a question about water turbulence....

[Mnx4]

I read the article recently dealing with water physics and how sometimes the water does not properly flow through the heatsink system. I was thinking that for increased turbulence, one could create a heatsink who's contents was hollow except for a small rotor/fan. I'm talking about a propeller that moves because the water would be forced inside the heatsink, by means of the inward flow, and be moved a little faster because of the water being moved out. Would this not work? here's a little diagram. Please help me cause I would really like to perfect this idea (if it is actually a good one). Please give me any input at all. Thanks. Mnx4

PS. I would have posted the diagram on the page, but i really dont know how to do it.. it someone could help me about that.. thanks :D

[nikhsub1]

Well you may be on to something but that particular design has some flaws. First of all it is best to have the water enter the block in the center, right above the cpu. With the paddle in your design it is fixed (center) where water should enter. Second, having a moving part in the block can only cause more problems with failure, pieces breaking off etc. This of course would ruin anyones day. The new water blocks are getting wise to the "turbulence" factor and some new ones that are coming out look really good. Creating obstacles or ridges in the channels does a great job of achieving this. If I had the tools to make my own block I would make the channels similar to the barrel of some guns with the twisted ridges throughout. It is almost as if the water is on a cork-screw rollercoaster. Good thought though!

[Greedy Guido]

If the flow through the block is more then 0.02 L/s (20mL per sec) then it will be turbulent. This assumes the the hydraulic diameter is approx 7mm and turbulent flow is at Reynold's Number >4000. In most cases turbulent flow will start to occur at Reynold's Number >2000 due to bends in the channel.

You should not have to add anything to the channel to get turbulent flow.

[UserName]

Greedy Guido;
Of what value would you give Mikes article?

[Greedy Guido]

Who is Mike.

[Greedy Guido]

Read the article and it was almost like being back at uni. One thing that it does not allow for is the fact the most water block fittings come in at a sharp right angle.

This theory is based on flow over a flat surface and holds true. It should be noted that the transition zone is for Reynolds Number between 2000 and 4000.

However if you have high velocity water (high Reynolds number) entering a water block first at right angle this helps to quickly establish full turbulent flow (if turbulent flow had not already established). This can be tested by adding a clear top to a water block and injecting a small amount of dye using a syringe (this tends to be a standard uni experiment to see the effect of reynolds number).

This can also be seen in industry where the fluid that is being pumped is a slurry (slurry is a liquid containing suspended solids). If you place a sharp bend in the pipework or have a sharp entry, then this area is quickly eaten out by errosion due to turblence. In slurry pumping this types of bends are to be avoided.

So I am not sure that it is necessary to add these devices to create turbulent flow. This would be an interesting experiment. In my opinion most water blocks on the market would benifit little by doing this. I COULD BE WRONG.

[BillA]

not so sure this is worthwhile, but

"eaten out by errosion due to turblence"
no
perhaps a terminology thing

most slurry lines are eroded due to sliding abrasion, also in the ells due to impingment plus sliding abrasion
turbulence is needed to maintain the solids in suspension, but is not the cause, per se, of erosion

the erosion due to cavatation is quite different

be cool

[Aesik]

I'm working on the follow on article still (getting sidetrack by a different article first) that will deal with some real world numbers. But just to toss out a few numbers right now since you guys have brought it up, turbulent flow should be brought about in most systems by having a flow of 30-35 gph (as Guido previously noted, this is for a Reynolds number of 3-4000). Any obstruction in the path of the flow, i.e. angles, corners, and especially a pump impellar will help induce turbulence. However, there is a big catch to this. Any such obstruction will also increase the head loss and reduce the overall flow. It must also be noted that turbulent flow in and of itself will increase the total drag that the fluid exhibits on the system and thus also reducing the overall flow rate. So just like with many things in life, it becomes a balancing game.

I'll be making another post shortly asking for some real world numbers from all the water coolers out there to help get a feel for where everyone's current systems stand. Hopefully we can gather enough information to see where current systems need improvement.

[The Overclocker]

i would just get a 'normal' maze pump and put twisted wire in it to disrupt water flow

[UserName]

Are you all saying given any system, if the flow is fast enough, then turbulence occurs naturally?

Are there simple ways to establish the Reynolds number for a given block?

[Aesik]

If your flow is fast enough, that alone will trigger turbulence.

