flow v.s head

[fhorst]

Hi All,

I'm getting a bit confused about the flow vs head thing.

What's better, buld a Wc setup with lots of restrictive parts to build up the pressure, or make it as smooth running as possible?

I read about placing pumps in series, then first going to the CPU before going to the rest. some preffer to place a rad after the pump and then to the CPU, some say, place it after the CPU, to give the CPU more head. (as the rad will cost some)

what about this setup:
one 1/2 (19mm) hose, with first a T 19-4-19mm for my GPU, followed by a T 19-4-19mm for my HDD, then again a 19-4-19mm for my North bridge, and finaly go with an elbow to my CPU (13mm). and the same for the outlets.

This will clean up my hoses (now looks like spagetty), and will give the best flow.
Maybe a 5 or a 6 mm for my GPU, as it produces also quite some heat, but you get (I hope) the idea.

Only thing I'm OC'ing is my CPU (2.8@3.5 ghz)

Sure my head on my cpu will be less, but the overall flow will be great!

I have 2 L20 pumps, one will push, the other will pull. The extra pump will compensate (I hope) for the head loss.

As the water flows quicker throug my radiators, they will perform a lot better (I know, if the water would be in a low felocety in the rad, the water has more time to cool down, but the disorption of heat by the rad will be worse, as a better flow will improve that.)

At the end the water temps will rise, with a high and a low flow. Only difference will be that with a high flow, the heat transportation will be better, and the water goes through the rad every x seconds.

[jaydee]

Head is the amount of pressure the pump can provide. Generaly you want as little flow restriction as possible. But if you have a flow restrictive system then you wnt a pump with a higher amount of head pressure to push the water through the restricted area and maintain a decent flow rate.


So IMO you want to try and build as least restrictive system you can.

[Zogthetroll]

basic way to think of it, any restriction in the system results in head loss, which really takes away from the head (or pressure) supplied by the pump. this will result in a drop in flow rate. For some answers on actually setting up a system and the effects of different types of fittings (avoid T's and elbow fittings like the plague) check out THIS article. should explain some of the basics for you. also, there were a couple of threads here very recently about dual pump setups, you might do a search for those.

[fhorst]

Thanks for the replys.

I did not read that article jet.

This guy ends up with a nice clean system. If I would only cool my CPU, it would be "simple".

This is what I have, and what I want to cool:
Hydror L20, (**2x**)
BlackIce Pro 13mm in/out
DTec Rad 13mm in/out
Astec rad 8mm in/out
Whitewater CPU block 13mm
DD Z 875chipset block 13mm
Innovatek Graph-O-Matic 8mm
Innovatek Graph-O-Matic Gforce4 10mm
Koolance Hard Drive Cooler 6mm (**3x**)
Dangerden Reservoir round, 10mm
Astec reservoir round 7mm

CPU P4 2.8, (OC'ed to min 3.5Ghz)
GPU G-force4 (Not OC'ed
Nothbridge
3x HDD

Going from the CPU to the NB needs to be quite a loop, if I want to use a serial setup.
Same goes after the NB to the GPU.
Then I need to split fo het to my HDD's... in total 6 hoses (3 in, 3 out)

This is why it looks like spaghetti.

I want to reuse as much as possible (i paid for it....), as long as it is usefull.

My hose with the T's was a bit simplefied. this is what I'm building:
http://www.liquidninjas.com/photopo...o.php?photo=982

It will act as a reservoir, and as a way to distribute the water.

As I want to reuse as much as possible, I want to use all 3 radiators, split the tubing in 13/13/8mm, and make a second reservoir, where the water from the rads flow into.
This will acht as feeding res for the other pump. (first one gets feeding from the first reservoir)

I'll be using 20mm hose, from the res to the pump. That should do the trick (I hope)

So, if I get it correctly, the only reason to gain more head, is to overcome the waterflow loss that is coursed by the restriction of all the items used (waterblocks, hoses e.g.)

At a certan point a waterflow increase won't add anymore performence gain so then a higer pressure between the copper of the block and the water will give some performence gain.

I did read in a article that the minimal temp diference between the water and the copper wil be with a hig flow about 2 degrees celcius. Adding extra pressure will gain you an other 1 degree.

