Products > Thermal Imaging

Specification of a thermal camera for PCB repairs - some thoughts

(1/10) > >>

Fraser:
I have just been using one of my thermal cameras to repair a laptop and it occurred to me that it may be useful to start a thread discussing the minimum specification of thermal camera that is practical for use in repairing modern PCB’s. I will provide my view on the matter but welcome others participation as newbies to thermal imaging may be uncertain of what is required for PCB repair work.

Before we get into specifications etc, it is worth stating what a thermal camera offers by way of PCB repair and disagnostics.

The thermal camera can only show the user temperature information, be it a spot temperature or a temperature differential. As such it is well suited to spotting parts of a PCB that exhibit one of the following symptoms:

1. A component or area on a PCB that is running hotter than expected (e.g. a regulator with a shorted output)
2. A component or area on a PCB that is running cooler than expected (e.g. a circuit that has lost its power supply so not working)
3. A component that is pulsing on and off when it is supposed to be constantly on (commonly called hiccuping due to overload or the opening and closing of a Polyfuse due to overload)
4. Low or no thermal activity in a microprocessor indicating a potential “SLEEP”,  “HALT” or “RESET” state.
5. PCB tracks becoming thermally visible due to high current flow through them (often caused by a shorted component or track “down stream”)
6  Capacitors becoming hot (Often a sign that the capacitor is failing or in distress)
7. Excessive temperatures found on a heat-sink (either due to an inadequate heatsink or a fault causing excess dissipation from an associated component.
8. Excessive heat build up within an equipment case (often caused by inadequate ventilation, obstruction of air flow or a failing fan)
9. Components running unexpectedly hot. (May be due to designer underrating of component or a fault causing higher than normal current flow through the component)
10. Batteries becoming unexpectedly hot during charging (potential battery issue or a failure in the charging circuit - where Li-Ion is involved, this is serious !)
11. Heat visible from a connector (this can be evidence of an underrated connector, for the current it is carrying, or a connection with a higher resistance than normal due to contamination or corrosion. Some GPU PCB’s suffer from this on power connectors.)
12. Part of a circuit warm when it should be switched off (this can indicate failure in a power rail control, such as a series MOSFET short. This can lead to battery operated equipment discharging batteries even when switched off.

Well that will do for my list of potential issues that a thermal camera may assist in diagnosing. There are bound to be other examples so feel free to share your experience here. My following posts will discuss thermal cameras and their specifications.

Fraser

Fraser:
Thermal Camera specifications and how they effect use for PCB thermography…..

As detailed above, a thermal imaging camera offers the user an insight into the thermal domain of a PCB or piece of equipment. Such insight can greatly aid diagnostics and provides a better understanding of electrical activity within a system. The thermal camera only provides information on thermal energy however and it is for the user to interpret what it shows on its display. This can be easy or challenging to the user, dependant upon experience and the nature of the fault. For a user to be able to interpret the thermal domain of a device under test (DUT) the thermal camera needs to be of adequate performance to both detect and display a thermal anomaly, if present. This post will detail the effect of specifications on a cameras usability for PCB repair work. It is not a discussion of the best camera specifications or best “bang for buck”. That is for the reader to consider. I offer only an insight into the effect of certain specifications on a thermal cameras capabilities.

So on with the specifications of interest to us in PCB thermal profiling and repair……..

RESOLUTION

I have placed this specification at the top for good reason. It is often used by manufacturers marketing teams to try and impress the potential purchaser of a thermal camera. There is no argument against “more pixels is better” in the world of thermal imaging as cameras generally have relatively low resolution compared to modern visible light digital cameras. The problem is, more pixels = more expensive and the increase in price can be exponential with resolution. Hence this thread discusses what you actually need for PCB work rather than what you think you need or want ? It should really be stated that a very badly designed higher resolution camera may offer poorer imaging performance than a well designed lower resolution camera. There is more to thermal imaging camera performance than purely pixel count.

