Another smartphone thermal imager from i3

[schlafli]

How secure is the USB connection? Does it fall out or wobble at all?

The Opgal one seems like it has the edge there with the snap-on aluminum bracket.

It's sits flush with my phone when plugged in and is quite sturdy. With a case on the phone it might be a bit difficult.

When I bought the camera i3 system threw in the accessory kit for free, so I did get the nice snap-on bracket too. The bracket uses a standard tripod mount and comes with a mini-tripod and monopod. It's a bi-stable bracket that uses magnets to hold it in either position. I think they could have made it much more useful if it was a friction hinge that let you angle it any position.





Another thing I've been meaning to mention: The thermal expert comes with a 6.8mm f1.3 lens while the therm-app TH/optional lens is f1.4.

[schlafli]

Mini rant time. 

I've been trying for the past few weeks to get access to the SDK without any luck. I3 seem to have changed their tune recently and are not offering the SDK or Windows drivers unless you get a special "development edition" thermal expert. An SDK and Windows support were one of the main reasons I bought the Thermal-expert over the Therm-app.

It's quite annoying since that means I can't actually get any raw data from the camera. The Android app does have the functionality to export thermal data but that simply gives a normalized unsigned 8-bit image with a temperature lookup table.

Another reason I was looking forwards to windows support is because the performance on my phone can be a bit sketchy and the battery life is pretty poor. (this is of course my phones fault)

I'm going to keep pestering them and hopefully they'll see the light, especially since all their competitors (Seek, Flir, Opgal) do provide an SDK. If not, I'll have at least learned a lesson about being explicit about clarifying features before I buy something...

[Fraser]

Sorry to hear this. Ensure that you tell them of your membership and postings on this forum. (Provide the URL to this thread). Advise them that you intended to use to SDK to show this community the benefits of their camera over others and its excellent adaptability. I would hope that a manufacturer of a new product would want their unit discussed in a positive light on forums such as this. Deciding to not provide the SDK to early adopters is a recipe for negative views of the OEM and the product. A tad silly me thinks.

Keep trying. Offer to sign an NDA etc, they may change their mind.

Hope you get a good result.

Fraser

[schlafli]

Sorry to hear this. Ensure that you tell them of your membership and postings on this forum. (Provide the URL to this thread). Advise them that you intended to use to SDK to show this community the benefits of their camera over others and its excellent adaptability. I would hope that a manufacturer of a new product would want their unit discussed in a positive light on forums such as this. Deciding to not provide the SDK to early adopters is a recipe for negative views of the OEM and the product. A tad silly me thinks.

Keep trying. Offer to sign an NDA etc, they may change their mind.

Hope you get a good result.

Fraser

Thanks for the tips 

I've already sent them an email today pointing out the competitors' SDK. If they don't budge I'll try the things you mentioned.

-Michael

[schlafli]

Good news! Looks like I may be getting access to a version of the SDK after all. It may be a while yet but once I get it, expect loads of raw images

I don't know if this is just me being naive, but I assume there would be a lot of people interested in getting one of these cameras if i3 system took an approach similar to Flir with their F1G2. I.e., push the development platform aspect of the device and release an SDK for Linux/Windows/Android. Thoughts?

Update: I should be getting the Windows version of the SDK at the start of next week. I'll need to dig through tomas123's posts on creating Flir tools compatible images   

[alanpro123@hotmail.com]

How do you purchase this camera? I contacted them online and recieved not asnwer.

[schlafli]

How do you purchase this camera? I contacted them online and recieved not asnwer.

I initially tried the form on the website a few months ago but got no response, so I'd avoid that. I then contacted them at the email address show here: http://www.i3-thermalexpert.com/contact

They're pretty good at responding to email quickly (usually within a day).


