Is the Nightdriver A-11 camera a double resolution Raytheon BST?

[ArsenioDev]

Hey everyone, I recently got a hold of a bunch of thermal cameras for futzing around with and noticed in testing that the A-11 seemed to have MUCH lower noise performance and better resolution, but I'm not completely sure.
First image is everyone's favorite, the Cadillac, second one is the Nightdriver.
I also have one non-starting gen 1 cadillac and another gen 2 that will not power up, still need to find that one fraser thread on the servicing of these.

[Spirit532]

There was only ever one commercial BST sensor made, a 320x240 array from Raytheon.
AFAIK, there are a few revisions and camera designs that result in lower noise and higher sensitivity. The resolution is always 320x240.

[Ultrapurple]

The first image appears to be a very low thermal contrast scene, so the camera is struggling to detect anything.

The second image has much higher thermal contrast - a warm hand against a cold background - which is much easier for any thermal camera to see.

Whilst I am not calling into doubt the OP's assertion about different sensitivity and noise, two identical scenes, preferably low thermal contrast, would be a better way to demonstrate the effect.

[ArsenioDev]

Update: At the behest of posts in thread, I've gone with two low contrast thermal scenes of a single sheet of printer paper placed over the lens for a flat field noise reference. The cameras are side by side on my workbench in an air conditioned lab space, powered up at the same time and left to sit for 5 mins to stabilize temperature.

[Fraser]

The Raytheon core came in different versions throughout its life. 320 x 240 pixel resolution but with different image processing electronics packages. To illustrate this, the Thermal Eye 300A was the original ‘analogue’ model and the Thermal Eye 300D was the later version with much better ‘Digital’ image processing.

I suspect you are effectively comparing an ’Analogue’ model with a later, more refined, ‘Digital’ model that carries out better image processing and presents lower noise to the user. That is, of course, assuming that the more noisy camera is in good condition with an I damaged chopper wheel membrane. Damage to that chopper wheel can produce a poor image output.

Fraser

[Bill W]

There are several control options on the Raytheon BST core allowing the setting of AGC (active, fixed etc) and DC grey level for no signal.  The system builder could access these and set them up as desired.

This could give a very different 'feel' to the image

Bill

[ArsenioDev]

Fraser, that was my thought until I noticed the different IFOV with the same lens stack from the other camera. Bonus is all my chopper wheels are in FANTASTIC shape. Grabbed a spare athermalized stack I had sitting around and there's DEFINITELY a big difference in the physical size of the FPA. It's like the difference between an APS-C and Full frame sensor, the A-11 takes the WHOLE lens projection and when correctly focused can even see the physical structure vignetting of the lens frame. Images in order are test setup, cadillac and A-11.

[Fraser]

Hmmmm interesting. I have only seen the standard 320 x 240 pixel Raytheon BST FPA.

Bill W may know more about what BST sensor FPA’s were produced.

Fraser

[Bill W]

I am not aware of anything other than the 320x240 48um sensor ever being made.

The later (= 205 digital)  kit had a 2x digital zoom (NTSC) and a PAL zoom option.  It also had several steps of frame averaging available.

Bill

[ArsenioDev]

So nobody knows what the hell this sensor is? Neato! Unless the other sensors are already zoomed in for some reason... Does anyone know how to count lines of video with a 1054Z?

[Ultrapurple]

I can't comment on counting video lines but I can offer the following thoughts.

When I view the output of a 320x240 camera on my living room HD TV I can very clearly see each individual camera pixel because the camera electronics reads each one twice to produce the 640x480 NTSC video, which is then upscaled by the TV processor. The result is a very, very clear fixed pattern.

640x480 resolution sources may also have a slight residual pattern but it is much finer and far less distinct.

If you photograph the 320x240 displayed on the large screen you can easily count the pixels (and usually spot the bad ones even if masked by the camera).

Another easy way to determine if the camera is 320x240 or 640x480 is to focus on a sharp, reasonably high contrast edge sloping at 45°. A soldering iron is too much; a warm phone on a colder background is enough. Shoe those images on the big screen and the reduced jagginess from 640x480 will be clear.