So what resolutions of thermal imaging camera are common in the marketplace and what are my thoughts on each ? See below:

a) 16x16 pixels (such as the IRISYS Redeye 6)

With only 16 x 16 pixels present on the sensor array, the ability of the camera to resolve much detail on a PCB will be severely limited. The manufacturer will likely employ interpolation to increase the presented image to something more reasonable, such as 128 x 128 pixels. It should be understood that interpolation does not improve the RAW resolution so cannot really pull more detail out of the original data from the sensor array. I personally consider this resolution too low for PCB work as the image provides little to no thermal scene context for interpretation by the user.

b) 32x32 pixels and 49x49 pixels (for example the IRISYS Redeye series)

As stated previously, low resolution thermal sensor arrays can provide a thermal image but they normally lack the detail needed by a user for context within the thermal scene that is a PCB. Whilst 32x32 and 49x49 pixels is significantly better than the lower resolution sensor arrays, it remains too low for many PCB thermal analysis tasks and far better options exist for not a lot more investment by the user.

c) 80x60 pixels (for example the FLIR Lepton 2 core)

In the past, 80x60 pixel sensor arrays were considered a serious entry point into thermal imaging that was actually useful. Interpolation was still employed to make the displayed image more acceptable to the user but in this case the amount of RAW thermal data from the sensor array was adequate for some understanding of the thermal scene and interpretation by the user. The FLIR Lepton 2 was a very popular thermal imaging core that proved 80x60 pixels was a viable resolution for many non-demanding thermal imaging tasks. Such a low resolution is still far from optimum for creating easily interpreted thermal images, but when combined with a visible light camera scene overlay (MSX or Thermal Fusion) the context of the scene became clearer to the user. Sadly the use of a visible light image overlay on such a low resolution sensor array image can be difficult when working on a PCB at close range due to parallax error. There is no doubt in my mind that a 80x60 pixel thermal camera may be used for diagnostics on a PCB, but it is more challenging than when using higher resolution cameras as the exact source of thermal energy is not always obvious. Such low resolution cameras also need a suitable lens system to improve their usefulness in PCB work and this will be discussed later.

d) 96x96 pixels (HikMicro produce such a sensor array)

This resolution is a relatively recent addition to the market and is found in some low cost thermal cameras. It is common for cameras using this sensor array to interpolate the RAW data and state a 240x240 resolution in the specification. This is an old trick from the early days of thermal cameras and if the use of interpolation is not made clear to the user, it is a deceptive practice. Some manufacturers employ both Interpolation and “Super Resolution” to the image presented to the user. It should be understood that true Super Resolution is a technique that makes use of the natural hand shake of the user. It is rendered ineffective if the camera is rigidly mounted on a stand or tripod. Such a camera would then rely on the interpolation image enhancement only. 96x96 pixels is in the same category as 80x60 in my opinion. It can be used fir PCB thermal analysis, but if the super resolution mode is not effective, the displayed image remains relatively low resolution for the user to interpret. Such a camera would certainly be useable for PCB work though.

e) 120x90 pixels (common in many entry level budget cameras that use the Guide Sensmart TIMO imaging core)

Another relatively recent resolution sensor array offers 120x90 pixels and this can produce pretty decent images when interpolation is also applied to the thermal scene data. This resolution is most definitely useable for PCB thermal imaging when searching for thermal anomalies such as previously detailed. Whilst the images will lack fine detail, they are adequate and may be interpreted without too much difficulty. This is especially so if the lens system is suited to PCB work. (Close focus lenses) in my opinion, 120x90 pixels is the lowest resolution that I would recommend for PCB work. It is by no means optimal, but having used it myself with good success, this resolution is worth considering if working to a tight budget.

f) 160x120 pixels (this resolution has been around for many years with many imaging cores available)

The 160x120 pixel resolution is well known to those of us who have been involved in thermal imaging for the past three decades :) It used to be the popular “entry point” for microbolometer based thermal imaging systems. This resolution, when used in a well designed camera system, offers decent thermal imaging that is most definitely adequate for PCB repair work. I have used this resolution for PCB repair work many times with reasonable ease. Once again the ease of use is often dictated by the optics of the system as this effects the resolvable detail, as will be discussed later.

