The WPS opensource project

[davidzo]

Hello testing guys,

I want to present the WPS to you today.
WPS means "Wasserkühler Prüf Station", what means something like "Waterblock Testing Station".

It is a fully atomated testingdevice which you can connect to your computer over a serial bus cable. The design and The Software has been developed in the kaltmacher Forum as an OpenSourceproject:
http://www.kaltmacher.de/viewtopic.php?t=6130&highlight=wps

It is driven by a small Infineon Microprocessor on a 2layer Platine and a Firmware by Chrisbe
- connected over a serial port to the PC, Software that can regulate all values and gereates graphs automatically.
- 4x thermal sensors with a industrial precision of less than 0.1K (or you can connect also better or less better ones, but of course they cost much more ore less)
- DFK (Flowcontrol with exact flowmeasurement in l/h)
- 3x Fanbus regulated by software over the pc (you can easily test 12V, 7V, 5V)
Forum for the WPS:
http://heikop.no-ip.org/phpbb/index

It should make the testing easier and better because the precision comes also with the number of tests you do. The more testloups with remounting the block you make, the more accurate the final results are (guaranteed that the equipment is good).

Sadly i can't upload any pics, because i have a strange error: wps.jpg:
Not a valid image.

[killernoodle]

Would there be any way to incorperate a very powerful pump to the loop and allow changes in tubing size? It would be nice if there was a test bed that was scaleable from a lower flowing system and a higher flowing system. Perhaps incorperate some kind of valve to lessen the pressure of the pump to that of a smaller pump? Or go even easier and use a 12v pump and change the voltage. I think it is silly to base all waterblock testing on a single loop design.

The real-time graphing seems to be a great idea.

And your second link is dead.

[Incoherent]

Thanks for starting this thread davidzo.

Could you get more specific about some things. If I am interpreting correctly the sensors are DS18S20 ONEWIRE digital sensor at 9bits. This is not at all high resolution although I am not suggesting it is inappropriate for the purpose. Can it be upgraded to something more accurate. (I am not confusing resolution and accuracy BTW, one limits the other.)

The FET heater, I like the idea. How is it controlled? What is the feedback if any. If none why not, how do you maintain a constant power output.

Flow control and measurement. DFK, is that a make of flow sensor, how does it work? Flow restriction etc.

Any pressure measurement possibilities?

Is the fanbus regulate-able by water temperature via the software or is it just the 5V,7V,12V steps.

Is there a die simulator attached to the Heater?. Size, material, sensor location relative to die surface. Waterblock mounting system?

How is the whole calibrated? Is there a procedure? I think this would be necessary if results from this are to be comparable between systems.

It should make the testing easier and better because the precision comes also with the number of tests you do. The more testloups with remounting the block you make, the more accurate the final results are (guaranteed that the equipment is good).

I do agree with the first part of this statement. The second part I am not so sure, an improvement in a given mount suggests that the previous mount's data is not accurate. Still, thats a generic problem.

[BillA]

lots of software can control the hardware, LabVIEW, TestPoint, etc.

it is the construction of the hardware and the specific procedures that are of interest

what is the heated area ?
how uniform is the heat flux ?
how are secondary losses addressed ?
how is the ambient temp influence controlled ?
how is the coolant temp controlled ?
how is the flow rate controlled ?

[davidzo]

fixed the second link.

the electronic is compatible also to highflow, of course. I like also the idea of a regulated pump. Actually that is what kaltmacher uses with their tests. They have an Timmer Cooplpump AP1500 which is a special model of Oase pumps. It can be regulated from 12V to 24V (10-19Watts). Watercoolplanet.de measured its power with 2m 10mm id tubing and got araound 7,8l/h and a actual head of 2,83m. For Europeans this is a very powerful pump. (no BillA, i don't sell this product, its from the concurrence ;) )

@BillA: If you would have read the links, you would have known that there is no diesim, socket, tubing etc. included, its only the electronics.

Because there is not socket and system included, the results of different WPSs cannot be compared. Its only the electronics, that make it easier for users or testsites to compare their waterblocks.

