The short answer
Start with the photons your experiment actually delivers.
A camera cannot recover photons that never reach the sensor. Before comparing megapixels or frame rates, define the wavelength, expected signal per pixel, exposure time, optical magnification, and smallest feature you must resolve.
For dim, short-exposure imaging, high QE and low read noise usually dominate. For bright samples, read noise becomes less important and pixel size, sensor area, speed, and dynamic range often decide the result. The right camera is therefore application-specific—not simply the model with the largest headline number.
The governing principle
Choose the camera that maximizes signal-to-noise ratio while sampling the optical image correctly. More QE helps the signal. Less noise protects weak signals. Pixel size decides how that signal is distributed spatially.
01 — Quantum efficiency
QE measures photon conversion—not image quality by itself.
Quantum efficiency is the percentage of photons at a particular wavelength that generate detected photoelectrons. If 100 photons reach a pixel and the QE is 80%, the expected signal is about 80 electrons.
Signal electrons = incident photons × QE
Check the curve, not only peak QE
Peak QE may occur near 550 nm while your signal is at 405, 785, or 1000 nm. Compare QE at your emission or illumination wavelength.
Include the complete optical path
Filters, objectives, windows, fiber coupling, and spectrographs reduce the photons that reach the detector. Camera QE cannot compensate for severe upstream losses.
Use QE to manage dose
In live-cell fluorescence, higher QE can produce the required signal with less excitation power or a shorter exposure, reducing phototoxicity and bleaching.
02 — Read noise
Read noise matters most when the signal is small.
Read noise is the uncertainty introduced while converting the pixel charge into a digital value. It is normally specified in electrons RMS. Unlike photon shot noise, it is generated by the detection chain rather than by the arrival statistics of light.
When each pixel contains hundreds or thousands of signal electrons, a difference between 1 e⁻ and 2 e⁻ read noise is often negligible. At only a few electrons per pixel, it can be decisive.
0.97 e⁻
sMAX04BM HDR 11HL read noise
1.07 e⁻
sMAX04BM 12-bit CMS read noise
4.06 e⁻
sMAX16AM 12-bit HCG read noise
4.59 e⁻
sMAX16AM HDR16 read noise
Always confirm the read-noise value for the actual gain mode and readout speed you intend to use. The lowest advertised value may not apply at maximum frame rate or maximum dynamic range.
03 — Pixel size
Pixel size is a sampling decision before it is a sensitivity decision.
Larger pixels intercept more light at a given sensor irradiance and often offer greater full-well capacity. Smaller pixels provide finer sampling or a larger field of view for a given sensor size—but can spread a diffraction-limited spot across more readouts.
Larger pixels
- More photons collected per pixel at equal irradiance
- Often higher full-well capacity
- Fewer pixels across the same optical feature
- Useful for very dim signals or long focal-length systems
Smaller pixels
- Finer spatial sampling at the same magnification
- Potentially more resolution when optics support it
- More flexible digital cropping and registration
- Useful for bright samples and wide-field detail
Object-space sampling
Effective pixel size
Camera pixel pitch divided by total magnification.
For diffraction-limited microscopy, a practical target is usually about two to three pixels across the smallest resolvable feature. Oversampling far beyond this adds data and distributes photons without creating new optical information.
04 — Put the specifications together
Signal-to-noise ratio is the common language.
A simplified camera SNR calculation reveals which noise source controls your image. Use electrons—not digital counts—whenever possible.
Detected signal
S = photons × QE
Convert expected photons per pixel into signal electrons using QE at the relevant wavelength.
Total noise
σ = √(S + B + D + RN²)
Combine signal shot noise, background, dark electrons, and read noise in quadrature.
Image quality
SNR = S ÷ σ
Compare candidate cameras under the same exposure, optics, wavelength, and sampling conditions.
Interactive SNR lab
Compare real catalog cameras
Camera A
sMAX04BM
95% @ 560 nm · HDR 11HL · 0.97 e⁻
- Signal
- 23.75 e⁻
- Total noise
- 5.34 e⁻
- Pixel pitch
- 6.5 µm
- Sensor
- GSENSE2020BSI
Camera B
sMAX16AM
73.94% @ 600 nm · 12-bit HCG · 4.06 e⁻
- Signal
- 18.49 e⁻
- Total noise
- 6.16 e⁻
- Pixel pitch
- 9 µm
- Sensor
- GSENSE4040
This comparison uses each sMAX camera's published peak QE. It does not substitute peak QE for a full spectral curve. Dark current and pixel-area irradiance effects are omitted here so the controls isolate QE and readout-mode noise.
Interactive sampling lab
Match pixel size to your microscope
Object-space pixel
108 nm
Airy radius estimate
280 nm
Pixels across radius
2.58
sMAX04BM · 6.5 µm pixels
GSENSE2020BSI · 2048 × 2048 · 4.2 MP
The glowing spot represents the 0.61λ/NA Airy-radius estimate relative to the sensor's object-space pixel grid. Aim for roughly 2–3 pixels across the smallest resolvable detail; confirm with your complete optical model.