g) 256x192 pixels (very common on modern thermal camera releases from Asia)

The 256x192 pixel sensor array is a relatively new release to the market and this is because China started mass production of microbolometer sensor arrays at this resolution. The choice of resolution was likely the result of balancing resolution, resultant die size, production yield and cost. As China has become a powerful influence on the budget thermal imaging equipment marketplace, it is no surprise that 256x192 pixel thermal cameras are now very common. In my opinion, this resolution is an excellent choice for PCB repair work as it appears to provide the best balance of resolution and cost for some very useable thermal imaging. I have no hesitation in recommending a decent 256x192 pixel thermal camera for PCB work. Lens choice must also be considered but this will be covered later.

h) 320x240 pixels (this is a “Standard Resolution” in the thermal imaging industry that has been around for decades)

320x240 pixels is QVGA and has met the demanding needs of Industry, Fire fighters and the military for many years. In recent years we have seen QVGA+ in the form of 336x256 pixels and 400x300 pixels as enhancements on the standard QVGA resolution. At 320x240 pixels the thermal scene is easily interpreted by the user due to the scene detail captured providing good context. This is where the “if you can afford it” recommendation comes in. I personally like to use thermal cameras that are QVGA or better resolution as the imagery is a pleasure to interpret and decent cameras produce crisp, low noise imagery at this resolution. Sadly the increased size of the microbolometer die over a 256x192 pixel microbolometer, combined with lower production numbers, means a QVGA thermal camera may cost significantly more than a 256x192 pixel mass produced model. For PCB thermal analysis, QVGA and QVGA+ is a joy to use but appropriate optics are still required.

i) 640x480 pixels (often thought of as a Gold Standard in thermal imaging and less common due to high cost)

Whilst it is true that 640x480 pixel imaging sensor arrays offer excellent thermal imagery for the user, it often comes at high cost. The relatively low production numbers of the VGA sensor array and associated cameras tends to keep retail prices high. As such, a user needs to determine whether the higher resolution and associated cost is truly justified in their use case. Whilst the military may have good reason to need a VGA sensor array in their long range thermal targeting systems, do you really need such for just PCB repair work ? I would say no. If your budget is such that a VGA PCB thermal analysis camera may be easily purchased, that is great and you will like the imagery that such produces….. provided the optics also suit the task at hand ! More on that later. VGA thermal cameras used to be rare indeed. With advancements in production techniques and die yields the VGA thermal camera is more common these days. It remains a much more expensive camera for the reasons already mentioned but is to be found in thermal CCTV cameras that offer wider fields of view whilst offering similar image detail to that of narrower field of view QVGA models. Bargains can be found on the secondary market but VGA cameras and cores are not something I feel is necessary for most PCB thermal analysis tasks. There will be exceptions however, such as in Science Labs etc, but they are in the minority in the context of this thread.


There are sensor array resolutions higher than 640x480 pixels but I have decided to ignore those here as they are too specialist and expensive for the intended readership of this thread.

SEEK Thermal produce an unusual 200x150 pixel resolution sensor array that is to be found in many of their products and the products of those OEM’s who buy SEEK Thermal cores. That resolution falls between the 160x120 pixel and 256x192 pixel sensor arrays. Given a choice, I would choose the 256x192 pixel sensor array over the SEEK Thermal product.

As a footnote to this post……..

Be wary of products that appear to offer surprising resolution at unusually low cost. For many years there have been manufacturers who will use a low resolution thermal sensor array and apply interpolation to its output so that higher resolution may be claimed in the specifications ! The use of interpolation without it being clearly stated is deception. Some manufacturers also provide the LCD display resolution in the specifications rather than the true thermal sensor resolution in the hope of tricking the buyer. Note that it is normal for a manufacturer to upscale a thermal image to fit a nice high resolution LCD display…. For example a 320x240 pixel thermal image may be upscaled to a 640x480 LCD panel. That is very different to using a 96x96 pixel sensor array and stating a thermal resolution of 240x240 in advertisements ! HikMicro have the ECO range of cameras that do this BUT they make it clear in adverts and specifications that the true sensor array resolution is 96x96 pixels.
Be careful…... If it looks too good to be true, it often is where new thermal imaging equipment is concerned.

