Interferential scanning grating position sensor

operating in space at 4 K
Guy Michel, Kjetil Dohlen, Jerome Martignac, Jean-Claude Lecullier, Patrick Levacher,
and Claude Colin

An interferential position sensor for operation in space at a deep cryogenic temperature 4 K is derived
from a commercial sensor. The application is for the Spectral and Photometric Imaging Receiver
submillimetric imaging Fourier-transform spectrometer on the Herschel space telescope. This sensor is
used to control the displacement of the interferometers moving mirrors and to sample the interferograms.
This development addresses the following points: minimization of the effects of cooling critical optical
parts, introduction of a fully redundant focal plane, selection of optoelectronic components efficient at 4
K, and design of a cryogenic preamplifier. 2003 Optical Society of America
OCIS codes: 120.0120, 120.4120, 120.6810.

1. Introduction

A Spectral and Photometric Imaging Receiver

SPIRE is one of the three focal plane instruments of
the European Space Agency Herschel mission. The
instrument is dedicated to imaging spectroscopic
studies of deep extragalactic and galactic sources.
The satellite consists of a 3.5-m telescope passively
cooled to 80 K and a liquid-helium cryostat that surrounds the instruments. The SPIRE is an imaging
Fourier-transform spectrometer that operates at 4 K
and covers the spectral range from 200 to 670 m
with a maximum spectral resolving power of 1000 at
200 m.1 Working at such a long wavelength
should in principle relax the requirements on the
sampling accuracy of the interferograms. Nevertheless, the large dynamic range of the interferogram
peak at zero path difference has a strong effect on the
required sampling accuracy since the goal is to recover the spectral continuum with high precision.
The sampling accuracy has been specified to better
G. Michel guy.michel@obspm.fr and J.-C. Lecullier are with
Observatoire de Paris, Laboratoire dEtudes Spatiales et
dInstrumentation en Astrophysique, 5 place Jules Janssen, 92195
Meudon Cedex, France. K. Dohlen, P. Levacher, and C. Colin are
with Laboratoire dAstrophysique de Marseille, 2 Place Le Verrier,
13248 Marseille Cedex 4, France. J. Martignac is with the Commissariat a lEnergie Atomique, Service dAstrophysique, Batiment 709, Orme des Merisiers, 91191 Gif sur Yvette, France.
Received 10 March 2003; revised manuscript received 8 July
2003.
0003-693503316305-09$15.000
2003 Optical Society of America

than 0.1 m over a stroke of 4 cm. Besides, the

transducer has to comply with the operational mode
of the SPIRE, a so-called fast scanning interferometer. The interferogram is time sampled,2,3 and, as
samples are collected at fixed time intervals, each
sample is associated with its position. Usually an
oversampling factor is applied depending on the velocity fluctuations of the controlled drive mechanism.
Then the interferogram is converted by an algorithm
into an interferogram resampled at equal path difference intervals prior to fast Fourier transform.
Among the candidate position sensors investigated
for this application were the laser interferometer, the
linear variable differential transformer LVDT, and
the interferential scanning grating ISG. They
were compared with respect to resolution, linearity,
power dissipation, and quality of translation required
from the scan mechanism.
The laser interferometer is excellent in terms of
accuracy and linearity but it must be custom designed and requires significant development for the
qualification of optical parts such as the monomode
diode laser light source.
The LVDT has been flown on space missions in a
cryogenic environment at 4 K.4 It is a robust sensor
with no active component at cold temperature.
However it is marginal for the accuracy we need and
has poor linearity even for Fourier spectroscopy at a
moderate resolution. Besides, for a stroke of 4 cm, it
is somewhat bulky and requires a long moving core to
be held with critical tolerances. Redundancy is difficult to apply because of the reduced space available.
The ISG has the required accuracy and linearity
1 November 2003 Vol. 42, No. 31 APPLIED OPTICS

6305

Table 1. Comparison of Sensor Options

Parameter

SPIRE
Req.

LVDT

ISG

Laser
Interf.