The general procedure was something like this:

• Tell them you're interested and want the prices/brochure
• Once you know what you want, send them an email asking for a proforma invoice for the items. Make sure to include name, phone number and shipping information.
• Once you get the invoice and are happy with it: print it, sign it, scan it and email it to them
• Transfer the money in the currency specified on the invoice. In my case it was USD.
• Wait 1 week
• Play with your new toy

This is important so I'll mention it again: When you transfer the money, make sure that they receive the amount specified in the currency specified.  I used ukforex and was very happy with the cost/speed. I ended up paying £10 over the mid-market rate and i3 had the money in their account less than 10h after I opened a new account on UKForex.
For comparison, my bank had an estimate of 2 weeks for the transfer

[encryptededdy]

3 updates to the Thermal-Expert.

• They now take PayPal, so you can pay with a credit card and it's no longer a pain in the ass to deal with the international bank transfers
• Linux SDK is available as well
• There is a "Thermal-Expert Pro" which is calibrated up to 250C instead of the normal Thermal-Expert's 150C

Oh, and this is the big one - there's a new Thermal-Expert V1 (standing for VGA-1, the current Thermal-Expert has been renamed to the Q1, for QVGA-1)

It uses a 640x480 17um sensor and has a 30Hz option! Price is currently unknown, I've emailed them for more information.

If this is <$3000 I'm pretty sure it's going to be the cheapest 640x480 thermal camera out there by a huuuuggeee margin, especially if we consider someone like FLIR's cheapest 640x480 camera, the T620, costs $20,000+ (yes this has way less features but fundamentally the sensor won't be too different)

[frenky]

Nice...
640x480 with 19mm f/1.0 lens would be really awesome.

This should really shake the TI market. I just hope Flir does not buy them out...

p.s.

Your link does not work, so here is the correct one: http://www.i3-thermalexpert.com/thermal-expert-v1/

[frenky]

I believe Flir Vue 640 is the cheapest (3000$) VGA thermal module.
http://1uas.com/FLIR-VUE-640-Slow-Version?search=flir%20vue&sort=p.price&order=ASC

[encryptededdy]

Ah yes, that was it. I remember mentioning it in some other thread. I think the problem with the Vue is that it's analog video out only. The Vue Pro solves this, however it doesn't offer any sort of radiometric data / temperature measurement.

[Ben321]

Also not to be confused with the FLIR i3 (what's with these name collisions?)

A 100 maximum temperature is rather low.

Radiometric (temperature measurement via LWIR detection) capability probably maxes out at 100degC (212degF), just like with the FLIR One. However its ability to see LWIR radiation probably goes well above that. While a radiometric app that displays temperature probably will artificially limit the range by indicating that it maxed out at 100degC, a 3rd party app that can access the raw pixel data from the sensor probably would be able to detect LWIR radiation from much hotter sources, even if it couldn't accurately assign a temperature to it, as the per-camera calibration values needed to calculate this are probably proprietary, and the only app that will have access to these values will be the official one, which will NOT allow direct access to the raw data by the end user, just like with the FLIR One SDK sample app. However, once temperature is above about 1000degF, then sensor's output may overload the ADC (analog to digital converter) causing it to wrap around back to digital output of 0, as I found it does on the FLIR One when accessing the raw 14bitsperpixel data through the FLIR One SDK sample app (which causes some of the hottest objects, like the heating element on a stove to actually appear colder than room temperature, and there is a sudden change from white to black on the image where this transition occurs).

[Ben321]

So this is called an i3 thermal imager? Is there any relationship between this device and the FLIR i3, i5, and i7 devices? Did FLIR have any hand in making this device?

[Ben321]

Also why does this have a price twice that of the Seek, even though it has the same resolution of 384x288?

[frenky]

i3 ThermalExpert is made by Korean company i3system. It has no connections with Flir i3 thermal camera.
It also does not have the same reasolution as Seek which has 206x156 themal sensor.

It's closest competitor is Therm-App which has the same resolution but is a bit more expensive.
http://therm-app.com/product/therm-app-device-with-19mm-lens/

[Bill W]

Ben321 wrote:

Radiometric (temperature measurement via LWIR detection) capability probably maxes out at 100degC (212degF), just like with the FLIR One. However its ability to see LWIR radiation probably goes well above that.

Hotter things simply give off more LWIR, not some different sort of LWIR.  See Wikipedia on Planck curves and Black Bodies.  The problem is sensor saturation.

