![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
calculating optimal mass flow rate?
- Thread starter THRESHIN
- Start date
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
mass flow calculations of that sort need a target temperature delta.
since, in watercooling, you want that delta as small as possible and the delta gets smaller with increasing flow, you want the flow as high as possible up to the point that heat dump from the pump starts getting out of hand.
since, in watercooling, you want that delta as small as possible and the delta gets smaller with increasing flow, you want the flow as high as possible up to the point that heat dump from the pump starts getting out of hand.
DFI Daishi might be referring to this equation:
Q = cm(DeltaT)
Q = heat added
c = specific heat capacity
m = mass
DeltaT= temperature change
(divide the eq by time to get your rate..)
c is constant...
If you want your DeltaT to be as close to 0 as possible, you'll need mass to be as high as possible...And since you divided by time to get your rate, you'll need your flow rate to be as high as possible (eg. grams/sec).
Q = cm(DeltaT)
Q = heat added
c = specific heat capacity
m = mass
DeltaT= temperature change
(divide the eq by time to get your rate..)
c is constant...
If you want your DeltaT to be as close to 0 as possible, you'll need mass to be as high as possible...And since you divided by time to get your rate, you'll need your flow rate to be as high as possible (eg. grams/sec).
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
yeah....good old Q = Mc(delta)T, then drop in a time/flow factor......
then there's the performance curve on the blocks and rads always showing reduced thermal resistance with increasing flow rates.
then there's the performance curve on the blocks and rads always showing reduced thermal resistance with increasing flow rates.
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
accuracy? that has no place here!dotK said:
though if you feel like dropping me a hint, i would always like to know more about the nitty gritty.
well the idea behind it is this - what kind of a pump do we really need for a watercooling setup? obviously it varies with the setup. for example, i want to try and design a system that uses one pump and one reservor and splits into two lines to cool the cpu and video card(s) seperately. the flow rejoins in the reservoir. simple, but it prevents dumping warm water on the video card after it leaves the cpu. all that is required is a couple Y-conenctors and some extra tubing.
even without that, once we know an optimal mass flow rate for handling a certian amount of energy it would make it much easier to select a pump more cost effectively. once the flow rate is known, the next step would be to calculate the head required by the pump. that would be calculated using darcy's equation for energy loss and calculating the loss due to friction. since its likely to be laminar flow in a watercooling setup, it can be easily calculated from the hagen-poiseuille equation. combining the result would give us the required head from the pump. using this information, i would like to find out if it is really worthwhile to get a $60-80 pump or if a more simple pump would suffice. if not, its still fun for me and i wouldn't mind finding out exactly what is needed and what is overkill.
darcy's equation for energy loss is the energy loss due to friction of the fluid. since some of it is greek alphabet, please excuse me if i'm not using the proper variables but they are defined.
h(loss) = f x L/D x v^2/2g
where
h(loss) = energy loss due to friction
f = friction factor
L = length of flow in stream
D = pipe diameter
v = average velocity flow (find this from the mass flow rate)
i'd get into the other one but its been over 2 years since i took my fluid mechanics lol. i've gotta read a bit more of my text to refresh my memory before i start trying to get any answers.
what we should end up with here is an exponential curve where our delta T gets closer and closer to 0 but will never quite reach it. i'd like to find out the optimal point where the difference in adding a higer flow rate beyond what you already have makes a miniscule difference.
now off to find some of that data on the waterblocks daishi mentioned....you guys got the grey matter working, so thank you all!
even without that, once we know an optimal mass flow rate for handling a certian amount of energy it would make it much easier to select a pump more cost effectively. once the flow rate is known, the next step would be to calculate the head required by the pump. that would be calculated using darcy's equation for energy loss and calculating the loss due to friction. since its likely to be laminar flow in a watercooling setup, it can be easily calculated from the hagen-poiseuille equation. combining the result would give us the required head from the pump. using this information, i would like to find out if it is really worthwhile to get a $60-80 pump or if a more simple pump would suffice. if not, its still fun for me and i wouldn't mind finding out exactly what is needed and what is overkill.