Stroke cm
Resolution nm
Linearity
Power diss. mW

4
100
104
1

4
300
2.5 103
0.5

4
10
105
1

4
1
Perfect
1

Req., requirements; Interf., interferometer; diss., dissipation.

but it is not commercially available as a cryogenic

space-qualified subsystem. This sensor requires a
reasonably good quality of the translation stage residual acceptable rotation in the minutes-of-arc
range, but this has not been found to be too constraining. A similar moire -type scanning grating
sensor has been developed for the 4 K space instrument, the Far-Infrared Absolute Spectrophotometer
FIRAS.5
Table 1 summarizes the trade-off among the three
options. Although the LVDT is insufficient in terms
of resolution and linearity, the two remaining candidates require important space qualification efforts.
We found that investing in the modification of an
existing commercial ISG transducer would be less
challenging and more adapted to our budget than the
development required for the laser interferometer.
We confirmed this decision when we cooled the original commercial transducer. Fringes were still
present down to 50 K although with a severe loss in
contrast and amplitude. This gave us the idea of
modifying the commercial model to make it work
with acceptable performances for our project.
Compared with the Far-Infrared Absolute Spectrophotometer project in which the optical signals were
fed in and out of the instruments by fibers, we preferred having active components at 4 K, which avoids
the use of special radiation shielded feedthrough connectors for the optical fibers in the cryoharness. For
reliability we introduced full source and detector redundancy at the sensor focal plane. The sensor is
compact; it includes a small footprint optical head
and a moving scale grating of 5-cm length.
This transducer gives only the relative position.
It is associated with a reversible counter that yields
the incremental position by fringe counting. To get
the absolute position, this counter has to be reset at
a well-known position of the interferometer translation stage. For the SPIRE this is done by a short
stroke a few millimeters LVDT placed just before
the zero path difference, providing a start position
fiducial mark. For such a short stroke, the LVDT
has the required resolution and its volume is acceptable.
In Section 2 we describe the principle of the ISG
sensor and give the theoretical background required
to understand the effects of cryogenic space operation
of the sensor. In Section 3 we explain the modifications applied to the commercial sensor to make it
useful for the SPIRE. In Section 4 we discuss various factors that influence the transducers resolution.
6306

APPLIED OPTICS Vol. 42, No. 31 1 November 2003

Fig. 1. Sketch of the Heidenhain LIP 401A sensor.8

2. Principle of the Interferential Scanning Grating

Sensor

ISG sensors have been used for a long time to control

the ruling of gratings6,7 and to measure the displacement of machine tools. Here we present some of the
theoretical background and important features of the
commercial model8,9 that we selected that are essential to analyze and counter the effects of cooling the
sensor to liquid-helium temperatures.
A.

Description of the Sensor

The ISG sensor Fig. 1 consists of a short reticle

grating and a long scale grating of identical ruling
period . The reticle grating is located in an optical
head together with collimating optics, a LED light
source, and photodetectors. Light that emanates
from the LED, located in the focal plane of the plano
convex collimating lens, undergoes three successive
diffractions: first, at the transmitting reticle grating order i; second, at the reflecting scale grating
order j; and third, at a second passage through the
reticle grating order k. The result of these diffraction processes is to spread the light into a number of
beams, each identified by the sequence of diffraction
orders, i, j, k, exiting at an angle i j
k n, where is the wavelength and n is the
composite diffraction order. Figure 2 illustrates
these diffraction processes. Certain of the beams

Fig. 2. Equivalent optical diagram showing the three successive

diffractions, twice at the transmissive reticle grating and once at
the reflective scale grating. Order combinations whose direction
corresponds to one of the three detectors are labeled, and thick
lines and boldface indicate the principal beams.

