[SOLVED] Mystery "Raytheon" Camera

[Abbott242]

Hey all! I just bit the bullet and bought an allegedly raytheon 320x240 core and control board off eBay. Hopefully it arrives soon! In the meantime, I'm trying to find more documentation on the model to help me in my testing/potential repair. I've dug through this forum and elsewhere, but just can't find anything. If anybody has ideas on what it is or where to start, that would be greatly appreciated. Thanks!

[Fraser]

Not Raytheon....... Made by ISG using an ITC microbolometer for their Elite range of Fire Fighting cameras. I will link to forum posts on this core in a minute.

Fraser

[Fraser]

Here you go...

https://www.eevblog.com/forum/thermal-imaging/mystery-imaging-core-fraser-has-been-buying-toys-again-)/msg2799504/#msg2799504

Fraser

[Abbott242]

Oh fantastic! That is great to hear, far more information that I had hoped for - you may have just saved my life.  I'll dig for more information on the model, do you have any advice for getting the thing running? I'm planning to hook it up to a Adafruit mini NTSC TFT display (https://www.adafruit.com/product/911) and possibly add custom overlay graphics. Any pinout / power supply information would be greatly appreciated, I don't want to fry this thing by accident. But that really puts my mind at ease, knowing a little more about it. Now I'm really excited!

[Fraser]

PM sent 

[Abbott242]

Quick update / call for help - it finally arrived! I've powered it up and connected it to a mini analog display, but unfortunately was met with a vague blob instead of an image. The image overlay seems okay, as does the boot-up sequence - but the thermal image itself is terrible. It does respond to heat changes in it's fov, by shifting the brightest portion of the blob around, but there's no objects visible. Not sure what would cause this, I've played around with the focus of the lens but to no avail. I've attached some photos of damage to the shutter and possibly the detector itself?  If anybody has a clue as to what might do this, it would greatly help. Note that the only four wires I've connected are 12V power and video out/ground. Meanwhile, I'll keep fiddling. At least it turns on!

[Fraser]

I am not sure what to make if this core and its current condition. The fault looks unusual and the damage to the FFC shutter and microbolometer window looks like deliberate vandalism. I cannot think of a common situation where such damage woukd occur naturally, except maybe if the camera lens were hit by a projectile that drove shattered germanium into the FFC shutter. (The lens you have fitted may not be the original)

I am sorry to say that the combination of physical and electronic damage make this core a very sick patient that may not be economically repairable  Did the seller detail the faults with this core ?

Fraser

[Fraser]

Looking at your cores produced image I am leaning towards microbolometer damage. There is a dead pixel column and the vacuum may have been lost if the microbolometer was struck on its face and a gas seal cracked. I have been told that a microbolometer that loses its vacuum mat still ‘function’ but is almost blind 
the damage around the lens mount is also not typical of what is normally found on these cores. This core has the appearance of a scrap unit that has had some very unpleasant ‘adventures’ before coming to you. Scrap from a disposals skip maybe ?

Fraser

[Abbott242]

I think you're right about the lens not being original - I finally managed to get some images out of the core, by unscrewing the lens entirely and holding it in from of the sensor by hand. It focuses about half an inch away, well beyond where the intended threads stop. However, it's as if I'm looking at a clear image through a mesh screen - there is a stationary, grainy texture covering everything. The edges of the image are darker too. Moreover, the shutter has yet to move except on startup, and after prolonged use the center of the image becomes "whited-out". I'm sure much of the image quality issues are due to the my janky connection to a TV, upscaling, and the fact that I'm holding (not very steadily) the lens. The focus issue can be fixed with an adapter, but the grainy images  and progressive white-out are a little much to stomach.

The seller mentioned that he could not test the core, and that it might have damage. He did not specify any more than that, and failed to mention the obvious damage the she shutter, etc. I guess that's what I get for buying a cheap, used camera though. It looks like it has had a rough history for sure.

I do hope that it hasn't lost vacuum, that would be very bad. I'm unclear if the dead pixels are an artifact or not however. An awful self-portrait is attached.

[Fraser]

That is an interesting development. At least you know that the lens needs to be adapted to focus correctly and you have got an image out of the camera, that is positive news  The correct FFC flag operation is very important. If it does not operate when the system wants to do an FFC calibration you will get some very strange images as the camera will capture the viewed scene instead of a blank flat field flag and the scene becomes the new flat field correction ! You say the FFC flag operates at start, does it operate once or several times and then stop to produce the thermal image ? An FFC flag event should occur approximately every 120 seconds during normal operation.

