Working out flow-rates from BillA's test results
(<-- I'm grinning and blushing at the same time)
In order to answer the flow rates and hence pump selection question, I'll post a summary version of a post I made over at Overclocker's Australia on this subject, which basically explains how to work this stuff out.
-------------------------------------------------------------------------------
Bill's data typical consists of two important graphs. The "flow vs C/W" graphs, and the "flow vs pressure-drop" graphs.
Looking at the White Water review we find the following:
Flow vs Head-loss (Pressure drop)
Flow vs C/W
Where it can get tricky is attempting to predict where one will fall in terms of flow rate. Every 0.01 C/W difference correlates to anywhere from 0.5-1.1C full-load temperature differences depending on how aggressively one over-clocks/over-volts their CPU.
In order to answer the flow-rate question, we must have a pump's PQ (flow vs pressure-drop) chart handy. For the Eheim 1046, 1048 and 1250, these charts look like the following:
So how does one predict how much flow they'll get? First it's important to understand the pressure drop increases roughly proportional with the square of a flow rate increase. If you attempt to double the flow rate, you will enounter 4 times the pressure-drop resistance. If you half the flow rate, you get one quarter of the pressure-drop resistance. This is nice, because it means that even fairly restrictive blocks can still benefit from decent volumetric flow rates because the pressure backs off quite rapidly if the flow rate is dropped by even smallish amounts.
The best place to start is to first figure out how much flow you have through your pump, and then your pump/heater-core.
For my Eheim 1250, it pushes 14lpm through 2m of Tygon tubing alone, and 12lpm through 2m of tubing and my "Big Arse" heater-core which has 1/2" fittings of 11mm inner diameter. ie. the heater-core has fairly low restriction fittings on it.
Looking at the 1250's PQ curve above we see that 14lpm translates to about a 0.85m pressure drop, and 12lpm translates to about a 1.2m pressure-drop.
Now the pressure drop of a full system is the sum of it's individual pressure drops. In order to work with pressure drops as a result of differing flow rates, we need to bring them all back to an arbitrary reference flow rate. Let's choose 10lpm as a fairly convenient flow rate.
At 10lpm, the pump + 2m of tubing exhibits 0.85m / ((14/10) * (14/10)), or 0.85m / 1.96, or 0.43m of pressure drop. Much of this pressure drop resistance will actually be due to the pump's barbs/fittings as a pump's PQ curve is calculated without the barbs in place.
At 10lpm the pump + 2m of tubing + the heater-core exhibits 1.2m ((12/10) * (12/10)), or 1.2m / 1.44, or 0.83m of pressure drop.
ie. At 10lpm, the heater-core itself offers a pressure drop resistance of 0.4m (being 0.83 - 0.43 = 0.4)
The above gives us a good idea of what the pressure drop resistances of various items in our system.
Okay, so how about if we now plug the White Water block into the system?
At our 10lpm mark, the White Water offers around 3.25m of pressure-drop. In our hypothetical full system, the total pressure-drop at 10lpm is 0.83 + 3.25, or around 4.2m of pressure-drop.
Working out some data points to plot against the 1250's PQ graph.
10lpm => 4.2m PD
9lpm => 3.4m PD
8lpm => 2.7m PD
7lpm => 2.06m PD
6lpm => 1.51m PD
Looking at the 1250's PQ curve, we can see that it's going to push somewhere between 6 and 7lpm. So let's work around there:
6.5lpm => 1.77m PD
Actually this pretty much meets up with the 1250's PQ line almost exactly.
So we can predict that the 1250 will push about 6.5lpm in our hypothetical setup.
In real life, I measured 6.7 in this exact setup. The differences are probably due to my measuring errors, but the theory and the in-practise values do seem to meet up quite nicely within a 3% margin of error.
So with an Eheim 1250 you'll be at or slightly below the 0.18 C/W mark.
With an Eheim 1048, it works out to around 4.9lpm. With an Iwaki MD-20RZ you'll get close to 10lpm (haven't done the exact math at this time).