Flir E4 Thermal imaging camera teardown

[Fraser]

@Unicorn5

The E4 uses a small aperture lens and a fixed focus. The large depth of field produces a reasonably focussed image between 0.5m and infinity. Being fixed focus, it is a compromise solution and does not generally produce the sharpness of focus achievable with a larger lens capable of being focussed. The down side of a lens that can be focussed and that has a larger aperture is the small depth of field that it exhibits so you have to re-focus often if changing the camera to target distance by very much.

The E4 does not contain a 'rubbish' lens so improvement in clarity may be possible with the use of auxiliary optics. This is what I do for close-up work and the focus is good.

To gain very crisp images with the E4, at most distances, a lens with adjustable focus would be desirable. How well the current E4 lens performs if its focus is adjusted for each distance, I do not know. It is optimised for a large depth of field not fine focus.

[pomonabill221]

My first post here but I have read EVERY page of the great thread!
It's YOUR fault that I bought an E4 
AND I should have it tomorrow (Friday the 13th!) 

A friend of mine also bought on and should have his today (he ordered it before I did by a week).

I am going to upgrade both of ours AFTER I make SURE nothing is wrong with them or show any anomalies like banding, etc., like I have read in this forum.

I have a 3d printer and got the lens adjusting tool .stl file so I am going to print a couple for us.
We are in a paranormal investigation group so we'll see if these E4 [8's] help... they should.

I am also going to use it for thermal mapping my house's heat leakage (a 1911 craftsman with shiplap wood siding), and I want to see how my attic insulation is working.

There are many uses I am thinking of to use the E4[8] with as well, but this thread made me bite the bullet.
I got a good price for it also... 895 free shipping (about %10).

Thank you Mike for your hard work coming up with the very simple hack!

[dustout]

Every single camera in the i-series has a 320x240 detector that is cropped, in software, to either 140x140, 100x100, or 80x80.

Assuming you're a good hacker, you could probably get access to all that 320x240.

It is possible that they use reject sensors with defects for the lower-res products, and/or possibly only calibrate them for the resolution they're sold as - I think the latter is fairly plausible if cal takes appreciable time as that equates directly to cost. thermalguy could probably shed some light on this.

I did some more experimenting with my i3 resolution. If I tell it use resolutions above 120px it maxes out there.  Oddly I cannot get it to actually go up to 140x140 as setting it to that just maxes it out at 120x120.  Digging around in files I noticed a file

Code: [Select]

/FlashFS/system/fpga_isc0601_120x120_c16.bin which makes me think the max resolution on this thing is a bit more hard-coded.  No matter what I set the resolution to above 120x120 the web interface shows a "Sub Area A_W1" of 14400px (which would be 120x120), 432 nops, 2 cols, 2 rows and "Area A_3" of 1600px. I'm not sure what that is.

Oddly enough the original fpgaResolution setting in the service config is 144 so I would think it WOULD at least go up to at least 144x144, but it will not.

I wonder if it would be possible to drop in a higher resolution fpga bin file.

[Fraser]

I did some low temperature tests this evening.

The target was a pair of diecast boxes that have been painted with Kylon paint. The Emissivity is supposed to be 95% according to FLIR.

One box was cooled to +7 Degrees Celsius and the other to -8 Degrees Celsius

The attached pictures show the direct comparison between my Industrial PM695 and the E4.
The PM695 permits entry of ambient temp, distance and humidity. the E allows only Reflected temp. All settings were entered from readings taken with a suitable thermometer/humidity meter.

A contact thermocouple was used to measure the surface temperature of the boxes. The -8 Degrees Celsius test had to be carried out quickly as the frosting that developed degraded the accuracy of the readings (as Mike warned)

The E4 appears to have accuracy issues as the temperature drops below zero Degrees Celsius. The error is similar to that experienced outside when I measure a wall temperature. I am not in the habit of measuring low temperatures with my TIC so I am not at all bothered by this minor issue. I can always use the PM695 anyway.

Note that when a thick frost had formed on the -8 Degrees Celsius test box, both cameras produced readings of around 0 Degrees Celsius even though the box was still well below that temperature. The emissivity was totally changed by the layer of water.

Please note that this was not a 'lab conditions' test and some small error will be present. What was apparent though was the size of the E4 error when reading a -8 Degrees Celsius test piece.

