WC simulator discussion.

[Since87]

I brought up the idea of developing a watercooling system simulator in this thread and rather than continue threadjacking there, I'm hoping to continue the discussion here.

Quoting a post from gmat in the earlier thread:

Originally posted by gmat
That's the problem. If you count on only one person to fill the data, the DB won't be nearly useful before 1 or 2 years, and i'm optimistic.
One needs the curves (actually, the data behind them) for every pump, piece of tubing, rad, fan, and waterblock around. The software can interpolate between curve points, using linear or more elaborate techniques, but too many missing points and you'll get inconsistent results.
That said, many curves show a nice inverse exponential look, which can be neatly approximated. But one needs to consider each case (pump, rad, etc) and each subcase to match the math functions as exactly as possible. Not a short work either.
Also, some parameters are very hard to code, like turbulence behaviour for example. This is the domain of finite elements calculus, and out of reach for a small java applet...
We need to define exactly what to expect from that piece of software.

I think a very useful simulator could be developed without modeling every conceivable detail down to the part per million.
For example, modeling a pump as crudely as shown below, appears unlikely to result in a final temperature error greater than 0.1C. (The simulation is a simple parabolic curve.)



Of course if you stack up enough 0.1C errors, the total error becomes significant. More accurate models would certainly be desireable, but even crude models would allow people to make vastly better informed decisions than is common now. Even if such a simulator is not very accurate at predicting actual die temperature achieved, it could still be very useful in doing comparisons between different sets of components.

I certainly don't see a need for finite element analysis in order to produce a useful simulator. The relevant effects of turbulence in the water are, for the most part, already contained in BillA's dP vs flowrate and "C/W" vs flowrate curves.

In terms of the components, I envision something like this:

1) Fans - skip em, just allow a CFM input. Maybe report the pressure drop required to get that air flow rate with the selected rad. (There are a bazillion different fans. Their PQ curves are ugly. Actual performance depends greatly on shrouding, pushing vs pulling, etc.)

2) Pumps - simulated PQ curve are fairly easy to generate. I could set up a spreadsheet in an evening or two that would allow me to quickly come up with simulator equations for pumps that are substantially better than the one shown above. Power put into the water is another issue. Would have to be guesstimated for inline setups unless testing was done. Probably 80% of manufacturers spec'd max power could be used for submerged. A couple dozen pumps ranging from Eheim 1046 to Iwaki MD20-RZ and RX would be a worthwhile starting point IMO.

3) Waterblocks - List all of the commercial blocks known. If Bill hasn't tested it, report something like, "The manufacturers of this block are losers and have not provided the data necessary to simulate the block."

Of the blocks Bill has reported data for, it appears that simulating the relevant curves is usually easy. MCW462 types are the only ones which deviate substantially from simple exponential curve.

4) Rads - Biggest problem area. Curves can be very ugly. No data for most popular models - Chevette and Camaro heatercores, BIX, BIM? Perhaps an algorithm for determining the relevant data for Fedco heatercores, based on the commonly listed dimensions, could be determined. (Maybe Fedco could provide such?)

5) Tubing etc. - Allow four selections:
Typical 1/4"
Typical 3/8"
Typical 1/2"
Typical 5/8"

Eventually add a section allowing a person to detail his tubing/fitting setup? (More detail than I'd likely bother with, but if someone was motivated to generate that level of modeling...)

6) Heatpath through CPU pins to MOBO - use a guesstimated C/W.

Aside from the gaping hole, (data for popular rads) I think something fairly useful could be put together in a reasonable period. Ideally it would be a work in progress; components could be added as data became available, and errors in the results tracked down and corrected.

Seems worth consideration to me, but I'm not much interested in arguing about it. I can do the simulations I'm interested in, easily enough in Excel. Considering the generally dismissive response I've seen to the idea so far, I only posted this, because I said I would. I'm not much interested in pushing string.

[nicozeg]

Excellent project! That could be a very good tool to answer most noob questions seen around.

