Cheap new basic thermal camera on eBay - Multicomp / Uni-T + teardown

[Fraser]

I just found this guide to Bilinear (with mention of Bicubic) interpolation and, as I thought, my description, above, of interpolation is wayyyy over simplified 

https://learn.adafruit.com/improved-amg8833-pygamer-thermal-camera/the-1-2-3s-of-bilinear-interpolation

[jeremy]

Perhaps what they do is some sort of motion-based superresolution interpolation. Basically, it uses multiple images with very slight movement of the camera to interpolate far greater detail. This paper covers one type of it pretty well (Serious Maths(TM) warning), although there is an example "pretty picture" example reconstruction from a FLIR camera on page 23: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.2.7447&rep=rep1&type=pdf

[Fraser]

That is an interesting paper. Thank you. I fear that the Uni-T UTi80P that this Multicomp camera truly is, would not employ super resolution as the price point is too low and the requirement for a decent processor would drive them towards simpler interpolation techniques. Getting super resolution on much more expensive cameras remains a luxury, or additional cost option. A pity as that mode has much to offer. My Agilent/Keysight 160 x 120 pixel camera has Super resolution on it to improve its imagery. That camera was made by NEC Avio who make some very nice cameras.

I really should look deeper into recognising the sort of image enhancement and filtering that modern thermal cameras are using. Sadly manufacturers tend to hide such fine detail behind fancy names and give little detail of what is actually done to the raw image. Clever people on this forum can recognise certain tell tale signs of filtering, edge enhancement and interpolation, but not me

Fraser

[Vipitis]

I am currently doing a deep learning course and might finish it was a homework project. There are already three ideas on my mind but only one of them is Pseudo NLP focussed. Doing image upscaling for thermal imaging based on compressed sensing doing to the sparsity of microbolometers might be an idea. However I think it has been done before. I have seen published research on it a while ago, maybe I dig them up in a "scientific publications on thermal imaging" thread I wanted to start a few times.

2x "super resolution" and even 4x will usually look alright, since that's what we are used to with visible light color cameras. They prototypically use a 2x2 CFA (color filter array) with a R-G;G-B pattern. Meaning 75% of R and 75% of B pixels are interpolated (=4x super resolution in pixels, 2x in resolution). Non DRS microbolometers have an effective pixel area smaller than then pixel pitch. It's not quite 25% but it's not more than 85% either. So you could easily make up values for the "missing" space in between. So thermal cameras should always give you more pixels than you see. In visible light you have some overlap and correlation with R G and B, but it's usually not looked at(for highlight recovery it is for example).

Upscaling and downscaling a digital image has been a task for a long time. There is plenty of different approaches (recently learning based stuff like Topaz for example), but it's all math. Doing it based on the physics of a sensor and maybe the point spread function of the optics... Is only something you find in scientific research for like space telescopes or Mars rovers.

Doing it based on sparsity might work. But it requires a random sample, so your array won't do it. But the odd dead pixel would. And maybe even randomly angled lines was like a wavelet.

[Fraser]

A small epilogue to my Multicomp thermal camer purchase.....

This diminutive camera has proven to be an excellent addition to my tool kit. I have been working on a number of old laptops that I am refurbishing for my lab use. Some of these were bought faulty and then repaired. I found the little Multicomp camera to be soooo convenient and quick to use when wishing to see what thermal activity, if any, there was on the motherboard. You grab the camera from the tool kit, push the power button and have a thermal image in seconds  This little camera helped me to diagnose three laptop motherboards that were suffering from failed DC-DC converters and failed bypass MLCC's. I would normally have used my FLIR E60+ camera, and indeed I would have, if the Multicomp camera had not been up to the task. The E60+ does produce a far superior thermal image but I was interested to see if the budget Multicomp camera was a useful tool or more of a toy. I am pleased to say that it has more than proven its worth to me in real world use.

The camera does benefit from the use of a supplementary close-up lens when being used on PCB's. The common 12mm diameter ZnSe CO2 laser cutter lenses are adequate for the task. I purchased some accessory lenses for mobile phones as a potential lens mounting system but I have yet to take that project forward. The phone kits 'Macro' lens looks perfect for the role as its lens holder is 13mm diameter. The lenses are attached to the phone, or the Multicomp thermal camera, using a ferrous ring attached to the phone/camera and a magnetic ring on the lens. Whilst this is a very neat lens mounting system, it has a potential issue for use on modern mobile phones and even thermal cameras. That issue is the use of a magnet as modern auto focus and stabilized cameras use magnetic flux to function and the FLIR Lepton core uses a magnetic FFC flag acuator. I need to see whether the strong magnet on the supplementary lens effects the FFC solenoid operation once the magnet is attached to its 'keeper' ring.
We shall see. These supplementary lenses cost less than £5 for a set so are very reasonably priced. The lens barrels are aluminium and the lenses are retained by screw mounts so may be easily dismantled for fitment of another (ZnSe) lens.