You can of course use a smaller screen than a living room TV but it's remarkable how much clearer the effects are when 'in your face'.

[Bill W]

The analogue video will be either NTSC or PAL.  Both are around 64µs line time and easiest to tell apart by the field time.  NTSC 16.7ms, PAL 20ms.

Your problem is to work out how many sensor pixels go into a line, and whether the two fields are the same.

The following might work:

At turn on the BST takes about 30 seconds to stabilise in temperature.  During this time the sensor pixels should be very visible as a 'mesh' like mess on the image.  You should also see the current consumption swing around in sympathy.

If you can grab a single video line you may be able to see the individual pixels on it and measure them in time. 
The expected answer, for NTSC 320x240 full frame, is around 52/320 µs

If you have a dead pixel, you could simply measure that.

Bill

[Ultrapurple]

Bill is right of course, though given that the OP is in the USA and he mentioned that at least one camera was from a Caddy, I suspect the chances of the output being 'PAL' (ie 625 lines) is fairly remote.

[wadizuwaro]

Fraser, that was my thought until I noticed the different IFOV with the same lens stack from the other camera. Bonus is all my chopper wheels are in FANTASTIC shape. Grabbed a spare athermalized stack I had sitting around and there's DEFINITELY a big difference in the physical size of the FPA. It's like the difference between an APS-C and Full frame sensor, the A-11 takes the WHOLE lens projection and when correctly focused can even see the physical structure vignetting of the lens frame. Images in order are test setup, cadillac and A-11.

From the 2 pictures, the A11's pixels are even bigger than the cadillac's, means it don't have more resolution, however the much wider FOV is a mystery...
Can you take pictures to show us both sensor's physical look side by side?

[ArsenioDev]

To answer Wadizuwaro, the A-11 is the one with the smaller pixels, larger are the Cadillac. Externally the sensor packages look basically the same, I'm COMPLETELY mystified as to what the heck is this sensor. the resolution is DEFINITELY higher as when held at an angle, the stairstepping of a line is MUCH smoother on the A-11.

[Spirit532]

Easy answer. One is stuck in a zoomed-in mode, the other is not.
Again, only one 320x240 BST sensor was ever manufactured and put into production. The lenses did vary throughout the years.

[ArsenioDev]

Is there a way to unlock the zoom of the Cadillac sensors? 

[Bill W]

Yes, in theory.  Realistically you'd need the Raytheon factory software.

The BST 'kit' used RS232 (or RS485 - you could choose) to control the 'kit'.

It was not unknown for them to change their mind from time to time, which must be what yours has done.  It was a return to factory to fix it.
There was also an option to use a different calibration table when zoomed, which might explain any video artefacts that you see.

Makes sense now, I had also misread which image was which, and assumed the Cadillac was working normally.

regards
Bill

[ArsenioDev]

Bill, the cadillac imager are indeed working normally, crosschecked with existing footage online and it's totally correct so it's not zoom locked on any of my three units. My puzzlement increases again, so it's time to break out the scope

[Fraser]

See here...

http://www.nightdrivingsystems.com/main/l3.html

Note the comment in the last line. No details on the higher resolution unit sadly.

Fraser

[Fraser]

Just to throw some more information into the discussion...... I understood the Cadillac De Ville camera to be a 320 x 240 pixel unit but see that the same core is used in the Thermaleye 4000 camera and in that unit only 160 x 120 pixels are offered to the owner !   Hmmm I am wondering if Cadillac De Ville cameras are indeed set to X2 zoom and truly provide only 160 x 120 pixel images ? I have always thought the Cadillac de Ville camera produced an inferior image to that found on the Argus 2 camera that uses the same BST sensor.

http://wanderlodgegurus.com/database/House/Cameras/4000B.pdf

http://www.americaninfrared.com/ProductDetail.asp?ID=41

I will need to do some more digging on the topic.

