Hey guys. I'm trying to find a commercially available thermal test vehicle with built in temperature monitoring to test thermal interface solutions. I need something fairly high power, like 100W minimum over a 20mm x 20mm area. Is there anything out there? All I have been able to find are tests done on proprietary equipment.
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Looking for high power TTV for TIM testing... 2
- Thread starter TurboLS
- Start date
-
1
- #2
I have no clue what you are talking about or what the abbreviations stand for. If you want to elaborate a little I/we may be able to actually help.
Keith Cress
Flamin Systems, Inc.-
Keith Cress
Flamin Systems, Inc.-
- Thread starter
- #3
Sure, no problem. TTV stands for thermal test vehicle and TIM stands for thermal interface material. In product development, we have to test our TIMs in as close to real conditions as possible. In my case, I need socketed CPU that consumes a minimum of 100W of power so that it generates enough heat for me to see the difference between my material and competitors. It can be a fully functional CPU, or one that is defunct and can be hit with clock signals to just get hot. As long as I know the power consumed and the temperature at the interface, I'm good.
Ah I see now. Thanks.
Let me suggest a new DUO CORE Intel processor. Why? Because over the years motherboard CPU temp monitoring has gone from nil to absolutely great! After no sensing, came socket sensing where some temp sensor was incorporated into the CPU socket. It measured the bottom of the chip carrier and only allowed vague inference of the heat sink performance.
Next they decided to put the sensor onto the actual IC. This would then be read by circuitry on the motherboard and dealt with by the BIOS. This was much closer to the right solution. The problem though was variation between motherboards and the characteristic of the substrate sensor. This meant the actual temp was +/-15C which is easily the difference between disaster and happy-happy-joy-joy. This caused a ton of overclockers to make claims of Intel because they "never reached the temperature limit" but the CPU still died. In reality they had exceeded the limit and fried the CPU but the temperature monitoring system was just flat wrong.
With the shrinking of the die and the inclusion of multiple cores exaserbating the whole heat problem Intel decided to fix this niggling problem once and for all. Clever really, because if the overclockers have a 'good' tool they won't be frying their CPUs and making claims because NOBODY wants to fry their CPUs.
So they included a die sensor on each core of a multicore CPU and they included the A/D on each die to read the core temperature and they even calibrate them.
All you need to do is read a certain register in the CPU to get the latest instantaneous calibrated CPU die temperature.
The computer I'm writing this on has this feature. It's a Duo Core 6700. I run a small program that constantly shows both core temps. Just by grabbing a window and wiggling it I can see the resulting near instant temp rise that is caused by the extra processing it takes to process the translated image. The instant I stop wiggling a window the temp drops back. A lot of fun actually.
The only gotcha is that a lot of these transition motherboards still try to read the CPU temp via the motherboard scheme and so are reporting bazaar temp values that cause things like over heat alarms to go active. Just be aware.
Anyway if you use a motherboard and one of the latest CPUs with the temp reading utility you will have a nice accurate reading of the die temp. Add a system watt-meter. Write some software to change the CPU loading automatically. Log the test regime result. Voila!
Keith Cress
Flamin Systems, Inc.-
Let me suggest a new DUO CORE Intel processor. Why? Because over the years motherboard CPU temp monitoring has gone from nil to absolutely great! After no sensing, came socket sensing where some temp sensor was incorporated into the CPU socket. It measured the bottom of the chip carrier and only allowed vague inference of the heat sink performance.
Next they decided to put the sensor onto the actual IC. This would then be read by circuitry on the motherboard and dealt with by the BIOS. This was much closer to the right solution. The problem though was variation between motherboards and the characteristic of the substrate sensor. This meant the actual temp was +/-15C which is easily the difference between disaster and happy-happy-joy-joy. This caused a ton of overclockers to make claims of Intel because they "never reached the temperature limit" but the CPU still died. In reality they had exceeded the limit and fried the CPU but the temperature monitoring system was just flat wrong.
With the shrinking of the die and the inclusion of multiple cores exaserbating the whole heat problem Intel decided to fix this niggling problem once and for all. Clever really, because if the overclockers have a 'good' tool they won't be frying their CPUs and making claims because NOBODY wants to fry their CPUs.
So they included a die sensor on each core of a multicore CPU and they included the A/D on each die to read the core temperature and they even calibrate them.
All you need to do is read a certain register in the CPU to get the latest instantaneous calibrated CPU die temperature.
The computer I'm writing this on has this feature. It's a Duo Core 6700. I run a small program that constantly shows both core temps. Just by grabbing a window and wiggling it I can see the resulting near instant temp rise that is caused by the extra processing it takes to process the translated image. The instant I stop wiggling a window the temp drops back. A lot of fun actually.