I think that the piping is not difficult to detail. Lots of people still use ¼ id hoses in long loops, with some tees and elbows in between (sure not this forum readers ), and that account for a big pressure drop. All the numbers for that are currently known and could be added easily. Sure there’s lot of cases where this is the biggest source of pd.

[gmat]

Actually i'd be able to come up with something. With Java, and MySQL.
The problem is, i'm 100% occupied by work currently. Maybe later when things slow down...

For the pump curve, having some points would be neat - then using parabolic interpolation (for example) the simulated curve would match the real one closely.

[jaydee]

3) Waterblocks - List all of the commercial blocks known. If Bill hasn't tested it, report something like, "The manufacturers of this block are losers and have not provided the data necessary to simulate the block."

I don't think any single person on the planet knows exactly how many blocks are out there. I seen one website that had over 100 different commercial one's listed. I doubt BillA will ever see all the blocks available so these results will be limited on what he has done. But this could be a good thing. If this works and people actually use it as a reference, then the manufactures that have not submitted a block to BillA will be more likely to. Then again the majority of water coolers never come to this site before buying, so who knows how it will work out.

Maybe come up with the 10 best of everything. 10 best fans to use, 10 best blocks, 10 best rads, 10 best pumps, ect.... In order to have good usable data each part needs to be considered including the fans being used. One 50CFM fan may not provide the same static pressure of another 50CFM fan. I think the fan is extreamly important and should not be left out. In other words narrow it down to the better parts that are readily available and easy to aquire. Leave out all the "junk" parts.

Good Luck!

[BillA]

so many different aspects to a "Watercooling Calculator" that its a bit difficult to address
particularly with different people talking about different aspects of different components

but it is doable, if someone has GREAT stamina

Since87, if you are going to sparkplug this thing
suggest getting JoeK to setup a separate heading under which there could be 'topics/components' such that the info relevant to a subject was in the same thread

for example rads, characterize them as the mfgrs do:
core type - thickness factor - frontal area - dissipation vs. air velocity
they have this info, we might have to derive it - but such is not really difficult (but rather time consuming)

fans are totally known

wbs are dealt with as a 'black box', just use the test results
(BTW, there are more mfgrs having some testing done than is now apparent, watch OCers)

hose is no biggie

pumps are a HUGE problem, as a 'good' calculator would flag 'bad' selections
how to do ?
how does one convince someone that a pump delivering 15% of its free flow rating is a terrible choice ?
(when such choices are quite the norm today !)

lots of work in this project

[gmat]

Originally posted by unregistered

how does one convince someone that a pump delivering 15% of its free flow rating is a terrible choice ?

Maybe by displaying the flow rating (in %) next to the pump icon... I have a few ideas that i need to formalize. I'll do it this WE as until friday i'm at work...

[Alchemy]

Wow. Very ambitious. I like it.

I'd go about it this way:

1) Get numerical PQ curves for all the major components - WB, rad, pump - and figure out how to combine the data to determine the pressure drop and flow rate of the system. This might take some iteration for a three-curve problem, but I'm fairly sure it can be done by numerical means. I don't know much programming, but I hope any app you're talking about can run a Newton's method pretty easily.

This will solve the flow rate of the system.

I'd ignore effects from tubing and take care of them later with a correction factor. You *need* to run experiments to test these things for at least a few combinations.

2) Get water-side hA-Q diagrams for the radiator and WB, and air-side hA-Q diagrams for the radiator. Input the Q for the water from (1) and the air velocity from whatever fan you're running. You might find it best to have a list of hA values for each fan, since different fans will have different airspeeds depending on the geometry of the radiator.

This will generate the heat transfer coefficients (hA) for the WB, water-metal interface in the rad, and metal-air interface in the rad.

3) Solve for the overall heat transfer coefficient UA using the hA coefficients from above:

1/UA = 1/hA (rad-water) + 1/hA (rad-air) + 1/hA (rad-cpu)

4) Using UA from (3), the ambient temperature, and the heat output of the CPU (I'm ignoring the pump here b/c it requires running a couple of the above steps again, probably for minimal benefit). From that you can tell:

Heat = UA * (Tcpu - Tambient)

Solve for Tcpu and you will know the temperature of your processor.