It is also worth being a little curious about any thermal camera that has a 1:1 aspect ratio sensor array. Most modern microbolometer arrays have the common 4:3 aspect ratio. Anything different to that may suggest the use of a less common sensor array type or technology. HikMicro are supplying 96x96 pixel arrays for their economy product lines. Their choice of 1:1 aspect ratio is interesting and they must have their reasons for such. The number of dies per wafer could be a factor in their decision.

Fraser:
Lens System, including field of view, minimum focus distance and quality

It is important to not become fixated on a cameras resolution and ignore other important specifications that effect performance when viewing a PCB. An important part of a thermal cameras design is the optical block. The optical block is the group of lenses that sit in front of the sensor array and illuminate it with the thermal scene. The quality of lens elements, lens block design and field of view will all influence the image produced by the thermal camera. Let us look at the lens block in terms of what is desirable for PCB thermal analysis work.

a) Field of view

Field of view (FOV) is a very important specification when comparing thermal imaging cameras, especially cameras with different resolution sensor arrays. The FOV of the lens block has a direct influence on the detail visible on a PCB at a specified distance and this should be well understood by those considering a purchase. I am going to avoid maths here and go with simple cases that illustrate my point well.

Let us take a “Standard camera” with 320 x 240 pixel microbolometer being illuminated by a lens block that provides a horizontal field of view (HFOV) of 50 degrees. This means that the 320 pixels are each taking a share of the 50 degree HFOV. This will mean that each pixel sees 0.156 degrees of the horizontal scene (50/320=0.156). A 50 degree HFOV lens is quite wide angle and Industrial QVGA thermal cameras often use a 24 Degree HFOV lens. In that case each pixel sees 0.070 Degrees of the horizontal scene (24/320). So we can see that by halving the HFOV we gain double the amount of available detail in the scene whilst losing half of the horizontal scene coverage. Losing coverage area of a scene is sometimes acceptable as multiple images may be captured for full coverage. The level of “granularity’” in the scene data can be more important than scene coverage. Bare this in mind with what follows :)

If we take our “Standard camera” with its 320 x 240 Pixel sensor array and 50 degree HFOV lens block and compare other cameras of lower resolutions with it, what do we find ?

Case 1.

A camera with a 160x120 pixel sensor array and 50 degree HFOV lens is compared to the “Standard camera” at the same viewing distance from a PCB. We see a thermal scene on both cameras that covers the same area of the PCB. We note that the 160x120 pixel camera is showing less detail (granularity) in the image and this is because it has half the number of pixels in each vertical column and horizontal row of the sensor array. This is a quarter of the number of imaging pixels covering the same area of the PCB. The difference is definitely noticeable but the components on the PCB will still likely be imaged if producing heat.

Case 2.

A camera with a 160x120 pixel sensor array and 25 Degree HFOV lens block is compared to the “Standard camera” at the same viewing distance from a PCB. We see a thermal scene on both cameras but the area of the PCB displayed on the 160x120 pixel camera is a quarter of that being displayed on the 320x240 pixel “Standard camera”. The level of detail (granularity) within both cameras scenes is, however, the same. In this case the scene detail has bee prioritised over scene area on the 160x120 pixel camera. We see one quarter of the PCB area but obtain the same scene detail as a camera with four times the number of pixels. This trade-off means that a cheaper camera can still resolve the same level of detail in a scene as a more expensive camera that has higher resolution.