Fraser

[Fraser]

I have been pondering this Cadillac camera resolution issue and I do recall reading an article that stated that it used a 120 pixel height frame that was “doubled”. I thought that was a reference to an interlaced image but now think it could well be a reference to the camera being operated in 160 x 120 pixel zoomed X2 mode. The camera output would still upscale to 320 x 240 pixels for display. This is starting to make sense to me now. There were issues with the use of thermal imaging on vehicles, even back in the era of the BST sensor arrays. There may have been restrictions on the resolution permitted for such a camera being provided to the public. The Cadillac De Ville pixels always looked prominent and relatively large to my eyes but I never really studied that cameras images in any detail. I have two but they are not in use.

So, if I am correct, and the Cadillac De Ville is operating as Spirit suggested, in X2 Zoom mode, then the A11 is operating like the Argus 2 camera in its full 320 x 240 pixel glory and hence the much better images produced. Analysing the output image could get a little confusing as upscaling to 320 x 240 output will be present.

People, including me, may have read that the camera used a 320 x 240 pixel sensor and assumed that it was used at this resolution. As we know, windowing is common on thermal imaging cameras to deliberately limit the resolution for whatever reason. It must be remembered that in the late 1990’s the military in the USA had concerns about the public having access to decent thermal imaging systems. I can see the Cadillac De Ville camera being subject to many discussions with the relevant authorities on its deployment and capabilities. Unlike the Autoliv cameras, it had no protection against redeployment. I seem to recall the later Honda version of the L3 BST camera producing superior imaging. That camera has protection against redeployment on another platform. I suspect it was permitted to operate at its full resolution as a result.

https://www.eevblog.com/forum/thermal-imaging/honda-megane-night-vision-camera/

https://www.eevblog.com/forum/thermal-imaging/honda-night-vision-camera-teardown/

Interesting stuff 

Fraser

[Bill W]

The following may be of interest;

https://web.archive.org/web/20030618095648/http://www.raytheoninfrared.com/productcatalog/index.asp

Leading to:

https://web.archive.org/web/20030615211129/http://www.raytheoninfrared.com/admin/file/SPIE4360-211finalauto.pdf

Suggests there was an optics change during the Cadillac builds, '25% smaller window'.  Change of f number or change of focal length ?

Might even explain some of the chopper wheel variations

[Bill W]

Bill, the cadillac imager are indeed working normally, crosschecked with existing footage online and it's totally correct so it's not zoom locked on any of my three units. My puzzlement increases again, so it's time to break out the scope

Can you measure horizontal field of view for whatever camera / lens combinations make sense ?
The lens images from the link in the post above may help identify them.

As a guess, lenses will be near f/1, so diameter and focal length will be near equal.
From that you can work out the field of view at the sensor, in mm wide, and then see if it makes sense with 48µm pixels.

Bill

[Fraser]

Further to Bills comment. There were two versions of the Cadillac De Ville camera that are easily differentiated. The Mk1 was a bit of a disaster. It had an oval lens barrel with a Germanium window sealing it. When the window got broken by a flying stone water ingress followed. The unsealed lens barrel then allowed water onto the electronics package. The Window was not user replaceable. The Mk2 camera was a superior design that incorporated lessons learnt from the Mk1 issues. It has a sealed lens assembly so if the window fails, only the objective lens behind it gets wet/dirty. The window is also user replaceable by easy removal of the front window retaining ring. Both my cameras are Mk2 units. Working Mk1 units are quite rare and, in my opinion, best avoided.

I note from the documentation that I can find on the Cadillac De Ville night vision system that no true resolution is stated. The cameras operating thermal resolution seems to be deliberately omitted. Even Raytheon/L3 appear to omit the resolution for that particular camera. We know that it uses their 320 x 240 sensor, but that means little without knowing how the sensor is operated, full resolution, or windowed.

I also note that Bill has advised that the Zoom setting was a ‘return to factory’ setting. This makes me think that Windowing (electronic zoom) was a method of deliberately crippling the cameras resolution to meet either Authorities or OEM requirements. Most modern cameras provide electronic zoom on a simple command or I/O pin. No return to factory required. Some BST cameras that used the same core offered X2 zoom to the user via a button operation. Those cameras were already operating at 320 x 240 pixels so no OEM or authorities were involved in crippling the resolution in that particular deployment.

Fraser