View Single Post
  #63 (permalink)  
Old 27th December, 2004, 07:10 AM
Kaitain's Avatar
Kaitain Kaitain is offline
Join Date: September 2001
Location: MK10, UK.
Posts: 4,372
Send a message via MSN to Kaitain Send a message via Skype™ to Kaitain

Originally Posted by gizmo
Yes and no. Depends on the radiator design. I use regular car heater cores, which are a flat tube arrangement. The pressure drop across this kind of rad is very low. I have two of these in my system in series right now, and they work very well. My main concern with running two rads in parallel is that he'll end up in a situation where he isn't getting maximum cooling because the water isn't staying in the rad long enough. Of course, valves could be used to help this situation, but we will get to that in a moment.
This is the problem with concept design. You get very fast heat transfer if the water flows quickly through the radiator. You also get as effective a cooling job if the water flows very slowly through the radiator but spends long enough in there that it still cools down despite the lower efficiency. The former is a more elegant solution, but the latter is rather easier to build.

I have a tendency to err on the side of caution where radiator pressure drops are concerned - the design for most "for overclocking" radiators is a single-pass, many-leg radiator with somewhat tight bends, and a not particularly large pipe diameter. They're small and neat, but promise to provide quite some impediment to flow.

The type of radiator gizmo has specified is already two-pass, in effect, two radiators in parallel in the same package. Being of a flat profile, its wetted contact area is rather greater than that of round-tubed variations (though it will have a higher pressure drop per metre pipe than a round tube). In addition its liquid capacity will be rather high. I rather suspect that on entering the radiator and separating into each of the two passes, the water slows to a snail's crawl and spends a very long time being cooled at low efficiency - remember, car heater cores are designed to remove many kilowatts of heat at high efficiency. In effect it's a reservoir with cooling abillity I have occasionally wondered about using a central heating radiator for the same reason...

Originally Posted by gizmo
Aye, there's the rub. Each valve is a flow restriction, and it is significant (more so than a turn or two of tubing). I would think the reduced flow from all those valves would be even worse than the pressure drop from the rads in series.
Possibly if you had an awful lot of them, but I rather doubt it. I posted this up back in 2002 in answer to another question, when I had the relevant book to hand. It uses equivalent pipe diameters, so is rather easier to calculate than by trying to work out the acceleration on the fluid at each corner and the coefficient of discharge of every valve. It's empirical, so just gives a rough answer...
Originally Posted by Kaitain in 2002
Pressure drop in pipes due to friction is given by:

Pf =8 f (L / di ) .(pu^2)/2

Where Pf = pressure drop in N/m^2
f = friction factor
L = total pipe length in m
di = pipe inside diameter in m
p = fluid density in kg/m^3
u = fluid velocity in m/s

The bits you'll need:

Friction factors: found from a nasty looking graph. A typical value is 0.0035

p for water is approx 1000kg/m^3

Method of equivalent diameters: For each of the following fittings, multiply the number of equivalent pipe diameters by the inside diameter of the pipe and add this to the length of your tube. Use the total sum in the above equation:

45deg std elbow = 15
45deg long radius elbow = 10
90deg std radius elbow = 30 - 40
90deg long radius elbow = 23
90deg square elbow = 75
Tee - entry from leg = 60
Tee - entry into leg = 90
Union and coupling = 2
Sharp reduction (tank outlet) = 25
Sharp expansion (tank inlet) = 50
Gate valve (fully open) = 7.5
Globe valve (fully open) = 300
Plug valve (fully open) = 18

(Numbers after the = are the number of equivalent pipe diameters). This method assumes that the pressure drop due to each of the above fittings is equal to that caused by an equivalent length of the straight tubing.

Whatever value you get for the pressure drop is the pressure head your pump needs to provide to make the system flow.

Shout if you need more info.

(Source: Coulson and Richardson's Chemical Engineering volume 6, ed 3, R K Sinnot, 1999, Butterworth Heinemann)
As you can see, a 180 degree std radius U bend (most common in made-for-watercooling radiators) gives a pressure drop of up to 80 pipe diameters per bend. A well chosen valve results in a pressure drop of a tenth that.

I'm not clear on whether this is meant for turbulent or laminar flow - I suspect it's for laminar, but that the modifications for turbulent flow would be in scale. I don't have the relevant book handy though.
It is by coffee alone I set my mind in motion...
Reply With Quote