Signal-to-Noise Ratios (SNRs) of Resonon Hyperspectral Cameras
By Dr. Rand Swanson, CEO - December 12, 2023
The Signal-to-Noise Ratio (SNR)
is a commonly used metric when one is evaluating the performance of a hyperspectral
camera.
Our previous post covered the different variables at play, some common
approximations made, and the different means available to change the SNR of hyperspectral
data.
This post covers the modeled
SNRs for each of Resonon’s Hyperspectral Imaging Sensors. In the
plots that follow, the SNRs are determined for each spectral channel of the
camera, resulting in an SNR plot that is a function of wavelength.
In these data, the hyperspectral
imaging sensors are modeled as sampling the signal from a solar-illuminated, perfectly
diffuse (i.e. Lambertian)
object with 100% reflectivity, resulting in the spectral radiance profile shown
in Figure 1. The integration time used in the modeling (the inverse of which represents the maximum
frame rate) and any default spectral binning are provided for each camera. The
integration time for each hyperspectral imager is chosen such that the strongest channel is
at approximately 90% of the pixel full-well depth.
Figure 1: Spectral radiance as a function of wavelength from a solar-illuminated Lambertian surface with 100% reflectivity.
Pika UV
Figure 2: SNR for the Pika UV with an integration time of 21.7 milliseconds and default spectral binning of 4.
The
falloff of SNR for the Pika UV hyperspectral imager observed at short wavelengths is largely due to
the weak solar illumination at these wavelengths.
Pika L
Figure 3: SNR for the Pika L with an integration time of 18.5 milliseconds and default spectral binning of 2.
The SNR profile of the Pika L hyperspectral imager is
similar to the solar spectrum. A weak detector response at longer wavelengths
contributes to small SNRs at these wavelengths.
Pika L-F
Figure 4: SNR for the Pika L-F with an integration time of 11.6 milliseconds and default spectral binning of 2.
The SNR profile of the Pika L-F hyperspectral imager is very similar to that of the Pika L.
Pika XC2
Figure 5: SNR for the Pika XC2 with an integration time of 19.25 milliseconds and default spectral binning of 2.
Like
the Pika L, above, the SNR profile of the Pika XC2 hyperspectral imager is similar to the solar
spectrum. This camera also has a weak detector response at longer wavelengths.
Pika IR and Pika IR-L
The
Pika IR and the Pika IR-L hyperspectral imagers use the same detector. While they do not use the same
exact gratings, the SNR vs. wavelength for the two cameras is almost identical. As
such, they are both represented by the following plot.
Figure 6: SNR for the Pika IR and the Pika IR-L with an integration time of 5.32 milliseconds and no binning.
The strong spectral features in the Pika IR and Pika IR-L SNR plot are due to the
shape of the solar spectrum in this wavelength range.
Pika IR+ and Pika IR-L+
The Pika IR+ and the Pika IR-L+ hyperspectral imagers use the same
detector. Like the Pika IR and Pika IR-L, these two cameras do not use the same
exact gratings, but the SNR vs. wavelength for the two cameras is effectively the same.
As such, they are both represented by the following plot.
Figure 7: SNR for the Pika IR+ and Pika IR-L+ with an integration time of 19.3 milliseconds and no binning.
Like the Pika IR and Pika IR-L above, the
strong spectral features in this SNR plot are due to the shape of the solar
spectrum in this wavelength range.
Pika SWIR
The Pika SWIR hyperspectral imager uses a cryogenically-cooled MCT (Mercury Cadmium Telluride) detector.
Figure 8: SNR for the Pika SWIR with an integration time of 11.5 milliseconds and default spectral binning of 2.
Like Pika IR-series cameras above, the strong spectral features in this SNR plot are due
to the shape of the solar spectrum in this wavelength range.
SUMMARY
As has been shown, the peak SNR value of a
hyperspectral imager does not tell the whole story. The spectral response of
your application, the spectral radiance of your illumination source, and the
specific hyperspectral imaging sensor and framerate you use all determine the SNR
of your collected data.
We hope that by providing you the SNR
plots for each of our hyperspectral imagers, we are enabling you to make a more
informed purchasing decision.
To learn more about modeling SNR in hyperspectral imagers, please see this post.
To explore all of the specifications of each of our hyperspectral imagers, please visit our cameras page.
Finally, if you have any questions about what
you’ve read or if you’d like us to scan some of your samples for you to
see the actual data output, please contact our Sales Team.
Dr. Rand Swanson, CEO
Dr. Rand Swanson, CEO of Resonon
Rand Swanson is the Chief Executive Officer and co-founder of Resonon. Along with overseeing company management, Rand applies his decades of expertise in optics to optical design and radiometric modeling for commercial and R&D projects.
He considers himself fortunate to collaborate with the talented and dynamic Resonon team and to focus his efforts on hyperspectral imaging—a technology that is simple in concept, complex in execution, and applicable to a vast array of uses.