While a radiometric app that displays temperature probably will artificially limit the range by indicating that it maxed out at 100degC, a 3rd party app that can access the raw pixel data from the sensor probably would be able to detect LWIR radiation from much hotter sources, even if it couldn't accurately assign a temperature to it, as the per-camera calibration values needed to calculate this are probably proprietary, and the only app that will have access to these values will be the official one, which will NOT allow direct access to the raw data by the end user, just like with the FLIR One SDK sample app.

Unlikely. To get a good noise free image the scene dynamic range will be quite limited.  A large dynamic range setting would push the NeTD/MDTD up to 500mK or more.  There is nothing proprietary about the calibration, just whether the manufacturer can be bothered plotting the curve all the way.  Thermal sensors are linear in energy, so it is just maths to convert that scale to temperature.

However, once temperature is above about 1000degF, then sensor's output may overload the ADC (analog to digital converter) causing it to wrap around back to digital output of 0, as I found it does on the FLIR One when accessing the raw 14bitsperpixel data through the FLIR One SDK sample app

Well before that I expect, more like 150°C as I suggest above.  Higher temperature ranges need special low gain modes in the sensor to avoid saturation.  These may have their own factory calibrations or might be simply mathematical relations (like 1/8th gain and sort out offsets with a shutter calibration).

What you saw on the FLIR is either a palette that is black = overload or else simply bad coding.  It is not the the ADC, an ADC will sit at 16383 when overloaded but when the gain/offset corrections are applied to 16383 that may wrap back if the code is not written with very hot things in mind. Some of the FLIR Twitter images of the K2 fire camera at the FDIC show the black-very hot artefact too, with a flame measured at >120°C.

[Ben321]

Ben321 wrote:

Radiometric (temperature measurement via LWIR detection) capability probably maxes out at 100degC (212degF), just like with the FLIR One. However its ability to see LWIR radiation probably goes well above that.

Hotter things simply give off more LWIR, not some different sort of LWIR.  See Wikipedia on Planck curves and Black Bodies.  The problem is sensor saturation.

While a radiometric app that displays temperature probably will artificially limit the range by indicating that it maxed out at 100degC, a 3rd party app that can access the raw pixel data from the sensor probably would be able to detect LWIR radiation from much hotter sources, even if it couldn't accurately assign a temperature to it, as the per-camera calibration values needed to calculate this are probably proprietary, and the only app that will have access to these values will be the official one, which will NOT allow direct access to the raw data by the end user, just like with the FLIR One SDK sample app.

Unlikely. To get a good noise free image the scene dynamic range will be quite limited.  A large dynamic range setting would push the NeTD/MDTD up to 500mK or more.  There is nothing proprietary about the calibration, just whether the manufacturer can be bothered plotting the curve all the way.  Thermal sensors are linear in energy, so it is just maths to convert that scale to temperature.

However, once temperature is above about 1000degF, then sensor's output may overload the ADC (analog to digital converter) causing it to wrap around back to digital output of 0, as I found it does on the FLIR One when accessing the raw 14bitsperpixel data through the FLIR One SDK sample app

Well before that I expect, more like 150°C as I suggest above.  Higher temperature ranges need special low gain modes in the sensor to avoid saturation.  These may have their own factory calibrations or might be simply mathematical relations (like 1/8th gain and sort out offsets with a shutter calibration).

What you saw on the FLIR is either a palette that is black = overload or else simply bad coding.  It is not the the ADC, an ADC will sit at 16383 when overloaded but when the gain/offset corrections are applied to 16383 that may wrap back if the code is not written with very hot things in mind. Some of the FLIR Twitter images of the K2 fire camera at the FDIC show the black-very hot artefact too, with a flame measured at >120°C.

I'm pretty sure there are no corrections applied to the RAW image. In the FLIR One SDK sample app, the 14bit raw image is the raw linear image that represents the detected radiant energy, not the conversion to temperature values. Yet this raw image has the same problem I just described.

[encryptededdy]

Got a response from i3system - the 640x480 unit will be approx $3500 (not final) and the 30hz will be slightly more.