darcy's equation for energy loss is the energy loss due to friction of the fluid. since some of it is greek alphabet, please excuse me if i'm not using the proper variables but they are defined.
h(loss) = f x L/D x v^2/2g
where
h(loss) = energy loss due to friction
f = friction factor
L = length of flow in stream
D = pipe diameter
v = average velocity flow (find this from the mass flow rate)
i'd get into the other one but its been over 2 years since i took my fluid mechanics lol. i've gotta read a bit more of my text to refresh my memory before i start trying to get any answers.
what we should end up with here is an exponential curve where our delta T gets closer and closer to 0 but will never quite reach it. i'd like to find out the optimal point where the difference in adding a higer flow rate beyond what you already have makes a miniscule difference.
now off to find some of that data on the waterblocks daishi mentioned....you guys got the grey matter working, so thank you all!
alright! found something!
http://www.frozencpu.com/images/products/detail_secondary_hires/ex-blc-270_12.jpg
looks like its somewhere between 1.5-2.0 gpm where more flow doesn't help too much. well for this waterblock anyways, i'm sure it varies slightly per block but its still a good target area. make it say double that for a safety factor (ie. your pump motor starting to die out and begins to turn slower.) and all that is really required is about 240 gph. even if its only 300, pretty simple to get. as for the head required, i think thats where the money goes.
read a little and turns out i should be using either darcy's OR hagen-pouseuille's euqation. they both come to the same answer. since the flow should be laminar in any W/C setup, the only friction will come from friction of the fluid itself.
hagen equation is this:
h(loss)= 32uLv/yD^2
where
h(loss) = energy lost due to friction
u = dynamic viscosity (at 30 C its about 0.0055 Pa.s, i'm using a chart)
L = length of flow (tubing)
v = average velocity of flow (get that from pump flow rate, 240 gph here)
y = specific weight (should be greek gamma, but oh well)
D = diameter of tube
so with that the loss due to friction can be calculated and taken into account. so the head needed from is simply the total vertical height the pump needs to move the fluid plus the head lost due to friction. so if a typical W/C setup has a total height of say 0.75m, simply add the friction loss to that and voila - some nice pump specs. within a reasonable safety factor of course. it should not be taken as the answer calculated. so if the answer was say 2m, make it 3 or 4m.
i'm GUESSING here from this that all that is really needed for a typical W/C setup is a pump capable of 250 gph at a pressure head of only 3 meters. maybe even two, i don't want to guess at the length of flow path at the moment. i want to take the time to estimate with some tubes and have an educated guess.
http://www.frozencpu.com/images/products/detail_secondary_hires/ex-blc-270_12.jpg
looks like its somewhere between 1.5-2.0 gpm where more flow doesn't help too much. well for this waterblock anyways, i'm sure it varies slightly per block but its still a good target area. make it say double that for a safety factor (ie. your pump motor starting to die out and begins to turn slower.) and all that is really required is about 240 gph. even if its only 300, pretty simple to get. as for the head required, i think thats where the money goes.
read a little and turns out i should be using either darcy's OR hagen-pouseuille's euqation. they both come to the same answer. since the flow should be laminar in any W/C setup, the only friction will come from friction of the fluid itself.
hagen equation is this:
h(loss)= 32uLv/yD^2
where
h(loss) = energy lost due to friction
u = dynamic viscosity (at 30 C its about 0.0055 Pa.s, i'm using a chart)
L = length of flow (tubing)
v = average velocity of flow (get that from pump flow rate, 240 gph here)
y = specific weight (should be greek gamma, but oh well)
D = diameter of tube
so with that the loss due to friction can be calculated and taken into account. so the head needed from is simply the total vertical height the pump needs to move the fluid plus the head lost due to friction. so if a typical W/C setup has a total height of say 0.75m, simply add the friction loss to that and voila - some nice pump specs. within a reasonable safety factor of course. it should not be taken as the answer calculated. so if the answer was say 2m, make it 3 or 4m.
i'm GUESSING here from this that all that is really needed for a typical W/C setup is a pump capable of 250 gph at a pressure head of only 3 meters. maybe even two, i don't want to guess at the length of flow path at the moment. i want to take the time to estimate with some tubes and have an educated guess.
zer0signal667
2[H]4U
- Joined
- Jan 11, 2002
- Messages
- 3,847
THRESHIN said:
1- There is no vertical head if your loops is continuous, only frictional losses.