Calculating the Reynolds number for flow through a straight, smooth-walled, tubular shaped pipe is quite easy. Throw in other geometry factors and then things start getting significantly more difficult.

Generally speaking, any variations of geometry will help increase or 'trigger' turbulence. There are however exceptions, such as the scales on a shark which actually help to reduce turbulence and create laminar flow (thus reducing drag). This principle has been applied to all sorts of things like wet suits, airplane wings, etc. and is not something we want to do in water cooling.

[Les56]

Dunno what it means but you can use the sub-menus in Kryotherm* to predict the flow behavior in various channel configurations.
For water at 20c, my interpretation suggests that Flow will become Turbulent(ReynoldsNo.> 2,000) at a flow rate of ~ 0.09cubic metre per hour in a chamnel 6mm wide x 10mmm deep eg Maze 2.1**
http://www.jr001b4751.pwp.blueyonder.co.uk/Laminar.jpg
http://www.jr001b4751.pwp.blueyonder.co.uk/Turbulent.jpg

* http://www.kryotherm.spb.ru/soft.htm
** http://forums.overclockers.ws/vb/showthread.php?threadid=51894

[Greedy Guido]

Originally posted by BillA
not so sure this is worthwhile, but

"eaten out by errosion due to turblence"
no
perhaps a terminology thing

most slurry lines are eroded due to sliding abrasion, also in the ells due to impingment plus sliding abrasion
turbulence is needed to maintain the solids in suspension, but is not the cause, per se, of erosion

the erosion due to cavatation is quite different

be cool

Straight sections of the pipeline wear due to sliding erosion (there are three types or erosion wear, sliding bed, low augular impact and high angular impact). However, at bends the wear is due to firstly impingment (where fluid hits the bend face)and directly after the bend wear occurs due to turbulence (this can be on the opposite side of the pipe). The method is still due to impingment but this is caused due to turbulent flow (usually low angular or high angular impact).
Another plave wear is encountered in slurry lines is at the pump discharge where the pipe changes size. There is no change in direction but the sudden expansion causes localise turbulence and hence wear. There are many good texts on this but the best I have found is Solid-Liquid Flow Slurry Pipeline Transportation by Edward Wasp.

The other to forms of wear are Abrasion (these are crushing, grinding and low stress) and corrosion. These definitions are directly from SVEDALA Slurry Pump Handbook.

Les56,

At 6mm by 10mm the characteristic dimension should be 7.5mm by my calculations. This is known as the Hydraulic diameter or radius and is what should be used to calc Reynolds number as stated in most text books for non circular channels.

I must admin that I used a higher temp for the fluid but this would change Reynolds number only by a small amount.

[Greedy Guido]

To calculate Re for a rectangular section use the following

Re = (r x V x D)/m

Rh = A1/Wp

D = 4 x Rh

V = Q/A2

Where

r is the fluid density in kilograms/metre3 (kg/m3)
=998 kg/m3@20°C
V is fluid velocity in circular pipe section in metres/second (m/s)
D is diameter of pipe in metres (m) or equivalent pipe diameter
m is viscosity in pascal seconds Pa.s
= 0.0010049 Pa.s @ 20°C
Rh is hydraulic radius
A1 is area of rectangle channel in metres2 (m2)
Wp is wetted perimeter of rectangle channel in metres (m)
A2 is area or pipe with diameter equivalent to calculated D in metres2 (m2)
A2=(p x D2)/4
Q is the flow rate in metres3/second (m3/s)

For the channel that is 6mm by 10mm then the follow applies.

For Re =2000 the flow rate is 0.012 l/s or 0.0433 m3/s

[Aesik]

.012 l/s=1.2e-5 m^3/s or 11.1 gph

A Reynolds number of 2000 would seem a bit low to me for turbulent flow. Most current texts will call 2300 the beginnings of turbulence, with 4000-6000 for fully developed turbulence.

[Les56]

For me,with the rise in Nusselt No. from 3.25 to 22.21 also being shown ,the Kryotherm result is the more attractive:

0.1 m3/h = 100 l/h = 26 g/h = 0.44 g/m ?? Hope I have conversion correct and not 2 or 3 orders of magnitude out?


