NeutronOptics - Optional High Resolution CCD

Large pixel CCD and megapixel CMOS cameras for neutron imaging.

Camera makers often emphasise the advantages of megapixel CMOS cameras, which are by far the most popular and usually the most suitable for commercial applications of optical imaging; they are also less expensive to make. CMOS cameras have advantages, but also disadvantages.

More pixels mean smaller pixels. Light gathering capacity depends directly on pixel area, and this is more important than small differences in quantum efficiency. Resolution in radiography is determined in practice by the resolution of the scintillator (50-100µm for ordinary 6LiF/ZnS), beam collimation and the distance between object and scintillator, while an optical camera is limited in principle only by the wavelength of light (~0.5µm).
CMOS cameras have faster readout and lower readout noise, but at the expense of higher thermal electronic noise (dark current) which becomes important for the long (seconds or minutes) exposures often needed with neutrons. CMOS technology is however developing rapidly, dark current has been reduced, and efficiency increased with back illuminated chips
Both CCD and CMOS cameras can use Peltier cooling to reduce thermal noise, but because the total noise is lower for CCDs, and readout is 16-bits, the dynamic range is higher. High dynamic range means that with a single exposure both low and high intensity features can be seen, by choosing the perceived intensity range of 256 levels out of a possible 65,536 (16-bit). The dynamic range with CMOS is usually lower, especially with high gain and longer exposures.
CMOS readout is faster because circuitry is simpler and usually 12-bits rather than 16-bits, and pixels cannot usually be binned before readout, as with a CCD to increase sensitivity and reduce noise. But again, technical advances have increased CMOS dynamic range to 14-bits or even 16-bits in some cases.
Electron or photon multiplication is sometimes used to boost low light signals, with "single quantum" detection claims. But a single neutron capture produces up to 160,000 photons; even if only a small percentage are captured by the detector, further multiplying photons does not improve statistics. Multiplication is useful when imaging light, or with x-ray scintillators, but electron or photon multiplication also increases noise and reduces dynamic range.
An effort has been made to combine the low noise qualities of CCDs with the fast readout of CMOS, resulting in "scientific CMOS" (sCMOS), again mainly used for biological imaging. But sCMOS is also useful for neutron imaging with high flux sources when very short exposures are required.

Both CCD and CMOS cameras are optimized for imaging directly with light. The largest markets are for consumer cameras, industrial and security applications, and biological science, all of which have different requirements to neutron imaging. Expensive cameras are produced for biological science, but such cameras have few advantages for neutron imaging. We use cameras designed for amateur astronomy, where the technical requirements are closest to those needed for neutron imaging (longer, low noise exposures with high dynamic range). See Why buy a camera instead of making one. Finally, neutron imaging has a radiation background that eventually damages or destroys the detector, and otherwise negates many of the advantages of expensive cameras.

Choice of CCD and CMOS detectors.