In the above simple examples we can see that lens choice is important. Whilst a manufacturer may opt for a 50 degree HFOV lens with a certain resolution of sensor array, a particular application may benefit greatly if the same resolution of sensor is chosen, but with a narrower field of view lens, to provide more scene detail. The FLIR E8 has an HFOV of around 45 degrees for general observation work using its 320x240 pixel microbolometer. My FLIR E60 has an HFOV of around 25 degrees with the same number of pixels observing the scene. Whilst the E60 covers less of the thermal scene area at a given distance, it provides greater detail (granularity) in the displayed thermal scene. Supplemental lenses are available for the E60 that can half or double the field of view but, just as in the examples above, there is an effect on the scene coverage and scene detail.
I hope this makes sense !

b) Minimum focus distance

A lens block will provide a certain field of view, as already discussed. It will also have a specification for minimum focus distance. This is the minimum distance between the cameras lens and a given target at which good focus may still be achieved. The situation is somewhat complicated by the fact that some thermal cameras use a “Fixed Focus” lens, whilst others use a “Manual Focus” lens. There are also true “Auto Focus” lenses available. With “Fixed Focus” lens blocks the lens is actually focussed at the hyperfocal distance that provides acceptable focus between a stated minimum distance and infinity. Such fixed focus lens systems can be very convenient for a user as no focus adjustment is required. That said, these fixed focus systems are a compromise and do not provide the best possible focus at all points of their range coverage. Some fixed focus lens systems may be manually adjusted to favour focus at closer distances than the manufacturer intended. A manual focus lens system requires the user to manually adjust the lens block for optimum focus of the target. Whilst this may be less convenient for the user, it can mean sharper images are produced as the focus is optimised for a particular targets distance from the camera. Manual focus lens systems will also have a specified minimum focus distance but users may find that closer focus than specified is possible. With simple screw in lenses the lens is wound out of the lens holder on a fine thread and the length of that thread often dictates the closest focus before the lens barrel falls out of the lens holder, or hits an end stop to prevent such ! Auto focus lens blocks come in various designs that will not be detailed here. They have a specified minimum focus but their distance detection systems may complicate matters for PCB use. The focus servo system may take its focus point data from a distance detecting sensor, such as ultrasonic or infrared but modern systems tend to use image detail based focus detection. Such systems can struggle in low contrast thermal scenes. For PCB repair work, it would be best to avoid auto focus lens focussing systems unless they may be set to a manual focus mode. The minimum focus distance is an important specification when selecting a thermal camera for PCB repair work but there are ways to adapt a camera to focus closer than the minimum focus distance. These will be discussed later.

The distance between the target and the camera is an important consideration. Just as was detailed in the discussion of lens FOV, the distance from the target effects the amount of available thermal detail (granularity) captured by the camera. In a given situation, if you half the distance between the camera and target, you quadruple the image detail provided in the captured scene but observe one quarter of the original area. This only helps if the camera can actually focus on the target at half the original distance however ! Doubling the distance between the camera and the target has the opposite effect….. a quarter of the detail in the captured image but four times the area covered in one scene capture. For the reasons detailed, it can be an advantage to get as close as practical to a PCB containing modern miniature SMD components if the greatest detail is desired from a relatively low resolution sensor array. The desired distance may well be closer than the cameras specified minimum focus distance, and be warned that a poorly focussed thermal image is a most undesirable situation ! We can add a close focus capability to a camera with nothing more complicated than a single lens element placed in front of the cameras standard lens. For this reason, do not discount a particular camera model because it has, for example, a 30cm minimum focus distance.

So how do we produce a close focus accessory for a standard thermal imaging camera. Well there is plenty of information about this on the EEVBlog thermal forum but in précis, you buy a ZnSe Planar Convex or Meniscus lens that is designed for use on a CO2 laser engraver and mount it in front of the cameras normal lens. You are effectively giving the camera a reading monocle ! Such “close-up” lenses are also used in visible light photography for Macro photography. The ZnSe lens elements are inexpensive and available in several diameters and focus distances. With regard to diameter, it is important to select a lens element that is large enough diameter to avoid vignetting. Focus distance is a mater of preference. I personally use 100mm, 63mm and 50mm focus distance lenses for PCB work. The lenses focus distance specification will approximate to the focus distance that is achieved for the camera when the lens is in use. The depth of field is shallow so the correct working distance is important. With such a lens fitted, a thermal camera normally incapable of focussing closer the 1m can focus on a PCB at just 50mm distance ! This technique works for both fixed focus and manual focus lenses but I find that manual focus lenses still provide the best image clarity. Thermal camera manufacturers have caught on to the market demand for close-focus lenses and several produce official close-focus lens accessories for their cameras. DIY close focus lens mounts are also common and various 3D printable designs may be found on Thingiverse etc. The lenses are common on eBay and often come from China at low cost.

