EEVblog #669 - FLIR TG165 Thermal Imager Teardown

[Dave Turner]

On a different issue did my eyes deceive me or are the two traces from the ARM chip above/between ST (first 2 characters of the chip description) curved lines rather than 45 degree angled traces as the rest of the board?

If so, why would that be?

[SeanB]

Maybe there is a trick to dismantle it. E.g. Slightly heat the window and suck it off.

Alexander.

Most likely, like any modern touch screen with thermoplastic adhesive. Would be easy to do with any cell shop tool set. Now we see how it is put together I am sure I could open it with only a small screwdriver set, a hair drier and a little work with a spudger. You probably would be able to put the Dave tested unit together with a little bit of work with 2 part polyurethane flexible adhesive, like Pratley flexiseal ( should be available in Bunnings in AUS) to replace the rubber seal and hold the pieces together.

[bktemp]

Accuracy was my initial guess. But the FLIR one is specified at +/-3°C. That is good enough, the TG165 is only slightly better. The extended temperature range makes sense. Can the FLIR one only measure up to 100°C or is its accuracy only specified upto 100°C?
50° FOV gives a spot size of about 2:1 when avaraging all pixels, but you could use digital zoom, using only a few pixels to get a similar 24:1 spot size. Or do I miss something?

I dont't get it, why they didn't simply built a standalone camera around the lepton core like Mike did? With automatic shutter it should be accurate enough for the lower price segment.
Adding the thermopile sensor with all the manually adjusted lasers only adds more costs to the system. Without that they could sell it quite a bit cheaper.

[bktemp]

On a different issue did my eyes deceive me or are the two traces from the ARM chip above/between ST (first 2 characters of the chip description) curved lines rather than 45 degree angled traces as the rest of the board?

If so, why would that be?

I think those are the differential pair going to the micro USB connector. USB requires some special layout techniques like cotrolled impedance.

[nitro2k01]

Indeed, pins 70 and 71 are the USB pins. You don't need curved traces for USB 1.0 (12 Mbps) but that's probably what their EDA package does by default when specifying matched length traces, so they just went with it.

[ElektronikLabor]

I know Artlav mentioned this already; I just want to show you the shutter so you don't need to search for it:

[tszaboo]

Actually the STM32F103 has to run at 48 Mhz for the USB, so the 8 Mhz main crystal is PLL-ed up.
The small thingie with the label is in fact the 32Khz RTC crystal. I've seen similar windowed on an older sytem, it was relatively hard to get a replacement for it with the same footprint. I'm not sure at all, why they would use that old construction on a new design. (although this looks smaller than the one I had)

I believe the unpopulated connector which you called a power connector is some sort of communication interface. If you look closely, id goes through several transistors marked with Qn, and it goes to pin 51,52,53 and 54 which is "SPI2_NSS / I2S2_WS / I2C2_SMBA / USART3_CK/
TIM1_BKIN" and so on... So it is kinda like some sort of proprietary communication interface? I meal why would you connect them together with transistors? I'm curious. It is funny to see how they not populated the transistors but they did not bothered with the resistors. Hey! Extra 0.1 cents to save!
I believe the programming is done through TP10-13 on the side of the board. Of it is connected to the SDW port. It should be 4 pins, power, ground reset and a single wire to read write the MCU.

[ckambiselis]

Very nice tear down Dave, what is the range of the Lepton sensor, from how far can you distinguish for example a human? Are you thinking about selling it, maybe I can rehouse the electronics to a waterproof case and use it as a basic thermal imaging camera for our volunteer FD I'm part of?

[Fraser]

I am seeing a lot of comment regarding the use of a shutter and its purpose.

I read people calling it a calibration shutter and think it is worth stating what it is and why it exists.

The shutter that moves across the microbolometers field of view is used for Non Uniformity Corrrection, hence the name NUC shutter. It is not a calibration shutter as such.  The camera is calibrated at the factory and a calibration table loaded into its flash memory for system reference. As the camera and microbolometer change temperature, the calibration table provides correction factors for the data coming from the Read Out Chip. If this was not done, the microbolometer would be hopelessly inaccurate as it is basically like an array of thermistors.

The microbolometer pixels are unique individuals when it comes to their Delta temperature characteristics. Their characteristics and response curves are captured at the time of calibration and this forms part of the fixed calibration table.

Unfortunately, even the calibration table is not enough to tame the unruly microbolometer pixels. If the camera software detects a deviation in ambient temperature or excessive drift in the microbolometer pixels, it will initiate an NUC event. The NUC shutter is of a known temperature (measured by chassis or lens temperature). It is placed in front of the microbolometer and ALL pixel outputs are equalised against the shutter. Now this may be considered "calibration" but that isn't really its purpose as calibration is the factory process and NUC is just setting the nominal baseline of all pixels. To do this well, the OS needs to know what the shutter temperature is otherwise it would be reliant on equalising the pixel outputs on the average of the readings which isn't a great idea. If the shutter is at 26.3 C, the OS sets all Pixel drift compensation values to provide 26.3 C as the interpreted output from the ROC. Calibration ? not really, more compensation against a known reference. How accurate that reference is another matter. 