Fig. 3. Sketch of the two versions of the fully redundant focal plane. All the results cited in this paper were obtained with a the first
focal plane layout. The b second layout implemented in the qualified sensor is more compact and reduces vignetting. The Si detectors
are labeled 1, 0, 1 in reference to the orders of diffraction.

produced, referred to as principal beams and drawn

as heavy lines in Fig. 2, are of particular interest by
virtue of their pairwise identical optical paths. Interference between pairs of principal beams creates
the signal upon which this sensor is based.
As indicated in Fig. 2, all the principal beams have
a composite order of 0 or 1. To measure the interference between these beams, it is therefore sufficient
to use three detectors corresponding to the directions
0 and . Also, the principal beams are formed
only by the 0 and 1 orders of the reticle grating and
the 1 orders of the scale grating. The reticle grating is therefore designed to have approximately equal
efficiency in the 1, 0, and 1 orders, and the scale
grating, by virtue of its quarter-wave groove depth,
has zero transmission in the 0th order at the design
wavelength.
Other beams that exit the system into the same
directions as the principal beams have nonzero optical path differences. High-contrast interference can
occur between these beams only if the optical path
difference is smaller than the coherence length of the
light source, given by 2, where is the optical
bandwidth of the source. It can be shown that the
smallest optical path difference of nonprincipal beam
pairs are of the order of 60 m. To avoid spurious
interference patterns, it is therefore important to ensure that the source coherence length remains below
60 m, which for 0.8 m corresponds to 10
nm. This is one of the concerns for our qualification
since LED bandwidths tend to diminish at colder
temperatures, as will be demonstrated further on.
The collimating lens refocuses the principal beams
onto three photodiodes. With a 4-m grating period
and an 0.8-m wavelength, the principal directions
are separated by 11.3, translating into a separation
between detectors of 2 mm for a collimator focal
length of 9.8 mm. To avoid superposition of the LED
and the central detector diode, they are shifted apart,
away from the optical axis in the direction perpendicular to the diffraction.
Figure 3 shows two different versions of the focal
plane layout used in our modified sensor. Their

principal difference with the commercial version is

the doubling of source and detectors, which is required for redundancy. The two sets of LED and
detectors are labeled FP1 and FP2, and, to avoid
overlap, their off-axis distance is not the same. Figure 3a shows the first version, where both LEDs are
on the same side of the optical axis. All the results
shown in this paper were obtained by use of this
layout. As discussed in Subsection 2.D, the large
off-axis distance of FP2 introduces serious vignetting.
To reduce this problem, a different focal plane layout
has been implemented, as shown in Fig. 3b, where
the two LEDs are on opposite sides of the axis, allowing a 30% reduction in off-axis distance. As expected, this has all but eliminated the vignetting
problem. In both designs, the FP1 configuration
suffers from stray light on the central detector because of its proximity to the LED. Including a baffle
could reduce stray light, but this option was rejected
because it would be difficult to implement and because it was not required from a performance point of
view.
B.

Phase Modulation

Relative displacement between the two gratings in

the direction perpendicular to the grooves creates a
variation of the phase between interfering wave
fronts, hence modulation of the measured signals.
As shown in Fig. 4, a translation of the grating by a
distance x induces a phase shift independent of opti-

Fig. 4. Illustration of the phase modulation induced on the diffracted beam that is due to movements of the grating.
1 November 2003 Vol. 42, No. 31 APPLIED OPTICS

6307

Fig. 5. Reflection coefficient of the scale grating measured at

near-normal incidence in the 600 900-nm range.

cal wavelength of 2nx. Since interfering

beams have undergone diffraction of opposite orders,
n 1, the effective phase difference is 2, and so
the period of the measured signal is half of the grating period or 2 m in the present case.
The scale grating is built with a groove depth of
approximately 4 for 800 nm, minimizing the
zero-order reflected beam. Indeed, a phase grating
with a square groove profile, an aspect ratio of one,
and a groove depth of 4 has no even orders. Odd
orders are present, but only the 1 orders are significant. When cooled down to 4 K, the grating will no
longer be in this configuration, mainly because of a
shift in the LED central wavelength of some 30 nm
see Subsection 3.B. The effect of this change is
insignificant, however, as indicated by our spectrophotometric characterization of the zero-order reflectivity of the scale grating Fig. 5.
The reticle grating is a rectangular-groove transmission grating with a groove depth and an aspect
ratio designed to produce equal efficiency in the 0 and
1 orders and phase shifts of 0 and 60, respectively. Since this grating is used in double pass, this
leads to a phase shift of 120 between the output
signals from each of the three detectors. An electronic signal conditioner transforms these three signals into two signals phase shifted by 90, referred
to as quadrature signals. Besides determination
of the direction of motion, this also permits calculation of the exact position between zero crossings;
see Section 4.
C.

