Planck early results: first assessment of the High Frequency Instrument in-flight performance

Abstract

The Planck High Frequency Instrument (HFI) is designed to measure the
temperature and polarization anisotropies of the Cosmic Microwave Background
and galactic foregrounds in six wide bands centered at 100, 143, 217, 353, 545
and 857 GHz at an angular resolution of 10' (100 GHz), 7' (143 GHz), and 5'
(217 GHz and higher). HFI has been operating flawlessly since launch on 14 May
2009. The bolometers cooled to 100 mK as planned. The settings of the readout
electronics, such as the bolometer bias current, that optimize HFI's noise
performance on orbit are nearly the same as the ones chosen during ground
testing. Observations of Mars, Jupiter, and Saturn verified both the optical
system and the time response of the detection chains. The optical beams are
close to predictions from physical optics modeling. The time response of the
detection chains is close to pre-launch measurements. The detectors suffer from
an unexpected high flux of cosmic rays related to low solar activity. Due to
the redundancy of Planck's observations strategy, the removal of a few percent
of data contaminated by glitches does not affect significantly the sensitivity.
The cosmic rays heat up significantly the bolometer plate and the modulation on
periods of days to months of the heat load creates a common drift of all
bolometer signals which do not affect the scientific capabilities. Only the
high energy cosmic rays showers induce inhomogeneous heating which is a
probable source of low frequency noise.

Full text

Page 1

Astronomy & Astrophysics manuscript no. Planck2011-1.5_astroph
January 12, 2011
c ? ESO 2011
Planck early results: first assessment
of the High Frequency Instrument in-flight performance
Planck HFI Core Team: P. A. R. Ade47, N. Aghanim25, R. Ansari39, M. Arnaud35, M. Ashdown33,54, J. Aumont25,
A. J. Banday52,5,41, M. Bartelmann51,41, J. G. Bartlett1,31, E. Battaner57, K. Benabed26, A. Benoît26, J.-P. Bernard52,5,
M. Bersanelli15,20, R. Bhatia17, J. J. Bock31,6, J. R. Bond3, J. Borrill40,49, F. R. Bouchet26, F. Boulanger25, T. Bradshaw45,
E. Bréelle1, M. Bucher1, P. Camus38, J.-F. Cardoso36,1,26, A. Catalano1,34, A. Challinor55,33,7, A. Chamballu23, J. Charra25,†,
M. Charra25, R.-R. Chary24, C. Chiang11, S. Church50, D. L. Clements23, S. Colombi26, F. Couchot39, A. Coulais34, C. Cressiot1,
B. P. Crill31,43, M. Crook45, P. de Bernardis14, J. Delabrouille1, J.-M. Delouis26, F.-X. Désert22, K. Dolag41, H. Dole25,
O. Doré31,6, M. Douspis25, G. Efstathiou55, P. Eng25, C. Filliard39, O. Forni52,5, P. Fosalba27, J.-J. Fourmond25, K. Ganga1,24,
M. Giard52,5, D. Girard38, Y. Giraud-Héraud1, R. Gispert25,†, K. M. Górski31,59, S. Gratton33,55, M. Griffin47, G. Guyot21,
J. Haissinski39, D. Harrison55,33, G. Helou6, S. Henrot-Versillé39, C. Hernández-Monteagudo41, S. R. Hildebrandt6,38,30,
R. Hills56, E. Hivon26, M. Hobson54, W. A. Holmes31, K. M. Huffenberger58, A. H. Jaffe23, W. C. Jones11, J. Kaplan1,
R. Kneissl16,2, L. Knox12, G. Lagache25, J.-M. Lamarre34,?, P. Lami25, A. E. Lange24,†, A. Lasenby54,33, A. Lavabre39,
C. R. Lawrence31, B. Leriche25, C. Leroy25,52,5, Y. Longval25, J. F. Macías-Pérez38, T. Maciaszek4, C. J. MacTavish33,
B. Maffei32, N. Mandolesi19, R. Mann46, B. Mansoux39, S. Masi14, T. Matsumura6, P. McGehee24, J.-B. Melin8, C. Mercier25,
M.-A. Miville-Deschênes25,3, A. Moneti26, L. Montier52,5, D. Mortlock23, A. Murphy42, F. Nati14, C. B. Netterfield10,
H. U. Nørgaard-Nielsen9, C. North47, F. Noviello25, D. Novikov23, S. Osborne50, C. Paine31, F. Pajot25, G. Patanchon1,
T. Peacocke42, T. J. Pearson6,24, O. Perdereau39, L. Perotto38, F. Piacentini14, M. Piat1, S. Plaszczynski39, E. Pointecouteau52,5,
R. Pons52,5, N. Ponthieu25, G. Prézeau6,31, S. Prunet26, J.-L. Puget25, W. T. Reach53, C. Renault38, I. Ristorcelli52,5,
G. Rocha31,6, C. Rosset1, G. Roudier1, M. Rowan-Robinson23, B. Rusholme24, D. Santos38, G. Savini44, B. M. Schaefer51,
P. Shellard7, L. Spencer47, J.-L. Starck35,8, P. Stassi38, V. Stolyarov54, R. Stompor1, R. Sudiwala47, R. Sunyaev41,48,
J.-F. Sygnet26, J. A. Tauber17, C. Thum29, J.-P. Torre25, F. Touze39, M. Tristram39, F. Van Leeuwen55, L. Vibert25, D. Vibert37,
L. A. Wade31, B. D. Wandelt26,13, S. D. M. White41, H. Wiesemeyer28, A. Woodcraft47, V. Yurchenko42, D. Yvon8, and
A. Zacchei18
(Affiliations can be found after the references)
January 12, 2011
Abstract
The Planck High Frequency Instrument (HFI) is designed to measure the temperature and polarization anisotropies of the Cosmic
Microwave Background and galactic foregrounds in six wide bands centered at 100, 143, 217, 353, 545 and 857GHz at an angular
resolution of 10?(100GHz), 7?(143GHz), and 5?(217GHz and higher). HFI has been operating flawlessly since launch on 14 May
2009. The bolometers cooled to 100mK as planned. The settings of the readout electronics, such as the bolometer bias current,
that optimize HFI’s noise performance on orbit are nearly the same as the ones chosen during ground testing. Observations of
Mars, Jupiter, and Saturn verified both the optical system and the time response of the detection chains. The optical beams are
close to predictions from physical optics modeling. The time response of the detection chains is close to pre-launch measurements.
The detectors suffer from an unexpected high flux of cosmic rays related to low solar activity. Due to the redundancy of Planck’s
observation strategy, the removal of a few percent of data contaminated by glitches does not significantly affect the sensitivity. The
cosmic rays heat up the bolometer plate and the modulation on periods of days to months of the heat load creates a common drift of
all bolometer signals which do not affect the scientific capabilities. Only the high energy cosmic ray showers induce inhomogeneous
heating which is a probable source of low frequency noise. The removal of systematic effects in the time ordered data provides
a signal with an average level of noise less than 70% of our goal values in the 0.6–2.5Hz range. This is slightly higher than the
pre-launch measurements but better than predicted in the early phases of the project. This is attributed to the low level of photon
noise resulting from an optimized optical and thermal design.
Key words. Methods: data analysis – Cosmology: observations
1. Introduction
Planck1(Tauber et al. 2010a; Planck Collaboration 2011a)
?Corresponding
lamarre@obspm.fr.
1Planck (http://www.esa.int/Planck) is a project of the
European Space Agency (ESA) with instruments provided by
two scientific consortia funded by ESA member states (in par-
author:J.-M. Lamarre,
jean-michel.
is the third-generation space mission to measure the
anisotropy of the cosmic microwave background (CMB).
It observes the sky in nine frequency bands covering 30–
ticular the lead countries France and Italy), with contributions
from NASA (USA) and telescope reflectors provided by a collab-
oration between ESA and a scientific consortium led and funded
by Denmark.
arXiv:1101.2039v1 [astro-ph.IM] 11 Jan 2011

