Hi - I operate a publication which conducts reviews of computer hardware components, and one aspect of this involves the evaluation of different thermal solutions for processors, motherboards, videocards, etc.

We're considering the use of thermal Imaging in conjunction with thermocouple temp. measurements for evaluating the effectiveness of the various heatsinks designs, primarily to illuminate design flaws, overheating, poor joint conductivity, etc. The test products would range in size from 8cm to 15cm on average, while the largest would be no more than 30x24cm in area. The maximum temperature gradient in a test object is likely to span 20-to-150 Celsius or less.

Of the thermal cameras in our budget, Flir infraCam SD & Fluke Ti10, both have a very small resolution of 120x120 or 160x120 pixels that concerns me, and minimum focal distance of between 30cm and 15cm, respectively, that seems acceptable.

My question is twofold;. Are either of these units appropriate for this kind of up close application?

Secondly, computer components are manufactured from a range of metals, plastics, glass fibers and ceramics of varying surface attributes. Some components are matt plastic, others shiny metal, some coated with paint, plating/anodizing, stickers. Is it possible to image a diverse scene of materials and surface texture and generate an accurate thermal picture overall? There seems to be a potential problem with accurately measuring temperature because of emissivity variations...?

On a final note, the results generated would be presented to readers in image format, so we're seeking reasonable visual and temperature accuracy of a degree or so.

Oh, and is it possible to remove the watermarking on the image generated by the Flir camera?

Thanks,
M

I would say that the cameras you mention are not suitable for that application (and the cameras within this price range). You will need Lens options not available for the cameras. Typical accuracy of the cameras is +or-2 degrees or 2%, not a degree or so. Variations in emissivty will be an issue. Given all of this I would suggest you hire in the services of someone that has the correct camera and lenses, and that is capiable of dealing with the emissivity issue.

 Yes, the Flir watermark can be removed on the camera, but it is not easy. Within the software it can be removed easily.

 

www.thermalvision.ie

Something to consider is that many of these manufacturers may themselves already be using IR as part of their quality control process.  Depending on the material you are looking at, the conditions of your test and the type of equipment and personnel employed, you may end up with a very different idea of what is going on than what the manufacturer is seeing with a calibrated unit, set up by a qualified thermographer in an ISO process.  This might cause problems if you draw conclusions from your images with a unit that is either operating in the wrong range of wavelengths or not properly corrected for the characteristics of the materials - especially if it leads your readers not to buy said component.

For your consideration.

-John

Hello M:

The first question you have is about those two imagers…I agree with Bob, your budget won't allow for a unit that you really need to publish results in a magazine, since you are going to need want nice-looking imagery. You could rent one or hire the work out. If you want to purchase equipment for a long-term project, you will have to spend much more.

As to your second question…there are serious issues regarding sensitivity, reflection and spatial resolution with this application. We have imaged down to 15 microns size, but I do not trust the temperature readings. Qualitative results often serve our client’s purposes. The catch is, if you treat the surface to even out the emissivity, you are changing the thermal characteristics of the target(s). You will have to measure the emissivity of each object and calculate the temps for each separately. However, there are also issues with thermocouples affecting these readings as well, because they themselves act as heat sinks.

To answer the third question...no, off-the-shelf IR imagers are not accurate enough to measure one degree (F or C) and produce repeatable results on tiny objects, but many can discern temperatures differences in tenths of degrees.

---

Greg Stockton
Stockton Infrared Thermographic Services, Inc. and United Infrared
(800) 248-SCAN (toll-free)
(336) 498-GREG (voice)
(336) 689-3658 (cell)
www.StocktonInfrared.com
www.UnitedInfrared.com

Hi M,

I looked at my opened PC.

The image was taken with a camera with 160x120 and 20mm lens. Note the difference in detail.

Dominique Akermann

DA Mechanical Services

www.damechanical.com

  

M -

Greg's comments about accuracy and the impact of emissivity on the actual temperatures, as well as on the perceived temperatures, are correct.  He and I have discussed this in the past.  Let me suggest a few things for your consideration:

One - as said by him and others, you are not going to get 1 degree accuracy.  I doubt that you really need that level.  I think (guessing on my part as to the real issues) that what may be of greater interest is variation and anomalies.  That is, if a heat sink is not uniform across its plan area, then there is a problem. 

Two - you are definitely going to have problems in IR imaging, esp. trying to measure temperatures, with the large variation in emissivity that you will encounter among the materials.  This could possibly be addressed by altering the suface conditions of the materials under test, with the understanding that the alteration will affect the actual temperature.  A way around this, if it is needed, is to calculate the temperatures that were occuring before the emissivity alteration.  This would involve thermal modeling of the component under test under both altered and orginal emissivity conditions.  The altered emissivity condition would be used to validate the thermal model, which should match the measured results.  The model would then be adusted to match the original emissivity, and recalculated to give the temperature conditions of the original.  (Modeling of this type is a service I can provide and would be glad to discuss with you.)

Three - The size of the IR array will directly affect the perceived quality of the printed IR image in a publication.  You will have to decide what is acceptable and what is affordable.  I, as well as many others reading this board, could provide off-site testing for you if you decide that that makes sense.  Many of us have larger array sizes.  If nothing else, going that route as a first step would let you develop approaches and decide on the value of it to your publication without a capital investment or an investment in training.  Another route to consider is using multiple images from the camera and combining them into a mosaic.  This may have issues at the size and configuration of objects you are looking at. 