of a millimeter would have any significant influence

on the interferometer performance.
The system is also insensitive to aberrations in the
collimating lens. The ray tracing of Fig. 6 indicates
good overlap of the two interfering beams. Figure 7
shows that, although each beam suffers from approximately 100 wavelengths of coma a, the difference
between them b is flat to the numerical precision of
the calculations. This feature rules out any reduction of contrast that is due to thermally induced aberrations in the optical system.
D.

Sensitivity to Angular Alignment

The sensor is sensitive to relative rotation of the gratings about the direction normal to their plane. Such
rotation produces a shift of the diffracted beams in the
focal plane of the collimator lens and hence, by virtue
of the van CittertZernike theorem, results in reduced
fringe contrast. For a rectangular pupil mask, the
contrast is described by a sinc function, and the first
zero occurs when the rotation is such that the two ends
of a reticle grating groove coincide with neighboring
grooves in the scale grating. Shortening the height of
the gratings at the cost of reduced throughput therefore makes the system less sensitive to rotation.
Reduced height of the gratings can be achieved either voluntarily by changing the size of the mask
placed at the reticle grating, or involuntarily by vi-

Common-Path Interferometer

An important feature of the ISG sensor is that the

interfering beams follow almost identical paths
through the optical system. This is illustrated in
the ray tracing of Fig. 6, where the ray paths of two
principal beams are superimposed. In particular,
since the distance between the reticle grating, where
beam separation and recombination occurs, and the
scale grating is small, the effect of optical defects in
grating substrates are negligible unless they suffer
from severe small-scale defaults such as bubbles in
the reticle substrate. With a separation between
the two gratings of approximately 0.5 mm and a maximum angular beam separation of 23, only 4 variations in the optical path on the scale of a few tenths
6308

Fig. 6. Ray tracing of the ISG sensor showing the main parts of
the system and two different principal beams. Apart from a slight
difference in path between the two gratings, the two beams follow
identical paths through the system.

APPLIED OPTICS Vol. 42, No. 31 1 November 2003

Fig. 7. Wave-front plots generated by ray tracing showing a the

wave-front error of a beam after double passage through the system and b the difference between two interfering wave fronts.
Although each beam suffers from approximately 100 wavelengths
of coma, the difference is flat to the numerical precision of the
calculations.

Fig. 8. Results of contrast measurements for each focal plane in

the first version compared with best-fit sinc functions. FP2 is four
times more tolerant to rotation than FP1 because of vignetting.

gnetting the beam as it passes through a baffle mask

placed within the system. The latter case was implemented in the first version of our modified focal plane,
shown in Fig. 3a. Figure 8 shows the results of contrast measurements for each focal plane in this system, clearly illustrating the sinc nature of the contrast
reduction and the large difference in rotation sensitivity that is due to vignetting. FP2 is four times more
tolerant to rotation than FP1, as a result of a fourfold
increase in vignetting. The amount of vignetting has
been found to increase sharply at an off-axis distance
of approximately 2.5 mm. With the new focal plane
layout shown in Fig. 3b, where the off-axis distance of
FP2 is reduced from 3 to 2.1 mm, further measurements have shown that the vignetting, and hence rotation sensitivity, of FP1 and FP2 are nearly equal.
Ray tracing of the system illustrates this situation see
Fig. 9. Although the resulting sensitivity to rotation
is compatible with the expected performance of the
scan mechanism, eventually, pending measurements
of actual mechanism performance, it might be necessary to adjust the sensitivity by changing the grating
mask. Such adjustments would act equally on both
FP1 and FP2 sensitivity.
3. Development of the Sensor in View of Its
Application to the Spectral and Photometric Imaging
Receiver Instrument

To meet the cryogenics and space requirements,

many issues had to be solved, namely, the severe loss
of contrast upon cooling, the complete loss of sensi-

Fig. 9. Ray tracing of the sensor head illustrating vignetting for

off-axis distances of 1 and 3 mm.