Page 2

2 The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
857GHz with high sensitivity and angular resolution from
31?to 5?. The Low Frequency Instrument (LFI; Mandolesi
et al. 2010; Bersanelli et al. 2010; Mennella et al. 2011) cov-
ers the 30, 44, and 70GHz bands with amplifiers cooled to
20K. The High Frequency Instrument (HFI; Lamarre et al.
2010; Planck HFI Core Team 2011a) covers the 100, 143,
217, 353, 545, and 857GHz bands with bolometers cooled
to 0.1K. Polarization is measured in all but the highest two
bands (Leahy et al. 2010; Rosset et al. 2010). A combina-
tion of radiative cooling and three mechanical coolers pro-
vides the temperatures needed for the detectors and optics
(Planck Collaboration 2011b). Two data processing centres
(DPCs) check and calibrate the data and make maps of the
sky (Planck HFI Core Team 2011b; Zacchei et al. 2011).
Planck’s sensitivity, angular resolution, and frequency cov-
erage make it a powerful instrument for galactic and ex-
tragalactic astrophysics as well as cosmology. Early astro-
physics results are given in Planck Collaboration (2011h–x).
The goal of this paper is to describe the in-flight perfor-
mance of the HFI in space and after the challenging launch
conditions. It does not attempt to duplicate the content of
the Planck pre-launch status papers (Lamarre et al. 2010;
Pajot et al. 2010), but rather presents the operational sta-
tus from an instrumental viewpoint. These results propa-
gate to scientific products through the data processing re-
ported in the companion paper (Planck HFI Core Team
2011b) which describes the instrumental properties as they
appear in the maps used by the “Planck early results” com-
panion papers. This paper focuses on the ability of the HFI
to measure intensity without any description of its perfor-
mance in measuring polarization, which will be reported
later.
Section 2 summarizes the instrument design. Section 3
focuses on early in-flight operations, the verification phase
and the setting of the parameters that have to be tuned in
flight. Section 4 addresses the measurement of the beams on
planets and the disentangling of time response effects from
the beam shape. It also presents the best current knowledge
of the physical beams resulting from this work. The effec-
tive beam obtained after data processing are to be found
in Planck HFI Core Team (2011b). Sections 5, 6 and 7 are
dedicated to noise, systematic effects and instrument sta-
bility respectively. A summary of the HFI in-flight perfor-
mance and a comparison with pre-launch expectations are
presented in section 8.
2. The HFI instrument
2.1. Design
The High Frequency Instrument (HFI) was proposed to
ESA in response to the announcement of opportunity for
instruments for the Planck mission in 1995. It is designed to
measure the sky in six bands (Tab. 1) with bolometer sen-
sitivity close to the fundamental limit set by photon noise.
The lower four frequency bands include the measurement
of the polarization. This sensitivity is obtained through a
combination of technological breakthroughs in each of the
critical components needed for bolometric detection:
– Spider web bolometers (Bock et al. 1995; Holmes et al.
2008) and polarization sensitive bolometers (Jones et al.
2003) which can reach the photon noise limit with suf-
ficient bandwidth to enable scanning great circles on
the sky at roughly 1rpm. They offer a very low cross-
section to cosmic rays that proves to be essential in this
environment and with this sensitivity.
– A space qualified 100mK dilution cooler (Benoît et al.
1997) associated with a high precision temperature con-
trol system.
– An active cooler for 4K (Bradshaw & Orlowska 1997)
using vibration controlled mechanical compressors to
prevent excessive warming of the 100mK stage and min-
imize parasitic effects on bolometers.
– AC biased readout electronics that extend high sensi-
tivity to very slow signals (Gaertner et al. 1997).
– A thermo-optical design consisting, for each optical
channel, of three corrugated horns and a set of compact
reflective filters and lenses at cryogenic temperatures
(Church et al. 1996). These include high throughput
(multimoded) corrugated horns for the 545 and 857GHz
channels (Murphy et al. 2002).
The angular resolution was chosen to extend the mea-
surement of the small scale features in the CMB, while
keeping the level of stray light to extremely low levels. At
the same time, at this sensitivity, the measurement and
removal of foregrounds requires a large number of bands
extending on both sides of the foreground minimum. This
is achieved with the six bands of the HFI (Table 1) and
the three bands of the Low Frequency Instrument (LFI;
Mennella et al. 2011).
The instrument uses a ∼ 20K sorption cooler com-
mon to the HFI and the LFI (Planck Collaboration 2011b;
Bhandari et al. 2000, 2004). The HFI focal plane unit
(FPU) is integrated inside the mechanical structure of the
LFI, on axis of the focal plane of a common telescope
(Tauber et al. 2010a).
The ability to achieve background limited sensitivity
was demonstrated by the ARCHEOPS balloon-borne ex-
periment (Benoît et al. 2003a,b), an adaptation of the HFI
designed for operation in the environment of a stratospheric
balloon. Similarly, the method of polarimetry employed by
the HFI was demonstrated by the Boomerang experiment
(Montroy et al. 2006; Piacentini et al. 2006; Jones et al.
2006). The HFI itself was extensively tested on the ground
during the calibration campaigns (Pajot et al. 2010) at IAS
in Orsay and CSL at Liège. However, the fully integrated
instrument was never characterized in an operational envi-
ronment like that of the second Earth-Sun Lagrange point
(L2). In addition to thermal and gravitational environmen-
tal conditions, the spectrum and flux of cosmic rays at L2
is vastly different than that during the pre-flight testing.
Finally, due to the operational constraints of the cryogenic
receiver, the end to end optical assembly could not be tested
on the ground with the focal plane instruments.
The instrument design and development are described
in Lamarre et al. (2010). The calibration of the instrument
is described in Pajot et al. (2010). The overall thermal and
cryogenic design and the Planck payload performance are
critical aspects of the mission. Detailed system-level aspects
are described in Planck Collaboration (2011a) and Planck
Collaboration (2011b).
2.2. Spectral transmission
The spectral calibration is described in Pajot et al. (2010)
and consists of pre-launch data, in the passband and

Page 3

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance3
Table 1. The HFI receivers. P stands for polarisation sensitive bolometers.
Channel
Central frequency
Bandwidth
Number of bolometers
100P
100
33
8
143P
143
32
8
143
143
30
4
217P
217
29
8
217
217
33
4
353P
353
29
8
353
353
28
4
545
545
31
4
857
857
30
4
(GHz)
(%)
Figure1. HFI spectral transmission
around, combined with component level data to determine
the out of band rejection over an extended frequency range
(radio–UV). Analysis of the in-flight data shows that the
contribution of CO rotational transitions to the HFI mea-
surements is important. An evaluation of this contribution
for the J = 1 → 0 (100 and 143 GHz), J = 2 → 1 (217
GHz) and J = 3 → 2 (353 GHz) transitions of CO is pre-
sented in Planck HFI Core Team (2011b).
3. Early HFI operation
3.1. HFI Cool down and cryogenic operating point
The Planck satellite cooldown is described in Planck
Collaboration (2011b).
The first two weeks after launch were used for passive
outgassing, which ended on 2 June 2009. During this pe-
riod, gas was circulated through the4He-JT cooler and the
dilution cooler to prevent clogging by condensable gases.
The sorption cooler thermal interface with HFI reached a
temperature of 17.2K on 13 June. The4He-JT cooler was
only operated at its nominal stroke amplitude of 3.5mm
on 24 June to leave time for the LFI to carry out a spe-
cific calibration with their reference loads around 20K. The
operating temperature was reached on 27 June, with the
thermal interface with the focal plane unit at 4.37K.
The dilution cooler cold head reached 93mK on 3 July
2009. Taking into account the specific LFI calibration re-
quirement that slowed down the cooldown, the system be-
haved as expected within a few days, according to the ther-
mal models adjusted to the full system cryogenic tests in
the summer 2008 at CSL (Liège).
The regulated operating temperature point of the 4K
stage was set at 4.8K for the 4K feed horns on the FPU.
The other stages were set to 1.395K for the so called
1.4K stage, 100.4mK for the regulated dilution plate, and
103mK for the bolometer regulated plate.
These numbers were very close to the planned operat-
ing point. As the whole system worked nominally, margins
on the cooling chain for interface temperatures and heat
lift are large. The Planck active cooling chain was one of
the great technological challenges of this mission and is
fully successful. A full description of the performance of
the cryogenic chain and its system aspects can be found
in Planck Collaboration (2011b). The parameters of the
operating points of the 4K, 1.4K and 100mK stages are
summarised in Table 2.
The temperature stability of the regulated stages has
a direct impact on the scientific performance of the
HFI. These stabilities are discussed in detail in Planck
Collaboration (2011b). Their impact on the power received
by the detectors is given in Sect. 3.3.1.
3.2. Calibration and performance verification phase
3.2.1. Overview
The calibration and performance verification (CPV) phase
of the HFI consisted of activities during the initial cooldown
to 100mK and during a period of about six weeks before
the start of the survey. The cooldown phase is summarized
in Sect. 3.1. The pre-launch value of the4He-JT cooler op-
erating frequency was used (see Sect. 3.2.2). Activities re-
lated to the optimization of the detection chain settings
were performed first during the cooldown of the JFET am-
plifiers, and again when the bolometers were at their oper-
ating temperature. Most of the operating conditions were
pre-determined during the ground calibration. The main
unknown was the in-flight background on the detectors.
The detection chain settings are presented in Sect. 3.2.3.
Other CPV activities performed are:
– determination of the detection chain time response un-
der the flight background
– determination of the detection chain channel-to-channel
crosstalk under the flight background
– characterization of the bolometer response to the 4K
and 1.4K optical stages, and to the bolometer plate
temperature variations
– checking the immunity of the instrument to the satellite
transponder
– optimization of the numerical compression parameters
for the actual sky signal and high energy particle glitch
rate
– various ring-to-ring slew angles (1.?7, 2.?0 [nominal], 2.?5)
– checking the effect of the scan angle with respect to the
Sun
– checking the effect of the satellite spin rate around its
nominal value of 1rpm.
On 5 August 2009, an unexpected shutdown of the4He-
JT cooler was triggered by its current regulator unit (CRU).
Despite investigations into this event, its origin is still un-
explained. A procedure for a quick restart was developed

