Engineering Qualification Model Robotic Arm Camera Lamp Performance

Test Conditions and Image Processing:

Test Report R.Tanner 4/21/97


Test Procedure:

The Engineering Qualification Model lamps were mounted in the Prototype model Robotic Arm Camera and the lamp pattern tested with the spectralon target. The images were taken in the black calibration box with a black cloth over the open panel. The light level in the box was very low, as lamp off images were almost identical to dark frames. Images were taken with the IMP BCE hardware ( PM boards ) and software.

Stray Light

The stray light test was done with the CCD window sealed with black tape. The lens was set to far focus position which is the most sensitive position for the light leakage from the lower lamps. Then images were taken with the lamps on and off. This quantified the light leakage from the back of the lower lamps into the interior of the camera. Several of the seams on the front of the camera had to be sealed to eliminate light leakage from the ambient light.

Flat Field

Each image was shutter corrected and a light off image at the same exposure was taken immediately after the light on image. Subtracting this image will eliminate the dark current and the ambient light in the box. The close up images were taken with the light from the lower lamps blocked.

Test distance 12.5 mm 29.7 mm 95 mm 95 mm 200 mm 300 mm
Lamps on Upper only Upper only Upper only Both Both Both
Approximate object to image ratio 1:1 3:1 8.5:1 8.5:1 17:1 25:1

Non-Ideal Test Parameters:

The range of focus for the PM RAC is limited to 12.5 mm to 95 mm from the face of the camera. This gives a slightly smaller FOV for the 200 and 300 mm distances ( 23.8 x 45.7 degrees vs. 25.5 x 48.7 degrees ).

The PM RAC currently has a KG3 filter glass for the CCD window which doesn't have the sharp cutoff in the red as the flight windows will. This will reduce the red and blue response after the sharp cutoff window is installed. The blue response will be affected because of the red leak from the blue filter will be reduced, as well as losing some of the blue light.

The spectralon is on the order of 99% reflectivity which will change the response for the close up flats where there is multiple reflections from the front of the camera and the spectralon. Actual usage will not have much reflected from the sample, and have some reflections from the scoop and light shield, especially the bottom which should be somewhat polished. Additionally, the samples at he closest focus will get some stray light from the lower lamps coming over the lip of the scoop, which can only be duplicated with the scoop in place. Additionally the new scoop is deeper, which means a 180 mm distance with the light baffling of the scoop should be taken to represent the back of the scoop.

Image Processing:

Stray light:

  1. Right and Left eye images from the IMP BCE software were combined into one array.
  2. The lamp off image was scaled by the ratio of the dark strips and subtracted from the lamp on image to eliminate the dark current.

The CCD was cooling down during this test, probably from running the focus motor. The scaled dark frame was slight above the light frames, probably due to the thermal transient. The result is the internal stray light numbers came out slightly negative. This indicates that the stray light shield is working well, as the unshielded numbers are about 800 counts per second. The processing was done with the IDL program:

/home/lpl/rtanner/IDL/racflat/racstraylite.pro

Flat fields:

  1. Right and Left eye images from the IMP BCE software were combined into one array.
  2. The lamp off image was subtracted from the lamp on image to eliminate the dark current and the ambient light contribution.
  3. A rough bad pixel correction was done to eliminate the vignetted pixels at the edges of the field and the bad column. The edge pixels were replaced with the adjacent row or column, and the bad column ( no 85 ) was replaced with the average of the right and left neighbor pixels.
  4. The image was also flipped to correct orientation.
  5. The distant frames were filtered with a 7 x 7 boxcar filter to smooth the contour lines. Without the filtering, the contour lines have so much noise that they are very wide and not able to be labeled. The close up 12.5 and 29.7 mm distances were filtered with a larger 11 x 11 window to eliminate the surface texture of the spectralon which is resolved at those close magnifications.
  6. The flats were normalized by dividing the image by the average of the central 10 x 10 pixels. For the close up flats this is not the maximum of the image, however the useful part of the frame for these distances is from the middle down, and they would be typically subframed to the lower part of the image. Also the spectralon is vertical, while the actual specimens would typically be at some slope, with the upper parts farther away from the light source.
  7. The counts per sec for the central 10 x 10 was also calculated.
  8. A contour plot of the normalized flat is plotted.
  9. The normalized flat is multiplied by 3000 to get a nominal exposure.
  10. The signal to noise ratio is calculated and a contour plot done.
  11. The average of the signal to noise ratio for the frame is calculated.
  12. The ratios of the normalized flats ( red/green, red/blue, green/blue ) are calculated to quantify the matching of the patterns for each color. Contour plots are made of these frames, and the average ratio calculated.

