2011-06-06
CCD cameras and digital Imaging
Pascal Chartrand
Modern microscopy depends on digital image capture, so we start there There are Various Types of Electronic Cameras: TV cameras of many kinds and cameras that use Charged Coupled Devices (CCDs). Thus, we will discuss: • CCD Camera Architecture • CCD Camera Performance • Acquiring a Digital Image
1
2011-06-06
Electronic Detection of Light depends on the photo-electric effect When a photon hits a metal or a “doped surface,” an electron is energized and therefore destabilized Its presence can be recorded by applying a voltage to pull the energized e- off as a “current” Detectors with no spatial discrimination • Photomultiplier tubes (LSCMs) • Photodiodes Detectors with spatial discrimination • Tube cameras - Vidicon, Newvicon, SIT, ISIT • Solid-state detectors – CCDs
The sensitivities of various electronic cameras: Video - CCD
2
2011-06-06
CCD cameras
Chip
camera
Photodiodes • A light-sensitive semi-conductor set up so incident light can knock electrons loose • A “loose” electron leaves behind a zone of positive charge or a “hole” • Electrons and holes can move in response to an electric field • Current is then proportional to the number of “loose” electrons, i.e., to number of incident photons
3
2011-06-06
For many LM applications, the best imaging device is a Charge-Coupled Device (CCD) A dense matrix of photodiodes with charge storage regions called “wells” or “pixels”. Stored charge can be transferred to a “serial register” where it is held until an amplifier reads out the quantity of charge accumulated in each pixel.
Image from Fundamentals of Light Microscopy and Electronic Imaging, Doug Murphy
Diagram of an individual CCD “Well” or “Pixel”
Hazelwood et al., in Shorte and Frischknecht 2007
full well capacity on the order of ~1000 electrons/µm2
4
2011-06-06
From Video Microscopy, 2nd Ed., Inoué and Spring
A metaphore for CCD camera readout
5
2011-06-06
Three Primary CCD Chip Designs
From Video Microscopy, 2nd Ed., Inoué and Spring
Parameters of Camera Performance •Quantum efficiency •Spectral sensitivity •Spatial resolution •Dynamic range •Signal-to-noise ratio •Temporal resolution (speed) •Linearity What is most important for your experiments?
6
2011-06-06
Spectral Sensitivity & Quantum Efficiency Quantum efficiency: percentage of photons hitting the photoreactive surface that will produce an electron–hole pair - It is an accurate measurement of the device's electrical sensitivity to light
Graph from microscopyu.com
Improvements in Interline CCDs
Single microlens added
Input light
Microlens
No microlens
Open window
Image from Butch Moomaw, Hamamatsu
7
2011-06-06
Backed thinned CCDs have higher QE
From Video Microscopy, 2nd Ed., Inoué and Spring
The idea of digital image sampling
Hazelwood et al., in Shorte and Frischknecht 2007
8
2011-06-06
Accuracy of digital sampling depends in part on “spatial frequency” of sampling
Nyquist-Shannon criterion and image sampling • How many CCD pixels are needed to accurately reproduce the smallest object that can be resolved by the scope? “When sampling a signal (e.g., converting from an analog signal to digital) the sampling frequency must be at least twice the highest frequency present in the input signal if you want to reconstruct the original perfectly from the sampled version.”
9
2011-06-06
Nyquist-Shannon Sampling
Optimal Pixel Size for Preserving Resolution - For a CCD camera, photons emitted by a single point source should be captured by at least 3 pixels
Objective NA
Airy disk radius (µm)
Projected size on CCD (µm)
Optimal pixel size (µm)
40x 1.3
0.26
10.4
3.5
60x 1.4
0.24
14.4
4.8
100x 1.4
0.24
24.0
8
Pixels range from 4 – 25µm. Around 7µm is now most common. Additional magnification can be used to optimize camera resolution for a given objective lens.
10
2011-06-06
Aliasing
Nyquist sampled
Undersampled
Dynamic Range •Dynamic Range is defined as the full well capacity divided by the camera noise •Large dynamic range allows discrimination between bright and dim parts of the specimen. •Smaller pixels generally have smaller full well capacity.
11
2011-06-06
Dynamic Range, bits and gray levels Bit = 2n gray levels. A 10 bit camera has 210 or 1024 gray levels
All camera have one goal: Separating SIGNAL from NOISE •
Signal is defined as the change in the state of a detector produced by photons from the object of interest.
•
Noise is defined as meaningless fluctuations in the signal. It is of three kinds: – Optical Noise is defined as any detected photon produced by stray light from the microscope or scattered from the object of interest – Camera Noise is defined as any change in the detector output not produced by photons from the object of interest. – Statistical Noise is defined as fluctuations in signal that arise from the random changes that result from inadequate sampling.