Page 4

4 The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
Table 2. Main operation and interface parameters of the cooling chain
Interface Sorption cooler-4He-JT cooler (4K gas pre-cooling temperature)
Interface4He-JT cooler-dilution cooler (dilution gas pre-cooling temperature)
Interface 1.4K cooler-dilution gas precooling
Temperature of dilution plate (after regulation)
Temperature of bolometer plate (after regulation)
Temperature of 1.4K plate (after regulation)
Temperature of 4K plate (after regulation)
Dilution plate PID power
Bolometer plate PID power
1.4K PID power
4K PID power
4He-JT cooler stroke amplitude
Dilution cooler4He flow rate
Dilution cooler3He flow rate
Present survey life time (started 6 August 2009)
17.2 K
4.37 K
1.34 K
100.4 mK
103 mK
1.395 K
4.80 K
24.3–30.7 nW
5.1–7.4 nW
270 µW
1.7 mW
3450 µm
16.19–16.65 µmole/s
5.92–6.00 µmole/s
29.4 months
and implemented in case the problem recurred, but it has
not. Six days were required to re-cool the instrument to its
operating point. The two-week first light survey (FLS) fol-
lowed this recovery, starting on 15 August 2009. The FLS
allowed assessment of the quality of the instrument settings,
readiness of the data processing chain, and satellite scan-
ning before the start of science operations. The complete
instrument and satellite settings were validated and kept,
and science operations began. All activities performed dur-
ing the CPV phase confirmed the pre-launch estimates of
the instrument settings and operating mode. We will detail
in the following paragraphs the most significant ones.
3.2.2.4He-JT cooler operating frequency setting
The
nominal value of 40.08Hz determined during ground tests.
Once the cryochain stabilized, the in-flight behaviour of the
cooler was found to be very similar to that observed during
ground tests. The lines observed in the signal due to known
electromagnetic interference (EMI) from the4He-JT cooler
drive electronics have the same very narrow width. The
long term evolution of the4He-JT cooler parasitic lines is
discussed in Sect. 6.
4He-JT cooler operating frequency was set to the
3.2.3. Detection chain parameters setting
The JFET preamplifiers are operated at the temperature
which minimizes their noise. This setting was checked when
the bolometers were still warm (above 100K) during the
cooldown, since the bolometer Johnson noise was then
much lower than the JFET noise. Optimum noise perfor-
mance of the JFETs was found close at 130 K, in agreement
with the ground calibration.
After ground calibration, the only parameters of the
REU remaining to optimize in-flight were the bolometer
bias current and the phase of the lock-in detection, which
slightly depends on the bolometer impedance. Fig. 2 shows
the bolometer responses for a set of bias current values
measured while Planck was scanning the sky. For this se-
quence, the satellite rotation axis was fixed. For each bias
value, the total detection chain noise was computed after
subtraction of the sky signal. Ground measurements have
shown that the minimum NEP and the maximum respon-
sivity bias currents differ by less than 1%. Because of its
higher signal-to-noise ratio, we use the responsivity to opti-
mize the bias currents (Catalano et al. 2010). The optimum
in-flight bias current values correspond to the pre-launch
estimates within 5%. Therefore the pre-launch settings, for
which extensive ground characterizations were performed,
were kept (Fig. 2). In a similar way, the lock-in phase was
explored and optimized, and again the pre-launch settings
were kept.
The optical background power on the bolometers is on
the low end of our rather conservative range of predictions,
even lower than expected from the ground measurements.
This is attributed to a low telescope temperature and no
detectable contamination of the telescope surface by dust
during launch. This should result in a level of photon noise
lower than initially expected and an improved sensitivity.
3.2.4. Numerical data-compression tuning
The output of the readout electronic unit (REU) consists of
one number for each of the 72 science channels for each half-
period of modulation (Lamarre et al. 2010). This number,
SREU, is the exact sum of the 40 16-bit ADC signal values
obtained within the given half-period. The data processor
unit (DPU) performs a lossy quantization of SREU.
First, 254 SREUvalues corresponding to about 1.4s of
observation for each detector, covering a strip of sky about
8◦long, are processed. These 254 values are called a com-
pression slice. The mean < SREU> of the data within each
compression slice is computed, and data are demodulated
using this mean:
Sdemod,i= (SREU,i− < SREU>) ∗ (−1)i
where 1 < i < 254 is the running index within the com-
pression slice.
Then the mean < Sdemod > of the demodulated data
Sdemod,i is computed and subtracted. The resulting data
slice is quantized according to a step Q fixed per detector:
(1)
SDPU,i= round((Sdemod,i− < Sdemod>)/Q)
This is the lossy part of the algorithm: the required com-
pression factor, obtained through the tuning of the quan-
tization step Q, adds some extra noise to the data. For
σ/Q = 2, where σ is the standard deviation of Gaussian
white noise, the quantization adds 1% to the noise (Pajot
(2)

Page 5

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance5
Figure2. Optimization of the bolometer bias currents. Vertical lines indicate the final bias value setting. These values
are shifted with respect to the maximum because a dynamic response correction has been taken into account. A bias
value of 100digits corresponds approximately to 0.1nA.
et al. 2010; Pratt 1978). In flight, the value of σ was deter-
mined at the end of the CPV phase after subtraction of the
signal from the timeline.
The two means < SREU> and < Sdemod> computed as
32-bit words are sent through the telemetry, together with
the SDPU,ivalues. A variable length encoding of the SDPU,i
values is performed on board, and the inverse decoding is
applied on ground. This provides a lossless transmission of
the quantized values. A load limitation mechanism inhibits
the data transmission, first at the compression slice level
(compression errors), and second at the ring level (Lamarre
et al. 2010).
For a given Q value, the load on each channel depends
on the dynamic range of the signal above the level of the
noise. This dynamic range is largest for the high frequency
bolometers because of the galactic signal. The large rate
of glitches due to high energy particle interactions also
contributes to the load of each channel. Optimal use of
the bandpass available for the downlink (75kbs−1average
for HFI science) was obtained by using initially a value
of Q = σ/2.5 for all bolometer signals, therefore includ-
ing a margin with respect to the requirement of σ/Q = 2.
The load on each HFI channel is shown and compared to
simulated data in Fig. 3. The increase of signal gradients
while scanning through the galactic center in September
2009 triggered the load limitation mechanism (compres-
sion error) and up to 80000 samples were lost for each of
the 857GHz band bolometers. Therefore a new value of
Q = σ/2 was set for those bolometers from 21 December
2009 onward, reducing the number of samples lost to less
than 200 during the following scan through the galactic cen-
ter in March 2010. An illustration of a compression error
loss is shown in Fig. 4. Thanks to the redundancy of the
Planck scan strategy and the irregular distribution of the
few remaining compression errors, no pixels are missing in
the maps of the high signal-to-noise ratio galactic center
regions. Periodic checks of the noise value σ are done for
each channel, but no deviation requiring a change in the
quantization step Q has been encountered so far.
3.2.5. Instrument readiness at the end of the CPV phase
The overall readiness of the instrument was assessed during
the FLS. This end-to-end test was completely successful,
from both the instrument setting and the satellite scanning
points of view. The part of the sky covered during the FLS
was included in the first all sky survey.

