Yet another cheap thermal imager incoming.. Seek Thermal

[Fraser]

I took some reference images using APK 1.3. Uninstalled 1.3 and ried to install 1.4. No dice. I even downloaded it again, edited its file name etc and it will not install on my MOTO G. No reason given on phone. Sadly this means I cannot provide pictures for a V1.3 Vs V1.4 comparison.

I did check the accuracy of the SEEK measurement by looking into the lens of my PM695. I know that the PM695 has a temperature stabilised microbolometer that runs at 30C. The seek correctly indicates +30C. I also noted some dirt contamination on the microbolometer or optics of my 695 that I will need to investigate. I attach the V1.3 images for interest. They show the PM695, its PSU and the view when looking into its lens. I also include a flat field image.

I can state that the image detail falls far short of what should be present at the cameras stated resolution. The power supply has a ribbed case and the ribs should be clearly visible. They are not. I cannot help wondering if the new 12um microbolometer is severely compromised in some way that effectively destroys detail in an image. My NEC F30S 160x120 pixel camera would wipe the floor with the SEEK so something is definitely up wrong. A very poor lens could also be to blame.

I will take some pictures with my F30S for comparison purposes.

Aurora

[arius]

Just updated my phone (nexus 5) to lollipop - Seek app not working.

[Fraser]

Some pictures from my NEC F30S for comparison to the SEEK images.

The F30S is a 160x120 pixel compact thermal imaging camera that originally cost around $5000.

It is not the most advanced camera in my collection but the nearest to the SEEK that I have.

I deliberately did not optomise the images with the manual mode as that would be unfair to the SEEK as that does not offer such. I actually overexposed the images a little by accident but they are respectable enough for these purposes.

Note the detail in the images. Who says you need MSX to see what you are looking at 

I am not trying to say that the $200 SEEK should be as good as the $5000 NEC F30S but the resolution of the F30S is lower than the SEEK yet it shows far more detail in the images. Hmmm something is very wrong with the SEEK in this respect.

Aurora
 

[eneuro]

@Aurora This nice +/-4*C gradient in attached file is at what estimated time after switch On and what ambient temperature could be when you were playing with this thermal dongle?

Just found another Seek 9mins "gardient demo" to see how strong smoothing is set in this oryginal software, but it looks like those nasty vertical lines are quite stable and it is interesting what is the reason for them-from hexagon black dots pattern one could expect something more noisy looking-no vertical lines patterns, but horizontal ones if any based on raw sensor data send via USB organized in rows and 207th addditional row data

There is no version info-in latest 1.4.x those vertical lines still such visible like in video above on Seek output? 

[efahrenholz]

Aurora,

The lens is quite small, certainly doesn't help sensor at all. We need lots of post processing to get a clear image. But I think we could get images like what flir have, but they are doing some voodoo to get those buttery smooth images.

[Fraser]

@Eneuro,

The SEEK had been on a few minutes but I will repeat the gradient test for 10 minutes and post an image for you. I will also detail the ambient temp.

Aurora

[Fraser]

A warning regarding covering the thermal camera lens with a thermally flat surface as a test for gradient or for compensation purposes. You can get narcissus effect. My F30S certainly could see its own microbolometer radiation.

[eneuro]

I will also detail the ambient temp.

It could be interesting mod to add.. additional external temp sensor close to shutter via drilled hole to monitor inside temperature of this dongle, so probably will do it when it comes if is it possible to add such thermometer inside this thing  to compare with ambient temperature changes and suspected temperature info send in each frame

[Fraser]

Some pictures of the SEEK using the F30S camera to look into its lens and at its case.

[Fraser]

I have taken a flat field image of a wall using the SEEK after a 20 minute warm up period and with an ambient temperature of 20C. The Gradient appears to have settled at 2C on my unit. Not too bad really and within most cameras +-2C tolerance . The reading of the wall temperature is also correct at 20C.

I also attach F30S pictures of the SEEK camera case after the warm up period. This shows the case temperature for reference.

Aurora

[Fraser]

An interesting posting on the E4 teardown thread:

https://www.eevblog.com/forum/testgear/flir-e4-thermal-imaging-camera-teardown/msg550382/#msg550382

Do not underestimate the importance of the optics in a good thermal camera. Also the pictures show just how poor the resolution of the FLIR One really is. No amount of filtering, interpolation and processing can add detail that was never captured. You CAN make a low resolution image look pretty though and that is what has been done in the FLIR One with the help of the MSX that provides context to the thermal blobs !

Aurora

[Fraser]

An interesting video reviewing the SEEK. Moving images certainly look better than still ones.



Aurora

[Fraser]

Thinking about the image noise, a microbolometer is an analogue device that is read by a ADC read out chip. A very important element of a conventional thermal imaging camera design is separation of analogue and digital power supplies to prevent digital noise getting into the sensitive analogue circuits around the microbolometer. The SEEK is a very compact and low component count design. I am wondering whether some of the noise coming out of the core is as a result of analogue supply contamination ?

Mike, if you get a spare moment, any chance of having a look at the core supplies for the presence of noise ?

Aurora

[mikeselectricstuff]

Thinking about the image noise, a microbolometer is an analogue device that is read by a ADC read out chip. A very important element of a conventional thermal imaging camera design is separation of analogue and digital power supplies to prevent digital noise getting into the sensitive analogue circuits around the microbolometer. The SEEK is a very compact and low component count design. I am wondering whether some of the noise coming out of the core is as a result of analogue supply contamination ?

Mike, if you get a spare moment, any chance of having a look at the core supplies for the presence of noise ?