After this, you'll need to run experiments yourself to determine what errors are generated and what correction factors are needed for specific arrangments. Use these correction factors in the appropriate places in the above steps and you will have a very accurate semitheoretical equation that should make everyone on this board (myself included) bow to your l33t skills.

BillA ain't kidding - this is not for the faint of heart.

Alchemy

[gmat]

Originally posted by Alchemy
run a Newton's method

Extactly what i was thinking about.
As always the theorical part of the program will be done in a few days.
The big and tiring part will be the user interface
Here's a brief - XML plugins for each component, Java / Swing GUI, hidden database (if needed, i'll see if we can solve everything with XML components)
The GUI will allow sequential placement of elements, ie pump->tube section->rad->splitter etc..
The sequential placement will be a grid-type, even columns being elements and odd colums being tube sections (or elbows, splitters...)
This is just a quick brainstorm. As i said i need to think carefully about it. One must not rush a software project at design stage

[hara]

I really think this idea is interesting !

[Alchemy]

Originally posted by gmat
Extactly what i was thinking about.
As always the theorical part of the program will be done in a few days.
The big and tiring part will be the user interface
Here's a brief - XML plugins for each component, Java / Swing GUI, hidden database (if needed, i'll see if we can solve everything with XML components)
The GUI will allow sequential placement of elements, ie pump->tube section->rad->splitter etc..
The sequential placement will be a grid-type, even columns being elements and odd colums being tube sections (or elbows, splitters...)
This is just a quick brainstorm. As i said i need to think carefully about it. One must not rush a software project at design stage

Taking tubing into account will complicate the problem several times over. I'd strongly suggest ignoring friction effects due to them. Unless you have a poorly-designed system like the mess mentioned in pHaestus's article, wall friction and form friction (due to bends, valves, etc.) should be negligible compared to the flow impedences caused by the rad and waterblock.

It can be done, of course, but if you want to make a grid with elements being added and flow calculations made, you're going to be going over your head, I think. At very least you'll be putting in a ton of extra work for very minimal gain in accuracy.

I'm working on my thesis right now so I can't guarantee much speed, but I'd be glad to help on this if you're interested.

Alchemy

[BillA]

do not ignore fittings, specifically single mitre 90s
treat them as 'equivalent lengths', not so difficult given all the other stuff

[bigben2k]

FWIW...

I say ignore the tubing too, but I'd add a recommended tube size, based on the expected flow rate, to minimize the pressure drop.

It's a simple calculation, and leaves few choices: 1/4, 3/8, 1/2, 5/8 (rare) and 3/4. If anyone wants to run a flow rate higher than norm, then the calc can simply state: "too high", i.e. do it yourself.

Bill's right, adding fittings can be simple. On the other hand, one could simply design a spreadsheet to download, so one can simply plug in their own numbers...

[Since87]

Wow, a lot more interest in this than I thought.

As I said in the other thread, I'm not at all qualified to design the user interface for something like this. (And what's being discussed is a lot more elaborate than I'd envisioned.)
I'd prefer to start off relatively simple with a well designed architecture that would allow the addtion of features later. Something like the Kryotherm TEC software where you can click on icons to bring up a 'subcalculator' and fill in the details. These 'subcalculators' could be added at a later date. I'm afraid that if the plan were too grand from the start, the project would die from participant burnout.

gmat,

As I said in the other thread, I'm not at all qualified to work on the user interface. It sounds like you are very well qualified for that, and even if you don't have time to do much coding for something like this, you could be a lot of help in designing a good architecture.

I can work on coding component models and the simulation itself in C. Would there be problems with accessing a C module from XML. I know absolutely nothing about XML.

