RAPID: A Revolutionary Fast Low Noise Detector

on PionierSylvain Guieu ESO / IPAG Jean Baptiste Lebouquin

Philippe Feautrier Gérard Zins Éric Stadler Pierre Kern

Alain Delboulbé Thibault Moulin Sylvain Rochas

… SOFRADIR, ONERA, LETI, …

Pionier Design

Niobate

PNR-MAN-OperationManual PIONIER Operation Manual

Revision: 0.1 Page 24/29

Figure 10 Detector RTD display in FREE configuration (top left, broadband light), in LARGE configuration (top right) and in SMALL (middle right). Bottom is a cut though the pixels after a proper alignment.

6.1.3 Procedure • In the RTD, zoom comfortably on the pixels for a better view. • Do a cut along the pixels. • Set fix vertical limit for the cut (0, 3000). • Generally, it is enough to adjust the vertical, the lateral and the focus. To do so:

o Do not consider the central pixels (only the left- and right-most pixels). o Try to maximize the flux in the pixel. o When observing with the prism, be careful to center vertically the spectra in the green

sub window. The prism LARGE disperses over 7 channels while the prism SMALL disperses over 3 channels.

• If you feel confident, you can touch the image tilt and the image magnification to better

project the spots in to the pixels. o A first inspection on the profiles should allow you to evaluate if the separation

between extreme outputs matches the pixels positions. o Adjust manually the magnification axis until satisfaction

Small Large WollastonFree

3 chan. 5 chan. 1 chan. 1chan.

Pionier throughput and limitations from star light to u/v visibility points

• Atmosphere ✤ Tip-tilt & Seeing: lost of injected flux -> Tip-tilt mirror + AO system. ✤ Piston: Fringes shuffling -> fast fringe scan for a fast OPD correction.

• VLTi ✤ Number of Telescopes (and Size) ✤ Tunnel and lab air thermal stability: -> zero thermal gradient ✤ Mirrors reflectivity ✤ Delay Lines: fast tracking, VCM, OPD models stability. ✤ Polarimetry: control of VLTi polarimetry from primary mirror to instrument for each lines. ✤ Guiding, IRIS camera -> Faster and more sensitive camera

• Pionier ✤ Injection: -> tip-tilt. ✤ Optical component: good transmission, zero cross-talk, multi wavelength. ✤ Polarisation: -> mono mode polarisation maintaining fibers, Niobate plates, Wollaston

prism to separate polarisation. ✤ Fringe Tracking: -> Fast fringe scan, fast responsive internal DL (piezo). ✤ Dispersion: -> increase of spectral resolution and spectral range (optical, j, h, k, …) ✤ Detector noise and quality ✤ Saoftware and hardware computing speed

Detection Speed MattersThe actual Pionier PICNIC detector, from the IOTA/IONIC-3 experiment, has a good noise <15e- !On sky operator has to adapt the scanning frequency and dispersion mode according to atmospheric conditions and target brightness to optimise SNR: !✤ Read out mode: DOUBLE (for very bright star), FOWLER (non-destructive mode reduce considerably

the noise to a few e-), ✤ # of reads per scan (256, 512, 1024, 2048, …), ✤ Piezo stroke length (40 to 120 µm), ✤ AC or ABCD integrated optic output (12 or 24) ✤ LARGE, SMALL or Free dispersion (5, 3 or 1 spectral chanels). !The scanning time is proportional to : #of optic outputs * #of dispersion channels * 1/#of read per scan * (read-out/pixel time) The SNR depend of : #of read per scan, read-out noise, seeing, #of good scan (-> Piston effect ~ Coherence time), …. !In bad or average condition the PICNIC is too slow, one has to sacrifice the spectral resolution, the Wollaston, to keep track of the fringes with a good SNR.

We need a faster detector with low noise capability.

RAPID

RAPIDRapid is born from a large collaboration within the “labex” FOCUS in which IPAG was involved to develop fast detectors for adaptive optics and optical interferometry. !Outer Electronic made by l’Onera, chip and proxy-electronics made by SOFRADIR, LETI, cryogen made by l’IPAG. !RAPID is intensively tested at IPAG on the interferometric bench named BETI. !It is a HgCdTe 320x256 Avalanche Photo Diode matrice made of 8 separated outputs. Adjustable multiplicative gain without additional noise (-7V reverse bias polarisation) !• Pixel size of 30µm • Almost flat Quantum Efficiency from 0.4 to ~3 / 3.2µm ! • Frame rates of 1600 Hz, full frame ! The fattest NIR detector ever made (i think). • Noise of ~2 electrons per frame ! !Operated in a compact Pulse-Tube Cryo-cooler at ~80K (first one at Paranal) -> No nitrogen re-feeling.

