|Dear authors, |
This paper presents a complete review of a flight detector onboard the Aeolus satellite. The analysis provides a deep understanding of both the ACCD working principle and the impact of the dark current on the wind measurement. The precise identifications of the relevant technological levers which could improve the gathered data as well as the next generation of instruments are highly valuable and represent an interesting opportunity for discussion as lessons learned. Moreover, the structure of the paper allows a good understanding of the challenges that must be faced to complete the mission requirement, and the explanations are sufficient to point out the underlying physics. It is a very nice paper.
Comments on this paper are mainly minor revisions or basic questions that could be answered to improve the global quality of the paper as well as the understanding for the non-specialist. I however have one main comment that needs to be clarified to deeply understand the role of the charge storage in the dark current.
As well described in the paper, the working principle of ACCD implies the accumulation of collected charges into a given column of pixels into a transfer row. The charge transfer in the memory zone follows.
1/ Does the charge transfer occur from the transfer row to the range gate #1 before being transfer from #1 to #2 until #25 or does the first transfer row is stored in range gate #25? This question points out the CTI issue in the memory zone that could be impacted by the radiations. Depending on the operation mode, DCNU could reveal the worst degradation for memory gates resulting from several transfers. Moreover, it appears that the charge retention time varies from a range gate to another. Has the range gate DCNU been investigated by looking at the dark signal with no transfer from the CCD?
2/ The author also well described that the memory zone plays a role in the global dark current. I agree with this statement, which is explained in the paper by the generation centers into the storage structure. However, regarding the charge retention time (i.e. depending on the range position) which can reach 0.4s, it appears that the leakage current from these storage nodes can impact the global dark current signal. The authors are invited to give their analysis regarding this leakage which could lead to a loss of collected charges and therefore implying a decrease of the signal and even partially compensate for the dark current increase from the CCD. As a complex structure, the quantification of such DC sources is not required. However, the paper could include a detailed description of the dark current sources in the ACCD to provide a better overview of the underlying mechanisms.
This part could include the CVF to account for the number of collected charges. It is specified later.
Retention time might be more appropriate rather than residence time.
I agree with the statement that the noise is dominated by the read-out noise. The authors are invited to specified “before the mission” or “before radiation exposition”.
Is this ACCD from Teledyne a COTS or a custom design? Can the reference of this ACCD be specified here?
The number of lost charges (CTI) depends on the collected charges. The authors are invited to specify “at the typical integration time” or “based on typical operation”.
The authors are invited to compare the pixel pitch to the required fringes resolution. Does smaller pixel pitch could work and what could be the limit (Full Well Capacity and/or sensitivity).
Charge accumulation can be preferred to signal accumulation.
The authors are invited to specify the operating mode of the CCD (rolling shutter) as well as how the memory zone is operated from a frame to another. Moreover, it looks like the limiting parameter in this application is the ADC frequency which implies storing several columns of accumulated frames in a memory zone. A brief description of this part would be appreciated.
Do the virtual pixels allow to monitor the applied offset?
What limits the size of the memory zone? Does a direct quantification of the transfer row possible? The authors are invited to specify the link with the readout circuit.
“the random as well as systematic error budget of CCD based measurements.” Noise or additional dark current shot noise could be preferred to random. Moreover, it does not appear clear to me what the budget means here.
“the noise contributions are related to the signal itself” Total noise comprises DC noise and read-out noise.
“The noise of Aeolus signals is dominated by the Poisson distributed shot noise as the levels for dark current and read-out noise” The dark current noise is also a shot noise. This sentence needs to be rephrased.
“Thus, the technique used for Aeolus is referred to as “quasi-photon” counting.” It does not appear that this device can integrate and measure a single photon. The noise is dominated by a readout shot noise, but photon-counting required other characteristics that are not met here.
Systematic errors could be corrected easily with an appropriate offset correction. The main issue may lie in the need to evaluate the DC along with the mission as well as the DC non-uniformity along with the pixel array. The authors are invited to introduce these concepts as well as the possible impact on the reduction of the dynamic range which might impact the minimum flux detection that could affect the mission.
“Dark current anomalies” Dark current increase and RTS are expected for this kind of mission. The term dark current increase should be preferred to Dark current anomalies where or not it has an RTS behavior or not.