Page 6

6 The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
0204060
channel
3
4
5
6
number of bits/sample
data
simulation
100143217353 545857thermos
Figure3. Load measured for each HFI channel on 16 July
2010. Simulated data for the same patch of the sky are
shown for bolometers. Channels #54 and higher corrre-
spond to the fine thermometers on the optical stages of
the instruments, plus a fixed resistor (#60) and a capaci-
tor (#61) on the bolometer plate.
2010/03/20 20:20:56
0100200300 400
Sample number
1.2 106
1.4 106
1.6 106
1.8 106
2.0 106
Amplitude (ADU)
Limits of the compression slice
DPU demodulated signal BC=25
Compression error
Figure4. Example of loss in one compression slice of data
on bolometer 857-1. Note the large signal-to-noise ratio
while scanning through the galactic center.
3.3. Response
3.3.1. Variation of the signal with background and with the
bolometer plate temperature
The optical background on the bolometers originates from
the sky, the telescope, and from the HFI itself. The oper-
ating point of the bolometers is constrained by this total
optical background, and the fluctuations of this background
have a direct impact on the stability of the HFI measure-
ments.
The power spectral density of each contribution to the
background is compared to 30% of the total noise measured
in-flight (NEP1column of Table 6). This specification cor-
responds to a quadratic contribution smaller than 5% on
the total noise.
The in-flight temperature stability of the HFI cryogenic
Table 3. Relative response deviation (in %) from linearity
for the CMB dipole, the galactic center (GC) and planets.
Saturation (Sat.) occurs for the Jupiter measurements at
high frequency.
DipoleGC
100 GHz
3.8 10−4
0.001
143 GHz
10−3
0.0017
217 GHz
8 10−4
0.003
353 GHz
6.4 10−4
0.007
545 GHz
< 10−4
0.01
857 GHz
< 10−4
0.1
Mars
0.01
0.02
0.05
0.06
0.08
0.06
Saturn
0.13
0.18
0.53
0.8
0.8
0.8
Jupiter
0.8
1.0
3.2
4.5
Sat.
Sat.
stages is discussed in Planck Collaboration (2011b). The
optical coupling of the HFI bolometers to each cryogenic
stage is shown in the left panels of Figs. 5 and 6 and in
Fig. 7. (The fact that the 100mK couplings all agree with
pre-launch measurements shows that no bolometers were
damaged during launch.) These couplings are used to cal-
culate the effect of the fluctuations of each cryogenic stage
on the bolometer signals. The right panels of Figs. 5 and
6 show the power spectral density (PSD) of the respective
thermometers scaled by the optical coupling factors for the
most extreme bolometers. The scaled PSDs of the thermal
fluctuations of the 4K and 1.4K stages are below the line
corresponding to 30% of the total noise of the correspond-
ing bolometer for all frequencies above the spacecraft spin
frequency.
The bolometer plate thermometers have a large cos-
mic particle hit rate (Planck Collaboration 2011b) be-
cause of the large size of their sensors compared to that
of the bolometers. Cosmic ray hits detection and removal
do not allow us to reach the thermometer nominal sensitiv-
ity, therefore they cannot be used to remove the effect of
bolometer plate temperature fluctuations on the bolometer
signal. Instead, the data processing pipeline (Planck HFI
Core Team 2011b) uses blind bolometers located on the
bolometer plate. The bolometer noise components are dis-
cussed in Sect. 6.
3.3.2. Linearity
The way a bolometer transforms absorbed optical power
into a voltage is not a linear process because both the con-
ductance between the bolometer and the heat sink, and
the bolometer impedance have a non-linear dependence on
the temperature (see e.g. Catalano (2008); Sudiwala et al.
(2000)).
The characterization of the linearity of the HFI detec-
tors has a direct impact on the calibration of the instru-
ment: strong non-linearity takes place during the galaxy
crossing for high frequency bolometers and during planet
crossings. An accurate absolute calibration is also neces-
sary for the CMB dipole. Finally, the energy scale of large
glitches can be corrected. The static response has been char-
acterized during ground calibration showing a small devia-
tion from linearity around a tenth of one percent for fainter
sources (few hundreds of attowatts) and around a few per-
cent for brighter sources like planets (Pajot et al. 2010).
The static response measured during the CPV phase agrees
with the ground estimate to better than 1%. Nevertheless,
the use of the static non-linearity determination does not
represent the true bolometric non-linear behaviour when

Page 7

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance7
Figure5. Left: coupling coefficients of the 4K stage. Right: scaled power spectral density (PSD) of the 4K stage thermal
fluctuations for the 100-1a and 353-5a-7a bolometers.
Figure6. Left: coupling coefficients of the 1.4K stage. The thermal emission in high frequency bands becomes too small
to be measured. Right: scaled PSD of the 1.4K stage thermal fluctuations for the 100-1a and 353-5a-7a bolometers.
Figure7. Bolometer signal coupling coefficients to the
100mK bolometer plate.
scanning through bright point sources like planets. The lin-
earization of the response done by multiplying the signal
times the gain (depending on the amplitude of the signal)
and convolving it with the temporal transfer function nor-
malized to 1 at the lowest frequency, is valid in the case of
small signals. However this is not the case for bright point
sources for which the estimate of non-linearity using the
static response may be incorrect by up to 40% in the ex-
treme case of Jupiter. For these sources we use a model
to correct the static results. The use of fainter planets like
Mars to characterize the beams minimizes this effect.
Table 3 gives the deviation from linearity for various
sources at the center of the beam for the bolometers at
each frequency.
3.4. Electrical crosstalk on HFI detectors
The electrical coupling of the signal of one bolometer into
the readout chain of another, or electrical crosstalk, was
measured to be less than −60dB for all pairs of chan-
nels during ground-based tests (Pajot et al. 2010). We per-
formed two tests in flight to verify this result, described
below.