What form do you see the component models taking? I'd envisioned defining a C function for each component. You'd pass a flowrate to each function and it would return a dP. (and C/W if applicable.) Not the most convenient scheme for adding models I imagine.

jaydee,

I think that if this thing got off the ground, it wouldn't take long to get around the forums that it was here. The people who really wanted to look before they leaped would find it fairly quickly. I would like to see the simulator allow for fan selection, but I think leaving that until after the 'wet side' was done, would increase the odds of something useful being completed.

Bill,

I'm not sure anyone has the 'stamina' to do the whole job by themselves, but with a good overall architecture, the job can be broken up into reasonable chunks and hopefully people willing and qualified will take on the subtasks. It may be possible to recruit some people who do not hang out here. Macklin over at oc-forums offered to do some coding when I was working on a TEC simulator. (Shortly afterwards I discovered the Kryotherm simulator and lost interest in reinventing the wheel.)

I'll contact JoeK about setting up a "WC calculator" topic. Definitely a good idea to keep everything in one easy to find place.

Alchemy,

You're over my head. I had pictured finding the intersection of the pump, and system PQ curves with a search algorithm:

1)Find the dP of the pump at X lpm and the sum of the dP's of the components at X lpm.

2a)If pump dP is greater than component-sum dP double the flowrate.

2b)Else halve the flowrate.

Repeat until desired accuracy is reached.

(Not well stated, but you probably see what I mean.)

There may well be much better methods. I don't claim any expertise in this stuff at all. From my own fiddling around, I've seen that even crude techniques can yield useful insight into the system interactions. I'm definitely interested in algorithms that simplify the coding.

[gmat]

What i had in mind (still need refinements, take note):
- Place elements in a GUI or a XML input file
- Solve the flow rate (by correlating backpressure vx flow data of each element)
- then solve the heat with flow / heat data
- finally with add up heat transfer coefficients and solve the temperature

Adding tubing is quite easy, it will be treated as strict equivalent height. Actually, one will just have to input approx. total length, # of elbows, tees and other fittings. All fittings will be treated has having a flat backpressure vs flow response (= constant backpressure) but if anyone has more data about this, it will be no problem.

Since87: i am a software engineer, that helps
XML is just a text format. The good thing is it's universal. You can export XML files from Excel for example. And parsing XML files in Java is about my day job.
For each component i'll make a 'black box' object model in Java.

I envision releasing first the XML file format so ppl can contribute easily, by adding components. You'll just copy the file in the appropriate directory, launch the software and you'll be able to include it in a sim run.
All i have to do then is to code the simulator itself (rather quick) and the GUI (rather long)
I'll get back there to make sure my maths are all right.

[redleader]

I'm studing programming now, so I won't be much use for a while, but I'd love to see how this works. Its sounding MUCH more interesting then writting C homework that calculates polynomial roots

[Since87]

Originally posted by gmat
What i had in mind (still need refinements, take note):
- Place elements in a GUI or a XML input file
- Solve the flow rate (by correlating backpressure vx flow data of each element)
- then solve the heat with flow / heat data
- finally with add up heat transfer coefficients and solve the temperature

gmat,

I look forward to hearing what you come up with as an architecture. I sent a PM to Joe today, about setting up a category for simulator discussions.

Originally posted by gmat

Adding tubing is quite easy, it will be treated as strict equivalent height. Actually, one will just have to input approx. total length, # of elbows, tees and other fittings. All fittings will be treated has having a flat backpressure vs flow response (= constant backpressure) but if anyone has more data about this, it will be no problem.

If I understand correctly, you're suggesting just using a constant for fitting headloss regardless of flowrate?

I think the PQ curves for fitting will generally follow the equation:

dP = Rf * Q^2

where Rf is a value I've been calling 'flow resistance'.

(If someone knows a proper name for this equation and/or more standard terminology for the value 'Rf' please let me know. 'Pipe length equivalent'? My background is electrical engineering and this equation is like a hydrodynamics equivalent to Ohm's law.)

BillA's waterblock data matches this equation very well. His rad data doesn't match so well. More complex equations will need to be determined for some rads.