This flow rate, probably coincidentally, perhaps does not contradict some of my own observations* using a Maze2.2 for Peltier cooling a CPU.

http://www.jr001b4751.pwp.blueyonder.co.uk/untitled2.jpg

* http://www.tekforums.co.uk/posts.php?threadId=1605&page=4

[Greedy Guido]

Fully turbulent flow is at Re>4000 but can occur as low as 2000. The range from 2000 to 4000 is the transition zone. At Re between 2000 and 4000 the flow can become turbulent (if it is not already) if something interups the flow (ie bend) and once it is turbulent it will stay this way unless RE drops below 2000.

However, as most blocks have a right angle entry, then this in most cases will cause the transition to turbulent flow if Re>2000.

Les56,

Do you even know what the Nusselt No. is used for. I was unaware that it was an indication of turbulent flow. From memory it is used for dimesionless analysis for thermal bounday layers. It is used as a scaling factors on models to calculate full size equipment. I don't think you known what your talking about.

While your at it throw in the Sherwood Number for dimensionless concentration gradient. What about the Biot Number, Bond Number, Eckert Number, Fourier Number, Grashof Number, Jakob Number, Lewis Number, Peclet Number, Prandtl Number, Schmidt Number, Stanton Number and Weber Number. I sure they helped me calculate turblent flow last time I designed a pipeline (sarcasm).

[Les56]

Greedy Guido
I am much more at home with dislocation dynamics than fluid dynamics.
Nusselt No is related to ease of heat transfer and as such I thought was very relevant. I do not need to understand to refer to work (ie Kryotherm) that I think contributes.However if you think they are frivoloous I will happily delete my posts.

[Greedy Guido]

The original topic was about turbulence and that is what I though the discussion was about. I still cannot see where the Nusselt Number fits in when talking about turbulence.

Can dislocation dynamics by applied to fluids as I though it was used for solids (usually metals). I am not familiar with dislocation dynamics as most of my work has been in fluid dynamics (slurry pumping) and some thermodynamics.

[BillA]

Guido, are you trying to prove your sig file ?

firstly, thanks for your clarification of the sloppy terminology contained in the "eaten out by errosion due to turblence" phrase
(and I did indeed suspect that my words would be found to be lacking, go back and check)

for someone with such a vastly superior amount of knowledge, I guess it's difficult to forgo the opportunity to tell someone that they don't known (sic) what they're talking about

may I ask what the partial pressure of oxygen is at the elevation of the stool you're perched on ?

why do I bother?
I cannot help but compare your comments to those of Mike Larsen, another extremely knowledgable contributor to this site

comeon, be a good sport - please DO explain ALL about
the Sherwood Number for dimensionless concentration gradient, the Biot Number, Bond Number, Eckert Number, Fourier Number, Grashof Number, Jakob Number, Lewis Number, Peclet Number, Prandtl Number, Schmidt Number, Stanton Number and Weber Number

do you even understand what you did NOT address ?

Les posted a graph with 2 curves, and he previously defined his channel width

why did you not apply your expertise to that DATA, looking for a correlation with turbulence ?

instead of picking your nose in public

BTW: Les is indeed quite knowledgable about the watercooling of TECs

be cool

[Aesik]

I fail to see why bringing the Nusselt number into this discussion is such a crime. The whole discussion stems from trying to improve the way water blocks work in the first place and the Nusselt number is a very integral part of that. Especially in turbulent flow! The Nusselt number used as a measure of the heat transfer on the surface of the flow and is directly affected by the Reynolds number when the flow is turbulent. Whether you use the Popov, Gnielinski or Dittus-Boelter correlations (among others as well).

You can talk theory and equations all you want, and drag every field of physics into the mess, but when it comes down to it we are all on the same team. A healthy discussion about heat transfer in cooling systems is going to involve more than just the Reynolds number. I see no reason to get nasty and attack anyone.

And yes Les, your conversions look right.

[UserName]

A know a few numbers that are pretty hot. All of them are fast, and a few of them are pretty turbulent too!

I would guess the whole theory of water blocks might take more than one thread. But like all things in life one thing touches another.

I guess it's really up to us to figure out if we want to do a whole text book here or just a paragraph. I think all the info was posted in good faith, but may be off topic. I suggest a time out for those who find themselves being negative.

I think the original post was very focused. After reading mikes article, do we need to worry about turbulence? our answer is not if you flow more than 35GPH thru a 3/8" opening straight, less if it's half.