Main comment: DC origin and impact on the measurement must be explained with the operation mode used by the instrument. A clear description of the DC accumulation over the SNR can highly facilitate understanding. More importantly, the accumulated dark charges are stored in the memory area for almost half a second (0.4s). such retention time makes the leakage current very important in this structure and can potentially lead to a reduction of the total stored charges implying a negative offset on the signal. Authors are invited to explain how important this leakage current is on the output signal and, if applicable, how it can to a certain extend compensate the DC.
Why CMOS imagers have not to be used for such applications? Space qualification history and legacy from other missions need to be highlighted.
A clear distinction between solar events, trapped particles, and cosmic rays should be made here. Cosmic rays are also referred to as heavy ions due to their mass.
SAA, a “region of particular interest” The authors are invited to briefly describe this intertest. Either it is for scientific purposes or a better understanding of the impact on electronics.
The Hubble dark current increase rather lies in the repetitive flyby of the area implying an increase of the deposited dose. Transient effects are photogenerated charges coming from the ionizing radiation during the flyby.
Dielectric materials could be preferred to oxide layers.
Vacancy-interstitial pairs, also called Frenkel pairs recombine at room temperature. It must be specified here that it is still the case at -30.
“Random Telegraph Signals (RTS)-noise” RTS is a signal – as specified in the name - not a noise. This terminology lies in the description of the signal from the mathematical point of view. Therefore, the preferred terminology is DC-RTS or RTS. Some papers used RTN for random telegraph noise when studying transistor RTS (trap-detrap mechanisms in the canal) and DC RTS (generation centers) at the same time.
Level should be preferred to state which rather lies in a defect configuration.
“In the framework of the Aeolus ACCDs development, proton tests (even at higher radiation doses as seen in-orbit) have been performed to evaluate the probability of occurrence of such hot pixels and RTS pixels at an operating temperature of -30 °C showing the presence of one-post irradiation RTS pixel. However, it has to be noted that the operation mode regarding the timing settings during 200 the tests were not fully comparable with the settings used in-orbit.” If neither results nor conclusions can be extracted from this test campaign, it should be withdrawn from the paper. It could eventually be mentioned in the discussion section.
“Transient radiation effects occur due to ionization-induced generation of charges within the CCDs and do not cause lasting damage. “I do not agree with this statement. SEE does result in remaining TID and DDD and can even lead to latchup.
“quite efficiently shielded from ionization damage.” The authors are invited to specified until which TID level. Globally, TID and DDD exposition are not reported in the paper.
“conventional thermal dark current in the CCD” lies in generation dark current in CCD in opposition to CIC. However, the CIC might be introduced as a field-assisted dark current source – trap-assisted tunneling - in comparison to generation dark current in CCD. Remain at the authors’ convenience.
Integration time could be preferred to timing settings.
RTS must be preferred to transient events.
“at measurement level.” This term must be specified> Does it refers to data processing or analog correction?
Lines 391: Transition must be preferred to dark signal spike/peaks.
Could also lie in a small temperature shift during the dark measurement and the collected data.
4minutes is very short to account for RTS. Most of the high RTS amplitudes cannot be seen in such a small period. Additional references on RTS in the CMOS literature might give a global overview of expected RTS behaviors.
RTS can show period lasting from seconds to hours and even days and this is much more important as the temperature decrease. This is the main reason why the correction is not possible.
“not perfectly linear.” The data are linear, but the interesting parameter here could be the total mean dark current over the pixel array. Moreover, DDD and TID ranges must be specified and compared to other results in the literature if possible.
“the hot pixel generation rate does not change with time” Unless for RTS pixels…
“RTS pixels show more than two levels.” Usually the case for DDD-induced RTS.
These on-flight observations are consistent with RTS observed in other imagers CCD/CMOS. References here could support these observations and confirm that these results are not surprising but highlight the need to perform more investigation on RTS in imager for space applications.
RTS behaviors present an infinity of different shapes that cannot be described. This is the reason why other studies use high-resolution imagers to outline major trends on thousands of RTS. See recent RTS studies in CMOS imagers.
CIC does exhibit RTS but with a much smaller RTS amplitude. The different RTS signatures between DDD-induced RTS and CICI-RTS can be made base on the RTs amplitudes. DDD-induced RTS amplitudes are more important. Reported results show large RTS amplitudes that can be linked to DDD RTS. Another method could be to use temperature dependence to extract activation energies. Again, the authors are invited to see the last published RTS studies on CMOS imagers.
Can a high-resolution imager with a smaller pixel pitch be foreseen to maintain sufficient sensitive pixels along the column while disabling RTS ones?