Aurora

Filtering the supply is relatively cheap - they do appear to have multiple supplies, each with their own decoupling, fed from a 4 way resistor network.
Also, with a single clock source, noise from the processor etc. is mostly going to be synchronous rather than random

[efahrenholz]

Woah. That photo of the beer crate through the e4 looked like garbage (even hacked to full resolution.) It looks quite a bit like our seek images. I am keeping in mind it was auto scaled, so of course noise will dominate in a situation where not enough thermal variance exists. Seek doesn't allow auto scaling of the scene. It prevents the palette from compressing beyond a certain span of temperature. So when we pan around a room with little to no difference in temperature, we only see solid color with an occasional outline of an object.

Miguevp is probably the only one getting decent images from the camera. Look back a few pages to his post. He has some very clean router images. Be also has some manual control sliders to set the range, so this will allow us to truly see how bad the noise of the sensor is. Obviously the lens is undersized, so post processing is needed. But whatever works for a pretty picture.

[tomas123]

I looked around for a scene without temperature differences and found the beer crate.
So I can show the differences in noise of Exx / Ex / One. (the scale is only 1 Kelvin)

I think, with real cold and hot spots the noise of SEEK is greater as the Flir One
... but I can't verify it.

I'm confused. The Lepton datasheet says 50mK, but I can't confirm it for the Flir One.
If Flir add noise with noise generator, than I can verify it with stacking of some images.

[marshallh]

Hi Everyone,

The Android orders are almost completely caught up and should hopefully be out of backorder within a few days. iOS cameras are currently on hold as we work through some additional, unanticipated requests from Apple. We are members of the Apple MFi Program which is the only authorized means to manufacture and distribute products for use with iOS. We are in direct contact with the MFi Program Team to provide all the necessary information required by Apple for compliance with program requirements. After being previously approved, our program plan is under review by Apple. We are eagerly working with them to complete the process as quickly as possible.

-The Seek Team

[eneuro]

Moving images certainly look better than still ones.

This is may be result of different human eye sensitivity to different color, so a few degrees temperature gradient can be easier notice at the end of Iron palette than at the beginning, where there is more blue channel less sensitive to our eye 
So, gradient can be present in images at different temperatures and the same, but more difficult to notice, especially when I'm not sure-they use 8bit (256) colors palettes ONLY? 
Can not prove this while in their software LUT directory there is complete garbage, and there is no classic LUTs but coffe cup with LUT applied 
I use 16bit Iron LUT so it will be interesting to see what difference it makes to output thermal LUT mapped RGB image.

Anyway it looks like inside this Seek dongle thing is quite hot >35*C around sensor, so this metal lens holder is cooler in typical 20*C-25*C ambient temperature and with hole in one side when shutter operates and pumps air at each calibration event, another side of sensor must be better cooled by lens holder and maybe this creates as mentioned earlier those gradients on calibration frames too.

Does lens used in Seek reverse image (flip) ?
It could be interesting to find out which corners looks more hot/cold on real hardware sensor block relative to the side shutter operates.
So, lets say we look at this Seek sensor from the top, we have shutter on lower 156 pixels side right where shutter operates and question is which part of the sensor senses this hotter gradient temperature? 
Can't do it now while haven't got this Seek in hands yet, however histograms from a few calibration frames from raw sensor data provided there, really looked like this calibration frame image standard deviation of pixel values is going down and values itself lower in time, so it canexplain that this gradient improves a litle bit and might stabilize at those +/-2*C at some conditions, but it looks very bad on output images even if it is temperature difference lower than 5*C   

[Fraser]

When thinking about the lens used in the SEEK and E4 we also need to recognise that they are nor Germanium but rather a compromise specialist glass Chalcogenide. The glass offers the ad advantage of lower cost but it does have image quality issues.

[jaybeez]

Obituary of Seek Thermal's President and CEO, Robert Lauren Acker. He passed away while jogging on November 10th 2014.

http://obits.oregonlive.com/obituaries/oregon/obituary.aspx?n=robert-lauren-acker&pid=173173002

[heavybarrel]

Hi Everyone,

The Android orders are almost completely caught up and should hopefully be out of backorder within a few days. iOS cameras are currently on hold as we work through some additional, unanticipated requests from Apple. We are members of the Apple MFi Program which is the only authorized means to manufacture and distribute products for use with iOS. We are in direct contact with the MFi Program Team to provide all the necessary information required by Apple for compliance with program requirements. After being previously approved, our program plan is under review by Apple. We are eagerly working with them to complete the process as quickly as possible.

-The Seek Team

My account was charged today for an IOS camera i purchased in the first few days of Seeks sales. I also saw their update,
hopefully they are shipping since they are now billing. Lets all cross our fingers!!

[mikeselectricstuff]

Just received a second unit - minimal gradient from cold, increasing to about 3-4 deg.C after 5 mins or so
When I get time I'll experiment with changing the temp of the lens holder.

[efahrenholz]

Holy crap.

Yes, Robert Acker is dead. He was found along a jogging trail. I don't know what to say. I wouldn't even begin to know how to proceed further with this topic...its extremely...unusual.

I wonder what this means for Seek. I surely hope someone will be able to take his place, but this must be a shock to everyone in the company and his family. At least he was able to see his idea take flight and be successful. That's the most important thing. He did something amazing.

[Fraser]

My sincere condolences to Roberts family and friends. A real shock to hear this very sad news. Aurora.

[cynfab]

I hope everyone here will join me in expressing sincere condolences, and sorrow to the family, friends and co-workers of Robert Acker.

Seek will endure and show that a dream can come true and make a difference in today's world.

The loss of a founder & CEO can be nothing but a shock to the company, I hope all who have ordered cameras will cut Seek a little slack, this can't be easy for them.