Seek Thermal: Hardware modifications to improve image quality

[frenky]

I have found out that i3 ThemalExpert won't be priced at 500$ but at 820$. (At this time I can't reveal the source or any more details...)

That price is a deal breaker for me so I decided to make some extensive modification to my SeekThemal module to improve it as much as possible.

The first mod will be to make a external shutter from a regular pcb. I will paint the copper side with matte black, since that worked good for me in the past.
And the second mod will be cooling the back of the sensor with the Peltier module. (I might even cut pcb in half to cool only the sensor part, but that is risky.)

Both mods should show a significant impact on the image quality.

External shutter should provide a much better isothermal plane so when we subtract pixel values we should get much cleaner image.
And the cooling of the sensor should prevent pixels from drifting too much.

The mechanism for shutter will look something like this:



Animation:


My progress so far:
I took a cheap solenoid, removed the core pin, and glued the magnet in the ink cartridge of a pen:

This is working good; when I apply voltage (5V) the cartridge with a magnet ejects out of a solenoid.



Next step is to glue on a pcb painted matte black...

P.S.:
Big thanks to Dave for giving us a subforum where we can share knowledge and help each other.

[frenky]

It's not pretty but it works.

[miguelvp]

Nice

[frenky]

Tnx.
It's a lot of fun making something mechanical since I mostly work with software...

[frenky]

This one did not work well, because friction between bearings and shaft is to big, but I thought I'd show it any way.
(I did lubricate it with lithium grease but it did not help.)

[Fraser]

Hi,

Whilst I appreciate what you are trying to do. Your current solution is somewhat different to the approach taken by the thermal camera industry. That is not to say your design will not work..... but most FFC shutter designs have low thermal and physical mass. That way they do a good job whilst requiring a light duty, low current actuator.

May I suggest you take a look at commercial actuators for a filter or lens as used in some CCTV cameras. Such actuators insert and remove the IR filter for day/night cameras.

I have seen these IR filter actuators on eBay and they are very cheap as they are designed for PCB cameras. You can even buy a complete Day/Night PCB camera and extract the actuator assembly. Then you just need to replace the IR filter with a suitable material for an FFC shutter flag.

I will add some links to the filter actuators shortly.

Fraser

[Fraser]

Here you go.... cheap as chips and a neat design.

http://www.ebay.co.uk/itm/171912913239?_trksid=p2060353.m1438.l2649&var=470829576710&ssPageName=STRK%3AMEBIDX%3AIT

http://www.ebay.co.uk/itm/401003344143?_trksid=p2060353.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

http://www.ebay.co.uk/itm/321492559019?_trksid=p2060353.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

http://www.ebay.co.uk/itm/262234538296?_trksid=p2060353.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

http://www.ebay.co.uk/itm/262234538299?_trksid=p2060353.m1438.l2649&ssPageName=STRK%3AMEBIDX%3AIT

You could probably get away with just spraying the glas IR filter with matt black paint, so no need to fit an IR opaque flag.

[frenky]

Big thanks for this. 
I think you are the top expert on this forum when it comes to thermal cameras inner workings so I really appreciate your input.

So does this actuators work like solenoid? While coil is energised the shutter is in position 1 and when you cut the power it goes to position 2?
Or it's more like latching relay? On every pulse it will toggle position?

The price is right  so I will order some from china, but continue working on my own design so I have more options...

[Fraser]

Hi Frenky,

The filter actuators can be either a toggle action or momentary, depending upon the particular design. It may be worth asking the seller before buying. Even if a toggle action, these units can usually be dismantled for modification. I regret I have no more data on the actuators.

They are certainly cheap enough for some experimentation.

Something to also consider when building an FFC shutter is any vibration caused by the actuator mechanism. The CCTV actuators cause little or no vibration when they operate.

Fraser

[frenky]

Tnx for the info. I'll do a video of received actuators when they arrive.

[frenky]

Today I got the actuators. I tried connecting the first to 5V and nothing happened... Then to 10V nothing... 20V, 30V, 40V, and then it started to smell... 

Went back to 5V and swapped the leads (red to - and black to +) and "click" it changed the filter. Swapped leads again and "click".
So every time you want to switch filters you must change the voltage polarity.

The smallest actuator is just perfect for Seek. I will be swapping its filter for a painted copper/alu plate in the near future.





P.S.
Fraser big thanks for pointing me into this direction.

[Fraser]

It was a pleasure to help.

As the flag is at ambient temperature you might not need to remove the filter. Just paint it.

Do all do the actuators toggle on supply polarity ? Maybe the flag mechanism can be modified ?

Looking forward to further updates on your project.

Fraser

[frenky]

Mechanism can slide out of the bulk housing so the module is even more compact.


Internals of one of the other actuators:


EDIT: Yes all actuators need polarity swap to change filter.