May be a great set of numbers would be flow rate and diameters to exceed to get turbulent flow, and another set with a bend, and another set with a right angle, and more if you think of them.

Being an ignorant cuss from way back I know how much people need to be told they are idiots,. If we all aren’t satisfied yet maybe we could start a new thread about idiots who use numbers.

[Aesik]

As a general rule, no, most water coolers will not have to worry about their system being turbulent or not. However, I think some would be very surprised to find out that their pumps are not moving anywhere near as much volume as they thought and their flow may not be optimal.

Once a flow is turbulent, it doesn't magically mean that all is well however. As turbulence increases, so does the heat transfer BUT so does the overall pressure loss (thus reducing the flow rate). We just can't have our cake and eat it!

There are just such a huge amount of factors that go into determining how efficient and how well and water cooling setup performs. I think that's what is so attractive to me about it. There is a ton of research that can be done on it and we are just scratching the surface. I'm working on the follow on article to the first, but it's getting put to the side for a short bit while I work on a different subject. This discussion has been good for me to see what kinds of questions people have and given me some ideas of what I'd really like to dig into and find some useful results.

[BillA]

while my intuitive understanding is ok (I guess), I can calculate the numbers for pipe flow;

is not turbulence also somewhat a matter of degree ?

so if there is a correlation between say the Re number and heat transfer rate,
would not more turbulence be beneficial ?
(things are being simplified here, I know)

and is not a distinction between the turbulent flow regime (within the tube or channel)
and the turbulent flow effect (at the wall/fluid interface)
worth making and considering separately ?
- certainally there different things one can do to affect them individually

please do correct the terminology (proper definitions enhance comprehension)
yea, you too Guido

be cool

UserName - slick numbers, had to laugh

[UserName]

It's nice to know my humor is appreciated by some.

AS mike's article pointed out there is a large increase in heat transfer as we move from laminare to turbulant, so i think most of us just want to make sure we are above that. A guideline list might make that easy and we could put in the water cooling faq's. I assume such a thing exists.

Now i know Bill wont be satifyed with that and we all want to go deeper.

More turbulance will impinge deeper into the boundary layer and cause more heat transfer. I think bill you are on to something with the wall fliud interface. If we assume turbulant flow, is there a shape of the wall that is better? I would assume smooth as the closer the turbulant flow gets to the wall the better the heat transer. Any other ideas?

How do we get turbulant flow to impinge deeper into the boundary layer. Or said otherwise, how do we minimise the boundary layer?

Is there anything more important to heat transfer to the water than the turbulant flow thru the block?

[BillA]

anyone have an image of the inside bottom of the original Senfu wb
(I just saw one but can't remember where)

the channel bottoms have undulations (call that macro)
and I have data on surface finishes (call that micro)
[ and smooth is not the right answer ]

be cool

as an aside:
I've said before, and still believe, that the Senfu kit was very well engineered

[UserName]

Your pic

http://www.anandtech.com/showdoc.html?i=1205&p=2

Why is not smooth the answer? I don't really care if it is ir isn't but i do care why?

[Aesik]

Heh, now things are starting to get heavy! Let me try to answer some of the questions first by using/stating just heat transfer and fluid dynamic theory. Then I'll interject some of my opinion and thoughts; I'd hate to get the two mixed up!

First off, the statements of fact, and hopefully I get these all right :)

Turbulence is a matter of degree. It's safe to say the higher the Reynolds number, the higher the turbulence.

There is a direct correlation between the Reynolds number and the heat transfer rate, WHEN the flow is turbulent. The Reynolds number can be used to calculate the Nusselt number which in turn is used to calculate the convection coefficient.

from BillA: "and is not a distinction between the turbulent flow regime (within the tube or channel)
and the turbulent flow effect (at the wall/fluid interface)
worth making and considering separately ? "

While you can consider them and their effects separately, a change in one will change the other. Thus they are pretty well tied together and must be treated together. Flow regime characteristics directly effect heat transfer, and the termperature profile of the fluid effects the flow (albeit at a much lesser degree than the flow effect the heat transfer).