12
2011-06-06
Noise 1. Statistical Noise: Random fluctuations in signal 2. Dark (Thermal) Noise
1.25 MHz/pixel
•Electrons pop out of the chip as it heats up •They build up with long exposure time •Can be reduced by cooling chip 3. Read Out Noise 10 MHz/pixel
•Errors as chip is read •Constant, regardless of exposure time •Can be reduced by reading the chip more slowly 4. Camera Noise (Dark + Readout) 5. Total Noise (Signal noise + Camera noise)
Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
Signal to Noise Ratio
SNR =
S N
SNR: S: N:
•
Improve the Signal by
•
Reducing the Noise
Signal to noise ratio Signal detected by the camera Total noise
– Collecting more photons from your specimen – Using a more sensitive camera – Amplyfing the signal, so long as it increases the difference between signal and noise – Reduce camera noise – Use frame averaging
13
2011-06-06
S/Ncamera = full well capacity / camera noise S/Nimage = signal / (signal noise + camera noise)
Signal to Noise begins with the Signal… Signal = I x QE x T S (electrons captured in pixels) = Signal I (photons/sec) = Input light level QE (electrons/photon) = Quantum efficiency T (sec) = Integration time
All signals have statistical noise associated with them
SNR =
Signal Signal
Think of it as counting radioactivity
14
2011-06-06
As the signal level increases, the S/Nimage gets better since the square root becomes a smaller fraction of the signal. This affects the precision of the data you can record.
5 /√5 = 2,24 10,000 /√10,000 = 100 100,000 /√100,000 = 316
Pictorial display of signal-to-noise ratio and its effect on image quality
15
2011-06-06
What can YOU do to increase the signal that the camera sees? •Brighter fluorophores •Align illumination optics (Koehler illumination) & bulb •Brighter objective lens (B = NA4 / M2) •Higher transmission objective lens (less aberration correction) •Minimize spherical aberration •Decrease specimen noise (mounting medium, BG fluorescence) •Decrease other optical noise (minimize reflective surfaces, use field diaphragm, work in dark, no dirt)
Low Light Level Cameras At low light levels, the signal approaches the noise level. Cameras employ different methods to improve the S/Ncamera. Low light cameras increase the signal, reduce the noise, or both.
Decrease Noise:
Cooling, Slower Readout, Frame Averaging
Increase Signal:
Higher QE, Bigger Pixels, Longer Exposure, Amplification
16
2011-06-06
Vacuum Chamber CCD Heat Sink
Window
Thermo Electric Cooler
Cooled CCD Images from Butch Moomaw, Hamamatsu
- Cooling the CCD camera reduces the dark current - Use thermo-electric cooling (Peltier effect) to reach -30°C to -40°C)
Graph from Video Microscopy, 2nd Ed., Inoué and Spring
Electron Multiplication (On-Chip Multiplication) - Signal amplification occurs in the gain register, before readout - Electrons are accelerated from pixel to pixel, generating secondary eletrons by impact ionization (A) to (B) - Camera cooled to -80°C (decrease dark noise) - However, signal amplification also increases dark noise
Image from www.andor-tech.com
17
2011-06-06
EM-CCD Performance
ORCA- ER 500ms, High Gain
EM-CCD 200ms, Low Multiplication
EM-CCD 200ms, Med Multiplication
EM-CCD 50ms, High Multiplication
Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
Tips on acquiring a digital image Determine optimal exposure time by looking at the gray scale values!! For fixed specimens, use full dynamic range. For live cells, collect as many photons as you can, considering temporal resolution, spatial resolution, bleaching, toxicity, etc. CCDs and PMTs are linear until saturated.
18
2011-06-06
Do not judge image quality based on how the image looks on the monitor
256 Gray levels
64 Gray levels
16 Gray levels
Images from Butch Moomaw, Hamamatsu
Image Scaling The monitor can only display 8 bits. Your image is probably 12 or 14 bits.
Scaling determines how the data are displayed on the monitor.
Actual gray values 198-1265 Displaying 0-4095
Actual gray values 198-1265 Displaying 198-1265 “Auto-scaled”
Actual gray values 198-1265 Displaying 236-546
Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
19
2011-06-06
Exposure Time 50ms
100ms
200ms
400ms
500ms
687
1051
1858
3260
3888
Image maximum grayscale value 12-bit camera maximum = 4095
Increasing exposure time increases signal while decreasing temporal resolution
Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
Gain Gain: the number of accumulated photoelectrons that determine each gray level step distinguished by the readout electronics Increasing gain reduces the number of photons /gray scale value Example: if your gain is 8 ē per gray value, a pixel signal of 8000 ē correspond to 1000 gray levels. Increasing the gain by 4 gives you 2 ē per gray value, and the same pixel signal results in 4000 gray levels
Increasing gain, Same exposure time Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
20
2011-06-06
Binning (2x2)
Read out 4 pixels as one Increases SNR by 2x Decreases read time by 2 or 4x Decreases resolution by 2x
Binning
25 ms, 4x4 bin Max 3493 50ms, 2x2 bin Max 2297 50ms, no binning Max 746
Images collected by JWS in the Nikon Imaging Center at Harvard Medical School
21
2011-06-06
Reading the manufacturer specifications Important informations are provided by the manufacturer! Photometrics CoolSnap fx
Reading the manufacturer specifications Camera noise (read noise and dark current) is given in electrons. Gray scale values are converted to electrons by multiplying by an electron conversion factor. The electron conversion factor = full well capacity / max gray scale value.
Photometrics CoolSnap fx
22
2011-06-06
23