[Fraser]

Thanks for the pictures.

If you add s small tension Spring you can hold the flag in one position with a single polarity pulling the flag into its second position. If there is a friction plate or 'bump' to provide a 'hold' in each position, remove it so that the flag can mode from one position to another easily.

The alternative would be an electronic toggle drive circuit but I prefer the hardware route myself. The only part to source would be a very light tension Spring.

Fraser

[frenky]

Tnx. I might try it in two different ways.

1. Add a spring to make it go back to the first position when the power is cut. Just like you suggested.
Then remove internal shutter and solder its wires to this new shutter.
If the resistance of internal coil is much higher then use transistor or mosfet to drive external coil. (So that I don't damage the circuit...)

2. Leave internals of the Seek intact and do the magic with the software (I prefer that).
The sequence of frames is like this: precalibration frame, calibration frame (image of the internal shutter), first thermal frame, second thermal frame...

When the pre-calibration frame shows up I would send signal trough serial port to attiny to close the external shutter.
Then keep it closed until the calibration frame and the first thermal frame are saved. Then send command to open external shutter.
This way I can decide on the fly if I want to use captured image frame as an external reference or just use calibration frame of the internal shutter.
Or even use both combined in some weird way. 

[frenky]

Two actuators connected to L293D and arduino...   

[Fraser]

Looking good. Nice snappy action on the actuators.

Fraser

[frenky]

Good news... I've connected SeekThermal to RPI2 loaded with Win 10 IoT core and out of the box it it recognizances it as "PIR206 Thermal Camera".
I believe this is a very good sign for putting existing SeekOFix code on RPI:

I a few months:

[miguelvp]

That's pretty cool, or hot depends on how you seek it

For one thing, it kind of shows that Win10 IoT is a viable solution for some applications on a small form factor.
And another thing it shows that you rock!
 

[frenky]

Tnx for nice words. 

Today I've put together a quick & ugly setup for external shutter just to see if it works.
From the code I send command via serial port to arduino to close or open shutter with L293D. I close it when frame ID1 comes and open it after first ID3.
I also made a few changes in code to always use external calibration from first frame ID3.

Setup:


Image of the reflectors (flood light) on the ceiling with internal shutter calibration:


And with external shutter:


As you can see there is less noise and better real temperature representation.
While Seek did sort-of fix that gradient it still shows up at low temp span. (Clearly visible on the first image bottom-right)

[frenky]

I'm not 100% sure what is going on but now I'm getting a lot cleaner images (with external shutter):

This was the image from few days ago:

And this is from today:




Most likely the temperature of the sensor influences the quality of the thermal images. I will test this by saving every 20th frame for a half an hour.

[Fraser]

Indeed, a significant reduction in the noise content. Had my SEEK produced such pictures I would have been more than happy with it.

Fraser

[frenky]

True...
This observation gives me hope that when I'll stick Peltier element on the back of the Seek the images will be low noise all the time.

[Fraser]

Hmmmmm, I presume you intend to stabilise the microbolometer at 30C and not try to cool it to a low temperature ? The designer will have intended the calibration tables to be valid at and around 30C, not at a low temperature of say 10C. You could try it but you may have issues with calibration and NUC. The FLIR E4 is only 'happy' once it's die temperature monitoring sensors indicate around 30C. It becomes less accurate as the die temperature moves away from that calibration temperature. My PM series cameras are Peltier element temperature stabilised at 30C.

As for why a microbolometer would be noisy at one temperature, and less so at another similar temperature, I cannot say. Microbolometers are, after all, just an array of VOX thermistors. Sure the VOX thermistor will have its own thermal noise but I would expect significant levels of cooling to reduce this significantly. FLIR have spent a lot of time improving the design of their microbolometers, and the processing software, in order to reduce noise. If cooling the microbolometer had a significant effect on the noise levels I would have expected them to cool the microbolometer long ago. Even my PM695 has a  temperature stabilisation Peltier element producing 30C at the die. The SEEK camera may have thermal issues however, such as an ROIC that produces noise that varies with temperature ?

Certainly worth investigating but I suggest you do not spend to much money on this avenue of investigation.

I am suspicious of the read out electronics and software processing in the SEEK. I can offer no explanation for why you suddenly had a decent lower noise image on your unit. Very interesting though.

Fraser

[Fraser]

I just had a thought.........

The microbolometer is an analogue device and it is essential that it, and the ADC has a nice clean power supply. The SEEK is a very compact device built with the minimum of components to maintain a low BoM. It would be well worth looking very carefully at the quality of the supply rails that are feeding the analogue sections of the camera. Any noise on these will be digitised and become part of the image. Unlike a boost converter frequency breakthrough which would likely produce banding, look for random noise. Such would be superimposed on the microbolometer data as random image noise. I can believe that such noise could vary with temperature or even the source of power for the camera.

Fraser