-
FLIR P640 thermal Image distorted vertical lines on screen
Posted by
radioorizont
on 14 Nov, 2021 11:54
-
Hi All,
Could you please advise what could be causing this issue where i see vertical lines on the screen when the hot subject is focused and the background temperature is lower?
The thermal sensor is not the problem as i swapped the thermal unit/pcb assembly with connecting flex cable and same issue occurs on a different sensor.
If i change the angle the vertical lines are not that pronounced, probably as the object is not in sharp focus, see difference between img 15 and img 17
If someone can explain this it would be amazing.
Thanks all.
-
#1 Reply
Posted by
Bill W
on 14 Nov, 2021 12:26
-
The sensors work by having one amplifier / integrator for each column, and it is activated for each row as it is read out.
When you have a hot object in a certain row it results in a lot more energy in the column amplifier and it slightly warms up when reading the hot pixel.
When it comes to do the following row it will be slightly out of calibration compared to how it was on the last 'flag' operation which of course was on a blank scene.
When this is also a colder area the differences show up more as there is less IR energy there. A very low noise mager like this (30mK) is ironically more susceptible as a 50mK imager would hide this in noise.
The image processing behind a lot of 'shutterless' systems will also result in the same effects although for different reasons.
-
#2 Reply
Posted by
Fraser
on 14 Nov, 2021 13:39
-
Bill-W,
Whilst I totally agree with your comments on these images, there is a little more to this story. The OP has managed to rebuild a working thermal camera using two faulty units, an SC640 and a P640. I am in communication with the OP via messaging but it will be far better to use this thread to help him.
My issue with the provided images is that these appear atypical of a SC6xx or P6xx camera as the column noise seems excessive in the example images. The SC6xx and P6xx are high performance imaging systems so something is not right here. My PM series cameras are older and less capable, yet do not display this column noise issue. Such is normally dealt with in the image processor sub system. I have asked the OP to detail exactly what parts were combined from the two cameras to create a working unit. We could be seeing either an issue in the video processor sub system or a calibration (NUC) issue. I would have expected dead pixels to appear if the wrong Calibration tables were in use though. It is a bit of a mystery to me at the moment.
If I saw a SC6xx producing these images when new, I would reject it as unacceptable column noise and a non uniform image response.
It is almost as though the cameras normally excellent image processing is not active
As we know, you have to be careful when transplanting PCB’s in these multi PCB cameras. Some do not play well together if mismatched. FLIR also program each PCB’s serial number into both the PCB subsystems and LiCO main controller. I would expect an error to pop up if the serial number mismatch was a problem though. When FLIR replace PCB’s in these cameras they also carry out a fresh calibration run to ensure all PCB’s work well together.
Fraser
-
#3 Reply
Posted by
Fraser
on 14 Nov, 2021 13:53
-
I believe Ultrapurple owns one of these cameras. He may be able to offer comment on the provided images and compare them to those produced by his camera in similar scenarios. His camera would be a good “baseline” standard to compare against.
Fraser
-
-
Just to get everyone updated, all the components are from the same camera, just transplanted into the donor case.
i did try the other microbolometer from the donor camera and it does work fine, but it exhibits the same visual issues with the vertical bar as the original.
See below first 2 pictures 0066 and 0068 taken at different angles of the same object, you can see one shows the lines and other doesn't.
That confirms Bill's theory that there is a column amplifier issue.
What i am hoping is that it's just a calibration issue which i could correct by recalibrating the unit, i should be able to do that from the service menu which i can access.
I also took some other random photos to show functionality, some with a hot soldering iron for a much larger temperature spam.
Cheers.
-
-
Update
I used different calibration files from another backup generously donated by a forum member, unfortunately it did not make a difference, except for showing a different dead pixel map.
I also replaced the main calib.rsc files from system folder, no change, so moved all original files back.
Tested 2 lenses, same issue.
For now i'm a bit stuck with image as it is.