[Hypernova]

Finally got mine, damn distributor took a month to get it in. Good thing they at least offer free calibration forever. I was getting worried.

Version 1.19.8
Cal: Nov 29th
Serial: 639086XX

Attached is KFC from last night

[Richard Wad]

I did some low temperature tests this evening.

The target was a pair of diecast boxes that have been painted with Kylon paint. The Emissivity is supposed to be 95% according to FLIR.

One box was cooled to +7 Degrees Celsius and the other to -8 Degrees Celsius

The attached pictures show the direct comparison between my Industrial PM695 and the E4.
The PM695 permits entry of ambient temp, distance and humidity. the E allows only Reflected temp. All settings were entered from readings taken with a suitable thermometer/humidity meter.

A contact thermocouple was used to measure the surface temperature of the boxes. The -8 Degrees Celsius test had to be carried out quickly as the frosting that developed degraded the accuracy of the readings (as Mike warned)

The E4 appears to have accuracy issues as the temperature drops below zero Degrees Celsius. The error is similar to that experienced outside when I measure a wall temperature. I am not in the habit of measuring low temperatures with my TIC so I am not at all bothered by this minor issue. I can always use the PM695 anyway.

Note that when a thick frost had formed on the -8 Degrees Celsius test box, both cameras produced readings of around 0 Degrees Celsius even though the box was still well below that temperature. The emissivity was totally changed by the layer of water.

Please note that this was not a 'lab conditions' test and some small error will be present. What was apparent though was the size of the E4 error when reading a -8 Degrees Celsius test piece.

Aurora I noticed this also several weeks ago, thanks for confirming I was not alone.
Outside temps were in the 40s F and my E4/8 was reading ground temperature 15-25 degrees less.
Granted surface ground temps are likely to be lower than air temperature where we are at, but it was definitely nowhere near accurate.
Where I am now is around 15 F, if you'd like I can post up some shots to compare.

[Wizzard]

Mine arrived today, ordered almost 3 weeks ago.

E4 1.1
639082XX
63901-0101
1.19.8

[tomas123]

All E4 are brand-new and calibrated.
I think they don't use cooled blackbodies for the low cost E4.

the calibration values are embedded in exif header (fingerprint like serial number):

Code: [Select]

$ exiftool -flir:pl* E4-8DC.jpg
Planck R1                       : 14871.23
Planck B                        : 1400.3
Planck F                        : 2.5
Planck O                        : -7575
Planck R2                       : 0.027825247

the my post here
http://u88.n24.queensu.ca/exiftool/forum/index.php/topic,4898.msg23972.html#msg23972

Code: [Select]

T_obj= B / ln(R1/(R2*(RAW_obj+O))+F)
  T_obj = object temperature in Kelvins
  RAW_obj = RAW value (pixel) (Voltage 16/14Bit)
  R1 Planck R1 constant
  R2     Planck R2 constant
  B    Planck B constant. Value range 1300 - 1600.
  F     Planck F constant. Value range 0.5 - 2.
  O    Planck O (offset) constant. Its a negative value.

[Zucca]

According to the Flir Ex Datasheet:

Temperature Range    –4° F to 482°F (–20° to 250°C)
Accurancy                        ±2% or 2°C

So if you guys are going with a Flir E4 to measure T<-20°C, you are out of spec and the Mr. Flir can do whatever he wants.

First I assume to read the accurancy specification like this:

IF -2°C<(Measured Temperature [°C] * 0.02)<2°C THEN Accurancy[°C]=+- 2°C ELSE Accurancy[°C]=+-(Measured Temperature [°C]* 0.02)

am I right? So at 250°C I am reading +- 5°C?

Thanks Aurora an error of -8.3°C-(-7°C)=-1.3°C is still in spec in my eyes.

So to me it sounds like the more you getting close to those range borders (far away from 20°C???  ) the more you will lose accurancy. It would be interesting to see this error amplitude vs measured temperature. I´m expecting now a good read between -10°C and 100°C (+-2°C). It looks like by reading this forum for about T<-10°C the Flir is out of spec... mah...
Interesting would be to see the accurancy in the high temperature zone >100°C, (if someone already did it sorry I didn´t read all the 153 pages).

