C
D
(for instance)
Coherent
Detection
Estimate of
complex V
Drawbacks:
 Requires more than one detector readout to measure a visibility
 Since the photometrys and OPD are changing due to the
atmosphere, these fluctuating components leak into the result.
 In order to reduce that effect, a faster readout speed is required,
reducing sensitivity in a NIR instrument.
Solution employed by the NFT: precise 50/50 beamsplitting through use
of the “Polarization-Based Collimated Beam Combiner” topology!
• Combines beams pairwise.
• Each telescope's light is split by polarization, to be combined with 2
other telescopes.
• Each combination produces 2 (or more) interferometric outputs based
on balanced combination: visibility estimates are immune to photometric
crosstalk!
Requires 3 essential stages:
1
Beam 1
Beam 2
Beam 3
Beam 4
H
Polarization
Reversers
(even channels only)
p
V
H
s
p
V
s
2
Polarization
Recombined
Beam
Polarization
Recombinaton
Stage
3
Detectors
Polarization
Analyzer
@ 45o
2-phase detection
configuration shown
Visibility
Estimate
+
-
Polarization Recombining Stage (PRS)
2
Consists of multiple Polarizing Beamsplitters (PBS)
 s polarization from telescope M is paired with p polarization of
telescope M+1 into same spatial mode but a different polarization (thus
a different mode)
 Thus the “polarization recombined beam” can proceed through various
(non polarized) optical elements and both waves are affected identically
 Only when they finally reach the polarization analyzer are the two
waves actually interfered and directed onto 2 (or more) photodetectors

Original
Polarization
H
V
2s
2p
V
H
3s
H
V
4s
4p
Telescope
3p Beams
Polarizing
Beam Splitters
Original
Polarization
Polarization
Recombined
Beams
1s
2p
V
V
2s
3p
H
H
3s
4p
V
V
Important point:
After the PRS, the “polarization recombined beams” no longer need to be treated
according to interferometric standards. OPD variations/instability, wavefront
degradation do not affect the visibility or rejection of photometric crosstalk!
3
Polarization Analyzer produces output beams implementing a
balanced beamcombiner. When their powers are detected and
differenced, photometric crosstalk is suppressed!
Detector A
Telescope 1
Telescope 2
Now tilt your
head by 45o!
Beam from the polarization
recombination stage (PRS)
has the two telescopes' light
in H and V polarizations
1s
2p
45o
Rotation
of
Coordinate
System
Detector B
Analyze the polarizations in the A
and B directions to get the
opposite interference phases with
equal photometric contributions!
Polarization Analyzer
in rotated system
IA
A = 1s + 2p
+
-
B = 1s - 2p
Detectors
IB
Differential
Amplifier
Immunity to photometric crosstalk demonstrated in experiment conducted at TU
Delft. Interference from He-Ne laser (left) detected using Polarization-Based
Collimated Beam Combiner” topology, with added incoherent light (“photometric
noise”) injected into each “telescope” beam from independently pulsating laser
diodes. Cancellation of photometric noise by ratio >> 100:1 (stable over weeks)
with intended interference signal retained.
Pulsating laser diodes
Injected “Photometric noise”
He-Ne
laser
Coherent “starlight”
split toward both
“telescope beams”
“Telescope beam 1”
“Telescope beam 2”
Piezo
Polarization
analyzer
(PBS) at 45o
and detectors
Each detector output (left) contains an equal combination of the two
photometric signals, but an opposite amount of the visibility signal!
A+B
(photometry
only) in red
A-B
(photometry
rejected)
in white
Shown: actual recorded scope data of detectors (left) and sum and
difference (right) while scanning the OPD. The “photometric noise”
was much (~ 10x) stronger than the “starlight,” but is completely
rejected in the difference signal. In locked mode (with feedback to
the piezo, not shown) the same photometric rejection is present.
Modification of
response by
1s
2p
inserting an
additional optical
element:
45o
Rotation
of
Coordinate
System
Polarization Analyzer in
rotated system
IA
+
-
IB
Original system: IA – IB = Re{V}
/4 plate
at 0o
45o
Polarization Analyzer in
rotated system
IA
Rotation
of
Coordinate
System
1s
2p
+
-
IB
Detection of quadrature phase: IA – IB = Im{V}
/2 plate
at 22.5o
1s
2p
45o
Polarization Analyzer in
rotated system
IA
Rotation
of
Coordinate
System
+
-
IB
Photometric detection IA = I1
IB = I2
Modification of NFT beam combiner in order to allow for (partial) detection
of quadrature visibility component and/or photometry
Basic implementation
IA – IB = Re{V}
Wollaston
Prism
Camera
Lens
Detector
Array
A = 1s + 2p
IB
B = 1s - 2p
IA
IB IA etc.
Augmented
system:
I'A – I'B = Im{V}
I''A = I1
I''B = I2
/2 plate
with wedge
/4 plate
with wedge
Proposed layout on
PICNIC detector:
I'B I'A
I'
etc.
A
Slightly deflected beams with quadrature or photometric
information detected on adjacent pixels, for instance
So, adding it all up:
The VLTI and its second-generation instruments (particularly
MATISSE) requires an updated fringe-tracker in order to
achieve its potential.