2- You're forgetting about the bulk of the restriction in the loop- the waterblock
Flow is laminar in the tubing, but is not within:THRESHIN said:
- Certain points of the pump
- Certain sections of flat radiator tubes
- Most regions of all modern waterblocks
Flow may be turbulant in some sections of the pump and radiator and highly turbulant in blocks. The majority of the pressure drop of a common waterblock lies within the jets, accelerator nozzles, changes in cross-sectional areas [baseplate plenum, mid-plate plenums, inlets, outlets] and turbulators. Friction within most regions of a waterblock is also moderate.
The added restriction of two Y-connectors (and their joining flow paths) would negate the benefit of not passing the "warmer" coolant over the GPU block. In reality, the temperature differential between coolant at various points in the loop is unsubstancial and the absorbed heat from the sort of heat sources we deal with is only a small fraction of the total capacity of the coolant. The Y-connectors also add a degree of routing complexity to an already complex system.THRESHIN said:
ah thank you zero. like i said, its been a while
as for the water block, the equation could be calculated using the minimum diameter in the system - the channel in the water block. sure might make it a little bit of overkill, but thats ok by me.
this makes me wish i had access to the machine shop back at college - i'd make my own water blocks in no time on the CNC mill and then finish them with the surface grinder! damn!
as for the water block, the equation could be calculated using the minimum diameter in the system - the channel in the water block. sure might make it a little bit of overkill, but thats ok by me.
this makes me wish i had access to the machine shop back at college - i'd make my own water blocks in no time on the CNC mill and then finish them with the surface grinder! damn!
zer0signal667
2[H]4U
- Joined
- Jan 11, 2002
- Messages
- 3,847
THRESHIN said:ah thank you zero. like i said, its been a while
as for the water block, the equation could be calculated using the minimum diameter in the system - the channel in the water block. sure might make it a little bit of overkill, but thats ok by me.
this makes me wish i had access to the machine shop back at college - i'd make my own water blocks in no time on the CNC mill and then finish them with the surface grinder! damn!
You're going to end up with such a rough approximation that I would venture to say it's not worth your effort to determine such selection criteria. The principle behind any decent block is some kind of extreme turbulence-inducing feature. I may be wrong, but I can't imagine that modeling this system with 100% laminar flow will yield anything close to realistic numbers. I'm intersted to see whether or not I'm right though... So have at it
DFI Daishi
2[H]4U
- Joined
- Feb 14, 2005
- Messages
- 2,839
just to point out the obvious........there are pretty well validated numbers for the performance vs fow and pressure drop vs flow on most rads and blocks.
the way that i have always considered going about calculations to determine exact performance etc is to pick numbers off of those flow vs head curves for all parts in the loop, add them (principle of superposition for curves), plot the result, then plot the flow vs head curves for any pumps of interest on the same graph and determine intercepts, and plug those flow values into the flow vs perfumace curves for block and rad to determine total thermal resistance for the loop.
this totally neglects the influence of tubing, though, which is where your fancy (to people who have never had to do a proper fluid dynamics course) equations come into play.
as previously mentioned, the difference in coolant temps from one point to another in most loops is really small. often smaller than standard PC thermal robes can resolve.
http://overclockers.com/topiclist/index31.asp#WATER COOLING
check out these rad reviews for an idea of the kind of heat dump from the pump that would start to be detrimental, and the flow rates when the rad doesn't benefit much from added flow. the other reviews aren't so great, but the rad testing is some of the best going.
