# My samples have very low signal, can this camera accurately image my sample?

##### THIS GUIDE IS FOR:

Biologists that use, or are interested in using, microscope cameras and don’t speak engineering

##### THIS GUIDE OFFERS:

Clarity on the relevance of camera specs to biological experimentation

## LISTED IN SPECIFICATIONS AS:

Hamamatsu listing Synonyms used by other vendors
Quantum efficiency
Dark current

Camera specs can give you an indication of whether or not the camera will be able to accurately image your sample, but image quality is dependent on the quality of your sample—as the commercials used to say, “actual results may vary,” and powerful, sensitive electronics cannot always makeup for high background or low contrast. With this caveat, the best indicators of camera sensitivity are a combination of quantum efficiency (QE) and camera noise.

### What’s QE?

QE is a measure of how many photons are converted to electrons by the sensor. QE is a function of the camera sensor materials and design and is dependent on the wavelength of the light being detected. In the camera specs, QE is presented as the percentage of photons converted into electrons—i.e. detected—at a particular wavelength, and manufacturers usually can provide QE versus wavelength plots.

Higher QEs generally indicate higher sensitivity, as the likelihood of detecting your signal is increased.

### EM-CCDs, QE and eQE

A common misperception is that EM-CCDs are really sensitive because they have a high QE. However, even though the conversion of photons to photoelectrons can be high at the pixel, the multiplication process (impact ionization) in an EM-CCD camera adds an extra noise factor that effectively lowers QE.

The origins of EM noise factor lie in the stochastic nature of the signal amplification process. We’ve written about this in more technical detail here, and the high-level summary is that the EM noise factor effectively lowers the QE across the entire spectra by 50 %. This reduction has been independently observed and noted in the literature by Huang, et al.,1 who find the lower read noise of EM-CCDs to be an advantage in only the lowest light level regimes.

### What’s ‘camera noise’?

Camera noise is the amount of noise introduced by the camera. It primarily consists of two components—read noise and dark noise.

Read noise is a term used to describe all the sources of noise associated with converting the photoelectrons in a pixel to a digital number. A large part of read noise happens when photoelectrons are converted into a voltage signal by the amplifier, and some occurs during the analog-to-digital conversion.

Lower levels of read noise are particularly important when signal is low, and part of the job of the camera designer is to optimize the system as a whole, balancing sensor performance and readout with read noise in the signal.

Read noise is typically presented as electrons, and is not affected by the wavelength of the signal. However, because the signal starts off as photons, and conversion of photons to photoelectrons, aka the QE, is wavelength dependent, it’s important to convert read noise from electrons to photons when comparing two cameras with similar noise specs but different QEs. When QE is higher and read noise in electrons is equivalent, the camera with higher QE will have lower read noise in photons, indicating the ability to image dimmer signals.

A last note about read noise specs—check whether the specification is “median” read noise or “rms” (root mean squared) read noise. The median value of read noise is typically lower than the rms, making it more attractive as a specification. And for CCDs, where read noise does not vary greatly from pixel-to-pixel, it is not terribly inaccurate. However, the different architecture of the CMOS chip—with conversion from photoelectrons to voltage occurring in parallel at each pixel—means that read noise does vary from pixel to pixel. For this reason, the rms read noise is a more meaningful metric than the median read noise.

#### Dark noise

Dark noise

Dark noise is statistical variation in the number of electrons thermally generated within the pixel in a photon-independent fashion, and is the electron equivalent of photon shot noise. Dark noise is calculated from the dark current:

$Dark\ noise=\sqrt{\mathstrut (dark\ current)(integration\ time)}$

Dark current, and therefore dark noise, are temperature dependent, with less noise at lower temperatures. For most biological experiments, dark current and dark noise are negligible over a typical exposure interval of less than five minutes.

Because dark noise is typically negligible, the main noise component coming from the camera that needs to be considered is read noise.

## BOTTOM LINE:

• Two specs are required for understanding camera “sensitivity”: read noise and QE.
• For EM-CCD cameras running in EM mode, the effective QE (eQE) is reduced by half across the spectra.
• QE is dependent on signal wavelength, read noise is not.
• For CCD cameras, rms read noise and median read noise are very close to the same value, but for sCMOS cameras, only rms read noise is an informative metric.
• QE and read noise need to be considered together, in the context of your sample and experimental question, when evaluating a camera for sensitivity. For example:
• Do you need to see the difference between two dim signals?
• Do you need to see the difference between bright signals with low contrast?
• Dark current, and the related dark noise, is negligible over typical exposure times for biological experiments.

## References

1. Huang, F. et al. Video-rate nanoscopy using sCMOS camera-specific, single-molecule localization algorithms. Nat. Methods 10, 653–658 (2013).
2. Moran, U., Phillips, R. & Milo, R. SnapShot: Key Numbers in Biology. Cell 141, 1262–3 (2010).