If you are working with a low flow, then that 1 degree gained by pressure will be a lot more, say 3 degrees.
But I don't know if that increased pressure will be better than a high flow..................

Did anyone messure? a block with a 300 l/h flow (flow is 450, but 150 goes to the GPU, NB and HDD's) and "no" pressure or a block with an 150 l/h flow with high pressure.
This is in theory the flow rates I will get with a paralel or serial setup. In the serial setup I did place the 8 and 6 mm's paralel, to sustain a reonaseble flow.

[Ewan]

You seem to be assuming that flow and head are two distinct characteristics. They are not. One (head) is a function of the other (flowrate) and can therefore be discarded.

Flowrate is the only measurement you need to worry about since it is what does the cooling. Whatever pressure the water is at, is completely unimportant.

Head means pressure and it's a stupid word.
Liquid flow through a system is motivated by a pressure difference. i.e. it's the pressure difference that makes the liquid flow form the high pressure place to the low pressure place. In a closed system the flow will go from the pump output to the pump input because the pressure at the pumps output is higher than at it's input.
Because of friction between the liquid and the pipe walls as the liquid flows through the pipe, there is always a certain resistance to the flow which will slow it down. This resistance is expressed as a pressure drop, i.e. the pressure differential you need from one end to the other to get the fluid to flow at a certain speed through the pipe.
The pressure drop gets worse if you have kinks and other obstacles which the fluid must negotiate as it goes through the pipe.
Pressure drop also gets worse as the flowrate increases, naturally because the frction force will be greater, because the fluid is rushing past the walls faster.
A watercooling pump is almost always a centrifugal pump, so it will pump as hard as it can to overcome the system pressure drop. So the flow of the fluid will be at a point where the system pressure drop is equal to the pressure that the pump is providing. The higher the flow, the less pressure it is able to provide, so an equilibrium is reached between the pump pressure and the system pressure drop. i.e. the flowrate will be such that the two are equal.
But the system pressure drop is not dependant on the pump, only the flowrate. Which means that if you put a pump which is capable of higher pressures, then the flowrate will increase so that the system pressure drop (which increases when you increase the flowrate) increases to match the pressure being provided by the more powerful pump.

Having said that, most water cooling pumps provide as much flow as their presure will allow them, so getting a pump with better "head" is generally the way to go. This is because watercooling pumps have such pitiful pressure (often less than 2m of water) that they are unlikely to ever put out their peak flow.

Another type of pump is the diaphram pump which will always put out a specific flowrate because it uses pistons to meter out a certain volume of liquid with each stroke. Such a pump will always produce the same flowrate regardless of the pressure drop that that flowrate generates, unless the pressure drop is so high that the pistons aren't able to provide that pressure, in which case the pump simply wont work. Such a pump would have no probelm with whatever restrictions were in a watercooling loop since they can put out pressure of well over 50m of water and in such a case you would get the one with the highest flowrate.

As I said at the start, the water pressure is not important to the water cooling, only flowrate. However flowrate is the same throughout the loop, whereas the water pressure will be greatest coming out of the pump and the least going into the pump. Where you put the pump in relation to everything esle is for this reason not important since the flowrate will be the same regardless of where you have the pump.

However heat transfer (which is the final goal) is helped by temperature difference between the water the whatever it is you transfering heat to/from. This means that the colder the water going into the CPU waterblock is, the better the cooling. Obviously the water will be coldest coming out of the radiator so the radiator should be positioned infront of the CPU waterblock. By the same token the water will be the hottest after it's gone though all the heating elements. That being all your waterblocks and your pump (the pump itself will add some heat). The water should enter the radiator when it at it's hottest so that it will be able to remove as much heat as possible. So the radiator should be positioned after the pump. Like this:
Pump -> Radiator -> CPU block -> other blocks -> back to the pump.
If you have two pumps have them right next to each other in series before the radiator. If you have them in a push-pull configuration then it simply means that you are adding heat from the 2nd pump to the water after the CPU block before the water goes to your other blocks. This means that the other water blocks wont work as efficiently becasue the water will be warmer.
Having two pumps in series doubles the pressure which should provide a roughly 40% increase in flowrate.