Username is correct in that by reducing the laminar boundary layer, heat transfer is enhanced. There are two main ways to change the boundary layer: change the flow, or change the physical channel characteristics. Increasing the flow rate will increase the Reynolds number and increase overall heat transfer. Creating wall geometry that disturbs and mixes the flow further induces turbulence. In this case, having a rougher wall is better. However (and this is a big however!) increasing the wall roughness also increases drag and head loss which reduces flow rate, which of course reduces heat transfer. The same thing follows for any change in the channel geometry such as turns or restrictions.

Ok, the above should pretty much be fact, not my opinion. So as not to confuse the two I'll now make a seperation in the post:

===Fact mingled with opinion from now on!!===

Now here is a loaded question from Username: "Is there anything more important to heat transfer to the water than the turbulant flow thru the block?" Well, I'd say the most important factor in heat transfer in a waterblock is going to be total flow rate. However, turbulence is deeply mated with the flow rate so you could argue that turbulence is the most important. I guess it really depends how you view it, but when it comes down to it they are so integraly entwined that it's hard to say one is more important than the other. Just to make things even more complicated, you could consider the case when the flow has induced turbulence, i.e. the flow would normally be laminar, but due to geometry or other factors the flow turns turbulent at a very low Reynolds number. To sum up this rambling paragraph, more flow=more heat transfer, more turbulence=more heat transfer and more flow=more turbulence.

One other thing to consider is that channels with extra geometry such as ripples, ridges or even just roughness also increase the surface area from which the heat can be transfered. But once again back to the double edged sword, extra geometry restricts and reduces flow.

Finding the balnce between channel geometry and flow rate is the real challenge. There are infinite combinations, thus infinite solutions to the problem. Finding which is best is our challenge!

Ok, quick sidenote to easily calculate the Reynolds number in a CIRCULAR duct/pipe/tube: (this is just my take on what greedy already posted, sorry if it may seem repetitive)

Calculate the average velocity:

V=4*Q/pi*D^2

Q=flow rate in volume/time (ie m^3/s or gph)
D=diameter of the pipe

Reynolds number:

Re=V*D/gnu

gnu is the kinematic viscosity of water, and for our temperature ranges would be 1e-6 m^2/s

The flow will start to turn turbulent at approximately Re=2300 unless it is otherwise induced due to the geometry of the channel.

The above correlations can be used for rectangular channels by substituting the pipe diameter with a hydraulic diameter:

Dh=4*b*h/[2*(b+h)]

where b=base length and h=height of the channel.

Personally, I can't stand to do calculations using English units and always do mine in metric. If you want to use English, some useful conversions are:

1m^3/s=2.12e3 ft^3/min=1.59e4 gal/min
1m^3=264.2 gal

[UserName]

If we assume a fixed pressure can we deduce what wall shape gives the best heat transfer?

How do we describe how turbulent the flow is and calculate its heat transfer ability?

[Aesik]

Quick answers before I head off to take care of my sick kids:

Yes, but it can be incredibly difficult for anything but a straight, smooth walled pipe. Anything else starts to get really complicated and enters the realm of using CFD (computational fluid dyanmics).

There are some simple (and I use the term simple VERY loosely) correlations for describing turbulent heat transfer in smooth walled pipes, otherwise See above ;)

[UserName]

Very sorry to here about your kids. I pray it's nothing very serious.

When you get a chance i would love to see the math that describes the very simple relationship between turbulance and heat conductivity.

[Aesik]

Ok! You caught me here still ;) Kids have foot, hand and mouth syndrome and just pray you never have to deal with it. While not serious in the long term, it's a bitch to deal with. Add to that some touches of a virus causing lots of throwing up and you have a fun time on your hands!

When I have a little more time, I'll show you some of the correlations involving the Nusselt number and how to calculate the covection coefficient.

[Greedy Guido]

Looks like I upset a few people.

The point is the average person will find little use for the Nusselt Number unless they are keen to perform some serious heat transfer calcs.

And yes, the Nusselt Number is related to the Reynolds number. I believe it is a function of the Reynolds Number (describes the flow) and Prandtl Number (property of fluid). It is used for a scaling foactor and after check with my thermo books at work it is also used to help calculate covection coefficients.

Nu= (h x d)/k

How is that going to help the average water cooling person. Once the average person knowns that he is in turbulent flow and therefore in an area with better heat transfer conditons what is he going to use the Nusselt number for. OK, maybe then they will be able to calculate forced convection coefficients but to go to the next step and actually calculate the heat transfer rate, I doubt it.