I managed to unlock the 300 to 2000c range which is amazing and works great.
The issue is also exhibited at the highest temperature range see below, that makes me believe it's something to do with how video processing from the sensor is processed by the camera, as Fraser mentioned.
Wonder if i can load a new video driver from the donated backup to replace original file, don't know what file to replace though.
-
#6 Reply
Posted by
Fraser
on 23 Nov, 2021 00:57
-
Hi,
Sorry but this is likely to be a bit beyond remote diagnostics
I cannot be certain of what exactly has gone wrong with the image processing in your camera. It is almost as if the image processing is in a default state and not applying corrections. In the case of a corrupted calibration I would expect more symptoms and loss of the dead pixel map would show many dead pixels on the screen. This is an unusual fault.
Have you tried testing the PCB’s out of the original camera in your revived camera ? I know that I warned of calibration issues and potential warnings coming from the operating system but such tests may produce useful data. You will also discover which PCB has the original ‘no display’ fault cause on it
You will not cause any calibration table damage by mixing the PCB’s. Sadly the main processor PCB may be the cause of the original fault and that is a pity as that PCB does an awful lot in the camera. I do not recall whether the image processor resides on another PCB though. It does in my PM series cameras.
I will help if I can, but I am “flying blind” here.
Fraser
-
-
Hi Fraser,
You read my thoughts
I did swap the boards and found out it's the BAPP board that has a fault, among other things it looks like it manages power to different areas of the board.
The other 2 PCB's were tested and work fine, although the main board does not boot due to me wiping the OS (stupid mistake)
Anyways that's another topic
I'm getting further with diagnosing the issue: today i did a temperature "drift compensation" test in the service menu and while doing that it looks like the video output is clean with no vertical lines. i took a photo to show that (below)
I was amazed to see the camera in it's full glory with no vertical lines while taking the same image which would normally present the lines due to high contrasts.
BTW the overall noise in the photo seems constant during this service stage process regardless of image pointed at due to the camera giving a more "raw" output" i assume.
Getting there i can feel it
-
#8 Reply
Posted by
Fraser
on 23 Nov, 2021 20:12
-
So is the camera happy with the working BAPP board fitted ? If so, and there are no other issues, you can likely leave the replacement BAPP board in the system. If power related, the BAPP board may be one that can be swapped out without a significant effect on the performance.
Can you post pictures of the BAPP PCB ?
Fraser
-
-
Hi Fraser,
The other BAPP board is the faulty one as o discovered after swapping all 3 boards. I Don't see a reason using it.
I m just using all the pcbs and components that came with this camera.
I attached the pic of faulty board for you.
Im still blaming calibration as the picture is good on the unprocessed image
Last night i pointed the camera at the clear sky and the image was uneven, it was clear in the middle but had a warmer circle around, that sort of tells me the cam is not performing temperature correction right. Btw it was freezing cold outside so the lens must have been cooling unevenly or cam had no chance to adapt to temperature change.
-
#10 Reply
Posted by
Fraser
on 25 Nov, 2021 12:46
-
Ah, I understand now. I thought that the BAPP PCB was causing the imaging fault and did not realise that it was the original issue with your camera before transplantation took place.
In general, service modes operate in a default raw mode in order to provide the tech with an unadulterated view of the cameras imaging system. Much like a ‘Safe Mode’ found in Windows. The minimum system elements are loaded to reveal any central system issues. Calibration is also carried out under this minimal system configuration as you do not want image enhancement algorithms effecting the calibration results.
A little ‘voice’ in my head is asking whether your camera has an image enhancement menu buried in some menu options. My PM695 has a menu that offers three strengths of noise reduction that the user may select from, depending upon need. The strongest noise reduction uses multiple frames to filter the fixed noise so is unsuitable for fast moving targets. The Pxxx series are very sophisticated cameras so may offer other image enhancement options and one may be causing your vertical line issues. I will need to study the user manual and come back to you on this.