Another possible option for closer focus capability is the manual adjustment of the fixed focus lens to permit close focussing, but this will be at the cost of normal distance focus unless the lens is returned to its hyperfocal distance after use on a PCB.

c) Lens block quality

It should now be clear to the reader that the lens is an important part of a thermal cameras design and errors in that part of a cameras design will impact imaging performance. We need to consider how much of an issue this truly is with modern budget thermal imaging cameras and their use for PCB observations.

In the early days of thermal imaging systems, the lenses were highly specialised products as they used an unusual lens element material that was hard to work with and special anti reflective coatings on the lens surfaces. The cost of these lenses was truly eye watering ! The material used to make the thermal imaging lens elements was/is Germanium that is grown as a single crystal before being cut to the required shape on a diamond lathe. The Germanium lenses made thermal imaging cameras inherently expensive beasts so research into cheaper alternatives began. Other materials, such as GaAs, ZnS and ZnSe may be used for thermal imaging lenses but each has its drawbacks. A new lens material and production method was developed and we now have the Chalcogenide IR moulded glass lenses that are found in modern thermal imaging cameras in the budget/semi-pro market sectors. The new lenses are made from a mixture of Germanium and other materials that may be hot moulded rather than cut on a diamond lathe. Both the raw material and the production process significantly reduced mass production costs for the industry.

Now whilst Chalcogenide IR glass moulded lenses (aka GASIR lenses) revolutionised affordable thermal camera mass production, it should be understood that this new “wonder material” is a compromise compared to a pure mono crystal Germanium lens. This is why Pure Germanium lens elements are still used in high performance professional thermal cameras. That said, Germanium lens systems suffer from issues if they get too hot as transmission reduces with increasing physical temperature. For PCB analysis use, I see no problems using the affordable Chalcogenide IR Glass moulded lenses common in budget thermal cameras as they still perform very well.

There have been lenses produced in the budget thermal imaging sector that did not impress me though. The FLIR Lepton uses a different approach to its lens construction. The Lepton wide angle lenses are created in silicon using the MEMS process, so are effectively printed lenses of the Diffraction type. I will be honest and say that I dislike those lenses. They were another approach to affordable thermal imaging lens production but they are, IMHO, inferior to Chalcogenide IR Glass lenses.

We have discussed lenses here but need to be aware that a lens block is only as good as it’s optical design and some manufacturers may make mistakes when designing a lens group for their lens block. Mistakes happen ! Do not assume that all Chalcogenide IR glass lens blocks will perform the same. It is best to see real world imagery from a particular camera of interest to assure yourself that the produced images are as crisp and well formed as you would expect. A poor lens design can cause poor image geometry and poor focus, amongst other issues. There is also the issue of the care taken at the factory to set the focus on a fixed focus lens camera. There can be improvement in image clarity on some cameras if the lens is focussed carefully by the user. Such corrective action should not be necessary in a properly set up product, but these are mass produced cameras and cores !

Fraser:
Minimum resolvable temperature difference (MRTD)

Much is often made of a thermal cameras sensors Noise Equivalent Temperature Difference (NETD) and this is a specification often used by manufacturers to “out-specification” their market competitors. In my experience NETD has been a much abused specification and it is actually quite hard for most end users to test in order to confirm the claimed performance ! Whilst it is true that newer generations of microbolometer sensor arrays improved the NETD figures through design and manufacturing improvements, I would suggest that users not get too “hung-up” on NETD figures. Professional thermal imaging cameras used to have NETD figures of 100mK yet those cameras were more than capable of use in most thermography applications in which a microbolometer based camera was acceptable. Exceptions are the specialist applications that need extremely low noise imagery and these tend to use cooled thermal imaging cameras. It is true to say that a camera sensor array with a genuinely lower NETD can better show very small temperature changes than one with a higher NETD figure but in most PCB analysis cases this is not a big deal.