Detector	Slim CCD	Slim CMOS	HiRes CMOS	Fast CMOS	Square CMOS	# FS60	#VS60	CMOS7.1	4/3" CMOS	APS-C CMOS	Full-Frame
Type	Interline ICX829	Pregius IMX174	Pregius IMX183	Pregius IMX432	Pregius IMX533	Interline ICX694	Interline ICX694	Pregius IMX428	Pregius IMX294	Pregius IMX571	Pregius IMX455
Resolution pixel	752 x 580	1920x1200	5472x3648	1600x1100	3000x3000	2759x2200	2759x2200	3208x2200	4144x2822	6248x4176	9576x6388
Image diag. mm	8 (1/2")	13 (1/1.2")	15.9 (1")	17 (1.1")	16 (1.1")	16 (1")	16 (1")	17 (1.1")	23(4/3")	28(APS-C)	43(35mm)
Image area mm	6.46x4.81	11.25x7.03	13.13x8.75	14.4x9.9	11.3x11.3	12.53x9.99	12.53x9.99	14.4x9.9	19.1x13	23.5x15.7	36 x 24
Pixel size µm*	8.6 x 8.3	5.86 x 5.86	2.4 x 2.4	9.0 x 9.0	3.76x3.76	4.54 x 4.54	4.54 x 4.54	4.5 x 4.5	4.6 x 4.6	3.76x3.76	3.75 x 3.75
Quantum effic*	~75%	~80%	~79%	~72%	~80%	~70%	~70%	~75%	~90%	~90%	90%
Fullwell e- **	~40,000	~30,000	~15,000	~80,000	~50,000	~20,000	~20,000	~20,000	~66,000	~50,000	~50,000
Read noise e- **	10	7	4	5	3	5	6	3	1.2-7.3	1.0-3.3	1.5-3.5
Dark c. e-/pix/s	<0.1@25°C	~1.0@45°C	~2.0@45°C	~2.5@45°C	0.001@-20°C	0.0004@-10°C	0.0004@-10°C	0.03@-10°C	0.002@-20°C	0.003@0°C	0.003@0°C
Peltier Cooling	uncooled	uncooled	uncooled	uncooled	Δ -35 °C	Δ -27 °C	Δ -35 °C	Δ -35 °C	Δ -35 °C	Δ -35 °C	Δ -35 °C
Read time (s)***	0.5	41 fps	18 fps	69 fps	20 fps	1 to 3	0.1 to 1	30 fps	16 fps	3.5 fps	0.5
A/D Readout**	16-bits	12-bits	12-bits	12-bits	14-bits	16-bits	16-bits	12-bits	14-bits	16-bits	16-bits
Binning h,v	x1 x2 x4 x8	software	software	software	software	x1 x2 x4 x8	x1 x2 x4 x8	sofware	software	software	x1 x2
Mount	CS-mount	C-mount	C-mount	C-mount	C-or F-mount	C-or F-mount	C- or F-mount	C- or F-mount	C- or F-mount	F-mount	F-mount
Trigger signals	Software	Software	Software	Software	Software	Software	GPIO	software	Software	Software	Software
Interface***	USB 2.0	USB3/GigE	USB3/GigE	USB3/GigE	USB2/USB3	USB 2.0	USB 2.0	USB 3.0	USB 3.0	USB 3.0	USB 3.0
Relative cost	1	1	1.5	1.5	2	3	6	4	3	5	8

The collimation and quality of your neutron beam-line will usually be the limiting factor for neutron imaging, not the camera.

Light capture per pixel is proportional to the pixel area and its quantum efficiency (at 525nm)

Fullwell Capacity is the number of electrons that can be stored without overflow (blooming)

Fullwell Capacity is also proportional to pixel area, but also depends on anti-blooming design

Noise can be "Dark current" due to thermal energy, or "Read noise" due to electronic readout

Dark Current can be reduced by cooling for long exposures. Modern Sony ICX CCDs have exceptionally low dark current.

Dynamic range is the ratio of Fullwell capacity to Noise, & is lower than the 16-bit (65,536) readout

Dynamic Range in Decibels DR= 20log (Fullwell capacity/Noise)dB eg typically DR= 5,000= 74dB

Dynamic Range is often exagerated by neglecting dark noise (true only for very short exposures)
     Note that electron- or photon-multiplied cameras generally have a low dynamic range eg DR= 1,500= 64dB
     A high dynamic range means that contrast between slightly different intensities is better, important for imaging
*** USB 2.0 is limited in practice to ~280Mbits/s i.e for a 2048x2048x16 bit camera to ~4 frames/sec (fps)
     USB 2.0 is also sufficient for the 1920x1200 12-bit Sony Pregius CMOS camera up to 10 frames/sec (fps) over >10m active cables
     USB 3.0 is limited in practice to ~4000Mbits/s i.e for a 2048x2048x16 bit camera to ~64 frames/sec (fps) with short ~3m cables.

In our "interline" CCDs, charge accumulated by photo-sensitive columns is quickly transferred to adjacent storage columns. The advantage is that smearing is avoided during readout, so a mechanical shutter is not needed. Some of the chip area is used for storage, but that is compensated by using micro-lenses over every pixel. Sony sensors are typically ~75% sensitive to the mainly green (540nm) light emitted by neutron scintillators, and at that wavelength there is little advantage for more expensive "back-illuminated" CCDs. This explanations is simplified, but summarises why we use "slow" readout CCDs for neutron imaging.

Our small cameras for fast beam alignment are very efficient; the chips are also small, limiting imaging area but permitting the use of inexpensive fast C-mount lenses. Our f/1.0 lens is twice as fast as an f/1.4 lens. More important than the efficiency of the detector itself, the efficiency of the camera is proportional to the ratio of the CCD to scintillator area.