The IDL program which was used was:

/home/lpl/rtanner/IDL/racflat/raccal.pro

Summary of Test Results:

Stray Light

Target: Less than 10 counts per second, approximately .5% of the typical responsivity at 300 mm ( 2000 ).

Results: Probably less than 2 counts per second, actual data was -1.5 to -2.8 counts per second..

Uniformity of illumination:

Target: Greater than 20% of central illumination in corners. Greater than 60% average over frame. For close distances, this is for the bottom 132 pixels of the frame.

Results:

Average / Corner illumination in percent of center 10 x 10:

color \ distance 12.5 mm, upper 29.7 mm, upper 95 mm, upper 95 mm, both 200 mm, both 300 mm, both
Red 84 / 58% 66 / 31% 71 / 23% 75 / 30% 75 / 29% 75 / 35%
Green 84 / 55% 67 / 30% 70 / 23% 73 / 25% 73 / 25% 73 / 30%
Blue 84 / 54% 67 / 29% 70 / 22% 74 / 25% 73 / 26% 72 / 32%

Signal to Noise Ratio:

Target: The signal to noise ratio should be greater than 200 average and 100 in corners.

The numbers for the two close up distances were calculated for the bottom 132 pixels of the frames.

Average / Corner signal / noise ratio for 3000 counts in the central 10 x 10:

color \ distance 12.5 mm, upper 29.7 mm, upper 95 mm, upper 95 mm, both 200 mm, both 300 mm, both
Red 274 / 227 241 / 166 250 / 144 257 / 163 259 / 160 258 / 177
Green 274 / 222 244 / 164 248 / 142 254 / 151 255 / 149 254 / 163
Blue 275 / 219 243 / 163 249 / 141 255 / 149 254 / 152 252 / 168




Responsivity:

Target: Responsivity should give minimum of 100 signal to noise ratio with ambient light from skylight with shroud in scoop. The number required will have to determined with roof top testing and ratio of earth to mars sky light. Or possibly testing in DISR 20" sphere with diffuse source and scoop and arm simulator.

Results:

Responsivity in counts/sec for the central 10 x 10 pixels:

color \ distance 12.5 mm, upper 29.7 mm, upper 95 mm, upper 95 mm, both 200 mm, both 300 mm, both
Red 23,264 142,572 25,389 55,603 13,900 6,185
Green 7,764 41,690 7,516 16,537 4,134 1,837
Blue 9,164 46,206 8,331 27,492 6,873 3,056

Color Balance Ratio:

Average Ratio / Worst Ratio of two colors over field:

color \ distance 12.5 mm, upper 29.7 mm, upper 95 mm, upper 95 mm, both 200 mm, both 300 mm, both
Red / Green 1.001 / 1.08 0.978 / 0.88 1.013 / 1.08 1.030 / 1.19 1.035 / 1.20 1.041 / 1.18
Red / Blue 0.996 / 1.10 0.987 / 0.90 1.008 / 1.06 1.027 / 1.22 1.044 / 1.22 1.053 / 1.25
Green / Blue 0.995 / 1.05 1.009 / 1.07 0.994 / 0.95 0.997 / 1.17 1.009 / 1.18 1.013 / 1/17

Data:

Contour plots of the flat fields for the 6 conditions , the contour plots of the signal to noise ratio, and the ratio between the colors are attached.



12.5 mm distance with upper lights only, approximately 1:1 imaging. The slanted island in the upper middle of the image is a small gouge in the spectralon.


This is the flats at 29.7 mm which is 3:1 imaging. The flat shows a high intensity near the top of the image. This typically won't be a problem as there will not be something up there in the scoop.


This is upper lamps only at 95 mm which is about 8.2:1 imaging. This is as far out as the PM RAC will focus.



This is both lamps on at 95 mm. The lower lamp is more directional and causes greater excursions in the ratio plots.



This is both lamps at 200 mm which is about 17:1 imaging. The camera is still focused at 95 mm.





This is both lamps at 300 mm which is about 25:1 imaging. The camera is still focused at 95.