For more information on what NETD actually is, see this page:

https://movitherm.com/blog/what-is-netd-in-a-thermal-camera/

I have deliberately titled this post “Minimum Resolvable Temperature Difference” as this is a specification that you will rarely find in the world of budget thermal imaging cameras ! It is what it says….. The minimum temperature difference in a scene that the user can see using the supplied camera SYSTEM. Note that his is a specification relating to the whole camera system and NOT just the specification of the sensor array tested under laboratory conditions for best possible NETD figure ! In other words, MRTD is more of a real world test specification than the “Laboratory test” world of NETD. Some camera manufacturers just take the best possible NETD figure from the sensor array manufacturers data sheets and paste that into their cameras specifications sheet. It is likely no testing of “System NETD” ever takes place on budget cameras. There are many things in a complete thermal camera system that can adversely effect the MRTD and it is good to know this when focussing on claimed NETD figures ! MRTD is the testing of a complete camera system , as supplied to the customer, to determine exactly what temperature difference is visible to a user with the provided lens block and display panel. This is a far more useful test of a thermal camera and I only mention it to highlight how relatively misleading NETD can be in a specification as it related only to the sensor array (in isolation) that is being illuminated by a quality F1 lens.

So is low noise and the ability to detect very small temperature differences important when working on a PCB for the purposes of repair ? Well yes and no. Much depends upon the specific scenario. When it comes to spotting things like overheating components of components that are just emitting a decent amount of thermal energy, high sensitivity to temperature differential is not so important. If, however, the user is trying to use low stimulation currents to detect a shorted component or PCB trace, the sensitivity of the camera to very small changes in temperature can become more important. In many cases, if a short cannot be identified at very low currents, the stimulation current is just increased carefully until the thermal camera can see the component or track becoming detectably warm. A classic case is failed MOSFETs on computer PCB’s that only display the slightest change in temperature due to their very low resistance to ground when failed. A sensitive, low noise, thermal imaging camera will detect the small “hot spot” on the MOSFET at a lower stimulation current than that needed for a lower sensitivity, more noisy, thermal camera. In real world scenarios I believe that any thermal camera with an NETD figure better then 80mK will be more than adequate for PCB repair work. It must also be remembered that tracing shorted tracks on a PCB is challenging if the track is within the many layers of the PCB. Higher stimulation currents may be needed to warm the copper pours above and below the track to show its presence within the PCB. Use of sensible stimulation currents will not cause damage to a PCB as the thermal camera will see the thermal energy well before enough heat is generated in the track to damage it or the PCB. Basically do not worry too much about the difference between a claimed NETD of 45mK on one camera and a claimed NETD of 60mK on another camera.

Fraser:
Temperature measurement range and and accuracy

Users need to consider the temperatures that they are likely to come across in their usage scenario as some thermal cameras may be incapable of measuring such temperatures. A classic comparison would be measuring the temperature of a micro-processor vs measuring the temperature of a vitreous enamel power resistor. The Micro-processor is likely to be running at a temperature below 100 Celsius, but a vitreous enamel power resistor is capable of operating normally at over 400 Celsius in free air. Whilst a thermal camera capable of measuring up to 150 Celsius will cover most likely temperatures on a computer PCB, it would be of little use if wishing to measure the operating temperature of a Vitreous enamel power resistor that is operating at around 250 Celsius. That said, sometimes it is enough to see that something is “HOT” in a system without actually knowing the current temperature…. Evidence of life in a system is such a use case. Be aware that if a thermal camera manufacturer states the maximum measurement temperature is 120 Celsius, it is highly likely that viewing targets in excess of that temperature, whilst not damaging to the camera, will not be displayed correctly and measurements may be flawed or not possible. Some thermal cameras start to display image artefacts when viewing a target that is at a higher temperature than the stated maximum measurement temperature of the camera. This is often due to reaching the end limit of the systems ADC that converts the microbolometer ROIC analogue output to digital data. Thermal cameras that are capable of broad temperature measurement ranges normally use multiple ranges that are manually or automatically selected to suit the expected target temperature.
Selecting the best thermal camera for a usage scenario is basically common sense. How likely will it be that high temperatures will need to be measured. If it is not likely then a single range (cheaper?) camera with a maximum measurement capability of 120 Celsius may be enough for you. If high temperature targets are expected and their measurement important, a higher specification camera with a high temperature range is required but expect the cost to be higher for a given resolution.