Fraser
-
-
Hi Fraser
I do have an image enhancement menu that can be off, low or high, that just reduces a bit the pixel noise but that's it. i preffer that off.
The camera doesn't seem to have advanced filters for raw image but i might be able to enhance it with config files from the p660 , I asked Ultrapurple if he can share his system files with me so i can perform some upgrades. For example i don't have thermal fusion or picture in picture in the toolbox menu, not sure if that was only a p660 feature, the user manual mentions it available throughout the range.
Apparently there's a frame averaging menu which i don't find on my model, that might help.
-
#12 Reply
Posted by
Fraser
on 25 Nov, 2021 13:30
-
Has Ultrapurple confirmed that his camera does not exhibit the same line effect on the image ? I would not expect such on this high end camera, but I do not own one to check for myself.
Fraser
-
-
He did say his does not exhibit the same, and i agree it's not normal. Doesn't bother me too much as if i change the camera angle slightly the vertical lines disappear, they only appear on objects with long vertical edges, wouldn't see it when camera pointed on a hot spot for example.
That just points me back to Bill's idea of the column amplifier circuit.
I don't have a service manual or expanded diagram for this so will keep on searching.
-
#14 Reply
Posted by
Fraser
on 25 Nov, 2021 14:17
-
Do you have the facilities to copy your working processor boards firmware onto the original processor board that you erased ? If so, that would potentially eliminate the processor board hardware as a potential cause of the issue.
Fraser
-
-
I connected with UART to the original camera and can get it into download mode over tfttp, then camera recognised as a usb network device, gets an ip, but when injecting the firmware through usb with fttpd32 it always fails after 3-4 sometimes 6mb, "ack timeout..." so couldn't yet figure out why that happened. Tried different packet sizes. Same subnet.
Asked around for the pinout to find out what are the jtag pins of the 12 pin connector, to try a direct jtag connection to the camera, but couldn't find that info yet.
I have the tx rx and ground pins but none of the other 9 pins.
I own a riffbox which can help with jtag
-
-
Managed to flash the os onto the other board , some task that was... and it doesn't boot, looks like that main board has a fault too and not just the BAPP board.
Same i2c error when checking boot up with putty, as it initially had hence using the case as a donor. I covered this camera boot error in another topic.
-
-
Another update
I used a great software called ThermoVision to connect to the camera
It has a menu where you can disable or change values (threshold) for "collum filter"
Once that's disabled, the vertical lines issue disappears, only thing is the image is noisy on the vertical plane.
Picture below is with that filter disabled.
I tried to edit the threshold but i probably don't know how to work that software well. it has a res-tree Tab/section where i can change the parameters but once i press enter, nothing changes.
Maybe someone knows where image.flow.digitalFilter.spatial.column data is read from in the camera file system and if there's a different way or trick to change the parameters.
Thanks.
-
#18 Reply
Posted by
Fraser
on 28 Nov, 2021 22:28
-
Interesting ! So this is looking more and more like a software induced issue rather than hardware as without the filter the column issue would logically be worse rather than better. It is as though the column filter is a ring like edge enhancement that induces the column issue. A column line issue is seen on conventional CCD imaging chips when they are optically overloaded. I regret that I have no in depth knowledge of the video processing used in these cameras…. That is the “special sauce” that FLIR jealously protects as their image reduction systems are very good and have been developed over many years since the original 1997 release of the PM570 camera with its Gen1 Microbolometer FPA.
There remains the possibility that the gain on the column amplifiers is being adjusted by the image processing system but I have no knowledge of how this is implemented and side effects of incorrect gain setting. The fact that both of your microbolometer PCB’s exhibit the same line issue would suggest that this is not an issue in the Microbolometer Read Out IC (ROIC) which is good news.
Bill might have some insight into whether column amplifiers are normally used as part of the image processing system by controlling their gain.