The only gotcha is that a lot of these transition motherboards still try to read the CPU temp via the motherboard scheme and so are reporting bazaar temp values that cause things like over heat alarms to go active. Just be aware.
Anyway if you use a motherboard and one of the latest CPUs with the temp reading utility you will have a nice accurate reading of the die temp. Add a system watt-meter. Write some software to change the CPU loading automatically. Log the test regime result. Voila!
Keith Cress
Flamin Systems, Inc.-
-
1
- #5
Great info, Keith. The only thing I would add is that the power per core should be consistent for all your tests or you won't have an apples-to-apples comparison of TIMs. That's one of the nice things about TTMs. Intel TTMs are difficult to get -- they don't sell them and only supply them to certain development partners. I don't know about AMD or any 3rd party hw.
There are many touchy variables in TIM testing: the attachment force, component height, planarity, and heat sink surface conditions have large tolerance ranges (even if you clean and re-use a sink, the behavior can change!). You'll need to run a lot of tests to get statistically significant results.
You can embed a thermocouple in the heat sink so you don't have to worry about sink-to-air test variations. Consistently locating and mounting this is a whole nother subject...
TIM mfgrs usually avoid these issues by testing to a standard like ASTM D374 and telling customers "your results may vary".
ko (
There are many touchy variables in TIM testing: the attachment force, component height, planarity, and heat sink surface conditions have large tolerance ranges (even if you clean and re-use a sink, the behavior can change!). You'll need to run a lot of tests to get statistically significant results.
You can embed a thermocouple in the heat sink so you don't have to worry about sink-to-air test variations. Consistently locating and mounting this is a whole nother subject...
TIM mfgrs usually avoid these issues by testing to a standard like ASTM D374 and telling customers "your results may vary".
ko (
- Thread starter
- #6
i would like to thank everyone for their input thus far. i have looked at the ASTM method, but the problem is that our solution cannot be simply wiped off, like a grease. This means that once we make a bond, it cannot be debonded (or at least not yet! that's another thing I'm working on). this is why i want to get the socketed version. if i bond the chip up, i can run my tests and then swap it out and use the same mobo for the next tests and so on.
I will definitely look into the Intel DUO core chips. So far, AMD's Athlon 64 FX 57 was the highest power chip that I could find, but maybe with the accuracy of that new Intel chip, I could see variations between TIM's even at lower powers.
I will definitely look into the Intel DUO core chips. So far, AMD's Athlon 64 FX 57 was the highest power chip that I could find, but maybe with the accuracy of that new Intel chip, I could see variations between TIM's even at lower powers.
- Thread starter
- #7
- Thread starter
- #8
- Thread starter
- #9
OK, I found some info on lid removal at overclockers.com, but now I'm thinking that the Intel chips may not be what i'm lookin for. Here's why. 1.) The fact that there is a dual core means that some temperature will go from die to die, making thermal measurements (of conductivity of the TIM and such) quite cumbersome. 2.) Lid removal is more of a pain than with the AMD chips. 3.) Lids are soldered to the die, as opposed to expoxied
Downsides of the AMD chips: thermal diode not necessarily placed directly over hottest part of chip.
In addition, for this idea to work, I have to be able to get the die surface completely clean so I can send them to be metallized with Ti-Ni-Au.
Downsides of the AMD chips: thermal diode not necessarily placed directly over hottest part of chip.
In addition, for this idea to work, I have to be able to get the die surface completely clean so I can send them to be metallized with Ti-Ni-Au.
It may take a lot of phone calls to Intel or AMD, but I wouldn't give up on the TTVs just yet. Maybe they have some older models available.
Maximizing power is not that important as long as you can get a good dT measurement. Thermal resistance (C/W) and impedance (C-in2/W) should be independant of power.
ko (
Maximizing power is not that important as long as you can get a good dT measurement. Thermal resistance (C/W) and impedance (C-in2/W) should be independant of power.
ko (
Why not just use a resistor attached to a piece of copper? Watlow and Omega make flat flexible heaters -- you can exceed their power specs if you provide enough cooling.
ko (
ko (
- Thread starter
- #12
In response to your maximizing power statement, I do need higher power because when you are trying to compare TIMs that are all really good, you have to crank up the temps and the only in-situ way to do that is with a high power CPU.
I'll look into the flexible heaters. One idea we had was to use ceramics, but that is too expensive because we have to fabricate, calibrate, and throw them away every time.
I'll look into the flexible heaters. One idea we had was to use ceramics, but that is too expensive because we have to fabricate, calibrate, and throw them away every time.
I like ko99's suggestion,(as I do most). Why not duplicate the CPU but use a heater element. Whatever you came up with would be vastly cheaper than $500 CPUs.
Keith Cress
Flamin Systems, Inc.-
Keith Cress
Flamin Systems, Inc.-
- Status
Similar threads
- Locked
- Question
- Question