So we have decided upon a suitable measurement temperature range that will be required of the thermal camera, but what about measurement accuracy ? Measurement accuracy is a bit of a minefield in the realm of thermal imaging. Even if a particular camera is as accurate at measuring temperature as is claimed in the specifications, there are other factors that will effect the accuracy of any measurements so we need to be sensible when looking at claimed measurement accuracy. Most thermal imaging cameras have a stated measurement accuracy of +/-2% or +/-2 Celsius. This is a very common specification and most thermal cameras that I have tested meet this specification (some only just though !). It is common for me to find thermal cameras from well respected manufacturers performing considerably better than their specification for measurement accuracy. Since the Covid-19 Pandemic a new measurement accuracy specification has become common amongst thermal cameras intended for human fever screening. The Worlds Governments decided that an accuracy of +/-0.5 Celsius was required when measuring people in the search for Covid-19 infections. A tighter +/-0.3 Celsius specification was also created for thermal cameras that use a Blackbody thermal reference to improve measurement accuracy. If a fever detection thermal camera system could not meet these specifications, it was rejected for use against Covid-19. For this reason these new, tighter, measurement tolerances can be seen on some specialist use cameras associated with measuring human targets. Note that the measurement accuracy is only applicable over a stated target temperature range, often +30 Celsius to +45 Celsius. For PCB use, the greater accuracy of these cameras is likely not useful due to their limited measurement range. That said, some surplus fever detection thermal cameras make very useful PCB diagnostic tools provided it is understood that some cannot measure outside +25 Celsius to +50 Celsius. Or even more restricted temperature ranges in some cases. Despite the measurement limitation, these camera still behave normally with regard to displaying a thermal image of the scene, with most having an upper display temperature limit of +100 Celsius or more.

So what can effect the measurement accuracy of a thermal imaging camera beyond the base accuracy of the camera itself ?

Common factors that effect measurement accuracy are…..

a) The selection of the most appropriate Emissivity on the camera for a given target. This is often not known by the user so a best guess is entered using Emissivity guides. A guess that is not correct introduces a measurement error.

b) Distance to target entry on the camera. The distance between the camera and a target is taken into account when a thermal camera makes a measurement. In the case of PCB imaging the camera is very close to the target so path compensation is less critical but a sensible entry is still needed in the distance to target menu.

c) Reflected temperature setting on the camera. The correct “Reflected Temperature” entry on the camera can be important when working with a target that may be being illuminated by an nearby thermal energy source, such as a heatsink. Ignoring Reflected Temperature can lead to measurement errors.

d) Additional lenses or materials in the optical path will degrade the cameras measurement accuracy unless these are compensated for. Examples would be the use of a lens protection film, such as Polyolefin film, or the use of a ZnSe close-up lens in front of the normal camera lens. Anything in the optical path that does not have perfect transmission at LWIR (microbolometer cameras) will have an effect on the cameras measurement accuracy.

A thermal camera user who is making measurements on a target needs to consider whether the camera is the best and most appropriate tool for the task. The cameras measurement accuracy is adequate for many tasks and very convenient, but a more accurate “contact” type temperature measurement sensor may be needed where greater accuracy is required.

Navigation

[0] Message Index

[#] Next page