In cases where the equation holds true, determining P vs Q for the loop is just a matter of adding up each component's Rf value and applying the equation to the sum.

Even if fittings don't match that equation perfectly, using that equation and an Rf value for the fitting, will yield more accurate results than a constant.

Originally posted by gmat

Since87: i am a software engineer, that helps
XML is just a text format. The good thing is it's universal. You can export XML files from Excel for example. And parsing XML files in Java is about my day job.
For each component i'll make a 'black box' object model in Java.

I envision releasing first the XML file format so ppl can contribute easily, by adding components. You'll just copy the file in the appropriate directory, launch the software and you'll be able to include it in a sim run.
All i have to do then is to code the simulator itself (rather quick) and the GUI (rather long)
I'll get back there to make sure my maths are all right.

So are you volunteering to be lead programmer?

Do you think that splitting the coding into seperate sections so that others can help is practical?

I don't understand this stuff enough to be able to picture what you are talking about. Unfortunately Excel 97 doesn't appear to support XML output. Time for me to upgrade I guess.

Can you post an example of an XML component model?

[saton472]

hey if you want to spend some money you could get inventer from the makers of AutoCad. I am learning it at school and found out that it has thermal and liqued simulators in it all you do is build the system inside the program and run the simulaters.

[Alchemy]

Originally posted by gmat
Adding tubing is quite easy, it will be treated as strict equivalent height. Actually, one will just have to input approx. total length, # of elbows, tees and other fittings. All fittings will be treated has having a flat backpressure vs flow response (= constant backpressure) but if anyone has more data about this, it will be no problem.

Tubing and fittings will have flow resistances that very with flow rate.

They will not have high resistance when compared to the WB and radiator, so you might as well assume these resistances are negligible.

The PQ curve for most flow restrictions is much more complicated than a simple square relationship.

Oh, and after some thought I believe determing flow rate won't require much calculation at all. Simply add the PQ functions for the flow restrictions and find where the point in that curve correlates with the PQ curve for the pump.

Quite easy, I think, to just subtract one function from the other and find the zero.

Alchemy

[Blackeagle]

Since87,

The rads you and I are sending to BillA can add a small bit in having some data, along with BillA's work with Big Momma, BIX ect., to reflect heater core & other rad performances.

Beyond that how to rate the many differant cores? 1 5/8", 2", 2.5" thicknesses. Single or double pass. # of flat tubes per pass.
This area is extremly variable.

Do we want to figure out a way to subdivied the heater cores?

Perhaps figure up the number of a rads tubes X tube length X thickness X2 to get the total surface area for cooling in a rad. But must also consider number of passes and tubes per pass to allow for head loss in narrow double pass vs wider single pass.

A interesting project.

edit:

A second thought: Could also rate the rads by totaling the crossection of the tubes in a rad X # of tubes, this would give a headloss due to total cross sections. Then factor in total length of a rad based on single or X2 for double pass of the number of tubes. The total cross section and total length would then give cooling effectiveness & headloss rate for each rad would it not, while also making the method of comparison managable.

[BillA]

Alchemy

do not assume fittings are negligible (this is a repeat message)
unless the system has no fittings
get a copy of Crane #410, but similar stuff is in the textbooks

a single single mitre 90 will have an effect, and is of no benefit to cooling

or perhaps you wish to show by calculation what 'negligible' means to you
- pHaestus' article is a good starting point, and I can give you some rad and wb flow curves if you cannot use those that are posted

Blackeagle
I have an assortment of rads of many different types that I can flow test
but they are of so many types that generalizing may be a stretch
but some in.³ factor could be ginned up
I can provide a matrix

[Since87]

Originally posted by Alchemy

Quite easy, I think, to just subtract one function from the other and find the zero.

Alchemy

Algebra, while relatively easy for a substantial percentage of people, is far from the easiest way to do this task on a computer.

Unless there is some public domain software library for doing symbolic math, I'd guess that something more brute force would be in order.