To perform a heat transfer calculation for even the most simple water block will mean performing a 3 dimensional heat transfer calculation. To perform this correctly by hand then you must set up a matrix similiar to a finity element analysis matrix used in 3 dimensional stress analysis.

I work for one of the largest engineering consultants in my country and I doubt that there is oone person within my company who could do this correctly (and that includes myself). The only person I ever met that could do this was a senoir professor at the uni I went to and he was a specialist in this field.

So, where am I heading with this.

All I want to point out that there is little point in telliing the general water cooling person out there to calculate any thing other then the Reynolds. All they need to do is calculate Re and ensure that they have good turbulent flow (get Re as high as possible without adding to much pressure drop).

This is what the original person wanted to know. I though it would be a good idea not to confuse them in thinking that turbulent flow was indicated by Nusselt Number.

I must apologise to Les56 for being so aggressive. It thats I get lecture regularly by my friend (and girlfriend) for giving to much info when a simply answer was needed and now I try only to answer the question and not throw in info that is not required.

I still think going to the lever of calculating the Nusselt Number is an overkill.

[BillA]

as much as I am enamored of CFD (or is it CFD software vendor's claims ?), Guido is quite correct in saying that few can use them
- which is why I've taken the empirical approach, testing

BUT, a candidate modification can best be made based on even a general understanding of the theory involved
- while I am forever being taken to task for being overly technical, I think that rigor in analysis does help

climbing on favorite soapbox:
from a WCing systems design veiwpoint it is not so difficult
- one needs to prioritize based always on the contribution to cooling

the pump's head/flow rate is what there is to work with, though it can be varied by pump selection
one should attempt to minimize all head losses EXCEPT those in the wb (where they "may" contribute to increased cooling)

then, for the available head/flow rate, the wb that performes most effectively under the resultant conditions can be selected
(the incremental contribution of the wb's flow resistance is added to all the other components for the actual flow rate)

the problem is that the wb mfgrs do not have (or disclose) their wb info
note that this is not a problem for those selling systems or kits, as they have presumably "designed" them for optimum performance within their size and/or cost limits

DIYers can presently only swap units to try and find "the best"
but since "the best" is quite dependent on the system parameters,
one size does not fit all

kit vendors should be able to provide a better performance package,
but then one is competing with a $200 package against the bull-s..t claims made by a supplier of a $120 package

and since the "reviewers" can't even measure the heat from a light bulb, . . . .

room for a good testing co here (pimp, pimp)

be cool

[Greedy Guido]

I totally agree with BillA about testing. It is the only accurate way of establishing true heat transfer rates for these types of water cooling blocks. Unless you have access to a Super Computer.

[ButcherUK]

Originally posted by BillA

as an aside:
I've said before, and still believe, that the Senfu kit was very well engineered

Not sure I'd go that far. The block design was good, but the tubing was too thin and restirctive, the rad is pretty weak, and they didn't warn about copper/alu corrosion issues. Maybe the weak rad is just given todays high-heat cpus, but the tubing issue was still there.

[Aesik]

Rather than me try to present the equations for the Nusselt number using such a limited text editor on a message board, I think I'll just refer you to a few websites:

http://www.egr.msu.edu/~somerton/Nusselt/
http://www.owlnet.rice.edu/~azul01/momentum.htm

If you are really a gluton for punihsment, here is a link to a pretty decent online text:

http://www.mas.ncl.ac.uk/~sbrooks/book/nish.mit.edu/2006/Textbook/

for some reason there isn't a direct link to chapter 13, which has some really good stuff:

http://www.mas.ncl.ac.uk/~sbrooks/book/nish.mit.edu/2006/Textbook/Nodes/chap13/node17.html

As you can see, it's not as simple as taking a set equation and "plugging and chugging" out an answer.

Another thing that is of great importance is to realize that most of these equations used to calculate the heat transfer characteristics are also assuming that there is either a constant heat flux along the entire path, or constant temperature. And of course we know that this case just isn't true and so some huge assumptions have to be made. These equations also assume that we are dealing with a straight pipe.

Without using extremely complex mathematical methods, the best case that can be modeled is a straight pipe with smooth walls and having a constant temperature or constant heat flux along the entire length of the channel (that means top, bottom and sides as well as along the length). I'm not trying to be a party pooper, but I spent a over 2 years studying heat transfer and over 2 more on fluid flow and I still have to have my stack of texts next to me to do any accurate calculations.