I bought my mobile phone sets from this UK seller but he has only gold coloured lenses remaining as I bought the last two black sets.

https://www.ebay.co.uk/itm/162554583353


Fraser

[Fraser]

Just tested the mobile phone lenses on the Multicomp camera.

1. The ferrous ring inner diameter does not fit around the Multicomp cameras raised lip surrounding the lens window so a double sided adhesive thin foam mount would be needed for the attachement as that is forgiving of the difference in height caused by the raised lip. Alterenatively the inner diameter of the ferrous ring could be increased slightly to clear the raised lip.

2. The 'Macro' lens causes vignetting due to its small diameter..... pity. The much larger 'Fisheye' lens has a large flare on the lens barrel so only minor vignetting occurs. This is the problem when mounting a suplemental lens in front of a wide angle lens. The lepton has a 50 Degree HFOV which mean that it can see the supplemental lens barrel unless the supplemental lens has a much larger diameter mount and barrel.

3. The supplemental lens mouting ring magnet does not adversely effect the Lepton FFC solenoid operation.

Fraser

[Fraser]

I am looking at another supplemental lens mounting system to see if it meets my needs. I have ordered the Kodak hone lens kit that comes with a novel retaining clip and two nice large diameter lenses….Wide angle and Fisheye. This lens system may be easily adaptable to holding a ZnSe lens that is large enough to avoids vignetting. The kit cost me less than £6 so not terribly expensive.
I have looked at, and own, the ‘clothes peg’ style of phone lens holder but I do not like the pressure that the clip puts on the screen and the thickness of the camera is limited by the clips maximum opening capacity.

I will advise on the Kodak clip system when it arrives. I need something quick and easy to put onto the Multicomp camera. Refined appearance is not an issue 

Fraser

[electr_peter]

FYI - MP780429 Multicomp Pro thermal imager is currently on eBay.com selling for 55GBP + ~22GBP shipping to EU + ~18GBP taxes = 95GBP total. Seller is based in UK. 17 units left. And it's gone.

This not endorsement, I am not related to the seller or eBay. To me seller seems legitimate plus I have eBay/credit card protection so it is worth a look.

Link - MP780429 Multicomp Pro thermal imager

[Fraser]

Brilliant deal 

Thanks for highlighting it electr_peter

There were a couple that sold for £40 earlier in the month. I have to wonder whether Farnell have been disposing of stock at their trade outlets ? Good news for us though. £55 for one of these little gems is brilliant value for money (if you can accept the 80 x 60 pixel resolution imagery in your application).

I have ordered another 

Fraser

[Fraser]

Just spoke to the seller. These units are listed as New(other) but he confirmed that they are new and unused

I will be permanently mounting one of mine on my BGA Rework station to monitor preheating temperatures and thermal coverage. It may also help with the main IR heating phase but only when the IR illuminator is in its ‘off’ period. These little units could be considered an IR thermometer ‘on steroids’  You can see exactly what you are measuring on the PCB.

Fraser

[Fraser]

When you consider what £54 buys you on eBay in the way of a new thermal camera at the moment…….

https://www.ebay.co.uk/itm/294260756039

A whole 8x8 pixels ! It puts this Multicomp deal into perspective 

Fraser

[kripton2035]

ordered one too ! thanks.
now waiting for a good lens to have some macro inspection possible...

[Fraser]

This little camera could form the basis of a fun series of projects to enhance its capabilities within the constraints of 4800 pixels 

A close-up lens is the simplest addition to impliment but there is great scope for creattion of lens mounts and different lens formats. A thermal microscope configuration maybe    Now 4800 pixels will be a limitation but if looking at a target only 5mm x 5mm that is plenty

Then there are are mounting options for use inspecting PCB's and image processing to pull greater thermal detail out of images. More advanced upgardes might include changing the lens fitted to the Lepton as the one fitted is far from optimal and degrades the image. A narrower field of view of 25 Degrees or less would improve the imagery produced. Bill_W is doing some work on Lepton lens upgrades. With these cameras being so well priced, there is less at risk when considering hardware modifications  That said, there is only so much that can be done to this camera before it makes better sense to buy a camera with higher perfromance at greater cost. It is a fun little unit to 'play' with though, and a good learning experience for those wishing to dip their toe into thermal imaging without spending a fortune.

Fraser

     

[Fraser]

Bill_W's thread on lens experiments with the Lepton core.....

https://www.eevblog.com/forum/thermal-imaging/flir1-g2-lens-experiments/msg3623452/#msg3623452

Fraser

[Fraser]

Decent CVD ZnSe lenses:

https://www.ebay.co.uk/itm/123079006101?hash=item1ca8149795:g:y~AAAOSwsWFg-P~j

Look at their various offerings in thier listings. Many diameters and focus distances. CVD lenses are superior optical quality to PVD. You need to select a lens diameter that is larger than the minimum required in order to make use of the better performing central portion of the lens and to reduce edge distortions.