Without an NUC shutter, the OEM needs far more sophisticated temperature compensation and microbolometer characterisation in the design. There are TICs that do not have an NUC shutter and my NEC AVIO cameras allow the user to set the minimum NUC period or even switch it off. Those cameras microbolometers are actively temperature stabilised though.

Without an NUC capability, or comprehensive microbolometer Delta T characterisation, the microbolometer will display signs of individual pixel drift that gets worse with time. As such the image can become 'mottled' in appearance and unreliable for measurement purposes.

The FLIR ONE is little more than a toy and would not reasonably be expected to provide industry accurate measurements. Now the TG165 is a strange situation. With the built in NUC shutter the LEPTON should have been able to provide microbolometer based temperature measurement. FLIR decided to use IR thermometer technology instead for the measurement function. I do not know their reasoning for this. The LEPTON internal calibration table may just be too crude for measurement purposes ?

Aurora

[mikeselectricstuff]

There  is some info on the Lepton's calibration stuff in the data sheet
https://docs.google.com/file/d/0B3wmCw6bdPqFdXI1bEFnOHdWZTQ/edit

[EEVblog]

I believe the unpopulated connector which you called a power connector is some sort of communication interface.

If it is then why are they using wide power-like traces?

[mikeselectricstuff]

I believe the unpopulated connector which you called a power connector is some sort of communication interface.

If it is then why are they using wide power-like traces?

The transistors look like an H-bridge connected to that 2-pin connector.
Sounder? (Fluke has an alarm mode I think)
Alternative shutter mechanism ? (but I'd also expect to see a temp sensor)

 

[Fraser]

@Mike,

Nice datasheet 

Interesting reading. This is the sort of microbolometer core data that normally requires signing of an NDA before release. Thanks for sharing the url with us.

I note the term Flat Field Correction is used so I appear to have my lines crossed between NUC and FFC. The shutter has always been an NUC shutter to me but I suppose the purpose and way that it works is the same so no great issue.

I note with interest that the LEPTON can be supplied with OPTIONAL temperature stabilised output. An interesting statement but I am not yet certain what they mean by that. I will have to read the datasheet properly to se if they explain.

Aurora

[eneuro]

There  is some info on the Lepton's calibration stuff in the data sheet

Nice fresh 1 week old only latest Lepton  44 pages datasheet
It should be easier understand Flir's patents now.

BTW: In your Flir One teardown 1st video maybe do you remember part numbers/manufacturer of this visual digital camera used together with Lepton module to see its basic specs?
Which distance might be between those cameras lenses holes?

I will try estimate based on known dimensions of Lepton module, but it is interesting how far away those modules are in Flir One, while it looks like there is some shift between those 2 captured IR & visual images and there is no additional optics, but probably software adjustments if any or assuming that such small distance does not affect too much those Flir MSX edge/contours overlays on output thermal image?
It is interesting if we tried to capture small high temp spot cleary visible is it any visible displacement in output image between its thermal image and those overlayed edges/countours?
I was exepcting some kind of additional optics to be able catch the same IR & visual image -not by shifted IR & visual cameras 

[mikeselectricstuff]

There  is some info on the Lepton's calibration stuff in the data sheet

Nice fresh 1 week old only latest Lepton  44 pages datasheet
It should be easier understand Flir's patents now.

BTW: In your Flir One teardown 1st video maybe do you remember part numbers/manufacturer of this visual digital camera used together with Lepton module to see its basic specs?
Which distance might be between those cameras lenses holes?

I will try estimate based on known dimensions of Lepton module, but it is interesting how far away those modules are in Flir One, while it looks like there is some shift between those 2 captured IR & visual images and there is no additional optics, but probably software adjustments if any or assuming that such small distance does not affect too much those Flir MSX edge/contours overlays on output thermal image?
It is interesting if we tried to capture small high temp spot cleary visible is it any visible displacement in output image between its thermal image and those overlayed edges/countours?
I was exepcting some kind of additional optics to be able catch the same IR & visual image -not by shifted IR & visual cameras 

I think the distance range is fixed on the standard F1 app but there is also a "closeup" app that allows the shift to be adjusted maually.

[Wilksey]

Was that donated or did you buy it Dave?

It's a lot of money to tear apart like that if you bought it!

We've seen you take it apart, but can you put it back together? 

[nitro2k01]

I am seeing a lot of comment regarding the use of a shutter and its purpose.

I read people calling it a calibration shutter and think it is worth stating what it is and why it exists.

