Flir E4 Thermal imaging camera teardown

[mikeselectricstuff]

Has anyone worked out how the 'manual' focus operates yet?

I made a tool to rotate the lens surround yesterday but I'm uncertain which way to turn it to get closer focus. I'm nervous that something will come loose or run off a thread etc.

Also, does the lens move relative to the bit that rotates? i.e. does the little lens 'poke/protrude' and the surround stay in the same place (if that makes sense)
Or does the whole assy move in and out together?

The lens is embedded in a bit of plastic. the whole thing turns. Screw it all the way out to take a look if you're curious. or watch the teardown.

Someone already asked how many turns are needed and nobody replied. So can I ask the same question? How many turns will I need to turn the lens for maybe 6 inch focus? ... and which way do I turn it?

Screw out for closer. Adjustment is fairly coarse - about 1 turn of useable range. Factory position on mine is about half a turn from fully in. To reset, focus on a distant object (roof edges are good)

[zapta]

Interesting read that implies that people will be upset paying more for un-crippled hardware (e8 versus e4 pricing).

Yet it is very common in the software industry.

For some reason it's feels different when it is a hardware product hardware.  With software you buy a license, with hardware you buy real stuff.

Software crippled hardware typically involved actual hardware differences.  For example, when Porsche sells a software option for a more aggressive sport mode, it also comes with a clock  ;-)

http://www.porsche.com/usa/models/cayman/cayman/chassis/sport-chrono-package/

[G0HZU]

Screw out for closer. Adjustment is fairly coarse - about 1 turn of useable range. Factory position on mine is about half a turn from fully in. To reset, focus on a distant object (roof edges are good)

Thanks!

I just tried turning it anticlockwise and it only needs to be turned a little bit as you say

See the image below for the first attempt at a tool. I made this on the T-Tech PCB router I have here. The next version will have a larger hole where the screw is and I'll make up a cone out of thin (tinned) brass sheet and solder it to the tool.

So hopefully it will be able to 'see' through the cone and allow focus and view at the same time

[G0HZU]

Here's an image of the second milled shape.

This one is the same as the first but with a much larger aperture. The idea is to solder a cone to it to act as a means to hold the shape/tool and also rotate it and see through at the same time.

[G0HZU]

    Ray Winstone:  Where's your tool?   


I'm now busy practising a bit of origami so I can make an accurate paper template to curl into the perfect cone shape

[G0HZU]

I'm now wondering if I could simply glue the top section of a stiff plastic bottle to the shape. This would give a cone shape for free and could actually look quite good if sprayed black.

Anyone got any suggestions of a suitable bottle or container?

Otherwise I've just got to find one in the house that fits (and drink/use the contents)

[Taucher]

seems like the lens-shape determines the keyform
attached an yet UNTESTED .stl for a mini-key

[G0HZU]

Whatever tool is used it has to be a perfect snug fit because I think the lens key holes will quickly wear out otherwise.

I managed to get it spot on and mine has no wiggle at all and also the round bits fit flush all round. See the image below.

What I would not recommend is anyone to try and move it with anything that is not a flush fit. I think I might look at using a different PCB material to FR4 to make it. Maybe FR4 will be too abrasive unless somehow polished smooth.

In the image below you can see I have already scratched the key area due to a few early attempts to move it with something unsuitable.

I've got some ceramic substrates here that are used up at many GHz because I normally use the PCB milling machine to mill exotic RF filters up in the microwave bands (for research)

I think I'll try making one tomorrow.

Can these lenses with the key body be completely removed and replaced? Mike's earlier post implied this was possible but I don't know if this needs a teardown to achieve.

[mikeselectricstuff]

yes, you can easily unscrew the lens right out.
As regards wear, I think plastic pins are probably the way to go, so tool wears before lens

[G0HZU]

Yes, I think caution needs to be exercised here.  Even a ceramic substrate may prove to be unsuitable. I have some old school GIL GML1000 PCB material somewhere that is (was) a very forgiving PCB material. It was brilliant for milling because it gave the tools a very long life.

Sadly it isn't made anymore but I think I've got a few sheets of it left somewhere.

[G0HZU]

The GIL GML PCB material is similar to PTFE but you can also get copper clad PTFE to use as a PCB material (and solder to it rather than glue to it). It's a bit like making a PCB out of a bendy cheese slice because the PTFE material is so soft and bendy.

I can mill these out in a few seconds each but I need to get the material correct because I fear the 10 little key slots will wear very quickly if the user started fitting/removing this lots of times to permit focussing the camera on a daily basis.

I think the ultimate solution will be something that fits in there permanently and 'another' device is fabricated that keys into it as a focus ring. So the 10 key slots will see far less wear that would otherwise occur from any clumsy removal/refitting.

Regards
Jeremy

[mikeselectricstuff]

It ought to be available as a spare part, as it is potentially subject to damage, whether it is or not is another matter...

I don't think the thread or lens format is anything standard.

[mikeselectricstuff]

Quick follow-up vid - close-up options and tripod mount

[ron]

If the lenses were available at reasonable cost, I would buy a spare to modify with a permanent focus ring assembly but I suspect the lens is unavailable, or available only at high cost. I will write to FLIR and ask 

don't forget to ask if E8 display covers are available too. 