http://www.swiftnets.com/
swiftech provides performance data for all of their blocks, and details the testing methodology they use (whether you agree with it or not) and until the contrast between the performance data that they provide and what an independant review of the apogee showed, their numbers have been pretty kosher.
the review mentioned: http://www.systemcooling.com/swiftech_apogee-01.html
i have no issue with different test beds shifting all blocks one way or another, however when the actual shape of the performance curve changes, and blocks move drastically relative to one-another........
the way that i have always considered going about calculations to determine exact performance etc is to pick numbers off of those flow vs head curves for all parts in the loop, add them (principle of superposition for curves), plot the result, then plot the flow vs head curves for any pumps of interest on the same graph and determine intercepts, and plug those flow values into the flow vs perfumace curves for block and rad to determine total thermal resistance for the loop.
this totally neglects the influence of tubing, though, which is where your fancy (to people who have never had to do a proper fluid dynamics course) equations come into play.
as previously mentioned, the difference in coolant temps from one point to another in most loops is really small. often smaller than standard PC thermal robes can resolve.
http://overclockers.com/topiclist/index31.asp#WATER COOLING
check out these rad reviews for an idea of the kind of heat dump from the pump that would start to be detrimental, and the flow rates when the rad doesn't benefit much from added flow. the other reviews aren't so great, but the rad testing is some of the best going.
http://www.swiftnets.com/
swiftech provides performance data for all of their blocks, and details the testing methodology they use (whether you agree with it or not) and until the contrast between the performance data that they provide and what an independant review of the apogee showed, their numbers have been pretty kosher.
the review mentioned: http://www.systemcooling.com/swiftech_apogee-01.html
i have no issue with different test beds shifting all blocks one way or another, however when the actual shape of the performance curve changes, and blocks move drastically relative to one-another........
Right on. The Reynolds number of most regions of a waterblock is typically in the hundreds of thousands.zer0signal667 said:
Decided to crunch the numbers for a typical tubing run. This is for one 1/2" ID (10 inches in length) section of tubing assumed to be reasonably straight and assuming a flowrate of 6 lpm and the coolant's dynamic viscosity of 800 10^-6 kg/ms.
Velocity is .789 m/s. Flow is turbulant with a Reynolds number of 125000. This is assuming no surface roughness, so very little surface friction and an unrealistically low Reynolds number. So, I was a bit off about my assumptions of the type of flow in tubing given reasonable (1+ lpm) flowrates. Flow should be turbulant at every point of a typical loop assuming the flowrate is greater than ~1 lpm.
ah more good stuff i assumed that the flow would be laminar so i did not bother to check the reynold's number. silly me.
so given that the flow is going to be turbulent, for just the head loss in the tubing the equation
h(loss) = f * L/D * v^2/2g
where f is the friction factor, for plastic/vynal tubing it is considered smooth and therefore a relative roughness factor is not needed. using the reynold's number from phide and a moody's diagram here the friction factor should be approximately 0.026-0.027.
only problem with this is the water blocks. they're copper and DO require the relative roughness factor and the difficult part is that its pretty hard to find roughness data for the internal channel of a water block which you have no idea of the surface finish other than what you can see by eye....
the good news is i might be able to get my buddy to machine some water blocks for me. now the next step is to come up with a design and figure out where the hell i'm going to get my hands on some copper....
i assumed that the flow would be laminar so i did not bother to check the reynold's number. silly me.so given that the flow is going to be turbulent, for just the head loss in the tubing the equation
h(loss) = f * L/D * v^2/2g
where f is the friction factor, for plastic/vynal tubing it is considered smooth and therefore a relative roughness factor is not needed. using the reynold's number from phide and a moody's diagram here the friction factor should be approximately 0.026-0.027.
only problem with this is the water blocks. they're copper and DO require the relative roughness factor and the difficult part is that its pretty hard to find roughness data for the internal channel of a water block which you have no idea of the surface finish other than what you can see by eye....
the good news is i might be able to get my buddy to machine some water blocks for me. now the next step is to come up with a design and figure out where the hell i'm going to get my hands on some copper....