RAPID

RAPID

RAPIDFlat Taken with the maximum multiplicative gain ~20

RAPIDPhoton transfert curve

Cubes of images are taken for different incoming fluxes (or exposure times). The slope of the signal variance function to the average of the signal gives us the conversion gain in e-/ADU

Figure 1: Filter H1 and H2 transitivity, measured on the IR spectrometer withan accuracy of 0.1% at worst, increasing with the wavelength

3 Stand Alone Detector Characterization

After the camera functioning properly and the VLTi software tuned to obtainedcubes of data by G. Zins, we put it front of the integrating sphere. The firstthings we had to characterize are the noise and the gain of the detector fordi↵erent amplifications and the thermal IR contamination.

We have first increased the exposure time (DIT) gradually with the detectorlighted by a constant flux of the integrating sphere. These allowed us to plotthe pixel’s response to electron flux and check the linearity regime and to drawa photon transfer curve (PTC). The PTC represent the noise variance functionto the flux, indeed, the measured signal on the detector is written as:

Stot

=1

kN

e

� + Soff

(1)

with Stot

the measured output signal (in ADU), k the conversion gain (ine�/ADU), N

e

� the number of generated electrons and Soff

is the o↵set sig-nal (in ADU, is the equivalent of BIAS for CCD). The noise on a pixel level isthe quadrature sum of two components :

�2tot

=1

k2⌘2R

+1

k2⌘2phot

(2)

With ⌘R

the noise associated to the readout channel (in e�) and ⌘phot

thephoton noise (in e�). Considering a Poisson noise, ⌘

phot

=pN

e

� so ⌘2phot

=k (S

tot

� Soff

), then :

�2tot

=1

k2⌘2R

+1

k(S

tot

� Soff

) (3)

The inverse of the slope of the variance (�2tot

) function to the average e↵ectivesignal (hS

tot

� Soff

i) gives us the conversion gain k.

3.1 Photon Transfer Curve, gain and floor noise

The first two PTC are plotted on figure 2 for a applied polar of 0 (e↵ectivepolar at -0.9V ! gain=1), 5v and 7.1V (the maximum polarization for this

2

Figure 1: Filter H1 and H2 transitivity, measured on the IR spectrometer withan accuracy of 0.1% at worst, increasing with the wavelength

3 Stand Alone Detector Characterization

After the camera functioning properly and the VLTi software tuned to obtainedcubes of data by G. Zins, we put it front of the integrating sphere. The firstthings we had to characterize are the noise and the gain of the detector fordi↵erent amplifications and the thermal IR contamination.

We have first increased the exposure time (DIT) gradually with the detectorlighted by a constant flux of the integrating sphere. These allowed us to plotthe pixel’s response to electron flux and check the linearity regime and to drawa photon transfer curve (PTC). The PTC represent the noise variance functionto the flux, indeed, the measured signal on the detector is written as:

Stot

=1

kN

e

� + Soff

(1)

with Stot

the measured output signal (in ADU), k the conversion gain (ine�/ADU), N

e

� the number of generated electrons and Soff

is the o↵set sig-nal (in ADU, is the equivalent of BIAS for CCD). The noise on a pixel level isthe quadrature sum of two components :

�2tot

=1

k2⌘2R

+1

k2⌘2phot

(2)

With ⌘R

the noise associated to the readout channel (in e�) and ⌘phot

thephoton noise (in e�). Considering a Poisson noise, ⌘

phot

=pN

e

� so ⌘2phot

=k (S

tot

� Soff

), then :

�2tot

=1

k2⌘2R

+1

k(S

tot

� Soff

) (3)

The inverse of the slope of the variance (�2tot

) function to the average e↵ectivesignal (hS

tot

� Soff

i) gives us the conversion gain k.

3.1 Photon Transfer Curve, gain and floor noise

The first two PTC are plotted on figure 2 for a applied polar of 0 (e↵ectivepolar at -0.9V ! gain=1), 5v and 7.1V (the maximum polarization for this

2

Figure 1: Filter H1 and H2 transitivity, measured on the IR spectrometer withan accuracy of 0.1% at worst, increasing with the wavelength

3 Stand Alone Detector Characterization

After the camera functioning properly and the VLTi software tuned to obtainedcubes of data by G. Zins, we put it front of the integrating sphere. The firstthings we had to characterize are the noise and the gain of the detector fordi↵erent amplifications and the thermal IR contamination.

We have first increased the exposure time (DIT) gradually with the detectorlighted by a constant flux of the integrating sphere. These allowed us to plotthe pixel’s response to electron flux and check the linearity regime and to drawa photon transfer curve (PTC). The PTC represent the noise variance functionto the flux, indeed, the measured signal on the detector is written as:

Stot

=1

kN

e

� + Soff

(1)

with Stot

the measured output signal (in ADU), k the conversion gain (ine�/ADU), N

e

� the number of generated electrons and Soff

is the o↵set sig-nal (in ADU, is the equivalent of BIAS for CCD). The noise on a pixel level isthe quadrature sum of two components :

�2tot

=1

k2⌘2R

+1

k2⌘2phot

(2)

With ⌘R

the noise associated to the readout channel (in e�) and ⌘phot

thephoton noise (in e�). Considering a Poisson noise, ⌘

phot

=pN

e

� so ⌘2phot

=k (S

tot

� Soff

), then :

�2tot

=1

k2⌘2R

+1

k(S

tot

� Soff

) (3)

The inverse of the slope of the variance (�2tot

) function to the average e↵ectivesignal (hS

tot

� Soff

i) gives us the conversion gain k.