[georgeteo]

Is the flowrate constant throughout the setup??

[jaydee]

Originally posted by georgeteo
Is the flowrate constant throughout the setup??

The only way the flow rate can change is if water is added or removed somewhere in the loop. Which shouldn't be the case. So yes, the flow rate stays the same once the system is up and running.

[fhorst]

Thanks for clearing that up.

2 pumps in a row, will be quite a line-up

Will the 2 L20 have a better performence that one L30? (or an emheim 1250)
L20 - 1.85m 700L
L30/emheim 2.5m 1200L

As I have 3 rad's, I'm thining about building a seperate box for the rads, and place the computer on top.

It must be possible to fit the pumps in there.

I found out (and was reading about it in forums) that sandwich a radiator gives the best performence. 2 fans on one side won't work as good as a sandwch. Don't know why, but that's how it is. (lower water temps)

The box will have a front of 2x 12cm and 3 8cm fans.
The rads will have a 12cm fan sucking out the air. All running at 7v, to keep the noise down.

Nice hobby... water cooling

[nikhsub1]

Originally posted by fhorst
...Will the 2 L20 have a better performence that one L30? (or an emheim 1250)
L20 - 1.85m 700L
L30/emheim 2.5m 1200L

In theory yes. Running the 2 L20's in series should yeild you a max head of 3.7M (1.85 x 2 = 3.7). It will not change the max flow rate but your flowrate will be higher due to the higher pressure. To double flow rates you need to quadruple the pressure.

[8-Ball]

Originally posted by nikhsub1
To double flow rates you need to quadruple the pressure.

Not necessarily true, I would've thought.

Surely it depends on the operating conditions.

8-ball

[pHaestus]

Flowrate is the only measurement you need to worry about since it is what does the cooling. Whatever pressure the water is at, is completely unimportant.

What flow rating is to be considered when purchasing a pump then? 0head flow (since pressure doesn't matter? That's a big step backwards as this number has little to do with real world performance. Flow and Pressure are intrinsically related (by a pump's P-Q curve) and this curve characterizes the pump's performance, To design a system to achieve a certain flow rate requires one to know the total resistance of the loop (the sum of each resistance). Then one can choose the pump to meet the design specs considering other concerns as well (heat added, size, noise, cost, availability).

I think that 2x flow means 4x pressure is a bit misleading as they should be related (roughly) by P = Q^2?

[Alchemy]

Originally posted by pHaestus
I think that 2x flow means 4x pressure is a bit misleading as they should be related (roughly) by P = Q^2? [/b]

Try something like P = Pmax * ( 1 - (Q/Qmax)^2)

Centrifugal pump PQ curves are usually darn close to parabolic, so if you only have a handful of data points, this equation is accurate enough.

Alchemy

[Since87]

Originally posted by pHaestus
I think that 2x flow means 4x pressure is a bit misleading as they should be related (roughly) by P = Q^2?

It seems to me that both statements say the same thing.

If, under operating condition one:

flowrate = 1 gpm, and pressure = 1 mH2O
and P=Q^2 -> 1=1^2

and under operating condition two:

flowrate = 2 gpm, and pressure = 4 mH2O
and P=Q^2 -> 4=2^2

The equation is more generally applicable, but nikhsub1's statement probably gets the point across to a larger audience.

Alchemy,

nikhsub1 and pHaestus are talking about the PQ curve of a cooling system excluding the pump, and you are talking about an equation for a pump's PQ curve.

Better quality pumps seem to be closer to parabolic than others. Danner pumps in particular have fairly linear PQ curves. I have no idea why.

The equation P=kQ^2 is just a simplification of the Darcy equation with unit conversions and other fixed values rolled into k, and the assumption that the f (friction factor) in the Darcy equation is a constant. It seems that frequently this assumption can be made and 'reasonably' valid results obtained. (As illustrated by the following graph.)



Measured data is from unregistered (BillA). (Eyeballed graph in the White Water case.)

The equation doesn't always hold though. The Becooling Slit Edge has a much 'flatter' PQ curve than this type of equation would predict. Again, I have no idea why.