[BillA]

davidzo, lets try to stick to testing
truly I am puzzled, yes I saw the boards and schematics and (download for) the software

but what is the heat source that is being controlled ?
I am quite interested in its characteristics

the brightest mind I know has built (I believe it is complete) what he calls a servo psu which is 'controlled' by selecting the wattage and the psu sets and maintains the voltage/current to maintain that power

and he is using mosfets as the heat source
this is most difficult and I'm attempting to see if your group has any insights into how to do this effectively

this statement is NOT helpful
"@BillA: If you would have read the links, you would have known that there is no diesim, socket, tubing etc. included, its only the electronics."
continue such and you'll get blasted again

does anyone in that group know something about making mosfets effectively simulate a CPU heat source ?

the big boys use special silicon, NOT mosfets

[davidzo]

Yes, a FET is used for heatsource. thats important, because the silicon can change its electrical behavior (don't know the right word) when it overheats and that can happen pretty often with a teststation (mounting issues f.e.). FETS are lowcostmaterial, and easy to change and identically fro their heating behaviour. I don't know why special silicon should be better for that, if you haven't your own wafer at home to produce new silicon if another one is overheated again. Maybe this would be the solution for bigboys, but opensource is not bigboys ;)
Imho FETs connected to a copperblock with a hardened iron dieplate are a very good solution. The dieplate should be very hard iron, because copper diesims change the form of the edges and corners with every mounting procedure and get round and not flat in quite short time. The conductivity plays alsmost no role when copper is used to spread the heat evenly to the irondie.
I don't know exact how the Fet is controlled, i'm not familiar with the electronics of the WPS but i know, that it can be precisely controlled from 10-200Watts.

@BillA, no i don't wanted to continue it, and i also don't know that i have been blasted before. Peace?

[BillA]

you perhaps do not understand my question

a CPU has a size and an assumed uniformity of heat flux (not true but a 'known' deviation)
do you know of a single FET with similar characteristics ?
I do not, no one I know does (the big boys)
- and if multiple FETs are being used, there are a whole bunch of new questions

apparently 'someone' has made an assumption
I am questioning that assumption
not about power or measurement, it is the heat source itself

"Imho FETs connected to a copperblock with a hardened iron dieplate are a very good solution. The dieplate should be very hard iron, because copper diesims change the form of the edges and corners with every mounting procedure and get round and not flat in quite short time. The conductivity plays alsmost no role when copper is used to spread the heat evenly to the irondie."
- this is conjecture (not the copper wear, a high and exacting maintenance item),
does data exist to validate the comparability of this type of construction to a CPU (equivalent) source ?

[davidzo]

ok, i just translated heat flux into german ;)
"Hitzefluß" sounds understandable to me.

The WPS uses a single FET. The heatflux and heat distribution allover the die is no problem, because the FET is not the die, it is only the heatsource. If you don't mean the room aspect of heatflux but the time aspect, i can only say, that as far as i know a FET has a very even heat dissipation which does not change time by time. As i said the die is not included in this projekt, so you can use Diesims from Copper, steel, Nickel, wood :D or whatever you want. The thicker the diesim is, the better the heat distribution allover the diesurface will be. Of course the diesim adds another isolating Layer, but i think when you don't compare results with other testbeds and other dies it doesn't matter.

[BillA]

clearly there is no cross-platform comparability,
that being the case one has to question the purpose of the activity ?

FETs I'm aware of (and I'm NOT expert is this field) have a wickedly concentrated heat source, as you say - put enough copper in there and it can spread; now the edge effects will eat your lunch

and as there is no definition to the die, there can be none wrt (with respect to) the temp measurement sensor position

sorry davidzo, this hardware and software does not appear to be useful

[Incoherent]

As i said the die is not included in this projekt, so you can use Diesims from Copper, steel, Nickel, wood :D or whatever you want. The thicker the diesim is, the better the heat distribution allover the diesurface will be. Of course the diesim adds another isolating Layer, but i think when you don't compare results with other testbeds and other dies it doesn't matter.

