RIDLNick Cox.
Fixed Pattern Noise Progress Report
Author: / Nick Cox / Date: / January 27, 2009Reviewer: / Date:
Printed on Wednesday, January 28, 2009LeachNoise.docprocedural nickname
RIDLNick Cox.
1.0Introduction...... 2
2.0Background...... 2
3.0Experimentation...... 3
4.0Mobile Testing...... 4
5.0Conclusion...... 5
6.0Tables...... 6
Fixed Pattern Noise Progress Report
1.0Introduction
This report outlines the severe image quality degradation in the form of fixed pattern noise that has affected the imaging systems used in the lab. It will cover the timeline over which the noise was introduced, characterization of the noise, measures taken to reduce the noise and identify the source, and further speculations on the root cause of the noise.
2.0Background
Dark images taken over the course of the past year have been processed using a subtraction technique as follows:
(ramp1_read2 - ramp1_ read 1) - (ramp0_ read 2 - ramp0_read 1).
Figure 1 and Figure 2 are two examples of this processing, and the differences are vividly apparent. By processing images in this manner from a range of dates, it was possible to pinpoint the introduction of the noise during the month of May, 2008.
Using the pixel readout times (/pxl horizontally and /pxl vertically), it was possible to estimate the frequency of the noise pattern to be approximately 20khz.
Figure 1.Dark image taken April 02, 2008 during dark current experiment does not show fixed pattern noise.Figure 2. Dark image taken November 21, 2008 during dark current experiment shows fixed pattern noise.
3.0Experimentation
The antenna picking up the noise has been traced to a loop of wires in the back of the
Leach electronics set. It has been shown that covering this loop with a shield such as aluminum foil eliminates the noise in the system,, however, it is desirable to locate the source of the interference and have a more permanent and steadfast solution.
A longer antenna was constructed to replace the loop of wires, and it was used to move the antenna away from the electronics set. This led us to believe that since the noise still existed regardless of distance from the Leach, that the Leach itself is not the source. We therefore resorted to other methods of diagnosing this ambient noise.
Placing the antenna across the threshold of the two lab rooms, crossing into A220 (the clean-room lab) had a significant effect on the noise pattern. It appears to increase the frequency of the fixed pattern as shown in Figure 3. However, once the lights were turned off, the higher frequency pattern disappeared. We were not able to show a similar correlation in the lab with the optics benches – as turning the lights off had no effect on the noise. Additionally, an antenna was constructed using an oscilloscope and we were able to find frequencies in the 20 khz range emitting from the fluorescent lights, with a direct relationship between distance from the lights and intensity of the noise.
Additionally, a cardboard box completely covered with aluminum foil, was used to shield the Leach from external sources of noise. With the Leach electronics and power supply on top of the optics table, the box was placed over each case one at a time, as well as both simultaneously while data was recorded. It was found that while the Leach electronics alone and both the electronics and power supply were shielded that no fixed pattern noise was visible. However, the fixed pattern was still visible when just the power supply was shielded. This reassuringly confirms that the Leach is not the source of the noise and that the pickup antenna is located in the Leach electronics alone, and not any circuitry in the power supply.
Figure 3. Noise image taken January 15, 2009 with antenna placed inside CIS A220 with overhead lights on.4.0Mobile Testing
In an attempt to localize the source of the noise, and eliminate equipment in our lab from
bearing sole responsibility, testing was done at several different locations around campus. It was found that the fixed pattern noise is in fact present in other buildings on campus, even those without substantial electronic or scientific equipment and that the strength of the noise decreases and ultimately disappears in proportion with distance from the Center for Imaging Science (#076). For instance, the noise pattern was very distinctly measured outside room #1200 of the JamesE.GleasonBuilding (#009), with an average standard deviation of 2.80. It is interesting to note that moving roughly 20 feet towards the doors leading to the Infinity Loop increased the visibility of the fixed pattern and the standard deviation.
Data was collected throughout various locations of the Center for Imaging Science (#076) that displayed varying levels of the fixed pattern noise. It is ever-present in the RIDL lab, while varying in strength on the 1st and 2nd levels and weakest in the basement by the east staircase and on the 3rd floor.
A relative strong noise level, with average standard deviation around 8.5 (compared to the ambient lab’s 15 – 20), was found in the College of Liberal Arts (#006) near room #1315. An equally visible pattern, with average standard deviation around 4.1 was found outside room #1117 in the Center of Bioscience Education and Technology (#075).
Buildings outside of this cluster such as the SlaughterCenter for Integrated MFG Studies (#078), the IT Collaboratory (#017), the George Eastman (#001), and the Student Alumni Union (#004) did not show any fixed pattern noise and therefore beg the question that the source is centralized in the near vicinity of the Center for Imaging Science (#076). The only outlier that may refute this claim is the JamesE.BoothFineArtsBuilding (#007A), which is in a direct line between #076 and #006, and roughly a part of the cluster, but did not show any fixed pattern noise during testing.
Figure 4: Campus Map indicating locations where fixed noise pattern was present (red dots) or not present (blue dots)5.0Conclusion
20 khz noise can be emitted from a variety of sources. It is part of the Very Low
Frequency band of the radio spectrum which ranges from 3 khz to 30 khz. The VLF band is used mainly in naval operations for communication with submerged submarines. The nearest of these high power stations is in Cutler, Maine, where VLF is transmitted with a full megawatt of power. Possible natural sources can be attributed to interference caused by solar flares, or high-energy gamma ray bursts from distant sources in the galaxy.
It is not clear at this point how to continue. We have eliminated all potential sources that we control in the lab to be the cause of the noise. There is a chance that faulty wiring within the walls and/or ceiling can be the source, and RIT Facilities Maintenance Services has been contacted about any changes in the building that may have occurred during the time frame in interest. Additionally, there is a sub-basement space below the lab which we have yet to gain access to that might yield some answers.
6.0Tables
Property / Value / EvidenceIntroduction of noise / May/ June 2008 / Analysis of archived data showed introduction of noise in system beginning around May or June 2008
Fixed pattern frequency / 20 khz / Comparing the spatial locations of the max and min of the repeating pattern with the pixel readout time
Noise is not attributed to: / Lights / Turning off all lights in lab does not remove the noise problem
Increased frequency pattern: / Lights on in A220 / Turning on lights and placing loop in A220 doubles frequency pattern
Noise is not attributed to: / Lab / Readings taken all around building 76 had noise pattern
Noise is not attributed to: / Leach / The ASIC electronics display the same fixed pattern noise
Noise is attributed to: / Loop / Shielding loop with aluminum foil removes pattern. Similar loop in ASIC???????
Noise is attributed to: / CIS or near vicinity / Testing at different locations around campus revealed strength of noise is centralized to building 76 and the near vicinity
Printed on Wednesday, January 28, 20091RIDL document template.docprocedural nickname