The NOVA consortium submitted a concept which meets all of
the requirements and achieves a high sensitivity (limiting
magnitude).

The proposal relies on standard optical components and
techniques, is easily understood and modelled.

The unique feature of the beam combiner has been tested in
the laboratory and (unsurprisingly) works.

A stripped-down version for only one baseline or otherwise
simplified could be easily and relatively cheaply built and tested
on the sky, for further verification.


So, where does ESO go from here?
The end
WRITE HERE!
Modulation of the OPD -II
Drawbacks:
• Requires more than one detector readout to measure a visibility
• Since the photometry (and OPD!) is changing due to the
atmosphere, this fluctuating component leaks into the result.
• In order to reduce that effect, a faster readout speed is required,
reducing sensitivity in a NIR instrument.
Example: the incoherent power spectrum (top) and coherently
integrated power spectrum (bottom) of 4 consecutive VINCI
observations (14 Aug. 2001) of eps sco at different framerates:
Good!
LF
Noise
More
White
Noise
Noisier
590 Hz: too slow. Spectrum is
broadened by atmospheric OPD
Just right! (For this
VERY bright star!)
3384 Hz: too fast. Detector noise
enhanced due to short exposure
Drift
Upward
Wiggle above 3
level, on fringe
Tracking
OK
Checking
Wiggle
Lost
Wiggle
Find
direction
of last
confirmed
tracking
Wiggle above 3
level, on fringe
Set limit
to drift
At limit of drift,
Reverse drift,
increase limit in
other direction
Drift
Down
Induced “telescope failures” simulated at these times, causing tracking failure
Sprial
search
pattern
invoked
following
detection
of tracking
failrure
Green
indicates
intervals
which have
been
“validated” as
acceptable
for use of
data
produced by
science
instrument.
See NFT main document or update document for explanation.
A laboratory setup of the concept of the Polarization-Based
Collimated Beam Combiner was carried out in order to verify and
demonstrate its performance goals.
Many thanks to the TU Delft applied optics department and
especially to the optics technician Thim Zuidwijk for facilitating.
Laser Diodes (source of “photometric interference”)
He-Ne
Laser
He-Ne
Laser
Polarization at ~
45o
LD
Detector assembly
rotates about beam
axis, nominally at 45o
LD
Polarizing
Beamsplitters
Photodiode
Aperture
Mask
Polarization
Analyzer &
Detectors
Photodiode
Photodiode
Photodiode
Scan
(for testing)
+
Control loop
filter integrator
Lock
HV
Amp
Photodiode preamps,
differencing circuit,
Actual application, using starlight!
Polarizing
Beamsplitters
Beam from
Telescope 1
Beam from
Telescope 2
1s + 1p
Polarization
Reverser
Vertical polarization
in VLTI tunnels
Polarization
Analyzer at 45o
1s
etc.
1s + 2p
2s + 2p
And/or VLTI delay
line, stroke = 120m
Real-time
Computer
HV
Amp
2s
Detector
Array
To next
combiner
Detector
Read-out
Electronics
The signal from just a single photodetector (without
balancing from the other one) showing the combined
waveform of the two laser diodes (incoherent source), one
at about 1KHz, the other at 4KHz. These added
“photometric fluctuations” are both highly rejected by the
differential amplifier
when the angle of the analyzer is
o
correctly set to 45 and the relative gains are set
appropriately.
He-Ne
Laser
LD

LD
Aperture
Mask
ATMOSPHERE
CORR
Hot turbulent air
from hair dryer
HV
Amp
(for testing)
+
Polarization
Analyzer &
Detectors
Photodiode

See movie 1
basic demo
Control loop
filter integrator
Photodiode
Photodiode
Photodiode preamps,
differencing circuit,
Lock
VPD = sin()
He-Ne
Laser
Added!
LD
See movie 2
dim beam
LD
Neutral
Density
Filter
x .03
Aperture
Mask
Polarization
Analyzer &
Detectors
Photodiode

ATMOSPHERE
Photodiode
Photodiode
HV
Amp
(for testing)
+
Control loop
filter integrator
Photodiode preamps,
differencing circuit,
Lock
VPD = sin()
He-Ne
Laser
LD
LD
Optical disturbance
following PRS does
not degrade visibility
or photometric
symmetry
PRS
See movie 3
wine glass
Polarization
Analyzer &
Detectors
Photodiode

ATMOSPHERE
Photodiode
Photodiode
HV
Amp
(for testing)
+
Control loop
filter integrator
Photodiode preamps,
differencing circuit,
Lock
VPD = sin()