Page 8

8 The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
Figure8. Top: Electrical crosstalk matrix Cij54x54 for all
bolometers, coefficients are in dB. Bottom: Distribution of
electrical crosstalk coefficients in dB.
3.4.1. CPV crosstalk measurements
During the CPV phase, we switched off each readout chan-
nel one at a time for ten minutes, and observed the impact
on all other channels. For each bolometer we collected about
660minutes of data.
The crosstalk coefficient between channels i and j is
expressed as:
Cij= ∆˜Vj/∆˜Vi,
(3)
where˜Viand˜Vjare the channel i and j voltages, corrected
for thermal drift. The crosstalk matrix and a histogram of
crosstalk levels are shown in Fig. 8. The crosstalk is mostly
confined to nearest neighbours in the belt, channels whose
wiring is physically close. The measured crosstalk level is
in good agreement with ground measurements, typically
< −70dB, and thus meets the requirement. A few of the po-
larization sensitive bolometer pairs show a crosstalk around
−60dB.
In the next section we see that crosstalk measurement
from glitches shows a much lower level of crosstalk. These
results suggest that the CPV test has measured electrical
crosstalk in current which is unrelated to the scientific sig-
nal.
3.4.2. Measurements using glitches
We used high energy glitches in one channel to study the
impact on the signal of surrounding channels. Thousands of
glitch events are collected for one channel, and the signals
of all other channels for the same time period are stacked.
The crosstalk in volts for individual glitches is defined as:
cV
ij= ∆Vj/∆Vi
(4)
where Viis the glitch amplitude in volts in the channel hit
by a cosmic ray, and Vjthe response amplitude of another
channel j. Then, for a pair of channels i and j, the global
voltage crosstalk coefficient is
CV
ij= median(cV
ij)
(5)
For SWB channels, in contrast with the CPV previous
results, no evidence of crosstalk is seen, with an upper limit
of −100dB. There are outliers in galactic channels because
of incorrect glitch flagging. A second analysis using planet
crossing data instead of glitches gave the same results.
Concerning the coupling between PSB pairs, we see
crosstalk around −60dB, in agreement with the CPV tests;
however, this is likely an upper limit because it includes
the effects of coincident cosmic ray glitches which produce
a similar effect but are not crosstalk.
4. Beams and time response
4.1. Measurement of Time Response
4.1.1. Introduction
The time response of HFI describes the shift, in amplitude
and phase, between the optical signal incident to each de-
tector and the output of the readout electronics. The re-
sponse can be approximated by a linear complex trans-
fer function in the frequency domain. The signal band
of HFI extends from the spin frequency of the spacecraft
(fspin? 16.7mHz) to a cutoff defined by the angular size
of the beam (14–70Hz; see Table 4 from Lamarre et al.
(2010)). For the channels at 100, 143, 217, and 353GHz,
the dipole calibration normalizes the time response at the
spin frequency. To properly measure the sky signal at small
scales, the time response must be characterized to high pre-
cision across the entire signal band, spanning four decades
from 16.7mHz to ∼ 100Hz.
The time response of bolometers typically is nearly flat
over a signal band from zero frequency to a frequency de-
fined by the bolometer’s thermal time constant, and then
drops sharply at higher frequencies. For the HFI bolome-
ters, the thermal frequency is 20–50Hz (Lamarre et al.
2010; Holmes et al. 2008), as noted in Lamarre et al. (2010)
and Pajot et al. (2010), however, the time response of HFI
is not flat at very low frequencies, but exhibits a low fre-
quency excess response (LFER).
We define the optical beam as the instantaneous direc-
tional response to a point source. Any sky signal is con-
volved with this function, which is completely determined
by the optical systems of HFI and Planck.

Page 9

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance9
Since Planck is rotating at a nearly constant rate and
around the same direction, the data are the convolution of
the signal with both the beam and the time response of
HFI. We separate the two effects and deconvolve the time
response from the time ordered data. This deconvolution re-
sults in a flat signal response, but necessarily amplifies any
components of the system noise that are not rolled off by
the bolometric response. This amplified noise is supressed
by a low-pass filter (Planck HFI Core Team 2011b).
4.1.2. TF10 model
The main ingredients of the time response are: (i) heat
propagation within the bolometer; (ii) signal modulation at
a frequency of fmod= 90.188Hz performed by reversing the
bolometer bias current; (iii) the effect of parasitic capaci-
tance along the high impedance wiring between the bolome-
ter and the first electronics stage (JFETs); (iv) band-pass
filtering, to reject the low frequency and high frequency
white noise in the electronics; (v) signal averaging and sam-
pling; and (vi) demodulation.
Because of the complexity of this sequence, a phe-
nomenological approach was chosen to build the time re-
sponse model. The time response is written as the product
of three factors:
H10(f) = Hbolo× Hres× Hfilter
Schematically, the first factor takes into account step
(i), the second factor describes a resonance effect that re-
sults from the combination of steps (ii) and (iii) , while the
purpose of Hfilteris to account for step (iv).
Detailed analysis and measurements of heat propaga-
tion within the bolometer have shown that Hbolois given
by the algebraic sum of three single pole low pass filters.
Explicitly:
?
with 6 parameters (a1,a2,a3,τ1,τ2,τ3).
(6)
Hbolo=
i=1,3
ai
1 + j2πfτi
(7)
Hres=
1 + p7(2πf)2
1 − p8(2πf)2+ jp9(2πf)
with 3 free parameters (p7,p8,p9),
(8)
Hfilter=
1 − (f/Fmod)2
1 − p10(2πf)2+ j(f/Ffilter)2
with one free parameter (p10). A total of 10 free parameters
describe this model, as indicated by its name. See Fig. 9 for
an illustration of the three components of the time response
model TF10 for a typical 217 GHz channel.
The parameter Ffilter characterizes the rejection filter
width and is kept fixed to 6Hz in the fitting process. Besides
the fact that this phenomenological model is physically mo-
tivated, this parameterization:
(9)
– ensures causality
– satisfies H(−f) = H∗(f)
– goes to 1 when f goes to zero (because we define a1+
a2+ a3= 1), while it goes to 0 when f goes to infinity
– includes enough parameters to provide the necessary
flexibility to fit the time response data of all 52 bolome-
ters.
0.0
0.2
0.4
0.6
0.8
1.0
1.2
Magnitude
0 20 40 6080100
Frequency(Hz)
-4
-3
-2
-1
0
1
2
3
Phase
Figure9. The amplitude and phase (in radians) of the
three components of the TF10 model of the time response.
The solid blue line is Hbolo(f), the dotted green line shows
Hfilter(f) and the dashed red line shows Hres. The solid
black line is H10(f), the product of the three components.
The vertical dotted black line shows the signal frequency
where the beam of a 217GHz channel cuts the signal power
by half.
4.1.3. Fitting the TF10 Model to Ground Data
To obtain the 10 × 52 parameter values, we used three sets
of pre-launch measurements. (i) The bolometer response
was measured at 10 different frequencies by illuminating all
52 bolometers with a chopped light source. (ii) Other mea-
surements were done using carbon fibers as light sources;
the latter were alternately turned on and off at a variable
frequency. (iii) The bolometer bias currents were periodi-
cally stepped up and lowered by a small amount. By adding
a square wave to the DC current, temperature steps are in-
duced, simulating turning on and off a light source (the
analysis of these data requires bolometer modelling).
None of these measurements was absolutely normalized;
all compared the relative response to inputs of various fre-
quencies. While measurement (i) only provided the ampli-
tude of the time response, measurements (ii) and (iii) pro-
vided both the amplitude and the phase. Note that for the
phase analysis, because of the lack of precise knowledge of
the time origin t0of the light/current pulses, a fourth fac-
tor exp(j2π∆t0f) is introduced in the expression of H10(f),
where the additional parameter ∆t0represents the uncer-
tainty in time.
Among the three sets of measurements, the carbon fiber
set covered the largest frequency range. Thus it was the best
for building the transfer function model described above
and for investigating its main features. However, it involves
uncertainties that could be resolved only with a very de-
tailed simulation of the set-up, therefore it was not used to
calculate the final set of parameter values. The final values
were calculated from data sets (i) and (iii), whose frequency
ranges are complementary, 2–140Hz and 0.0167–10Hz, re-
spectively. Since no absolute normalization was available,
the two datasets were matched in the overlap frequency
range.

