Yet another cheap thermal imager incoming.. Seek Thermal

[efahrenholz]

No matter what  anyone does, this sensor is a flop. I've heard plenty of people argue it's been a great little device. The quality is fair, considering the price point. However, the image quality and manufacturing issues make it a dud without using their proprietary processing. If anyone wanted to release a 3rd party driver to use with Windows, as was attempted, the real image that's there comes out screaming bloody murder. They promised an SDK. Where is it. The release of the next FLIR is just around the corner, and unless they do something revolutionary with the image, it's over. This is the Sega Dreamcast of thermal cameras.

I believed the hype, bought a device. Waited like 2 months to get it. Got it. Said 'wow this thing has issues.' Opened it, broke it and fixed it twice. Broke it a third time, couldn't fix. Never looked back.

[Uho]

Members who have broken their SeeK, do not throw it away. I need a tube with a lens. I broke her during the tests.

[Marco]

I do macro and telephoto lens for this project.

Are you talking about thermal lenses here or lenses for the optical sensor?

PS. why are these sensors so low FPS? Is it regulative restriction?

[marshallh]

It's due to the US export controls. Any device capcable of higher end quality must have a paper trail and someone is always responsible for it. In this case FLIR/SEEK are the parties responsible for making sure their products are always export controlled throughout their lifecycles.

Having fluid thermal imaging is a very powerful tool. US gubbmint does not want anybody else to be able to have the same toys they do.

[Uho]

I do macro and telephoto lens for this project.

Are you talking about thermal lenses here or lenses for the optical sensor?

PS. why are these sensors so low FPS? Is it regulative restriction?

Macro lenses are external. Prototypes have already been sent to Erik for testing. Telephoto lenses being developed as external. While they are rather large price. Perhaps I will change the lens to the sensor on the window. This will make the universal imager. It will be possible to put the lens with the right focus. But it is a complicated technology. Actual results will be in a few months. When Eric will send the test sample sensor.
 I asked the question at a low frequency. This is the answer Erik: I'll have to ask, but I don't think the pixel readout works above 9fps. The voltage offset on the amplifiers max out, same problem as the thermopile arrays.

[barehill]

Slow frame rate is due to thermal time constant of sensor, common to all of this type.

[Marco]

Their smaller arrays have 5 ms time constant ... then the higher density ones suddenly jump up to 100 ms.

[Trax]

I'm looking for a thermal camera for my Andoid LG p880 optimus 4x hd is the Seek Thermal sill recommended or should I wait for the Android version of Flir one that was announced on the CES?

[efahrenholz]

No, the framerate was calculated at about 36fps. The reason it appears so slow is actual several issues rolled up into one solution. First, ITAR states that certain types of imaging equipment require a paper trail. If the technology could enable the user the same or similar capabilities as NATO combatants, then it must be recorded who has it and where it came from. It is unlawful to internationally ship ITAR controlled equipment to another party who hasn't been licensed (essentially ultimately responsible for where it goes). The limit on framerate for LWIR cameras is 9fps. Once the camera exceeds this, it is now defined as capable of effectively tracking moving targets. If seek made it easy, or outright normal for the camera to exceed 9fps, they would be banned from selling the camera internationally direct to the door. It would have to go to licensed reseller, and people buying in those regions have to fill out paperwork to purchase.

The second reason they really benefit from the 9fps cap is the economic design of the sensor and lens. The lens is far too small to provide enough radiation real time. The noise floor of the sensor would swamp the image, so they stack several images together, using averaging, to eliminate the random noise of the noise floor, while the real data is usually fairly constant and averages into the dataset of each pixel.

By reducing 36 frames to 9, each 4 frames are averaged, giving an output of 9fps and keeping them capable of selling internationally door to door without extensive paperwork, and providing a usable image from a cheap product.

[eneuro]

No, the framerate was calculated at about 36fps.

Could you show us this calculation details to prove your statement? 

BTW: Seek App available from official release is sick, so unless someone write his own software for his specyfic task this toy can be useless and frustrating, but good news is it is easy write own software for this dongle 

[Trax]

I'm looking for a thermal camera for my Andoid LG p880 optimus 4x hd is the Seek Thermal sill recommended or should I wait for the Android version of Flir one that was announced on the CES?

push...

I read the thread but its never the less not clear to me how good or bad is this thing.
Sure in compaction to any professional hardware it may be crap,
but for a student that want to play around the standards are completely different than for a pro.
And for a student 200-300 bucks is a lot of money.

So any advice would be greatly appreciated.