Fig. 10. Measured contrast as a function of the temperature for

the original optical head circles and two versions of the modified
optical head: four cuts squares and eight cuts triangles. Broken and dotted curves show simulated contrast for a clearance of 0
and 2.5 m at the grating mount, respectively.

tivity of the silicon photodetectors below 50 K, and

the necessity to operate with a power dissipation of
only 1 mW at 4 K. The solutions to these problems,
involving mechanical modifications and replacement
of optoelectronic components, are described in the
following. For reliability we introduced full redundancy of the critical components by designing a focal
plane with two fully independent sections, each complete with photoelectronic components and associated cryogenic preamplifiers. Resistance of the
modified head to operation at 4 K has been verified by
the fact that no performance degradation was observed after at least ten cooldown cycles to 4 K. In
addition, a qualification program according to European Space Agency standards is in preparation.
A.

Effects on the Contrast of Cooling from 300 to 4 K

Both gratings are engraved on a Zerodur substrate

and therefore have identical coefficients of expansion
leading to a good match on cooling. The reflection
grating is mounted on the moving side of the interferometer drive mechanism, and held only in its center part by a fixture. It is almost free of additional
contraction that is due to its mount. Conversely, the
reticle grating is inserted into a stainless steel mount
with a tight fit. The cooling produces radial contraction forces that squeeze the reticle grating, leading to
a change in grating period, which has a strong effect
on the contrast. Calculations show that a contraction of 2 m over the full 6000 m length of the reticle
grating reduces the contrast to zero, and that the
functional form of the contrast variation with grating
contraction is that of a sinc function.
Figure 10 shows the measured contrast for the
original optical head as a function of temperature
circles. This curve can be compared with the simulated contrast, calculated with the radial stress derived from a formula for lens mounts10 in the case of
a clearance of 0 m broken curve and 2.5 m dotted
curve. The experimental curve is in good agreement with the simulated curves and falls to zero at
around 180 K, approximately midway between the
first zero of the simulated curves. The actual clearance is therefore estimated to be 12 m.
1 November 2003 Vol. 42, No. 31 APPLIED OPTICS

6309

Fig. 11. Actual modified optical head with eight cuts.

To achieve an acceptable contrast value, it is necessary to decrease the thermal squeezing. We found
that this is possible by machining longitudinal cuts
into the cylindrical grating mount, as seen in Fig. 11.
The resulting performance is plotted in Fig. 10 for
four cuts squares and eight cuts diamonds. The
eight-cut version has been retained. With the original optical head we measured a reduction of contrast
by a factor of 14 from 300 to 50 K, at which point the
silicon detectors stopped working. With the modified head, the contrast fell from 70% at room temperature to around 30% at 4 K, representing a contrast
reduction of a factor of 2.3. Following the mechanical modification, the optical head has successfully
passed the vibration qualification tests.
B.

Selection of Optoelectronic Components

Apart from certain heavily doped components, silicon

photodiodes generally stop working below 50 K because of a lack of carrier mobility.11 This was found
to be the case for the original silicon detectors, and it
was therefore necessary to identify new components
working down to 4 K. For this we tested a number
of components encapsulated in TO-18 packages by
measuring their response when dipped into liquid
nitrogen 77 K and liquid helium 4.2 K. The components were attached to a stainless steel tube together with an optical fiber bundle. We found a
component with a sensitivity loss of only 5% from 300
to 4 K S2386-18K. A large number of chips from
the same batch as those encapsulated were acquired.
For the LED we selected a GaAlAs LED OD880W with a peak emission at 880 nm. When
tested in the same way as the diodes, this component
was found to have an acceptable loss of optical power
of 15% from 300 to 4 K. The selection was made at
a diode power dissipation of 1 mW, corresponding to
the sensors power allocation. Spectra of the diode
have been measured with a grating spectrometer
Fig. 12, and the results are summarized in Table 2.
Variations of LED peak emission wavelength and
coherence length from 300 to 4 K are small and, as
explained in Subsection 2.A, they do not significantly
modify the contrast.
As seen in Fig. 12, the optical power of the diode is
stronger at 77 K than at 300 and 4 K. To make sure
that no sharp loss of optical power occurs around 4 K,
6310