Guido already made the point that if we are to really come close to modeling the true flow through a water block, a 3d analysis must be made. That include all sorts of really nasty integration, finite difference solutions and other assorted fun stuff. This is the very reason that CFD was developed in the first place.

I'll try to work up some really simple examples to at least show how all of this is tied together, but it will take a little time to get it into a form that can be posted here, so please be patient ;)

[BillA]

point conceeded Butcher, not quite "very"

but it took a while to do better
I have (from JoeC) one of their rads; and if one cuts away the excess sheetmetal it does rather well at low flow rates
- their design basis

ok Aesik (or Guido),
I didn't really want to do this, but now I have to ask

as near as I can determine there are 3 "good" candidates for this type of CFD
Flotherm
Icepak
Floworks (for Solidworks)

I've played with the demos some (except for Floworks) and can see a bottomless pit of thousands of hours (at least for me)
-- and THEN one starts to make and instrument models

I'm a bit more encouraged over Floworks given the relative simplicity of creating the Solidworks model
as well as it's much lower cost

Flotherm's latest control panel has some very spiffy design permutation checking capabilities, but even a "special deal" is $3K for 3 months
- Hell, it'll take me 6 months (or 12 !) to learn it

do you have any experience with these ? or others ?

be cool

[Aesik]

grr...nice double post...

[Aesik]

Bill, you'll get as many opinions as there are software programs out there ;)

Besides those you mentioned, there is also:

Softflo
Flow Pro
CFX

and more.

Honestly, I've never had alot of time to delve into any of these packages. Most of my past work has been done on high-conductivty composites and not on fluid based systems; but that's changing now! Through my school, my slave-driver heat transfer professor had us write all our code from scratch using Fortran using mostly finite-difference solutions. I've played around a little with Floworks and Flotherm. From my limited knowledge, Flowtherm is probably one of the more accurate packages out there but Floworks sure is nice to use because of the Solidworks interface.

The learning curve for using any CFD is STEEP, even with having lots of formal training in heat transfer and fluid dynamics. The hardest part is learning how to set up your boundary conditions to properly describe the flow. I'm not saying it can't be done, but I'd liken it to digging yourself out of an avalanche with a teaspoon. And at $3k for 3 months, that's a tad over my budget!

[Aesik]

Ack! How could I forget COSMOS/Flow. That's another heavy hitter.

[BillA]

HELLO ?

"high-conductivty composites" ?

I have a 2x2x.2 in. piece of "a proprietary patented aluminum composite" cold plate/wb bp
that I'm trying to figure out how to test

what can you tell me ?

it's pretty soft and supposedly can be soldered

be cool

[Aesik]

I worked mostly with carbon composites uni-directional fabric/resin form. Namely K-1100 which conducts about 3-4 times better than copper in it's pure form.

What exactly makes your sample a 'composite'? Any idea what's mixed in with the aluminum? I had a pretty wild setup for testing the thermal conductivities of solid thermal links and their interfaces at anywhere from cryogenic to above ambient temperatures. Unfortunately I no longer work for the space lab and they didn't let me make off with their equipment.

What exactly are you looking for?

[BillA]

isotropic carbon

[BillA]

more seriously,

I don't know what's in it, but the buzz word seems to be "alumina ceramic"
it's quite light, in fact I just dunked it to be sure it would even sink
and as I said, I can carve it with a knife

this seems to be different from the "other" aluminum composite going around that requires diamond tooling to machine

(both of these are upcoming watercooling products)

my preference would be to determine it's thermal conductivity directly - but I've none of that kind of equipment

and I'm not inclined to pay as it is not my product

any round-about easy way ?

be cool

[Aesik]

Just found this link over at [H]. Haven't had much time to go over it, but it looks to be a pretty decent text, especially to dl for free.

http://web.mit.edu/lienhard/www/ahtt.html

[Greedy Guido]

I have seen COSMOS/Flow in action as well as TRANSYS at uni (I think that is what it is called). These types of software take a long time to learn and even longer to set up the models so they are accurate.

They also require fairly powerful computers else you will be waiting days to get answers. They also cost a lot, in fact I think you have to donate a kidney to get a copy.

I have also seen a few of the cheaper packages but they tend to simplify the model and loss accuracy in doing so.