[Uho]

What will you watch to them? This is not an instrument for the measurement. Low sensitivity, a lot of noise, does not work with low temperatures. Poor software. If you want to see the hottest places - that's enough. But not anymore. For the price of $ 300 I like this one https://www.kickstarter.com/projects/ebeall/hemavision-smart-thermal-imaging-with-computer-vis/posts/1191118 .

[Fraser]

Based purely on price, I believe the SEEK to be a viable thermal camera for the purposes of experiencing what thermal imaging actually is. It is certainly far more enlightening than using a simple mono pixel IR thermometer ! You can pay as much for a decent IR thermometer so in those terms it is value for the money.

Why have I been so negative about the product of recent ? Purely through disappointment. I had high hopes (maybe too high) for what SEEK Thermal were going to produce. I was VERY disappointed that they appear to have 'dropped the ball' when it comes to community engagement but then that may have been part of their business plan. Make it, sell it and most users will be happy enough with its performance ? So in fairness, I was amazed at the retail price point, and likely expected too much because of the good pedigree of the lead members of the SEEK Thermal team (both ex Indigo thermal cameras). They broke new ground with a 12uM pixel FPGA. Previously not commonly available. Sadly the sensor/processing package is not producing the image quality common on more expensive cameras from the likes of FLIR. That does not make the SEEK rubbish though, just under developed.

There are two other budget thermal cameras due out soon. The FLIR One MkII and the Start-up Hema-Vison. From past experience the new 160x120 pixel FLIR will be an impressive, well developed product that meets its claims. The Hema-Vision appears to be an excellent piece of development work that is only limited by its resolution of approx 82x62 (?). I see the FLIR One MkII selling well, and I may buy one. The Hema-Vision is an unknown but I hope it does well.

If you wish to dip your toe into thermal imaging today, go for a SEEK and enjoy yourself.

If you catch the thermal imaging 'bug', like I did, you can always buy a better quality 'used' thermal camera when funds permit. I bought a fully functional FLIR PM575 Industrial quality thermal camera in pretty much 'as new' condition 3 weeks ago for ..... £150 (yes £150)  I just bought another PM575 in good condition and that one cost me a still very reasonable £600. The PM575's are amazing build quality with a 320x240 Gen 2 microbolometer running at 60 frames per second and with FLIR's excellent image processing capabilities.

I have often suggested to people that they think about buying a high quality 'Used' Thermal Camera rather than a 'built down to a price' new one. My PM575's originally cost over $50,0000 each and it can be seen in the quality of components used. Huge FPGA and large 50mm diameter Germanium Lenses.... lovely stuff and a joy to behold 

There are bargains to be had if you search for them. I should say that I am capable of repairing a faulty thermal camera so that removes some of the risk of buying used but PayPal also protects buyers from mis-described cameras. Do not buy used thermal cameras that are 'spares or repair' unless you know a lot about component level repair on 6 layer PCB's with high density component layouts. They are not simple to repair.

Hope this helps a little

Aurora

[Marco]

The limit on framerate for LWIR cameras is 9fps.

I'm going to take a wild guess and say that the "response time" on the larger density Heimann sensors is actually a mandatory crippling time.

[marshallh]

320x240 Gen 2 microbolometer running at 60 frames per second and with FLIR's excellent image processing capabilities.

Have you taken pictures of the internals? Very interested to see those.
My NEC F30S is still working, occasionally it has a temper tantrum and requires re-initialization but its great for PCB work. I am anxious to see if the new FLIR One 180x120 can be hacked to assist with my pcb testing usage.

[all_repair]

I'm looking for a thermal camera for my Andoid LG p880 optimus 4x hd is the Seek Thermal sill recommended or should I wait for the Android version of Flir one that was announced on the CES?

push...

I read the thread but its never the less not clear to me how good or bad is this thing.
Sure in compaction to any professional hardware it may be crap,
but for a student that want to play around the standards are completely different than for a pro.
And for a student 200-300 bucks is a lot of money.

So any advice would be greatly appreciated.

With so many new products coming on stream, with limited budget and no compelling applications, I shall invest on a gear that is not in a fast depreciation curve like a thermal cam now.  In my lab, thermal cam is now the no2 or no3 most frequently used tool.  First being the multimeter.  Of course, like a multimeter, I need to have one in my field bag.  Presently only seek can fit the bill.    Don't place your bet too high on the kickstarter thermocam.  It is quite certain to be much later than the promised date (unless they 3d print everything inhouse later), they may also never deliver at all.

[Fraser]

@marshallh.