19mm f/1.0 lens is included.

This is slightly less than the price of a FLIR Vue Pro, however the thermal expert v1 is radiometric / measures temperatures.

[Ben321]

Got a response from i3system - the 640x480 unit will be approx $3500 (not final) and the 30hz will be slightly more.

19mm f/1.0 lens is included.

This is slightly less than the price of a FLIR Vue Pro, however the thermal expert v1 is radiometric / measures temperatures.

$3500?! I thought the first post in this thread said that this smart phone only $500 (for a 320x240 version). Doubling the horizontal and vertical resolution (4x the number of pixels = 4x the price) only gets you up to $2000. There's no reason that this should cost more, unless of course this company had some shady "under the table" deal with FLIR to not compete with FLIR, and so i3systems had to make their price high enough that buying from FLIR would look like a bargain compared to buying from i3systems, and FLIR promised to pay i3systems a massive amount of kickbacks. In other words, price fixing.

[frenky]

You really should start reading posts that you quote...
In the first post you have resolution 384 x 288 not 320x240. 500$ price was just a rumour.

On the third page you have an actual price:  $880  ($820 + $60 shipping)

[Bill W]

Ben 321 wrote:

I'm pretty sure there are no corrections applied to the RAW image. In the FLIR One SDK sample app, the 14bit raw image is the raw linear image that represents the detected radiant energy, not the conversion to temperature values. Yet this raw image has the same problem I just described.

There are another set of corrections to get from 'digitised sensor video' to the 'linear in energy' image that you are describing as 'RAW', that is before you get to the simple curve to get to temperature.  I suspect maths overflow in these calculations for correcting the pixel gains and offsets that could cause FLIR's black hole, probably due to going over 32 bits.


Bill
www.fire-tics.co.uk

[Ben321]

Ben 321 wrote:

I'm pretty sure there are no corrections applied to the RAW image. In the FLIR One SDK sample app, the 14bit raw image is the raw linear image that represents the detected radiant energy, not the conversion to temperature values. Yet this raw image has the same problem I just described.

There are another set of corrections to get from 'digitised sensor video' to the 'linear in energy' image that you are describing as 'RAW', that is before you get to the simple curve to get to temperature.  I suspect maths overflow in these calculations for correcting the pixel gains and offsets that could cause FLIR's black hole, probably due to going over 32 bits.


Bill
www.fire-tics.co.uk

So there's no way to get the true raw digitzed analog signal from the FLIR One's ADC into your smartphone?

[drdespair]

Some advice needed guys, with Themapp being on discount now a kit with two lenses between Thermapp and Thermal Expert are almost identicaly priced at about 1200 USD, so which one should I pick? Lookintg at teardowns the Thermapp looks better constructed, there is some difference in the tempreture ranges they are calibrated , dont know if there is much difference in sensetivity. The Thermapp app looks better, seem to be able to set a palate range in the app. Any thoughts would be welcome! Primary use will be for outdoors.

[encryptededdy]

Some advice needed guys, with Themapp being on discount now a kit with two lenses between Thermapp and Thermal Expert are almost identicaly priced at about 1200 USD, so which one should I pick? Lookintg at teardowns the Thermapp looks better constructed, there is some difference in the tempreture ranges they are calibrated , dont know if there is much difference in sensetivity. The Thermapp app looks better, seem to be able to set a palate range in the app. Any thoughts would be welcome! Primary use will be for outdoors.

I assume you're purchasing the 9Hz version. If so, the intense rolling shutter effect on the Therm-App 9Hz may be annoying for outdoor use - check the last videos I posted on the Therm-App thread for an idea of what I mean. From what I can tell the ThermalExpert reads out the full sensor every frame which eliminates the annoying rolling shutter.

If you're using this for outdoor use then I would still highly recommend the Therm-App 25Hz. The extra frame rate really makes a huge difference.

[drdespair]

Yeah, I have seen the rolling shutter on the Thermapp, was not sure if the Thermal Expert had the same. Unfortunately the 25Hz is not in my price range, so will probably go with the Thermal Expert. Thank you!