[Fraser]

Thanks for the new video Mike. Very interesting. If FLIR have no answer to my tripod question I will modify one of the batteries as per your method. Thanks for testing the lenses. I agree that the lenses that I use (100mm FL) will be no good for the E4 field of view. I shall buy a 50mm FL from China.

Cheers

F.

[Richard Wad]

Great video Mike. Somehow I always end up with my jaw open by the end of the vid.
Love the new bootlogo 

[G0HZU]

A very good video However, I was a bit worried about the bit in the video that said the cal 'might' be adversely affected at close focus so I did a very quick and dirty test on my E4.

I have a programmable hotplate here that can be set and regulated with a little microcontroller inside. It goes up to >300degC but I set it to 60degC and taped a tiny thermocouple to it using some matt black tape to sandwich it between two bits of tape. I had this connected to a pretty accurate Digitron thermometer that I check against an ice bath now and again.

I tried the E4 at 1m and also really close up with the focus wound out and it's close enough for me at both settings. I could get typically +/- 1.5degC difference between the Digitron and the E4 and both the E4 and Digitron typically said around 62degC. This wasn't a very well controlled experiment but it has made me feel a bit better about the impact on the calibration.

It's nearly bedtime  and I'll repeat it tomorrow with a clearer head

[mrflibble]

Note that AFAIK there is no evidence that :
 - Lenses and sensors are not "binned" depending on their quality
 - The more expensive models could be going through more extensive testing / calibration to ensure their tighter specs

Regarding binning of sensors: would be interesting to get the bad pixel count for E4 vs E8.
The sensor on my E4 looks to have 62 bad pixels, with quite a few near the edges.

What's also interesting to see is the circular pattern in the calibration picture, since no doubt the black body they used is aaaalmost uniform but not quite.

Below are the calibration pic + extracted bad pixels. Notice the banding in the calibration pic. I normalized and truncated to 8-bits, but the banding is also there in the raw 16-bit.
There's also what looks like the correction factor per pixel, all with values between 0x1c and 0x29.

You can find the raw files in /FlashFS/system/maps/ . Those are all FFF type Flir files with raw data. The .gan and .crs is what I used for the bad pixel count etc.

[olsenn]

That calibration picture looks AWFUL! Is that supposed to be uniform black? Does the circular banding mean that the thermal photo quality/accuracy is awful everywhere but the center?

[mrflibble]

That calibration picture looks AWFUL! Is that supposed to be uniform black? Does the circular banding mean that the thermal photo quality/accuracy is awful everywhere but the center?

No. No. No. In that order. For a perfect calibration target it's supposed to be uniform gray.

Attached is the original cal picture before I normalized it. A nice uniform and uninforming picture.

The AWFUL cal pic is magnifying the grayscale as it were. Without it all you see is fairly uniform drab grey (see below), aka uniform temperature. But by exaggerating the differences you can see the circular shape, no doubt due to the non-ideal black body used during calibration.

[eliocor]

I think the map compensates for the lens vignetting, not for the supposed non-uniform black body...

[okent]

Another successful hack! 
Thanks to Mike and everyone here who did all the hard work for us noobs!

I thought I had it done right but noticed no real change.  Went back through the .cfg file and noticed I forgot to put in the serial number after running  crc01.   

Updated the file and did Mike's simple install and it's amazing now!  Huge difference.

If someone figures out a 30hz mod this would be perfect.

Thanks again.

[Taucher]

If someone figures out a 30hz mod this would be perfect.

Actually that one would be a real nasty problem as it would turn the camera to a "dual use" item that can be treated just like "arms/weapons" - so it would fall under trade restrictions and mean that any applicable government would have to take special care about the camera... guess how good that would be... 

Matter of fact: If you buy one of the higher spec'ed 30/60Hz cameras, then you'll be politely asked to explain(fill out a form) the security measurements against theft etc. - thus preventing some foreign intelligence person from easily stealing the camera, shipping/smuggling it to some "axis of evil" and turning the sensor into a missile guidence system, unwanted airborne defense or hell knows what else (yeah, what horrible threat such a thermal camera must be  ).
Well, this of cause assumes that a whole "enemy" country (usually that's one that didn't attack somebody else yet) will stay offline, dumb and will fail to produce any high-tech on it's own... but I guess the intention is to make it for them as difficult as possible (for a while)...

[Corporate666]

I am going to have a shot at making a lens adjustment tool maybe this week or next if I get a chance.  I'm planning to machine a cone out of aluminum and then have a UHMW (Delrin) base that mates with the lens... the delrin should be a reasonable mix between stiff but pliable that it doesn't harm the lens but is able to have small features machined in it.  I'll post results if they are halfway acceptable.

[Taucher]

Just some random notes on recent insights:

Hacking NK.bin (Windows CE base components - mostly uninteresting):
http://www.t-hack.com/wiki/index.php/NK.BIN_toolset
http://msdn.microsoft.com/en-us/library/aa449985.aspx (bintool.exe documentation)

Not documented in T559504$A but in http://support.flir.com/DocDownload/Assets/21/English/1557845$a40.pdf (tested+verified)
camera console: nuc.exe supports -n switch for calibration without triggering the shutter! - one can also pass a number (0...7) in order to specify the intregration interval. The number (n) is then converted to 2^n frames. Default is n=5 - resulting in averaging over 32 frames

EDIT:
http://qt-project.org/doc/qt-4.8/statemachine-api.html .. link regarding the MENU state machine