Page 10

10 The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
The fitting of the analytic expression given above to
the merged data was done in the range between 16.7mHz
and 120Hz. The 52 fits have a χ2/DoF distribution whose
mean value is 1.13, indicating that the model is adequate
to describe the data. The numerical values of p10displayed
a small spread, σmean< 6 × 10−4. This parameter was set
at its mean value to calculate the 52 covariance matrices of
the nine remaining parameters, which were useful in prop-
agating the statistical errors.
As described below, the time response thus obtained
was further tuned and checked using in-flight observations,
in particular signals produced by planets and by cosmic
rays (glitches).
An alternative model has also been defined, based on the
analytical expression of the steps (ii) to (vi) of section 4.1.2.
Based on a closer analysis of the electronics stages, this
model is more physically motivated than the 10-parameter
model. It requires only 8 parameters and provides better
results near the modulation frequency. Nevertheless the
model has not been used in the current data release. It
is only used as a benchmark, to check possible systematic
effects in the current release. Most of the effects of the dif-
ference between the models disappear when the data are
low-pass filtered.
4.1.4. Fitting TF10 to Flight Data
The planets Mars, Jupiter, and Saturn are bright, compact
sources that are suitable for measuring the beam and pro-
vide a near-delta-function stimulus to the system that can
be used to constrain the time response. During the first sky
survey, Planck observed Mars twice and Jupiter and Saturn
once (Planck HFI Core Team 2011b). During a planet ob-
servation, the spacecraft scans in its usual observing mode
(Planck Collaboration 2011a), shifting the spin axis in 2?
steps along a cycloidal path on the sky. Since planets are
close to the ecliptic plane, the coverage in the cross-scan di-
rection is not as fine as in the scan direction. In the case of
Jupiter and Saturn, each channel observes the planet once
per rotation for a period of approximately 6 hours (9 peri-
ods of stationary pointing, or "rings"). Because Mars has
a large proper motion, the first observation lasted 12 hours
(or 18 rings).
We use the forward-sense time domain approach
(Huffenberger et al. 2010) to simultaneously fit for Gaussian
beam parameters and TF10 time response parameters. A
custom processing pipeline avoids filtering the data. We
extract the raw bolometer signal and demodulate it using
the parity bit. We use the flags created by the time ordered
information (TOI) processing pipeline to exclude data sam-
ples contaminated by cosmic rays, and we additionally flag
all data samples where the nonlinear gain correction is more
than 0.1%. We use Horizons2emphemerides to compute the
pointing of each horn relative to the planet center.
The time domain signal from the planet is modeled as an
elliptical Gaussian convolved with the TF10 time response
as follows:
d(t) = H10? A(t)G[x(t);x0,?,θFWHM,ψ]
(10)
where the Gaussian optical beam model G is parameterized
as in Eqs. 9–11 of Huffenberger et al. (2010), except the
2ssd.jpl.nasa.gov/?horizons
0 00.050.05 0.10.10.150.15 0.2 0.2
0 0
0.10.1
0.20.2
0.30.3
0.40.4
0.50.5
0.6 0.6
0.70.7
0.80.8
0.90.9
1 1
time (s)time (s)
normalized signal normalized signal
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?? ? ? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?? ? ? ? ??
????
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?? ? ? ? ? ??
?
?
? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
???
?? ? ? ? ? ? ???
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??? ? ? ? ? ? ? ??
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
?
?
?? ? ? ? ? ? ? ? ? ???
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??? ? ? ? ? ? ? ? ? ? ? ???
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
???? ? ? ? ? ? ? ? ? ? ? ? ? ? ????
?
?
?
?
?
?
?
?
?
?
????????? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
?
?
?
?
?
?
????? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??????????
?
?
?
?
????????? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ????????????? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
? ? ?
?
?
?
?
?
? ?
?
?
?? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
?
?
?
?
?
?
?
?
?
???????? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
???
???
?
???? ??? ? ? ?
?
?
?
?
?
? ??? ? ???? ? ? ? ? ? ? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
???? ? ?? ? ??? ? ? ? ? ? ? ? ? ?
??? ?? ?
???? ?? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
?
?
?
?
? ? ? ? ? ?? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ??
?
?
? ???????? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ???? ? ? ? ? ???? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?
Figure10. Comparison of the impulse response of chan-
nel 143-2a (red curve) and a template made from stack-
ing glitch events (black curve). Noise begins to dominate
further in the timeline. The ringing observed at the mod-
ulation frequency is generated by the electronics rejection
filter.
planet amplitude is parameterized with a disk temperature
rather than a single amplitude:
A(t) = TdiskΩp(t)
Ωb
,
(11)
where Tdisk is the whole-disk temperature of the planet,
Ωpis the solid angle of the planet, which can vary signifi-
cantly during the observation, and Ωbis the solid angle of
the beam. Ωp is computed using Horizons, which is pro-
grammed with Planck’s orbit.
The free parameters of the fit are the six parameters
of the time response corresponding to Hbolo, the two com-
ponents of the centroid of the beam x0, the mean FWHM
θFWHM, the ellipticity ?, the ellipse orientation angle ψ, and
the planetary disk temperature Tdisk. The four parameters
describing the electronics are somewhat degenerate with
the bolometer part of the time response, and we fix them
at the ground-based values.
Because of the large nonlinear response and highly non-
Gaussian beams at 545 and 857GHz, we do not perform
fits to the planet data at these frequencies. Instead we rely
on pre-launch fits for the time response.
By taking the Fourier transform of the time response
function derived on planets, one obtains the system re-
sponse to a Dirac impulse. This response can be compared
to the glitches generated by cosmic rays that deposit energy
in the sensor grids.
The glitches detected by HFI are sampled with time
steps 1/(2Fmod). However, the glitches can be superresolved
in time by normalizing, phasing, and stacking single glitch
events (Crill et al. 2003). This gives glitch templates for
each channel (Alexandre Sauvé, private communication)
that are effectively sampled at a much higher frequency.
Figure 10 shows the comparison between a superre-
solved glitch template and the corresponding calculated re-
sponse. There is good agreement in general, but there are
discrepancies at high frequency (f > 100Hz). The physical
model for the electronics transfer functions briefly described
at the end of section 4.1.3 suppresses this discrepancy at
high frequency.

Page 11

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance 11
Each planet observation suffers unique systematic ef-
fects, so a comparison of the time response recovered on
each gives a good assessment of the effects of these different
systematics. Mars has a large proper motion, giving excel-
lent sampling in the cross-scan direction; however, there is
a known diurnal variability in its brightness temperature
(Swinyard et al. 2010). Jupiter has a large angular diam-
eter (48??) relative to the HFI beam size, and Saturn and
Jupiter are so bright that the HFI detectors are driven sig-
nificantly nonlinear (see Table 3). Nonetheless, we find that
the time response is consistent to 0.1-0.5% when recovered
from each of the planets individually as well as from all
planets simultaneously.
A further cross-check is done by stacking planet scans
to build a superresolved planet timeline. Time response pa-
rameters are fit to the superresolved planet using the as-
sumption of a near-Gaussian beam profile, and are consis-
tent with the first approach.
The in-flight time response differs from the ground-
based time-response by at worst 1.5% between 1Hz and
40Hz. We do not include this difference in the final er-
ror budget, because it is likely that the time response has
changed due to differences in background conditions.
4.1.5. Low Frequency Excess Response
The HFI bolometers show low frequency excess response
(LFER), (Lamarre et al. 2010). Though the planets are
bright, the short impulse they provide is close to a delta
function and the energy is spread evenly across nearly all
harmonic components. In combination with low frequency
noise, the measurements are not sensitive to frequencies
below ∼ 0.5Hz; so with planet observations alone we can-
not constrain an excess response at very low frequency. We
maintain the ground measurements as our best estimation
of the LFER. In the ground-based measurements the bias
step vs. carbon fiber differs by at most 1.5% at low fre-
quency, so we assign a systematic error of 1.5% for frequen-
cies below 0.5Hz.
For future data releases, we will use the difference of sky
signal between surveys to constrain the LFER.
4.1.6. Summary of Errors in the Time Response
As noted above, the data represent the combined effect of
the time response and the optical beam. The time response,
however, is not degenerate with a Gaussian parameteriza-
tion of the beam; the true beams deviate from a Gaussian
shape at the several percent level near the main lobe, while
time response effects tend to give the beam an extended tail
following the planet in the scan direction. The Gaussian as-
sumption could slightly bias the recovered time response;
however, any residual bias is captured in the measurement
of the post-deconvolution scanning beam (Planck HFI Core
Team 2011b).
Because of the high signal-to-noise ratio of the planet
data, statistical errors in the fit are small, so we assess the
systematic errors in the resulting time response by checking
the consistency of various methods of recovering the time
response. We fit to different combinations of planet data:
Mars, Jupiter, and Saturn data separately and all of the
data simultaneously to check for systematics resulting from
various planets. Additionally we compare the planet-fitted
time response with ground-based data and with the impulse
response from cosmic ray glitches.
Our final error budget is as follows:
– Low frequency (f < 0.5Hz): the errors are dominated
by the possibility of a low frequency excess response
below 0.5Hz at a level of 1.5%.
– Middle frequency (0.5Hz < f < 50Hz): We set an error
bar between 0.1% and 0.5% depending on the channel.
This error bar is set by the consistency in results from
different sets of planet data.
– High frequency: (f > 50Hz) Our empirical model of
the electronics in the TF10 model does not describe
the system very well at these frequencies, as shown by
some disagreement between the glitches and the TF10
impulse response. However, for this data release, the
low-pass filter applied to the data and the beam cutoff
reduce the importance of this frequency band.
The Planck scan strategy is such that the same region of
the sky is observed scanning in nearly opposite directions
six months apart. An error in the time response is high-
lighted in the difference of maps obtained from the first
six months and the second six months of the survey. This
difference map shows some level of contamination, in par-
ticular near the Galactic plane, where the signal is higher.
The same level of contamination is observed in simulations
in which the data are generated with a transfer function,
and analysed with a different one, in order to mimic the
uncertainties described above. With this technique, we val-
idate the error budget.
4.2. Optical Beams
The optical beam is defined (Sect. 4.1.1) as the instanta-
neous directional response to a point source. For HFI, the
optical beams for each channel are determined by the tele-
scope, the horn antennas in the focal plane and, for the po-
larized channels, by the orientations of their respective po-
larization sensitive bolometers (PSBs) (Maffei et al. 2010).
Model calculations of the beams are essential, since it was
possible to measure only a limited number of beams in
the telescope far-field before launch. The 545 and 857GHz
channels, which employ multimoded corrugated horns and
waveguides, were not included in this campaign. (The opti-
cal beam is related to, but is not the same as the scanning
beam defined in Planck HFI Core Team (2011b) and used
for data analysis purposes.) Tauber et al. (2010b) reported
the best pre-launch expectations for the optical beams, ob-
tained using physical optics calculations with CORRUG3
and GRASP4. Table 4 compares the calculated and mea-
sured (Sect. 4.1.2) beams for the single-moded channels (up
to 353GHz).
For these channels the pre-launch calculations of
FWHM and ellipticity and measured Mars values agree to
within a few percent. These differences are contained within
2.7σ of the data errors. The main source of discrepancy
could be a slight misalignment of the pre-launch telescope
model with respect to the actual in-flight telescope geom-
etry, which is currently being investigated (Jensen & et al
2010).
3SMT Consultancies Ltd. www.smtconsultancies.co.uk
4TICRA, www.ticra.com