[pHaestus]

I really like these graphs of Les's:



You can see the resistance curve of the system increase as flow rate increases, and you can see that this curve intersects with all of the pump P-Q curves. THIS is now useful info in selecting a pump. I will get around to testing a couple more loops and then writing that article up one of these days I guess...

[nikhsub1]

Originally posted by pHaestus
I think that 2x flow means 4x pressure is a bit misleading as they should be related (roughly) by P = Q^2?

For me, and I'm sure others, this formula "to double the flow you must quadruple the pressure" is a good 'rule of thumb' and not always dead accurate but pretty close.

[Ewan]

pHaestus

quote:
--------------------------------------------------------------------------------
Flowrate is the only measurement you need to worry about since it is what does the cooling. Whatever pressure the water is at, is completely unimportant.
--------------------------------------------------------------------------------

What flow rating is to be considered when purchasing a pump then?

My comment was in reference to the idea in the first post that the pressure of the water in the loop had some importance and that having pumps in various places somehow helped pressure.

When it comes to sizing a pump, while you can use a bunch of pump curves, it would seem to be superflious since watercooling is fairly well understood and you can use the rule of thumb "bigger is better" more effectively than pump curves. Pump curves are kinda useful if you happen to know your system resistance at varying flowrates. To get this information you have already had to purchase a pump and flowmeter. However pump curves are good for comparing one pump against another if you are unsure of what to get if you have anumber to choose between.

At the end of it all we're all stuffed because the best waterblock designs will be high pressure drop designs, and aquarium pumps just aren't suited for this.

I know that a few people are prepared to put some money into their water cooling. Why don't they get diaphram pumps instead?

[Skulemate]

Originally posted by Ewan
... I know that a few people are prepared to put some money into their water cooling. Why don't they get diaphram pumps instead?

Because that's not really needed. There's nothing in a typical water-cooling setup that a good industrial quality centrifugal pump cannot handle.

[fhorst]

Originally posted by pHaestus
THIS is now useful info in selecting a pump. [/b]

Hmm, it looks to me that the eheim 1048 gives 5L/m and the 1250 6L/m (at a higher presssure drop) (with the WW, big arse and 2m tubing)

only 20% gain with a double rated pump.....

Looks like I'm gonne stick with my 2 L20's

If one breaks down, can I replace it with a L30? Having a 700l and a 1200L pump... would one restrict the other?
It seems to be that I'll never will get the rated 700L from my L20, so placing a biger one beside it, sould not hurt (I think)

[PureSyN]

I have a ehiem 1048 that I bought to replace the swiftech mcp300 which started to leak cuz of housing cracks. I sent the pump back for warrenty and should be getting it back soon. I don't know whether they'll replace it with another mcp 300 or maybe they'll replace it with the mcp600, in any case is it ok to hook up different pumps in series? or will it put stress on the slower pump?

[Ewan]

It's fine to have different pumps in series

There's nothing in a typical water-cooling setup that a good industrial quality centrifugal pump cannot handle.

Watercooling pumps are nowhere near industrial quality though. Having a higher pressure output would allow for very restrictive waterblock designs, which should be more effective.

[PureSyN]

I have the swiftech mcw5000 water block for the cpu and the mcw 50 for my gpu, do you think it would make a heck of a lot of difference or should I just set the extra pump aside for a "leaky day" ?

[Skulemate]

Originally posted by Ewan
... Watercooling pumps are nowhere near industrial quality though. Having a higher pressure output would allow for very restrictive waterblock designs, which should be more effective.

That depends on what you mean by "water-cooling pumps." There are plenty of people around (myself included) that are using industrial pumps to power their systems... I've got an Iwaki 20RZT that can handle pretty much any block thrown her way.

[Ewan]

I've got an Iwaki 20RZT that can handle pretty much any block thrown her way.

Yeah, OK. That pump seems very suitable. Quite impressive in fact. Is it noisy?

[Since87]

Originally posted by Ewan
Yeah, OK. That pump seems very suitable. Quite impressive in fact. Is it noisy?

I can't say for the MD20-RZ, but my MD20-R is very quiet as long as the vibration is isolated. About as loud as an Eheim 1048?

Effectively silent when compared to my PSU fan.

[Skulemate]

My Panaflo L1As almost drown out the pump noise entirely. Considering its power it is a very quiet pump.