I would think that one of the goals should be cross comparability.
Of course, if this is just the electronics side of things then all well and good but I'd be concerned that the sensor type is not very flexible, they are relatively large and difficult to position in a manner condusive to repeatability between different units.
Is anyone who has one of these trying to do these kind of comparisons? What is their approach to the die?

I can say I am fairly impressed with the overall unit electronically, but the details are very awkward to glean due to my language limitations. I think the electro-thermal interface could be a problem for people trying to do comparisons.

I am working on something similar. I'm interested in other peoples approaches.

>> And OK I'll repeat the gist of everything everybody else says with almost the exact same words :) (cross-post with Bill, I am slow)

[BillA]

"Wasserkühler Prüf Station" is I think rather misleading
this is a data collection and presentation system

those things related to watercooling as such are not 'included'

a terrible repeat: if you cannot resolve to 0.01°C, with repeatability, you're pissing up a rope

pH, Joe Citarella, and myself all do so

[Incoherent]

a terrible repeat: if you cannot resolve to 0.01°C, with repeatability, you're pissing up a rope



For the record Bill I disagree...

With some qualifiers.

It's a matter of proportion, depending on what where and how you measure, scale and statistics.
This could lead to a new variation of your statement. If you can not obtain an effective resolution of 0.01°C, with repeatability, you're pissing up a rope.

0.01°C is pretty arbitary applicable perhaps to waterblocks, repeatability is the important part. The resolution is only important for establishing the detail of how repeatable the metrology actually is. You can obtain accuracy greater than your resolution.

He says blithely, teaching the fish how to swim...

Hmm, well, this is something for me to defend tomorrow.

[BillA]

I have no quibble with that change, the whole statement is rather general
the resolution is quite related to the magnitude of the change being observed

[davidzo]

the fet has a heatsource of almost 9*9mm, thats a small durondie, but not what i would call wickedly concentrated lol. repeatability is given in any way. just because there is no definition where to mount the sensor, its not mounted at the wrong place :D
I drille a hole into my copperdiesim and placed the sensor in it (you can make this sensors a lot smaller and faster with sanding off some plastic cover material). The copper diesim is connected to a real CPU, where i sanded the thin layer of plastiks away and soldered onto the copperlayers. That gives me an optimal mounting pressure and thats whats most important in testing. Iv you use sockets and mobosimulators from steel and unflexible cpus, you cannot simulate the Elasticity of the material itself and get the wrong pressure. Pressure differences kan easily make a 1-2K difference, whats much more than the differences you can get when the FET is only 9*9mm insted of the 11*11 of the AthlonXP and a 5mmthick isolated coppersim is between it... The cpu is and Palomino XP, which is a very good compromise in diesizes today. not as big as an Clawhammer but bigger than barton and prescott.
Thats only one solution. The project didn't included this standards, because diesims and cpus change time by time and this project should not be connected to any socket and any Diesize. Also the user should be free to measure the temp where he want and to measure with what he wants. for example you can also use 2, 3 or 4 tempsensors to read the cputemp if you would like it more.

[BillA]

is 9x9mm the package size, or the FET itself ? (have you seen a thermal map of a FET ?)
I think you are far off the mark with your diesim 'ideas'
you need data, then I'll rejoin such a discussion

in any case the heat source, cooling, etc. is not part of the "Wasserkühler Prüf Station"

perhaps i was too harsh when I said the project was 'not useful'
what this project is is a data acquisition system, which your group intends to be the DAQ part of an undefined watercooling test system
no problem
- but $300 on eBay could get me a board with greater capability I suspect, a different approach
I get into test equipment design only when I cannot find something existant to do the task

[davidzo]

the 9*9mm is the metal backplate of a fet. Show me an board from ebay which is better an delivers equal usability...
I paid less than 100EUR on my WPS, sensors included and got one premium software with it :). I don't think that you can get some existant testequipment for that price, that can do the trick so easily, generates graphs and everything for you, can regulate the fans, the heatsource etc. by software.

[BillA]

if you think for an instant that the heat flux is uniform off that FET, re-think it

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=42899&item=3842415953&rd=1
with docs - first one I looked at, probably many more
should work - and less than $100, for now

less than $100 for your WPS is good, I had imagined more than that