You could carry out a very simple test.... leave the lens off the core. Lay a piece of plain cardboard across the lens mount and let the camera start and settle for a minute or two. What does the image look like ? It should be a pretty low noise flat grey. Any defective pixels will be visible if the system does not have them in its dead pixel map. This simple test will give you an idea of the condition of the microbolometer and the image correction accuracy. The card acts like a flat field for the microbolometer so it does not matter if the FFC flag is not operating properly. The card is has virtually zero Delta T across its surface so the core produces a nice flat grey image

Fraser

[Abbott242]

On startup the shutter clearly goes back and forth maybe 5 times before the image comes on - it just never moves afterwards. After further fiddling with the lens, I've actually gotten a usable picture! I'll try your experiment and send a photo when I'm done. Do you know of any way to force a FFC event? Clearly the motor works, maybe it's some sort of software error?

[Fraser]

Was the cup you are holding hot or ice cold ? If hot, it is showing as cold so we have inversion taking place.

There are a lot of dead pixels showing on that image. That could be caused by a microbolometer PCB being attached to a core PCB set with which it was not manufactured and calibrated. The dead pixel map data would be invalid and so the dead pixels become visible. The other alternative is shock damage to the microbolometer that has caused pixels to become defective since the core was calibrated.

I am pleased to see that you have an image showing though. The FFC event is automatic on this core and I do not think you can initiate it yourself. I have not got comprehensive data on this ISG core though.

[Abbott242]

Here's a photo of the screen after 3min covered with paper, no lens. Note that the moire in the images hides some of the dead pixels, there are actually more. And immediately afterwards, an image of myself. There do seem to be a few dead pixels unfortunately, and the blackness around the edges isn't the lens. I suppose there's no way for me to handle the dead pixels, but that's not the end of the world. I would need to get the FFC working somehow, unless I just reboot the core every 3 min.    The cup was very cold, frozen soup.

[Fraser]

Well, we now know that the microbolometer still has a decent vacuum in it 

The ‘vignetting’ effect would normally be the result of under illumination by the lens where its output image circle is smaller than the microbolomter imaging area. You say that this is not the case though and your pictures support this. I am at a loss to explain this effect at the moment. Try moving something very hot like a cup of freshly boiled water into those dark corners of the image .... do you see anything ?

I am thinking you have a mismatched microbolomter PCB and core PCB set. This commonly happens when people are trying to make one good core out of a collection of faulty cores. Mixing PCB’s on thermal imaging cores has to be done with care and an understanding of any pairing that must be maintained to keep the calibration data valid.

Fraser

[ArsenioDev]

Yeah, this definitely looks like a bad matchup of gain tables and dead pixels.
I've got the follow on generation of these using L3 QVGA FPAs, they work damn good.
Rehousing three of em for friends to install on their vehicles to compliment NODs and NIR emitters for tactical driving.

[Abbott242]

Well, I've tried putting a cup of hot water in the corner, and noticed two things. Firstly, the image does seem the automaticly shift the environment to colder colours to compensate for the hot object. Secondly, the hot cup shone through the 'vignette'.  A mixmached core/pcb could explain the pixels, I suppose, but not the lack of automatic calibration.

After leaving the camera on for ~10min, the last ~5 of which it was covered with paper, the entire field of view is simply white (minus the black edges of course  ). The 'vignette' substantially worsened as the center whited out as well. Therefore, I don't think the camera is doing an FFC and not closing the shutter - it's just not doing FFC's at all. Also, the 'vignette' is clearly not mechanical, as it can be removed by hot objects and it's coverage changes with time. Also worthy of note: while the screen is whited out, there are 4 black pixels that cannot be removed with the hot water, while most others can. I assume that these are the only truly "dead" pixels, and that much of the rest is artifacting.

Would there be any way to somehow update or modify the gain tables?