APPLIED OPTICS Vol. 42, No. 31 1 November 2003

Fig. 12. Spectra of the selected LED at 300, 77, and 4 K with
constant electrical power dissipation of 1 mW.

we also measured the optical power at several temperatures between 120 and 4 K, while maintaining a
constant electrical power dissipation of 1 mW. The
result, shown in Fig. 13, is a slow optical power variation, which is of no consequence for the instrument
operation. These measurements have also shown
that the forward voltage of the LED increases rapidly
from 50 to 4 K Fig. 14. Knowledge of this behavior
is important to set the proper forward current and
monitor the temperature at the LED.
As seen in Fig. 3, the space available in the focal
plane of the sensors optical head is small. Therefore the LEDs and detectors are bare chips mounted
on a multilayer board of polyimide with a flexibleribbon connection to the cold preamplifier electronics.
The chips are gold bonded by ultrasonic thermobonding. The goal in the design has been simplicity and
robustness. Figure 15 shows how the different elements of the sensor are located within the scan mechanism.
Table 2. LED Spectral Parameters

Temperature
K

Peak
Wavelength
nm

FWMH
nm

300
77
4

876
858
852

72
51
42

Coherence
Length
m
10.7
14.4
17.3

Fig. 13. LED optical power versus temperature. The efficiency

of the diode peaks close to 85 K. At 4.2 K we lose only 15% of the
300 K emission. The horizontal line represents the optical power
at 300 K.

Fig. 14. LED forward voltage versus temperature.

C.

Cryogenic Preamplifier

A critical point is the design of the preamplifier that

carries the weak signals from the optical head inside
the cryostat to the warm electronics compartment
where the conditioner is located Fig. 16. The
length of the harness is many meters. It is therefore
important to minimize the sensitivity to perturbations generated by nearby instruments inside the cryostat. The adopted solution is to use a fully
differential transimpedance amplifier with a cold
electronics section located close to the optical head.
This includes two transistors as followers and associated feedback resistors for each photodiode,12,13
which provides a low-impedance connection to the
outside of the cryostat by twisted pairs in the harness.
For the cold electronics section, we had the choice
between two types of transistor operating at 4 K,
namely, the GaAs field-effect transistor and the Si
metal-oxide semiconductor field-effect transistor
MOSFETs.14,15 The GaAs field-effect transistors
are designed to operate in the gigahertz range, and at
audio frequencies they exhibit excessive 1f noise with
a breakpoint around 1 kHz in the case of the tested
models. Finally, we selected the enhancement-mode
Si MOSFET transistors 2N163, with a 1f noise
breakpoint at 100 Hz. The power dissipation for the
entire cold electronics circuit is 0.2 mW.

Fig. 16. Sketch of a transimpedance amplifier cryogenic preamplifier channel. D4 is a LED and D1 is a detector. The fully
redundant focal plane electronics contains two LEDs and six detectors associated with six identical preamplifier channels. VDD
and VSS are drain and source power supply voltages, respectively,
for the MOSFET transistors.

4. Sensor Resolution
A.

Theoretical Estimations

The resolution of this transducer depends on the following factors:

1. Engraving Quality of the Gratings
Inevitable random variations in groove spacing
set an ultimate limit to the resolution of the ISG
sensor. For double passage through the reticle grating, single passage through the scale grating, and an
averaging over the number NG of rulings involved, we
estimate this precision to be
x 2 2 2N G 12 5N G 12.

