-
Notifications
You must be signed in to change notification settings - Fork 3
Bench tests
This is a page describes some bench tests performed to assist us in determining the S4ACS capabiblities when using the SPARC4 CCD cameras.
This section describes the experiments made to test the performance of the acquisition system during the acquisition of images with high cadence. For that, the acquisition computer 2 together with the SPARC4 camera related to the r instrument channel were used. The camera was configured with a full-frame resolution, no binning, and a readout rate of 30 MHz, for different combinations of exposure times and number of frames. For exposure times greater than 40 ms, the frame-transfer option was used. For each combination, the time interval between the end of acquisition and the end of image writing to disk was calculated using the respective timestamps of these processes recorded in the log file. The results are presented in the table below.
| Texp (s) | #frames | Total Exp. Time (s) | Time to save the data (s) |
|---|---|---|---|
| 1e-5 | 500 | 5e-3 | 67 |
| 1e-5 | 1000 | 1e-2 | 134 |
| 1e-5 | 1500 | 1.5e-2 | 200 |
| 0.05 | 500 | 25 | 68 |
| 0.05 | 1000 | 50 | 133 |
| 0.05 | 1500 | 75 | 198 |
| 0.1 | 500 | 50 | 55 |
| 0.1 | 1000 | 100 | 101 |
| 0.1 | 1500 | 150 | 153 |
| 0.2 | 500 | 100 | 0.197 |
| 0.2 | 1000 | 200 | 1.36 |
| 0.2 | 1500 | 30 | 2.16 |
| 0.25 | 500 | 125 | 0.160 |
| 0.25 | 1000 | 250 | 0.168 |
| 0.25 | 1500 | 375 | 0.183 |
Based on these data, it can be observed that the time required for S4ACS to save all images to disk was approximately the same for the cases where texp=1e-5 s and 0.05 s. For the experiment with texp=0.2 s, this time reduced to a few seconds, while for the case where texp=0.25 s, this value was approximately constant for all values of the number of frames used. This result is related to the use of the camera frame-transfer mode and indicates that the average time required to write an image is 170 ms, added to some internal application delays.
To test the system performance even further, some experiments were done based on the previously presented results. These experiments were designed to provide a perspective of the system capability in acquiring images with a total exposure time of 30 minutes. The table below shows some experiments performed to meet this demand. The columns display an experiment identifier, the exposure time and number of frames used, the total exposure time of the image series, and the time required to save all images. For some of the experiments, the quantity of RAM memory required to store the images to be saved was larger than the memory made available by the computer. For those cases, the number of frames was reduced such that the maximum memory allocated by the application was 90%.
| Exp ID | Texp (s) | #frames | Total Exp. Time (s) | Time to save the data (s) |
|---|---|---|---|---|
| 1 | 0.1 | 6000 | 600 | 626 |
| 2 | 0.1 | 18000 | 1800 | --- |
| 3 | 0.15 | 9000 | 1350 | 475 |
| 4 | 0.2 | 9000 | 1800 | 8.49 |
For the experiment 1, the total exposure time is comparable to the time needed to save the images in the computer disk. For experiment 2, the amount of data allocated in the RAM memory was larger than 90% and, therefore, the experiment needed to be aborted. Experiment 3 shows that the system is capable of providing a total exposure time of 20 minutes, for a texp~0.15 s. Experiment 4 shows that the acquisition of high cadence image series with an exposure time of 0.2 can be done without extrapolating the acquisition system limitations.
In this section, some of the experiments previously presented were repeated using the software provided by the manufacturer of the SPARC4 cameras, named AndorSolis. Our intent was to provide a comparison of the performance provided by this software and S4ACS. This test was done using the acquisition computer 2 and the channel r of SPARC4. The camera was configured with a full-frame resolution, no bining, EM mode, and a readout rate of 30 MHz. The image acquisition was done using the kinetic mode and frame transfer options, for different number of frames and exposure times. In addition, the time to save the acquired data was calculated based on the "Created in" and "Last modification" timestamps found in the metadata of the generated FITS files. Table bellow presents the results obtained is this experiment.
| Exp ID | Texp (s) | #frames | Total Exp. Time (s) | Time to save the data (s) |
|---|---|---|---|---|
| 1 | 0.04 | 6000 | 240 | 240 |
| 2 | 0.1 | 6000 | 600 | 244 |
| 3 | 0.15 | 9000 | 1350 | 375 |
As one can see, AndorSolis needs less time to save the acquired images to the computer disk, when compated to the equivalent experiments done for S4ACS. Such a behaviour can be explained by the fact that AndorSolis groups the images in cubes of about ~500 imagens in a unique FITS file, whilst S4ACS saves each image in a separated file. On the other hand, based on experiments 1 and 2, this time is indenpendet of the chosen exposure time, which makes S4ACS faster than AndorSolis for acquisition of images series with exposure times larger than 0.2 s. This result is explained by the fact that, in S4ACS, the process of saving images occurs in parallel with the acquisition, whilst Andorsolis waits for the acquisition to finish.
In this section, we present some bench tests performed using the SPARC4 channel 2 to determine the largest number of exposures in an image series that can be acquired by the acquisition system, without extrapolating a percentage of 80 % of the RAM memory capacity allocated for the images pending to be saved. For that, the camera was configured with a full-frame spatial resolution and a readout-rate of 30 MHz, for different combinations of exposure times and total number of exposures. As the maximum number of exposures allowed in a single kinetic series for the used operation mode is 1500, the acquistion of larger image series were done adjusting the number of cycles. Table below presents the used operation modes that meet the allocated memory criteria, together with the total exposure time and the time needed by the system to write the data to the computer disk.
| Exp ID | Texp (s) | #frames | Total Exp. Time (s) | Time to save the data (s) |
|---|---|---|---|---|
| 1 | 1e-5 | 4500 | 0.045 | 611.18 |
| 2 | 0.05 | 5000 | 250 | 677.67 |
| 3 | 0.1 | 6000 | 600 | 600.04 |
| 4 | 0.15 | 9000 | 1350 | 452.91 |
These same tests were repeated using an exposure time of 1e-5 s and the spatial resolutions of 1024 x 1024 with binning of 2 pixels and 512 x 512 pixels with binning of 1 pixel. For both cases, the quantity of images allocated in the RAM memory did not exceeded de value of 50% of the RAM capacity, independently of the image series size.
In this section, we present the tests carried out to determine the maximum number of frames that S4ACS is capable to acquired in a single kinetic series, as a function of the images spatial resolution. For that, the camera was configured with different values of sub-image and binning. Then, the number of frames was incrementally increased until an error message was raised by SDK. The number of frames used for the iteration imediatelly before was considered the maximum number of frames for that mode.
| Sub-Image | Binning | #frames |
|---|---|---|
| 1024x1024 | 1x1 | 1500 |
| 1024x1024 | 2x2 | 5500 |
| 512x512 | 1x1 | 5500 |
| 512x512 | 2x2 | 22500 |
| 256x256 | 1x1 | 22500 |
| 256x256 | 2x2 | 91500 |
| Sub-Image | Binning | #frames |
|---|---|---|
| 1024x1024 | 1x1 | 29.000 |
| 1024x1024 | 2x2 | 110.000 |
| 512x512 | 1x1 | 110.000 |
| 512x512 | 2x2 | 450.000 |
| 256x256 | 1x1 | 450.000 |
| 256x256 | 2x2 | 1.800.000 |
For some reason, the experiments with a resolution 512x512 and binning=2 or below needed a large time interval to finish the execution of the code. Besides, the RAM computer memory achieves 99% even toughout I garantee that the exposure time is larger than the time to write the data to the disk.