The shutter that moves across the microbolometers field of view is used for Non Uniformity Corrrection, hence the name NUC shutter. It is not a calibration shutter as such.  The camera is calibrated at the factory and a calibration table loaded into its flash memory for system reference. As the camera and microbolometer change temperature, the calibration table provides correction factors for the data coming from the Read Out Chip. If this was not done, the microbolometer would be hopelessly inaccurate as it is basically like an array of thermistors.

...

That is all known (well at least to me, as well as Dave and Mike of course). However, the Lepton sensor contains new fancy, patented technology that allows it to NUC in software, as long as there is movement in the frame, without a shutter, because it can make assumptions about the data. The remaining reasons for having a shutter is if the camera is expected to look at static scene for extended periods (which makes sense for the TG165 since it has a tripod mount) or if absolute temperature calibration is needed (which is not the case for the TG165).

The FLIR ONE is little more than a toy and would not reasonably be expected to provide industry accurate measurements. Now the TG165 is a strange situation. With the built in NUC shutter the LEPTON should have been able to provide microbolometer based temperature measurement. FLIR decided to use IR thermometer technology instead for the measurement function. I do not know their reasoning for this. The LEPTON internal calibration table may just be too crude for measurement purposes ?

That is possible, however as I speculated in the other thread I suspect the issue is temperature range, as outlined in this PDF:

http://www.eevblog.com/files/FLIR-TG165/TG165_Comparison_FINAL.pdf

The TG165's thermopile sensor is rated -25 - 380 °C whereas the Flir One (and presumably the Lepton sensor itself) is rated for only 0 - 100 °C.

[EEVblog]

With the built in NUC shutter the LEPTON should have been able to provide microbolometer based temperature measurement. FLIR decided to use IR thermometer technology instead for the measurement function. I do not know their reasoning for this. The LEPTON internal calibration table may just be too crude for measurement purposes ?

Perhaps. But I think more likely because it essentially becomes a smaller and cheaper E4, and Flir didn't want that.
The Lepton doesn't provide the same extended temp range that the existing PIR sensors can offer.
Also, don't underestimate the market for people that are comfortable with their existing device, in this case the basic PIR sensor.
Could also be like why Fluke make an inferior averaging responding version of their 28-II meter, because some customers have procedures written for average responding meters and don't want to change anything.

[EEVblog]

That is all known (well at least to me, as well as Dave and Mike of course). However, the Lepton sensor contains new fancy, patented technology that allows it to NUC in software, as long as there is movement in the frame, without a shutter, because it can make assumptions about the data.

Yes, and that is why I didn't think twice that the Lepton might have had a built-in physical micro shutter (or in this case, a purpose designed clip-on attachment)

The remaining reasons for having a shutter is if the camera is expected to look at static scene for extended periods (which makes sense for the TG165 since it has a tripod mount) or if absolute temperature calibration is needed (which is not the case for the TG165).

I suspect that is the only reason left. Yet, once again, I didn't think this was a usage case for this device which is usually switch on, aim, press trigger, get measurement, switch off.

[Smokey]

They probably got the same guy to design the case that designed the Keurig coffee machine case.  The #1 tool you need to get into a Keurig is a dremel with a cutoff wheel.  And that's only if you want to do the minimal damage possible.

[miguelvp]

They probably got the same guy to design the case that designed the Keurig coffee machine case.  The #1 tool you need to get into a Keurig is a dremel with a cutoff wheel.  And that's only if you want to do the minimal damage possible.

That's why you get an AeroPress for around $30 and make the best coffee ever

[bitblt]

That TG165 teardown was one of the funniest EEVblog videos I've seen yet!  Absolutely hilarious watching Dave completely destroy (cosmetically) that brand new $500 thermal imager, all in the name of EE curiosity.  I could definitely hear high pitch frustration in his voice after the third fail.   

Then camera suddenly cuts to Dave with two shell pieces apart.  (I guess Dave edited out when he went medieval with a crowbar (prybar)?)  It was an entertaining and informative EEVblog teardown none the less.  Great job.   

Let me know if you decide to auction the TG165 carcass on ebay as "tested working, sold as-is".   I'll bite.

[tszaboo]

I believe the unpopulated connector which you called a power connector is some sort of communication interface.

If it is then why are they using wide power-like traces?

The transistors look like an H-bridge connected to that 2-pin connector.
Sounder? (Fluke has an alarm mode I think)
Alternative shutter mechanism ? (but I'd also expect to see a temp sensor)

 

That is absolutely correct, an H bridge makes 100% more sense.

[morpheus]

Wow.. Do you still need that thing? I dont think that i can afforn new one

[ddavidebor]

At 19:32 to the video, exactly when you say "yep" the sensor lens became blank and the image froze.
So it seems like it has a shutter!