Decent lenses do cost reasonable money so it may be best to avoid the really cheap offerings on eBay. That said, the 80 x 60 pixel resolution of this canmera will be very forgiving of lower qaulity lenses. ~$30 is about average for a decent 20mm diameter ZnSe CVD lens. Select a focus distance to meet your needs. I use 101.6mm and 50.8mm focus distance lenses. If you buy a decent 20mm diameter lens it will likely be useable on other, higher resolution, thermal cameras in the future. That said, a cheap ZnSe lens may still be the the best option if the budget is tight as it will still work OK on the Multicomp camera. The minimum diameter of useable lens will likely be 12mm and these are the cheapest.

Fraser

[kripton2035]

found these same lens on ali, almost the same price except no vat or customs possible surprise for EU shipping from Aliexpress
(as nothing is clearly explained on ebay...)
https://www.aliexpress.com/item/32234326702.html

[Fraser]

With regard to modifications and enhancements to the diminutive Multicomp thermal camera...

Some time ago I bought some Military surplus eyecup adapters that were designed to mount on a handheld military navigation and comms equipment. The Eye cup adapter is special in that it is equipped with a lens and also a eyecup 'iris' to prevent light leakage when not pressed to the eye (sniper countermeasure) I thought these unusual items would prove useful one day. Similar items may be purchased for modern DSLR cameras as well. As luck would have it, I knew where to put my hands on one of these military eyecup adapters amongst all my other 'may be useful one day stuff' ! It is a good match to the Multicomp camera display so I could make a thermal surveilance scope using it, or even a thermal microscope. The built in high quality lens in the eyecup adapter takes care of producing a well focussed image of the 320 x 240 pixel LCD panel to the user.

Fraser

[Fraser]

See this thread for some examples of different mounts that may be used to hold small thermal cameras for PCB inspetion etc......

https://www.eevblog.com/forum/thermal-imaging/review-dianyang-technology-ca10-circuit-board-thermal-analyzer-repair-rd/

I placed the Multicomp camera on the Gooseneck table mount and it looks perfect for PCB work when a Znse close-up lens is in use.
Picture attached.

Fraser

[Fraser]

I thought some real world images from my Multicomp camera when viewing a PCB might help others to understand the abilities and limitations of a thermal camera that uses the 80 x 60 pixel Lepton 2.5 core. When looking at the following images we need to keep in mind that the thermal data is only 80 sample points wide by 60 sample points high and the IFOV is less than great due to the relatively wide angle 50 Degrees HFOV lens.

I captured some images with a 20mm diameter ZnSe lens fitted. This lens has a stated focus distance of 4"/101mm.
Upon fitting the supplemental lens I realised that there was not a dramatic change in the close-up imaging performance of the camera. The depth of field is impressive with and without the Supplemental lens. For this reason, I also captured some images of the PCB without the supplemental lens to show the "RAW" cameras apability when working close to a PCB.

The test subject in the PCB imagery is a Haupppauge Video Capture unit and I include a visible light image to provide some context to the thermal images. I was interested in what parts, if any, of the Capture unit are working as the unit presents as totally dead. I captured an image as soon as power was applied and noted which IC's showed signs of life. More images were captured as the unit warmed up. Two IC's on the left of the PCB are clearly operating enough to generate some heat, though not excessive heat. The middle larger IC is also showing some evidence of thermal activity. There were no 'hot spots' of concern and I now know that there is some life in the PCB. Further tests with other test equipment will follow. Note: the reflective connectors in the images are reflecting my body heat so do not be confused by the 'hot spots' that they create.

I also captured an image of my small Butane gas blow torch just to see how the camera behaved with the ZnSe lens fitted. No problem with that at all. I imaged my hand as well 

On with the images and remember...this is a 80 x 60 pixel image and so will not impress on the fine detail front....I believe it would display hot SMT resistors of small size though.

Fraser

[Fraser]

The Hauppauge Video Capture Unit PCB.....

Supplemental lenses fitted.

[Fraser]

Continued....

[Fraser]

The Hauppauge capture unit using the same camera but "RAW" without a close-up lens fitted.

[Fraser]

Direct comparison of images captured with and without the supplemental close-up lens fitted to the camera. The first of each pair of images is without the supplemntal lens fitted.

Fraser

[Fraser]

My conclusion from this simple test was that the camera, as supplied, can image a PCB adequately due to its close focus capability.

I will do further tests that are more meaningful to determine whether a supplemental lens offers any benefits on this camera. Clearly for optimum focus at very close range, some form of additional lens would be of benefit to the imaging, but for basic PCB imaging it  seems to cope quite well without any additional optics.

I will do a 'fine wire' minimum focus distance check on the camera as well.

Fraser

[Vipitis]

Buy like 24 of them and put them in an array. Try to stich it together.