Fraser
-
#19 Reply
Posted by
Fraser
on 28 Nov, 2021 23:13
-
Some night time reading for you
This Patent discussed Row and Column noise reduction and how it is applied to the thermal camera system. The paper shows a Lepton core but it is likely your camera uses a similar technique.
https://patentimages.storage.googleapis.com/a6/40/79/d04523c7fab184/US9235876.pdfI am aware that the ROIC amplifier gain may be adjusted for a flat field but I have not studied gain adjustment for the purposes of noise reduction. I shall be studying this Patent to gain knowledge of Row and Column reduction. Patents can be a very useful source of information on thermal camera operation and design
Fraser
-
#20 Reply
Posted by
Fraser
on 28 Nov, 2021 23:34
-
A quick look at the Patent makes me think I was on the right track with your issue being an imaging processing problem. It does not appear that the Column amplifier gains are involved in the image noise processing stages. The “RAW” image information coming out of the Microbolometer is ‘warts and all’ but the various biases applied to the microbolometer are set to provide the optimum performance in the desired role or scenario. Different temperature ranges may be configured with appropriate biases applied. The “RAW” image undergoes ‘treatment’ by the video processing stages but the Radiometric measurement data is preserved. The Row and Column filtering is intended to greatly improve the image by effectively removing the horizontal and vertical line ‘noise’ as detailed in the a patent I have referenced. This noise reduction is ‘intelligent’ in that that it ‘studies’ the image data resented to it in the form of pixel values and makes appropriate adjustments based on adjacent pixel values. It is effectively ‘hunting’ for classic vertical and horizontal lines associated with rows an columns rather than straight edges in the thermal scene. I will need to study the Patent and think about this more. The old brain is tired and needs a rest.
-
#21 Reply
Posted by
Bill W
on 29 Nov, 2021 22:21
-
The Row and Column filtering is intended to greatly improve the image by effectively removing the horizontal and vertical line ‘noise’ as detailed in the a patent I have referenced. This noise reduction is ‘intelligent’ in that that it ‘studies’ the image data resented to it in the form of pixel values and makes appropriate adjustments based on adjacent pixel values. It is effectively ‘hunting’ for classic vertical and horizontal lines associated with rows an columns rather than straight edges in the thermal scene. I will need to study the Patent and think about this more. The old brain is tired and needs a rest.
So if what appears to the image analysis as a 'hot column' is a part of the real image, the software will try to 'fix' the whole column and cause problems to real bland areas where the 'hot column' is not ?
That sounds very close to the OP observations.
May be something of more use in shutterless applications as it is column patterns of divergence that tend to show up first, as the amplifiers slowly drift away from the last calibration.
-
-
Did a bit more work, In the calibration file i changed image settings of the digital filter from column threshold 60 to 15, left the thresholdPix at 120 default.
I did this for the -40 to 150 range (3 lines of code needing data changed)
The images are now much better with no vertical lines.
The only small thing is at high sensitivity with low temperature spans you can see some vertical noise
If i go up on threshold the vertical noise will be hidden but then the lines artefact gets introduced, it's a compromise.
I am happy with how the camera works now, i might look at other tweaks to further improve img quality.
Photos attached with outcome and with a noisy image when pointing at wall with slight vertical shape on left side (no temperature difference so gain at highest)
Very happy with how it looks considering the mod, left the 100 to 500 unchanged, might mod that too later, easy to revert back any time.
-
#23 Reply
Posted by
Fraser
on 29 Nov, 2021 23:29
-
It would be worth asking Ultrapurple what his cameras column filter settings are set to. It might be slightly different to your camera but will give you an idea of the ‘nominal’ values.
When buying a used thermal camera of this sophistication, it can be the case that it has been fine tuned in the configuration for a particular application and may not then produce good imagery in another application until a more general configuration is applied.
At the end of the day, as long as you are happy with the cameras performance, that is all that matters
Fraser