[Fraser]

Yes the FFC process appears to be defective in some way. Exactly how, I cannot say but if it is based on pixel outputs then the mixing of PCB's in the core may have really messed things up on the image management and processing front. The 'vignetting' effect is likely the result of incorrect gains in the NUC table. This is why I needed you to try the hot cup test in the corners. It proved that the dark corners were still responsive to stimulation

How to fix this issue ? Sadly when a core is made up of unmatched PCB's you are really facing a challenge. This is especially so when you do not have the service software for the core. You would have to reverse engineer the firmware in order to find the calibration tables etc and that is no simple task. The fact that the core operates the FFC flag at power on proves that the mechanics of the system are functional. You face a firmware issue that is likely responding to incorrect or 'out-of-range' input data from the microbolometer PCB. There is no simple fix for this situation. A microbolometer based core that is designed to operate with an FFC flag will not work correctly in the absence of those FFC events as the microbolometer warms up after initial power on and the FFC events are used to compensate for the natural 'warm up period' pixel drift as well as maintain a flat field.

Sorry that I do not have better news. On a positive front, you do have a decent lens and that is not a cheap item. You can also use the core to learn about the technology. You may wish to consider searching for another thermal imaging core or an MSA 5000 series camera though. There have been some on ebay.com recently at good prices.

Fraser

[Abbott242]

Well, that is unfortunate news for sure. I definitely don't regret the purchase, it's already been a ton of fun and a great learning tool. The way I see it, I have 3 options here.

1. Operate the camera in it's current state, and simply reboot whenever the image gets too bad (every ~5min). Maybe improve cooling on the core to lengthen useable time?

2. Attempt to reverse engineer the firmware and modify the tables - but I don't even know where to begin with this. I agree, this would be an enormous challenge.

3. Forgo the existing PCB and attempt to access the microbolometer directly. I mean, it has output pins - surely so could write code to operate the microbolometer? This seems easier than modifying the firmware.

Thoughts? While the lens is very nice, it would be a shame to let this go to waste. Some of the images I've been getting (when the lens is roughly aligned) are supprisingly useable, before the FFC issues gets really bad.

[MikeNC407]

I responded to your reply on my other thread already but I had one more thought; are you able to force an FFC event by tricking the automatic gain control into entering low-sensitivity mode?  All of my cameras will actuate the shutter once while switching from low to high gain and vice-versa.  Perhaps holding a hot soldering iron very near the lens so it fills the display could do it?

[Bill W]

3. Forgo the existing PCB and attempt to access the microbolometer directly. I mean, it has output pins - surely so could write code to operate the microbolometer? This seems easier than modifying the firmware.

You would essentially have to recreate the whole of the camera hardware & firmware.  What comes out the sensor is not much of an image, although VOX are better than ASi.  However the PCB set that you have has all the necessary circuitry, all joined up and is working.  So (in theory - it is still a huge task) you could write new code for the processor supplied rather than try to run the bolometer from the ground up.
To consider -
Can you get tools for the processor(s) ?
Has the sensor got any biases, and are they hardware or software controlled.
How to clock it
Once you have got to digital data in memory, you'd need the calibration and an output method.

[Abbott242]

Mike: I tried the soldering iron test this morning, unfortunately it didn't work. More over, I'm not sure how practical it would be to keep a hot pice of metal around to wave in front of the lens every few min  . Good idea though!

Bill: Those are some good points, I'm working on getting data sheets for the processors present. I've also reached out to my university, and they have a few people who might be able to help as well. I'll keep everybody posted!

[ArsenioDev]

OH! W/R/T the FFC flags not happening, swap the iris and FFC flag motor drive cables, I had an issue where it tried to FFC with the low gain iris, needless to say that didn't work too well.

[Abbott242]

Hmm, I can easily trace the flag motor cables, but not sure about an iris. I only see the flag in front of the microbolometer window, no iris. Maybe the FFC issue is because I'm just not providing power to a certain pin or something?

[ArsenioDev]

Interesting. I'd take that whole stack apart and ultrasonic clean it in isopropyl then airblast dry and clear. I had one of my cores covered in foam dust behaving weird, after cleaning it worked fine.

[Abbott242]

Well, I've got a bit of an update on this core. After some digging, I strongly believe that it's actually a ULIS UL 03 08 1 microbolometer, as indicated by the photos. I've found technical documentation for another ULIS model which appears to share the same pinout (UL 01 01 E), but the documentation for this model is behind a Baidu paywall. Therefore, there may be some small chance to fix this thing, as I now have the pinouts, example clock diagrams, etc. However, this is still a monumental task and probably not worth it. The bad news is, as I move the microbolometer, I hear motion inside- it sounds like a glassy or metallic object is sliding inside the vacuum chamber