I'm imaging  the marketing guys (they should just study economy on books and don´t mess around with specifications) discussing with the developers regarding what to put in the datasheet for temperature range and accurancy... poor developers, I was often on that side of the table.

Z

[daves]

Hello guys,

is flir.php (convert and exiftool) working for you ?
I have fresh new E4 (arrived in 4 days after purchase, from PASS UK) and I am receiving weird result.

It creates ir, raw, gradient, palette pngs. No one is real view. Ir and raw seems somehow damaged.

Any tips ?

FINAL SOLUTION: I had wrong version of convert, needs to be 16bit. SOLVED

[mkaufer05]

I received the Flir E4 two days ago, and I tried the hack for it, but it does not work. the firmware on it is 1.19.8. and has a  date of Nov-23-2013 as its date of it being installed with the Firmware  So unless Im doing something wrong, then the Hack for the E4 no longer works. 
If  I am doing something wrong then please let me know.

[mikeselectricstuff]

I received the Flir E4 two days ago, and I tried the hack for it, but it does not work. the firmware on it is 1.19.8. and has a  date of Nov-23-2013 as its date of it being installed with the Firmware  So unless Im doing something wrong, then the Hack for the E4 no longer works. 
If  I am doing something wrong then please let me know.

You are probably doing something wrong. Errors in creating the CRC are the usual issue

[tomas123]

is flir.php (convert and exiftool) working for you ?

Do you have used this version?
https://www.eevblog.com/forum/testgear/flir-e4-thermal-imaging-camera-teardown/msg342072/#msg342072
download the flir-script.zip

I think, you have trouble with the reverse byte order

Code: [Select]

  //16 bit PNG: change byte order
   $size=$exif[0]['RawThermalImageWidth']."x".$exif[0]['RawThermalImageHeight'];
   exec($exiftool." -b -RawThermalImage $flirimg | ".$convert." - gray:- | ".$convert." -depth 16 -endian msb -size ".$size." gray:- raw.png");   

[mkaufer05]

Well, I follow the instructions correctly.  when i use the Flir install program to install the FiF file it shows as camera been updated and everything O.K.. but still its not  hacking it.

[Mikesus]

The CRC is wrong.  Got a camera with the same version and it works fine.  You want to make sure you don't add any additional lines at the end.

[stefbeer]

I did a similiar but simpler test as Aurora, but I only have my hacked i3 for comparison. I turned on both cameras and let them stabilize for about 10 - 15 minutes with a closed shutter.
After that, first I took a picture of a radiator and tried to aim at the same measuring spot. Results:
E4: 35.7°C
i3: 35.7°C

Then I went outside and took a picture from the wall of our garage:
E4: -5.2°C
i3: -2.5°C

2.7°C temperature difference? I'd say that's a lot compared to the radiator measurement which was spot on.

Our thermometer shows 0.2°C outside but thats probably not very accurate, it's a cheap and very old wireless thermometer.

[ixfd64]

I found an interesting thread about the hack on a gun forum: http://www.northeastshooters.com/vbulletin/threads/showthread.php?t=231048

[Taucher]

Well, I follow the instructions correctly.  when i use the Flir install program to install the FiF file it shows as camera been updated and everything O.K.. but still its not  hacking it.

Take a look at Ez-CRC01 in my footer-  expecially made for people who fail to edit a .txt file

[mrflibble]

Right, video streaming time. I wanted to have my E4 in a setup in another room and still be able to check both the measurements and check the video. And preferably without having to walk over there all the time. I already had telnet + scripts to take readings and take snapshots and such. But no video until now. (Well not true, I ran mplayer over an X11 tunnel, but that was rather lame ).

So now I have a streaming server that can be connected to over the network. If you feel like it you could even make it remotely accessible because you want to check the thermal state of your beer before going home or something.

All you need is just one piece of software, and that is ffmpeg. Then setup a tcp server with ffserver (working example config attached, rename to ffserver-flir.conf). When that is running you send your E4 camera stream to this ffserver using ffmpeg.

Code: [Select]

# start streaming server
ffserver -f ffserver-flir.conf

# send flir E4 stream to ffserver
ffmpeg -r 25 -f video4linux2 -i /dev/video0 -vcodec copy http://localhost:8090/feed1.ffm

That's it. Now you can connect to the stream using your favorite media player or browser.