Page 12

12The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
Table 4. Comparison of pre-launch calculations and measured parameters for the HFI optical beams (band averages).
Standard deviations σ are computed as the dispersion between the Saturn, Jupiter, and Mars data for each given channel.
BandExpected
FWHM
[?]
9.58
6.93
7.11
4.63
4.62
4.52
4.59
4.09
3.93
Mars
FWHM
[?]
9.37
6.97
7.24
4.70
4.63
4.41
4.48
3.80
3.67
Mars
σFWHM
[?]
0.06
0.10
0.10
0.06
0.06
0.06
0.04
–
–
Expected
ellipticity
Mars
ellipticity
Mars
σellip
100P
143P
143
217P
217
353P
353
545
857
1.17
1.06
1.03
1.12
1.10
1.08
1.23
1.03
1.04
1.18
1.02
1.04
1.13
1.15
1.07
1.14
1.25
1.03
0.006
0.004
0.005
0.006
0.010
0.009
0.007
–
–
Figure11. The distribution of the HFI beams on the sky
relative to the telescope boresight as viewed from infinity.
Contours of the Gauss-Hermite decomposition of the Mars
data at 1%, 10%, and 50% power levels from the peak.
For the photometers containing a pair of PSBs, the average
beam of the two PSBs is shown.
Table 5 reports our best knowledge of the FWHM of the
optical beams for each channel. We stress that this table
does not provide parameters of the scanning beam of the
processed data, which accounts for the additional effects
of the instrument time response and of the time domain
filtering in the data processing (Planck HFI Core Team
2011b).
As reported in Maffei et al. (2010), the 545GHz and
857GHz channels are multimoded (more than one elec-
tromagnetic mode propagating through the horn anten-
nas) and their optical beams are markedly non-Gaussian.
The understanding of these channels through simulations
has progressed since Planck was launched, especially in
the characterization of their modal content (Murphy et al.
2010).
In Table 4 we compare pre-launch calculations of the
beams with the beams measured with Mars. Differences in
FWHM are less than 7%. We stress that this discrepancy
does not impact the scientific products of the Planck mis-
Figure12. The “dimpling effect” as seen at 545GHz (left
panel) and 857GHz (right panel). The grid spacing is 10?.
The color scale is in dB normalized to the peak signal of
Jupiter.
sion since the scanning beams are the ones to be used for
data analysis purposes. From an instrumental point of view,
the inflight measurements must obviously be considered as
the reference for the performance of these channels.
The development of the HFI multimoded channels ne-
cessitated the novel extension of previously existing mod-
elling techniques for the analysis of the corrugated horn
antennas and waveguides, as well as for the propagation
of partially coherent fields (modes) through the telescope
onto the sky (Murphy et al. 2001). Extensive pre-launch
measurement campaigns were conducted for all the HFI
horn antenna/filter assemblies (Ade et al. 2010). The HFI
multimoded channels are suitable for the scientific goals of
Planck. Nevertheless, for future instruments, more research
can be envisaged in this field. The characterization of the
modal filtering in the horn-waveguide assembly and the un-
derstanding of the coupling of the waveguide modes to the
detector need further theoretical and experimental study.
The similarity of the pre-launch expectations to our cur-
rent knowledge of the HFI focal plane (beams and their
positions on the sky) tells us that the overall structural in-
tegrity of the focal plane has been preserved after launch.
Furthermore, the optical beams as measured on Mars are
shown in Fig. 11 and can be compared with the equivalent
representations of the focal plane layout based on calcu-