 Feel free to contact me to discuss this further, if you wish. 

Jack

 

---

Jack M. Kleinfeld, P.E.
Kleinfeld Technical Services, Inc.
Bronx, NY
718-884-6644
JKEngineer@KleinfeldTechnical.com
come see what we can do for you: http://www.KleinfeldTechnical.com

Hey M,

I've messed around with this when I was building/overclocking my last rig; this shot is of a Zalman cooler mounted on an E6600 in an Antec 900 case.

 Unfortunatly the problem with imaging heatsinks is that they tend to be alluminium or copper which is a very tricky material to work with; I never managed to get measurements close to what the chips own sensors was reading even with the chip close to 70 degrees C.

 There has been some work done with putting thermal tracers into airflow to get a visulisation from that; sulfur hexafluoride works well due to being non-toxic, non flamable and with an emission peak within the range of the camera. Could be worth a look.

 I don't think you'll get better results with the camera then you would with a good thermocouple. I know a tech-mag in the UK who were messing round with IR cameras a year or so back. I don't think they got much beyond using them for case analysis.
 

---

Stephen Goodfellow

Level 2 Certified Thermographer
Product Development Enginer - Specalist Materials

Thanks to everyone for the frank and informative responses, most of my follow up questions have been anticipated and answered.

The general consensus seems to be that a thermal camera of the $5K class isn't ideal for this application with respect to resolution, the materials won't lend themselves to easy analysis without coating the surface or significant training, and the end result isn't going to be that accurate in any case because shiny metal is problematic.

Stephen, you've hit the nail on the head with your response, thanks for posting that image - more my language! The original idea was to offer an overall view of the hardware powered up and hot, not necessarily quantitative figures. Thermocouples are better for that. The shot of your computer shows the videocard PCB hot, the heatsink not, and the hot spot on the Zalman heatsink most likely just the fan's motor. The heatsinks on the VGA and CPU look to my untrained eye as ambient temperature....  not great for what we were hoping to do.

Incidently, it was talking to Zalman's heatsink designer that got me thinking about thermal imaging as a tool in hardware reviews. He uses a thermal imaging camera to discern how evenly heat is conducted along the length of heatpipes and the heatsink body, basically so there aren't large temperature variations all of a sudden. This is all done at the prototype stage of course, we've got to work with hardware in it's retail state. Coating it for better imaging just won't be feasible, nor is outsourcing going to be cost effective for us.

Cheers,
M

Two brief comments:

"The general consensus seems to be that a thermal camera of the $5K class isn't ideal for this application with respect to resolution..."

NOT necessarily so. It is the combination of the size of the detector array AND the lens that determines what you will be able to resolve. You can work the math for these lower cost cameras using the right lens and see that they can, in fact, work in some circumstances for these types of applications.

"Coating it for better imaging just won't be feasible..."

"Conformal" coatings of various types have long been used to increase the emissivity of board components. Perhaps these could offer viable options in your situation as well.

 

---

Thermally Yours,

John
ASNT NDT Level III #48166
The Snell Group
www.thesnellgroup.com
www.thermalsolutions.org
800-636-9820

I agree with John, but most cameras in this price range do not have a large selection of lenses available for them. It also depends on the minimum focus distance, what level of information you require, and what you are hapy to publish.

Dear M,

The number of pixels in an array does not necessarily determine the resolution of a camera. For example our EZTTHERM 880 is a 160x120 resolution camera but has a unique close focus capability that enables a user to essentially use the camera in a MACRO mode. At a 2 inch distance the spot size of the camera will be 200 microns. The attached spot size calculator computation is shown for your reference.

Our spot size calculator is available apon request so just email at astout@electrophysics.com and I'll forward the excel file to you.

Good luck,

Art Stout

VP, Business Development

Electrophysics

www.electrophysics.com

www.ircores.com

Hey M,

One of the big drawbacks of IR is you can only really compare like-for-like materials within the same image, but the average PC case has copper/aluminium heatsinks, plastic fans, plastic motherboard components, steel side panels...

As you guessed, the fan ooks warmer then the heatsink because its made of plastic.

The UK magazine that worked with IR cameras about two years ago was CustomPC. Unfortunately the article is long gone from their website but I know one of their writers so I might be able to source a copy.

This old digg reference gives you an idea of what they were doing; looking at a painted case from the outside to gauge the internal air temperature in different places.

It might be possible to take that one step further by opening up the case and using a thermal tracer in the airflow. There's an annoying trend in new cases to stick on as many fans as possible (like the new Antec 900) without any mention of how efficient the air is moving inside the case, might be a feature in looking at that.

Its something I meant to try a while back until I realised a small bottle of the tracer gas would cost the same as a new GeForce...the GeForce won out. So I don't have much idea how it might look or how effective it might be. It wouldn't give you anything quantitative, but it could be a good qualitative source.

Give me an email if you want any more info.

Regards,

Stephen Goodfellow.

---

Stephen Goodfellow

Level 2 Certified Thermographer
Product Development Enginer - Specalist Materials