(1)

With a 4-m average groove spacing and a 6-mm

width of the grating mask, the effective number of
grooves per grating is NG 1500. Assuming that
100 nm is due to the photolithographic process,
this leads to an estimated resolution limit of x 5.8
nm.
2. Incremental Position Resolution
A reversible counter operates from the zero crossing
of the two quadrature signals, v1 and v2. These signals, which are electrical voltage signals delivered by
the detection chain comprising Si photodetector, preamplifier, and signal conditioning circuit, can be written as

Fig. 15. Inside view of the scan mechanism showing the optical
head with its flexible-ribbon connection to the cold preamplifier
electronics. The scale grating on the moving part of the translation stage supporting the corner cubes is separated from the optical
head by a gap of 500 m. The total displacement is 4 cm.

v 1 v 0 sin4x,

(2)

v 2 v 0 cos4x,

(3)

respectively, where v0 is the signal amplitude, x is the

position, and is the grating period. Measurement
noise induces small fluctuations v in the signal, giving an uncertainty x in the determination of position. Differentiation of v1 gives dv1dx 4v0
1 November 2003 Vol. 42, No. 31 APPLIED OPTICS

6311

Fig. 17. Resolution measurement experimental setup. ADC and

PC stand for analog-to-digital converter and personal computer,
respectively.

cos4x, and so, at zero crossings, the position

error is
x 4 SNR,

(4)

where SNR of v0v is the signal-to-noise ratio. The

same result is valid for zero crossings of the v2 signal.
The SNR is a function of power dissipation since decreasing the power reduces the fringe amplitude.
While the power dissipation allocated by the instrument power budget is 1 mW at 4 K, the nominal
dissipation of the original sensor is 35 mW at 300 K.
Fortunately, for a constant SNR, we can trade-off
power versus electrical passband, and reducing the
band from 0 150 to 0 4 kHz will therefore not degrade the resolution. This is acceptable for our application for which the required band is 0 1 kHz.
The modified sensor has a measured SNR in excess of
1000 in the 0 1-kHz band at 4 K, and, with 4 m,
this leads to a resolution of approximately 0.3 nm.
Consequently the SNR will not be a limiting factor.
3. Fractional Position Resolution
The fractional position between the zero crossings is
determined by analysis of the quadrature signals.
The ratio of the quadrature signals equals the tangent of the phase angle:
v 1v 2 tan4x tan4x,

(5)

where x is the fractional position corresponding to

the modulo 2 phase angle. Hence
x 4arctanv 1v 2.

(6)

Since this sensor is interferential, the fringes are

high-purity sine waves. Therefore the evaluation of
x can be accurate. Nevertheless, some degradation of the accuracy is expected because of small amplitude and offset differences of the fringes during a
scan induced by imperfections of the translation quality of the drive mechanism. Simulations have
shown that the expected residual amplitude and offset differences of 10% lead to position errors in the
10-nm range. This effect is therefore expected to be
the limiting factor to the final sensor resolution. In
view of these resolution calculations, the sensor resolution might be estimated by root-sum squaring of
the individual estimations to 12 nm, which is well
within the instrumental requirement of 100-nm resolution.
B.

Experimental Verification

As shown above, the SNR, which is the only highly

temperature-dependent factor, is not a limiting factor
for resolution. Resolution measurements therefore
have been carried out at room temperature. Figure
17 illustrates the experimental setup. A commercial
translation stage was set to drive the scale grating at
the nominal SPIRE mirror velocity of 1 mm s1, and
a laser interferometer, used as a displacement reference, triggered the sampling of the three sets of
fringes at intervals of 4, where 780.878 nm is
the laser wavelength. Since the quality of the translation stage was much better than that expected from
the instruments drive mechanism, this experiment
should allow verification of the resolution limit set by
the grating ruling quality.
From the recorded data file, a program produced
the quadrature signals after offset correction, amplitude equalization, and phase correction of the fringes.
Position was determined from the quadrature signals
by use of zero-crossing detection and Eq. 6 for fractional position determination. In the flight instrument, the same program is executed in real time by
an onboard digital signal processing controller.
Figure 18 shows a Lissajous plot of the quadrature
signals left panel and resolution versus displacement obtained when we took the difference between
successive position measurements right panel.
The standard deviation is 8 nm, to be compared with

Fig. 18. Lissajous plot of the quadrature signals after processing the raw fringes issued by the transducer left and resolution versus
displacement right. The standard deviation is 8 nm.
6312

APPLIED OPTICS Vol. 42, No. 31 1 November 2003

the estimated resolution limit of 5.8 nm. These

numbers are in reasonable agreement, and the difference is probably due to a somewhat different grating ruling precision and to measurement noise.
5. Conclusion