Values available in your catalog
Complete MAX and sMAX series overview
Every MAX and sMAX series listed on the scientific-camera page is included. Models without both published QE and read-noise values remain available in the sampling lab but are intentionally excluded from numerical SNR comparison.
- Peak QE
- 95% @ 560 nm
- Pixel
- 6.5 µm
- Read noise
- 0.97–23.25 e⁻
- Resolution
- 2048 × 2048 · 4.2 MP
- Peak QE
- 73.94% @ 600 nm
- Pixel
- 9 µm
- Read noise
- 4.06–35.61 e⁻
- Resolution
- 4096 × 4096 · 16 MP
- Peak QE
- 90% @ 550 nm
- Pixel
- 9 µm
- Read noise
- Not specified
- Resolution
- 4096 × 4096 · 16 MP
- Peak QE
- Not specified
- Pixel
- 2.81 µm
- Read noise
- TBD
- Resolution
- 19200 × 12800 · 251 MP
- Peak QE
- Not specified
- Pixel
- 3.76 µm
- Read noise
- ~2.5 e⁻
- Resolution
- 14176 × 10640 · 151 MP
- Peak QE
- Not specified
- Pixel
- 3.76 µm
- Read noise
- ~2.5 e⁻
- Resolution
- 11648 × 8742 · 102 MP
- Peak QE
- Not specified
- Pixel
- 3.76 µm
- Read noise
- Not specified
- Resolution
- 9568 × 6380 · 61 MP
- Peak QE
- Not specified
- Pixel
- 5.94 µm
- Read noise
- Not specified
- Resolution
- 6064 × 4040 · 24 MP
- Peak QE
- 64.2% @ 595 nm
- Pixel
- 6.5 µm
- Read noise
- Not specified
- Resolution
- 2048 × 2048 · 4.2 MP
- Peak QE
- 93.7% @ 550 nm
- Pixel
- 6.5 µm
- Read noise
- Not specified
- Resolution
- 2048 × 2048 · 4.2 MP
- Peak QE
- 95.3% @ 560 nm
- Pixel
- 11 µm
- Read noise
- Not specified
- Resolution
- 2048 × 2048 · 4.2 MP
05 — Selection workflow
Make the choice in six defensible steps.
Define the signal
Estimate wavelength, photons per pixel, exposure time, background, and required frame rate.
Define the smallest feature
Use numerical aperture, wavelength, magnification, and desired field of view to set sampling.
Compare QE at wavelength
Ignore peak-only marketing numbers; read the spectral QE curve at your actual signal.
Compare noise in the real mode
Use the gain, bit depth, shutter mode, and readout speed you will operate.
Calculate SNR
Model at least a dim, typical, and bright condition, including background and dark current.
Check the complete system
Confirm sensor format, mount, data rate, trigger, cooling, software, and mechanical fit.
06 — Quick comparison
What should your application prioritize?
| Application | Quantum efficiency | Read noise | Pixel size |
|---|---|---|---|
| Very weak fluorescence | Prioritize high QE at the emission wavelength | As low as practical | Match sampling; consider modest binning |
| Fast live-cell imaging | High QE reduces illumination dose | Important at short exposures | Enough resolution without excessive data |
| Brightfield or inspection | Usually not the limiting factor | Moderate noise is often acceptable | Choose for spatial resolution and field of view |
| Astronomy / long exposure | High across the target spectrum | Low read noise and low dark current | Match image scale to seeing and focal length |
| Localization microscopy | High QE improves photon statistics | Low noise preserves dim localizations | Sample the PSF correctly after magnification |
Camera selection support
Send us your wavelength, optics, and expected signal.
We will help translate your experiment into a shortlist based on sampling, SNR, sensor format, frame rate, and integration requirements.
Frequently asked questions
A few important nuances.
Is the camera with the highest QE always the most sensitive?
No. QE tells you what fraction of incident photons become electrons, but total sensitivity also depends on read noise, dark current, optics, exposure time, and the signal level. A 95% QE camera with poor sampling or unsuitable noise performance may underperform a better-balanced camera.
Are larger pixels always better in low light?
Larger pixels usually collect more photons per pixel at the same irradiance, but they also sample the image more coarsely. They are valuable when light is scarce and resolution is not limited by sampling, provided the sensor size and optics remain compatible.
Can software binning make small pixels equivalent to large pixels?
It can combine signal from neighboring pixels and improve the displayed signal-to-noise ratio, but it does not recreate every property of a physically larger pixel. Whether read noise is incurred once or multiple times depends on the sensor architecture and binning method.
What specification is most commonly overlooked?
The QE curve at the actual wavelength. Peak QE is a headline number; your fluorophore, Raman band, UV signal, or NIR target may sit far away from that peak.
Technical note: the equations above are simplified for camera comparison. A rigorous model may also include clock-induced charge, fixed-pattern noise, pixel-response non-uniformity, excess noise, gain calibration, and the statistics of any image processing.