INCREASE HDR DYNAMIC RANGE?
With exception of the Arri Alexa 35, all cine cameras still only have a hardware-based (sensor) dynamic range (DR) of around 13 stops. However, using the log profiles of the camera manufacturers, it is possible to record up to 15 stops DR. However, this could change in the future, because Thomas Wocial, Konstantin D. Stefanov, William E. Martin, John R. Barnes and Hugh RA Jones have published a paper on the possibility of a new readout method for CMOS image sensors (CIS) entitled "A Method to Achieve High Dynamic Range in a CMOS Image Sensor Using Interleaved Row Readout". The "interleaved line reading" described therein led to significant increases in DR, although some work is still required before the process is ready for series production. However, manufacturers could quite quickly "upgrade" their cameras in terms of DR at no additional hardware cost, provided the controllers allow it. The results of the tests are more than remarkable!
STATE OF THE ART
There are a number of techniques for extending the dynamic range (DR) of CMOS image sensors (CISs). So far, these have been divided into seven categories:
1) logarithmic pixel response,
2) combined linear and logarithmic response,
3) well capacity adjustment,
4) frequency-based sensors,
5) time-to-saturation-based sensors,
6) global control of integration and
7) local control over integration
Further advances leading to increased DR could be achieved with improved dark current suppression, lower readout noise and multigain readout. The 4/5T pixel architecture is widely used in scientific imaging, but the choice of DR enhancement techniques is still severely limited.
A well-established technique for increasing DR is multiple exposure, which does not require additional circuitry. In the double exposure, two images are acquired with different integration times, Tlong and Tshort, with the DR extension corresponding to Tlong/Tshort. There is a signal-to-noise ratio (SNR) reduction in areas that correspond to saturation in the short exposure, since the photo signal is only sampled for a fraction of the total integration time. Using multiple shorter integrations can reduce the resulting SNR dip.
In a nondestructive readout (NDR), the signal is sampled many times during the integration time. The CIS is read out with an up-the-ramp scan, which is widely used for IR photodiode arrays for DR increase and cosmic ray suppression.
Now the team presented a readout scheme for CMOS image sensors with which, in principle, an arbitrarily high dynamic range (HDR) can be achieved without developing a new sensor.
A CIS115 was used, for which a line-by-line coded exposure was implemented. An increase in DR from 73.6 to 107.6 dB could be achieved, which corresponds to an increase of almost 50%! In addition, the signal-to-noise ratio (PSNR) was increased from 42.9 to 59.9 dB. Both remarkable values!
"We present a readout scheme for CMOS image sensors that can be used to achieve arbitrarily high dynamic range (HDR) in principle. The linear full well capacity (LFWC) in high signal regions was extended 50 times from 20 to 984 ke− via an interlaced row-wise readout order, while the noise floor remained unchanged in low signal regions, resulting in a 34-dB increase in DR. The peak signal-to-noise ratio (PSNR) is increased in a continuous fashion from 43 to 60 dB." (Abstract from the paper "A Method to Achieve High Dynamic Range in a CMOS Image Sensor Using Interleaved Row Readout")
In this way, the DR can be extended as required by reading out a line several times without negatively affecting the noise behavior. Among other things, this scheme also led to an increase in the centering accuracy in regions with a low signal level due to the single readout, while the linear FWC was expanded through multiple readouts. The order in which the lines are read out must be determined in advance, because this is the only way it is also suitable for scenes with minor temporal fluctuations.
As a result, the DR extension range achieved is not hardware-based and can therefore be increased to a user-defined level, but at the expense of a longer minimum integration time.
According to the results of the article, if one were to increase the number of reset and read samples from 50 to 200, the DR extension shown in the article to 120 dB would be achieved.
So the future of digital cameras does not necessarily require new sensors with new technologies; a significant increase in DR is already possible by changing the read-out. The question remains at which sampling rates which increase can be achieved. Use case tests are also still missing, but the proof has already been provided.
Personally, I have been working on and with imaging processes for over 10 years and am confronted with bottlenecks for electronic components on a daily basis. With the help of this method, especially for time-critical and scientific applications, e.g. in astronomy, tremendous increases in DR can be achieved. It remains to be seen whether and how this method will also be received and accepted by cine camera manufacturers.