Overall Accuracy, and Repeatability

[bigben2k]

Here we discuss Overall Accuracy, and Repeatability.

This may include, among other things, how all the measurements tie in together, as well as the error margins. It may also include details of the testing procedure.

[jaydee]

Opps, moved to the right section.

[Zhentar]

Edit: dammit jaydee, how am I supposed to respond to your posts if you move them?

[jaydee]

Originally posted by Zhentar
Edit: dammit jaydee, how am I supposed to respond to your posts if you move them?
Yeah, my bad. Everything I said though belonged in the forum spacifically made for what I said. I just didn't see it untill I hit reply and then went to the main section again.

[jaydee]

I would like to see this thread move forward. Would someone be inclined to to explain what temp measurements are crucial for water block testing? Then we can get into how to achieve those temps with some degree of accuracy.

What I have come up with is:

Inlet of water block
Outlet of water block
Watts used by the heat load be it a CPU or die sim
Die temp

What am I missing and how do these temps tie in together??

[Incoherent]

I would like to see this thread move forward. Would someone be inclined to to explain what temp measurements are crucial for water block testing? Then we can get into how to achieve those temps with some degree of accuracy.

What I have come up with is:

Inlet of water block
Outlet of water block
Watts used by the heat load be it a CPU or die sim
Die temp

What am I missing and how do these temps tie in together??

I'd say that it is possible to generate useful numbers with just Water in and die temperature at 0.1°C resolution, assuming that the watts, pump, coolant, hoses, TIM, mounting pressure, die dimensions, material and sensor location are constants between tests. This gives WB A is xx.x degrees better than WB B, on this setup

Adding flowrate and pressure drop measurement gives more info. A P-Q curve or K-factor for the block and an idea of how the WBs might perform in other setups.

The ability to measure Watts is a big jump because the data becomes more absolute. The big unknown is the TIM, but aside from that if the die material, dimensions and sensor location are known a fair bit can be gleaned. This kind of data can IMO be compared between test setups if die properties, sensor locations are known.

Water outlet temperature is not useful at all unless you have the resolution to make it meaningful. At least 0.01°C and even this is very noisy at high flowrates. If you can measure it accurately it becomes very informative, either as a measurement of heat-to-water, flowrate or coolant heat capacity, depending on what you already accurately know.

The huge fly in the ointment is the TIM. It is a totally variable parameter.
A fluxblock type arrangement gets around this problem but is not user friendly and adds another TIM interface noise to the data.

Other problems.
Accuracy.
Calibration.
Insulation.
Die surface degradation.

Some sort of calibration is absolutely vital. Even if it is just to compare how the sensors behave over a range of temperatures. Achieving the required accuracy depends on how much effort you are prepared to put into this and how much money you want to spend on equipment.

[Les]

Incoherent has it in a nutshell
I think his reply is succinct, accurate and comprehensive.

[bigben2k]

here here
10

[jaydee]

So to get usefull results we would like:

1) Water in temp
2) Die temp

Both atleast .1C resolution.

To take it a little further:

3) Flow rate
4) Pressure drop

Added plus:

5) Watt measurment

I guess the accuracy level of the equipment is up to the buyer.

[bigben2k]

Yeah. I'd stay away from #5, until you got the rest of it up and running.

#5 is going to require water temp in, and water temp out. I'm still figuring out ways to measure that secondary heat loss. My latest thought is to run a water block within a vacuum, and comparing the performance in normal atmosphere; aside from radiated heat, the comparison should indicate the actual difference. (inspired by another member here). Testing under those conditions though isn't obvious.