Code: [Select]

mplayer  -nocache   http://yourserver:8090/flir.asf

I tested it with 3 seperate machines all displaying the stream simultaneously, and no problemo.

There were some sneaky gotchas to get it working, but those are all neatly embedded in the config + command line. Re-ordering of certain options like "-r" is at your own peril.

Forgot to mention... You can also have it start ffmpeg automatically, so all you need is start ffserver. I'd advice starting things up seperately the first time though just so you can see all the debug output in case something doesn't go as planned. At any rate, to automatically start ffmpeg just add 1 line to Feed section, like so:

Code: [Select]

<Feed feed1.ffm>
File /tmp/ffserver_feed1.ffm
FileMaxSize 2048K
ACL allow 127.0.0.1
Launch ffmpeg -r 25 -f video4linux2 -i /dev/video0 -vcodec copy
</Feed>

And another thing I didn't mention is that this uses UVC so you will have to have kernel support for that, but that would seem rather obvious. Anyways, for most popular distros uvc should work out of the box. And another other thing ... this sends the RAW video stream directly to ffserver. This way ffmpeg does NOT do any encoding, nor does ffserver have to do any transcoding. So ffserver takes the raw stream, and directly encodes it (to a 512kbps asf stream in this case). If you want you can have ffmpeg do some initial encoding to get some compression, send it to ffserver and have that do transcoding before streaming. But the bitrate for the raw stream is pretty low so IMO the current setup makes more sense.

Doh! Also forgot to mention this ... you can check the server status by browsing to http://yourserver:8090/stat.html

[QuadESL63]

My first post on this magnificent forum....

My E4 arrived last week (from stock at a supplier in this country).
serial # 639065xx, SW 1.19.8
cal. date 9 Nov 2013

Thanks Mike for the tear-down video and for making the 320x240 resolution possible! This transforms an expensive toy into a very useable instrument.
I wouldn't have bought the E4 (with a limitation of 80 x 60 px.) otherwise.
Taucher: thanks for the extended menus.
And to all who contributed, thank you too most sincerely.

When I look with IrfanView at the image properties of an image, an "Exposure Time" of 1/28.1 seconds is indicated.
Anyone some thoughts on this?

Just fantasizing: assume the resolution of the microbolometer-ADC is 6 bit, then the following could be true:
* normal E4-mode: minimum temp span is 8.3 C. 1/64 x 8.3 C = 0.13 C
* extended menu (E8-mode): minimum temp span is 4 C. 1/64 x 4 C = 0.625 C.
This could correspond with the FLIR spec of thermal sensitivity < 0.15 C for the E4 and 0.06 C for the E8.
Or am I just imagining things??  (Or didn't I read the previous posts properly  )
I was just wondering where the difference in specs originated from.

Concerning the outdoor temp measurement of the brick wall of my home, I also experienced a much too low indicated temp...
Would be nice to know what I am doing wrong. Different settings of emissivity do not help and limiting the view to just a small part of the wall neither.
Sky indication is  <40C.

Jac

[mrflibble]

Just fantasizing: assume the resolution of the microbolometer-ADC is 6 bit, then the following could be true:

The ADC counts in the raw thermal image are 16-bit.

[QuadESL63]

thanks mrflibble, I was just babbling absolute nonsense.
6 bit could never span the whole range of -40.....+280C with the specified sensitivity...

J

[mikeselectricstuff]

Just fantasizing: assume the resolution of the microbolometer-ADC is 6 bit, then the following could be true:

The ADC counts in the raw thermal image are 16-bit.

I have a feeling that when I was looking at the raw stream on the scope it was 14 bits.

[Zucca]

2.7°C temperature difference? I'd say that's a lot compared to the radiator measurement which was spot on.

Well since both devices have the same accurancy +/-2% or 2º C, it could be that both are in spec since the difference is not bigger than 4°C.
For example the real temperature could have been -4°C.

See my previous post for details:
https://www.eevblog.com/forum/testgear/flir-e4-thermal-imaging-camera-teardown/msg346980/#msg346980

If I´m wrong let me know.

[tinhead]

that's crazy fast growing thread, within exactly two months you beated all other threads Mike!