3.1 Photon Transfer Curve, gain and floor noise

The first two PTC are plotted on figure 2 for a applied polar of 0 (e↵ectivepolar at -0.9V ! gain=1), 5v and 7.1V (the maximum polarization for this

2

Figure 1: Filter H1 and H2 transitivity, measured on the IR spectrometer withan accuracy of 0.1% at worst, increasing with the wavelength

3 Stand Alone Detector Characterization

After the camera functioning properly and the VLTi software tuned to obtainedcubes of data by G. Zins, we put it front of the integrating sphere. The firstthings we had to characterize are the noise and the gain of the detector fordi↵erent amplifications and the thermal IR contamination.

We have first increased the exposure time (DIT) gradually with the detectorlighted by a constant flux of the integrating sphere. These allowed us to plotthe pixel’s response to electron flux and check the linearity regime and to drawa photon transfer curve (PTC). The PTC represent the noise variance functionto the flux, indeed, the measured signal on the detector is written as:

Stot

=1

kN

e

� + Soff

(1)

with Stot

the measured output signal (in ADU), k the conversion gain (ine�/ADU), N

e

� the number of generated electrons and Soff

is the o↵set sig-nal (in ADU, is the equivalent of BIAS for CCD). The noise on a pixel level isthe quadrature sum of two components :

�2tot

=1

k2⌘2R

+1

k2⌘2phot

(2)

With ⌘R

the noise associated to the readout channel (in e�) and ⌘phot

thephoton noise (in e�). Considering a Poisson noise, ⌘

phot

=pN

e

� so ⌘2phot

=k (S

tot

� Soff

), then :

�2tot

=1

k2⌘2R

+1

k(S

tot

� Soff

) (3)

The inverse of the slope of the variance (�2tot

) function to the average e↵ectivesignal (hS

tot

� Soff

i) gives us the conversion gain k.

3.1 Photon Transfer Curve, gain and floor noise

The first two PTC are plotted on figure 2 for a applied polar of 0 (e↵ectivepolar at -0.9V ! gain=1), 5v and 7.1V (the maximum polarization for this

2

Without Multiplicative Gain With max Multiplicative Gain

RAPIDEfficiency

Map of conversion gain in e-/ADU (for max multiplicative gain: e.i. polar max)

RAPIDFloor Noise

Map of floor noise in e- (for max multiplicative gain: e.i. polar max)

RAPIDEntrance window & Filters

1E#24&

1E#23&

1E#22&

1E#21&

1E#20&

1E#19&

1E#18&

1E#17&

1E#16&

1E#15&

1E#14&

1E#13&

1E#12&

1E#11&

1E#10&

1E#09&

1E#08&

1E#07&

1E#06&

1E#05&

0,0001&

0,001&

0,01&

0,1&

1&

1,42& 1,51& 1,6& 1,71& 1,83& 1,97& 2,13& 2,32& 2,55& 2,83& 3,18&

Transmission&H1#H2#2K&

Transmission&2H1#2H2&

Transmission&2H1#H2#K&

Transmission&2H1#2K&

Transmission&2H1/2H2/K&

Tran

smiss

ion

Lambda µm

Filters cut after 2µm but transparent in optical. Aperture of f/1

GRISM GRISM+WOLL FREE WOLL

RAPIDStatus

We have currently two detectors. We have fully characterised the first one and made interferometric fringes on the new BETI bench. We have improved considerably the cryo design to lower the thermal background contamination (from ~100 e-/ms to 8 e-/ms). The detector performances are very good, however we have encountered unexplained instability on the cooling system (80k reached but with max power). !-> We have decided to delay the installation of RAPID on PIONIER which was first planned last December. But the software VLTi 2011 that come with RAPID is ready.

The second detector+cryo is currently in test and the cryo-cooler is stable. Our next slot to install the detector is in June.

RAPID

✤ Increase of the spectral resolution ✤ Go to J band ✤ Photometric channel ? ✤ Cold Optical ? ✤ Adjust Target/Calibration Signal with exp time ? ✤ ….. what else ?

What will Change

What RAPID+PIONIER could allow in the future

✤ No more K band integrated optic possible. ✤ Only 2 dispersion modes (FREE and LARGE(R) ~12 channels) -> Better Spectral

resolution ~70 ✤ Bonus of a shorter wavelength, a bit of J ? (we will see on sky, but it’s free) ✤ Wollaston can be used with and without spectral dispersion (and without cost) ✤ Only one read-out mode (with high dynamic). Change of multiplicative gain ? ✤ Better Sensitivity ?