[Since87]

Originally posted by Blackeagle

Perhaps figure up the number of a rads tubes X tube length X thickness X2 to get the total surface area for cooling in a rad. But must also consider number of passes and tubes per pass to allow for head loss in narrow double pass vs wider single pass.

We'll have to see from Bill's results whether we can come up with a mathematical model that yields the differences in flow resistance between our two cores fairly accurately. If we can, we may have a model that is generally useful for predicting the flow resistance of 2" thick Fedco HC's. It would be good to check the results for a third substantially different 2" Fedco HC to see how well the model held up.

It appears to me that water may only flow through 11 of the tubes in my heatercore. The top tanks don't seem to line up with some of the tubes in such a way that water would flow through all of them. Without taking the top tank off I can't tell for sure. I may end up removing the top tank after the testing though, depending on the results.

Bill,

I'll be interested in visible differences between the two that you notice.

[Blackeagle]

BillA,

Let me know what information on the cores we'll need and I'll get started at Fedco. Will take a bit of time to do them all, but I'll at least have a start when Gmat has the program ready so I can provide information on part of the cores at Fedco.

And thank you.

[gmat]

Originally posted by Since87

So are you volunteering to be lead programmer?

Since it's 100% my domain, i couldnt say otherwise.
So, it's a yes.

Originally posted by Since87
Do you think that splitting the coding into seperate sections so that others can help is practical?

Yes. I need to set up good foundations first, upon which people will add components.
There will be 2 axis of development:
- Java classes
- XML plugins
The Java classes will be set up on a CVS server, as an open source project. Indeed one needs to know about CVS and Java to get involved. I'll first review the code check-ins, then when the project has a good shape i'll let it go to anyone.
The XMLplugins can be developed by anyone, since it's only a text format (like HTML), and they're meant to be independant.
And example of plugin would be:

<simPlugin type="fitting" name="90° elbow">
<interpolation method="polynomial" order="2" />
<data>
10,20
15,30
20,50
...
</data>
</simPlugin>

This is just an example, the actual format will certainly vary from this.
The idea is to allow ppl like Bill, Since87 and others to design plugins with accurate data and interpolation, relieving the programmers from that task.
The Java part will be made of:
- A polynomial solver
- An iterative solver (for exponential interpolations)
- two XML parsers
- a GUI
The second XML parser will provide a mean of designing a layout without using the GUI -> feed an XML file describing the layout and you'll get the answers.
ex:

<layout>
<pump ref="e1048.xml" />
<tubing ref="10mm.xml" length="1"/>
<fitting ref="elbow90.xml" count="3"/>
<block ref="ddm3.xml" fittings="12"/>
<rad ref="bix.xml" fittings="12"/>
</layout>

The layout manager (and parser) will assume it's a closed loop.
lengths will be in meter, fitting sizes in mm.

All this will require a lot of work, i'll post news of my progress once something is up.

[Alchemy]

Originally posted by unregistered
Alchemy

do not assume fittings are negligible (this is a repeat message)
unless the system has no fittings
get a copy of Crane #410, but similar stuff is in the textbooks

a single single mitre 90 will have an effect, and is of no benefit to cooling

or perhaps you wish to show by calculation what 'negligible' means to you
- pHaestus' article is a good starting point, and I can give you some rad and wb flow curves if you cannot use those that are posted

Alright, clearly you want a program more precise than the one I imagine. That's fine.

My idea was to take the PQ curves for the various blocks and rads and combine them, then use the correlation with that curve to the pump curve to get the flowrate.

If you want to add more items in there, you'd want to design PQ curves for various fittings. Items that you probably don't have empirical correlations for - reservoirs and airtraps like the one in pHaestus's article where there is significant loss of energy due to chaotic flow - will have effects as high or higher than elbows and tees. These can't be correlated with Darcy or Moody.

So, does anyone have roughness coefficients for silicon tubing and form friction factors for specific fittings, then? They do vary by design, size, and manufacture, sometimes by quite a bit.

Alchemy