Page 13

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance13
Table 5. Geometric mean of the asymmetric gaussian
FWHM.
Bolometerbeam
FWHM
[?]
9.46
9.60
9.41
9.43
9.42
9.47
9.43
9.45
6.91
6.99
6.78
6.80
6.91
6.86
7.01
7.01
7.45
7.08
7.18
7.20
4.73
4.75
4.66
4.64
4.63
4.68
4.69
4.73
4.68
4.61
4.59
4.61
4.47
4.46
4.40
4.39
4.41
4.42
4.47
4.45
4.57
4.46
4.44
4.53
3.94
3.63
3.79
4.17
3.73
3.66
3.76
3.67
spectral Band
cut-on
GHz
84.9
87.0
86.5
84.4
84.4
84.4
84.9
84.9
120.8
120.3
119.8
119.3
120.3
120.3
118.8
119.3
120.3
120.3
120.8
120.8
184.0
183.9
182.5
189.6
188.6
189.6
182.5
182.0
189.6
189.1
191.1
193.1
310.9
310.4
323.5
313.9
302.3
299.8
300.3
314.4
310.4
312.9
326.1
318.5
466.1
464.5
467.6
479.2
748.1
736.5
747.1
744.1
spectral Band
cut-off
GHz
113.87
115.27
116.28
115.42
116.77
116.77
117.79
117.79
161.77
162.78
162.26
163.28
158.73
160.75
167.83
161.26
166.31
165.81
167.83
165.3
249.72
249.12
253.26
252.76
253.77
250.74
253.26
252.76
249.72
253.26
252.76
252.76
403.91
405.93
400.88
406.94
405.43
405.93
406.94
397.84
401.38
407.45
404.4
405.92
638.93
633.87
633.87
635.89
986.59
982.65
984.21
970.02
Name
100-1a
100-1b
100-2a
100-2b
100-3a
100-3b
100-4a
100-4b
143-1a
143-1b
143-2a
143-2b
143-3a
143-3b
143-4a
143-4b
143-5
143-6
143-7
143-8
217-5a
217-5b
217-6a
217-6b
217-7a
217-7b
217-8a
217-8b
217-1
217-2
217-3
217-4
353-3a
353-3b
353-4a
353-4b
353-5a
353-5b
353-6a
353-6b
353-1
353-2
353-7
353-8
545-1
545-2
545-3
545-4
857-1
857-2
857-3
857-4
lations in earlier papers (Maffei et al. 2010; Tauber et al.
2010b). A detailed account of the full focal plane recon-
struction can be found in Planck HFI Core Team (2011b).
There is a “dimpling” of the reflector surfaces from the
irregular print-through of the honeycomb support struc-
tures on the reflector surfaces themselves (Tauber et al.
2010b). GRASP calculations predict that this will gener-
ate a series of rings of narrow bright grating lobes around
the main lobe. Since the small-scale details of the dimpling
structure of the Planck reflectors are irregular, these grat-
ing lobes tend to merge with the overall power scattered by
the reflector surfaces (Ruze scattering (Ruze 1966)). Fig. 12
shows a HEALPIX (Górski et al. 2005) map of the first sur-
vey observation of Jupiter minus the second survey observa-
tion of the same sky region to remove the sky background.
We see the first ring of grating lobes as expected in the
map from all 545 and 857GHz channels, where the signal-
to-noise ratio on the planets is highest. The inner 15?of
the beam is saturated and does not appear in the map. At
857GHz, the discrete grating lobes appear at level below
−35dB with respect to the peak (∼ 30dB), and represent a
negligible fraction of the total beam throughput. The shoul-
der of the beam, extending radially to ∼ 15?, represents a
larger contribution to the throughput, ranging from 0.5%
to a few percent for the CMB and sub-mm channels, re-
spectively.
5. Noise properties
The Planck HFI is the first example of space-based bolome-
ters, continuously cooled to 100mK for several years.
Although the detectors were thoroughly tested on the
ground (Lamarre et al. 2010; Pajot et al. 2010), it remained
then to be seen how they would behave in the L2 space envi-
ronment. We describe here the noise properties of the HFI
in the first year of operation, focusing on the differences
between space and ground performance.
This section deals with the Gaussian part of the noise.
Section (6) describes the systematic effects that have been
analyzed in the data so far.
An example of raw time ordered information (TOI) is
shown in Fig. 13. The TOI is dominated by the signal
from the CMB dipole, Galactic emission, point sources, and
glitches. Therefore, the noise properties cannot be directly
deduced from the TOI. We first describe the general method
used to evaluate the noise, then we give general statements
on the noise properties.
5.1. Noise estimation
The Level-2 detnoise pipeline (Planck HFI Core Team
2011b) is used to determine a noise power spectrum, from
which one extracts the noise equivalent power (NEP) of
the detectors (see Planck HFI Core Team (2011b) for a full
description). The pipeline uses redundancies in the obser-
vations to determine an estimate of the sky signal, which is
then subtracted from the full TOI to produce a pure noise
timeline. The signal estimates are the integration of typ-
ically 40 circles of data at a constant spin axis pointing.
The average signal, binned in spin phase, provides an accu-
rate estimate of the signal. This signal as a function of spin
phase is then subtracted from the TOI. The residual is an
estimate of the instantaneous noise. Power spectra of this
residual timeline are then obtained for each pointing period
(see Fig. 14) and fit for the white noise level, i.e., the NEP,
in the spectral region between 0.6 and 2.5Hz. The lower
limit of 0.6Hz is high enough that the low frequency excess
noise can be neglected and the upper limit small enough to
keep the time response near to its value at low frequencies
(16mHz) at which the instrument is calibrated.
The noise is stable at a level better than 10% in the
majority of detectors. Exceptions are: (1) a few rings

Page 14

14The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance
Figure13. Examples of raw (unprocessed) TOI for one bolometer at each of six HFI frequencies and one dark bolometer.
Slightly more than two scan circles are shown. The TOI is dominated by the CMB dipole, the Galactic dust emission,
point sources, and glitches. The relative part of glitches is over represented on these plots due to the thickness of the
lines that is larger than the real glitch duration.
with unusual events that contaminate the measurement,
e.g., poorly corrected/flagged glitches or passage over very
strong sources such as the Galactic centre, especially at
857GHz; (2) a weak trend, smaller than 1% in amplitude,
that correlates with the duration of the pointing period
(an expected bias due to the ring average signal removal);
(3) bolometers affected by random telegraph signals (RTS)
(see next Section); and (4) some uncorrelated jumps in the
noise levels for about ten bolometers at the 30% level for
isolated periods of a few days. The overall result is that a
very clear baseline value can be identified and can be used
to determine the NEP of each bolometer. This is then con-
verted to NE∆T with the help of the flux calibration. The
NE∆Ts thus measured are given in Table 6. The quoted
uncertainties are derived from the rms of the NEPs in a
band around the baseline.
5.2. The noise components
The detector noise is described by the combination of sev-
eral components:
– Photon and bolometer noise, which appear as white
noise filtered by the time response of the bolometer,
the readout electronics, and the TOI processing.
– Electronics and Johnson noise, which produce noise that
is nearly white across the frequency band, but with a
sharp decrease at the high frequency end due to the
on-board data handling and the TOI filtering.
– The 4K lines (Sect 6), appearing as residuals in the
spectra.
– The energy deposited by cosmic rays on the bolome-
ters, which appears as "glitches", i.e., positive peaks
in the signal, which are removed by the TOI processing
(Sect. 6, and Planck HFI Core Team (2011b)). Residuals
from glitches appear in the noise spectrum as a bump
between 0.1 and 1Hz.
– Low frequency excess (LFE) noise, which is present be-
low about 100mHz.
The last three sources of noise are detailed in Section 6.
There is additional noise (of the order of 0.5% or less)
due to the on-board quantization of the data before trans-
mission. In general, the noise level, as measured by the
NEP, is between 10 and 20aWHz−1/2for the 100 to
353GHz channels, and between 20 and 40aWHz−1/2for

Page 15

The Planck HFI core team: Planck early results: first assessment of the HFI in-flight performance 15
Figure14. Typical power spectrum amplitude of bolometers 143-5 and 545-2. For the upper panels, this is the power
spectrum density of valid samples, after an average ring (the sky signal) has been subtracted from the TOI. Stacking of
the result for 200 rings is shown in the lower panel. Here, the instrument time response is not deconvolved from the data.
the 545 and 857GHz channels. It is in line with the ground-
based expectations and the lower estimate of the back-
ground load with a detector-to-detector variability of less
than 20% (see Sect. 8).
Due to the AC bias modulation scheme, the 1/f noise
from the electronics is aliased near the modulation fre-
quency where it is heavily filtered out. The benefit of this
scheme is visible on the noise power spectrum of the 10MΩ
resistor which shows a flat spectrum at the Johnson value
down to 1mHz, a tribute to the electronic chain stability.
At the present time, we assume that the LFE noise, not
observed in ground-based measurements, is mostly due to
the 100mK bolometer plate fluctuations. While drifts in
the 100mK stage that are correlated between bolometers
are removed, there are likely local temperature fluctuations
due to particle energy deposited close to each detector.
6. First assessment of systematic effects
6.1. 4K lines
The
(40.083373Hz) is phase-locked to the frequency of the data
acquisition (180.37518Hz) in a ratio of 2 to 9. EMI/EMC
impacts the TOI only as very narrow lines. Unfortunately,
in flight, unlike in ground-based measurements, these lines
are not stable. The 4K line variations are illustrated in
fundamentalfrequency ofthe
4He-JTcooler
Fig. 15. The variability of the lines is in part due to tem-
perature fluctuations in the service module of the Planck
spacecraft. Indeed, some of the variability was related to the
power cycling of the data transponder which, for stability
reasons, has been kept on continuously since 25 January
2010 (OD 258, Planck Collaboration 2011b, see Fig. 16).
6.2. Abnormal noise in the electronics
Of the 54 bolometers on HFI, three show a significant RTS,
also known as “popcorn noise.” These are 143-8, 545-3, and
857-4. Fig 17 illustrates their behaviour. The noise timeline
clearly exhibits a two-level system. The three RTS bolome-
ters in flight are the ones where RTS occurred most fre-
quently in ground measurements. However, in flight: 1) the
level difference is well above the noise (at least ten times the
rms); 2) the two-level system can be a three-level system
or even larger; and 3) the RTS is intermittent. For large
duration, it can be unnoticeable, especially for the 857-4
bolometer.
In an unrelated fashion, we see uncorrelated jumps in
the noise TOI of many bolometers at a rate of just a few
every year.