Theoretical and experimental developments have

been carried out for the implementation of a cryogenic interferometric scanning grating sensor. A
commercial transducer was stripped of its electronic
and optoelectronic circuitry and equipped with components that sustained operation at 4 K. The redesigned focal plane has full redundancy and a power
dissipation of only 1 mW. The sensor head was also
mechanically modified to reduce the loss of contrast
observed with the original setup, but all the optical
components lens, transmissive reticle grating, reflective scale grating of the commercial model were kept.
The modified sensor meets all the operational and
performance requirements of the HerschelSPIRE
instrument, and no performance degradation was
detected after more than ten cycles between 300
and 4 K.
This research was supported by the French Space
Agency CNES.
References
1. B. M. Swinyard, P. A. Ade, M. J. Griffin, K. Dohlen, J.-P.
Baluteau, D. Pouliquen, D. Ferand, P. Dargent, G. Michel, J.
Martignac, L. Rodriguez, D. E. Jennings, M. E. Caldwell, A. G.
Richards, P. A. Hamilton, and D. A. Naylor, FIRST-SPIRE
spectrometer: a novel imaging FTS for the submillimeter, in
UV, Optical, and IR Space Telescopes and Instruments, J. B.
Breckinridge and P. Jacobsen, eds., Proc. SPIE 4013, 196 207
2000.
2. J. W. Brault, New approach to high-precision Fourier transform spectrometer design, Appl. Opt. 35, 28912896 1996.

3. J. C. Brasunas and G. M. Cushman, Uniform time-sampling

Fourier transform spectroscopy, Appl. Opt. 36, 2206 2210
1997.
4. K. J. Wildeman, G. R. Ploeger, D. Snel, and J. J. Wijnbergen,
Grating drive for the short wavelength spectrometer in ISO,
Cryogenics 27, 68 72 1987.
5. K. W. Stark and M. Wilson, A mirror transport mechanism for
use at cryogenic temperature, in 20th Aerospace Mechanisms
Symposium, NASA Conf. Publ. 2423, 7395 1986.
6. G. N. Rassudova, Moire interference fringes in a system consisting of a transmission and a reflection diffraction grating,
Opt. Spectrosc. 22, 7378 1967.
7. F. M. Gerasimov, Use of diffraction gratings for controlling a
ruling engine, Appl. Opt. 6, 18611864 1967.
8. Exposed linear encoders: the interferential measuring principle with single-field scanning Heidenhain GmbH, Traunreut, Germany, 2002. Documentation available at http:
www.heidenhain.comexposedlinearencoderslip401A.
9. A. Spies, La ngen in der Ultrapra zisionstechnik messen,
Feinwerktechnik & Messtechnik 98, 406 410 1990.
10. D. Vukobratovich, Lens mounting, in Introduction to Optomechanical Design, SPIE Short Course SC014 SPIE, Bellingham, Wash., 2000.
11. B. Lengeler, Semiconductor devices suitable for use in cryogenic environments, Cryogenics 14, 439 447 1974.
12. C. L. Wyatt, D. J. Baker, and D. G. Frodsham, A direct coupled low noise preamplifier for cryogenically cooled photoconductive i.r. detectors, Infrared Phys. 14, 165176 1974.
13. B. B. Snavely and J. C. Yutzy, Impedance-transformation
circuit for operation at 4.2 K, Rev. Sci. Instrum. 38, 703704
1967.
14. A. Poglitsch, J. W. Beeman, N. Geis, R. Genzel, M. Haggerty,
E. E. Haller, J. Jackson, M. Rumitz, G. J. Stacey, and C. H.
Townes, The MPEUCB far IR imaging FabryPerot interferometer FIFI, Int. J. Infrared Millim. Waves 12, 859 884
1991.
15. T. J. Cunningham, R. C. Gee, E. R. Fossum, and S. M. Baier,
Deep cryogenic noise and electrical characterization of the complementary heterojunction field-effect transistor CHFET,
IEEE Trans. Electron Devices 41, 888 895 1994.

1 November 2003 Vol. 42, No. 31 APPLIED OPTICS

6313