There are some pictures of the PM series internals in these postings.

https://www.eevblog.com/forum/testgear/flir-e4-thermal-imaging-camera-review/msg307483/#msg307483

https://www.eevblog.com/forum/testgear/flir-pm695595-teardown-pc-connectivity-firmware-modifications/msg393754/#msg393754

https://www.eevblog.com/forum/testgear/flir-pm695595-teardown-pc-connectivity-firmware-modifications/msg605719/#msg605719

I have come to love the design now that I have reverse engineered it and understand its physical beauty. Sadly I am not so knowledgeable on the firmware side so that remains a mystery to me (for the moment).

I can supply higher resolution pictures direct if you want them. The PM series began with the gorgeous 550 cryo cooled camera and then the 570 uncooled version was released in 1997. All PM5xx series cameras that followed were based on the 570 chassis but had upgrades added, such as later generation microbolometers and visible light cameras. It was a case of .... 'if it ain't broke, why fix it ?' and that was a very good policy as the PM series have gone down in history as a resilient and dependable friend to the thermographer. Much loved by its user base.  These cameras set the benchmark for others to follow.

Aurora

[efahrenholz]

@Trax,

To be fair, the camera does show you hot and cold. Like anything new to someone (as it was new to me in terms of the flexibility of the app and etc), the honeymoon period does wear off. You start to notice things that you overlooked simply because you were thrilled you could see thermal. Once you try to get more technical with the camera, as we all have, suddenly these invisible walls come up and hit you. We initially reported a gradient. They said they would work on it. Took ages, then it was fixed. We decided to take a look and see why it even produced a gradient. Mike, myself and a few others determined it was a mix between the lens design and the actual centering of the barrel relative to the image plane. I favored more on the lens housing being to blame due to the size of everything. Just a bad design. Then we noticed that the emissivity simply didn't work. It was basically broken, so instead of fixing it they removed it. I guess the camera is good at knowing exactly what surface it's looking at. Magic. Then the fact that they haven't released the SDK...i mean come on...its the main thing a lot of us have been waiting on.

These problems are moot compared to this one though: sensitivity. If you look at a wall and there isn't a spot that is at least 5 degrees different, you can't see it. The developers refuse to snap the palette to the temperature span of the scene. If you are looking at a scene with about 5 degrees of variation across it (like inside a house), it will look mostly all the same. Maybe in the corners will you see different shades from heat loss (if it exceeds 5 degrees difference) , but seeing studs or the framing 2x4's you see in other camera images isn't happening here. I could sometimes see the framing boards in my ceiling but they were usually hot because I live on the third floor and the crawl space is very warm. But I had to hold the camera just right, with a hot source nearby, in order to force the temperature span to change the palette. That's pretty unacceptable in my book.

I mean yea it's a $200 camera... But it barely meets anyone's needs. I commonly compare it to a spork. It's a utensil that's very affordable and works very well as a spoon, but as a fork not so much. Seek is very affordable and can show you a hot coffee cup, but show it an ice cube and it freaks out.

[eneuro]

Poor software. If you want to see the hottest places - that's enough. But not anymore.

Students with basic C/C++ skils can write his own software with OpenCV accelerated by Nvidia CUDA and easy outperform FLIR postoprocessing capabilities, so many interesting problems can be solved with Seek toy  with custom software

I will publish in right time my research results with circular motion approach which require more processing power and it won't work on crappy phone, but who uses bloody phones for more serious work?

BTW: Do not wait for hacked Seek MCU firmware-nobody will publish unlocked version for well known reasons 

[Trax]

@Aurora
How do you get a used Thermal camera for £150 ?

@efahrenholz
So whats the smallest temperature difference you can see with the Seek Thermal?
what you are describing sounds really bad indeed.


I guess I should than wait for the next gen Flir One that has android support?

[Fraser]

Message removed by Aurora as it takes the thread way off topic.

[Trax]

@Aurora
Are you scouting ebay every day, or whats the trick to find such a good offer?

[Fraser]

Message removed by Aurora as it takes the thread way off topic.

[ebeall]

@efahrenholz, I read your comment on the fragility of the seek, how was yours broken?  I've heard even just a short drop onto carpet can kill random pixels. FWIW, the HTPA82x62 is very robust, it is considered practically impossible to break pixels by acceleration, according to the manufacturer. It also has the lowest temporal standard deviation (highest temporal SNR) of the sensors I tested and of course, needs no shutter.

[ebeall]

@Uho and @Marco, yes the frame rate is limited to 9fps, the readout cannot go any faster without saturating the amplifiers, and there is also a pixel time constant of about 100 milliseconds - you'd get really visible trails on images if the frame rate were to go faster with these pixels.