Y pressure drop / Rads parallel or series

[Kemist]

Hey guys

Im building an external rad box and i have two MCR-220 rads. As of right now im not sure how im going to plumb them. I will be running two 1/2 modded ddc pumps in series. Basically i am considering (A). Running rads in series or (B). running in parallel using a wye to split the flow into the rads, then from the rads running a hose from each to the reservoir (so two hose inlets into the res).

I know there is basically no way to know without proper testing which way would be better. I am wondering however if wye's are associated with large pressure drops? Right now im leaning towards going parallel rather than in series. While flow rate would be lower the Dt (air/water) of the water entering the second rad would be higher compared to the Dt of the cooled water entering the second rad in series. Since i will have two pumps i am assuming that the Dt component will be more important in lowering temps than the water's flow rate, since flow rate should be relatively high.

Just wondering everyone's thoughts on this and also that whole wye pressure drop question. Thanks

[redleader]

Given the sheer pumping head and radiator area, I'm guessing the difference will be immeasurably small as you will have extremely low radiator C/W either way. I'd route whichever way allows for easiest/neatest plumbing.

[greenman100]

Dt across rad will be less than a degree, more than mad eup for by increased water velocity of the seriec config


go series

[bobo5195]

water velocity really doesnt matter much.

As long as Mass flow water * Cp water is above the equivlant of air. Look at Bills graphs

[Cyco-Dude]

Originally Posted by greenman100

Dt across rad will be less than a degree, more than mad eup for by increased water velocity of the seriec config

exactly how will having a higher water velocity in the radiators help? wouldn't it be better to have the water moving slower in the radiators (i.e., spends more time in the radiators) to transfer more heat to the fins? or was that disproved?

[jaydee]

Originally Posted by Cyco-Dude

exactly how will having a higher water velocity in the radiators help? wouldn't it be better to have the water moving slower in the radiators (i.e., spends more time in the radiators) to transfer more heat to the fins? or was that disproved?

It was disproved.

[Long Haired Git]

Cathar tested dual BA radiators and found that serial was better in his rig on the day he did the testing. Its the case that parallel results in half the flow rate in each rad which does reduce performance, but it also reduces the overall restriction and thus increases overall flow rate to a small degree, but then you also have to factor in the impedence of the Y fittings.

The "Y"'s add restriction, most notably for 3/8" tubing and barb rigs. My modelling (cringe, I am crap at it) shows 1/2" barbs 30mm long with 8.46mm ID creates approx 20cm of head at 5m. 3/8" with 7.1mm ID Y is 40cm of head. I've assumed the Y "convergance" has the same head characteristics. Note also that my modelling results in barb conversion effects about 20% higher than noted in BillA's testing of HE rads with different barbs.

Time for approximator!
Swiftech MCP655 @ 5, Swiftech Storm block over 70W. In serial, I have 2m of tubing, whilst in parallel I have 1.2m outside of the two Y peices, and then 10cm runs from the arms of each Y to the radiators.
Each MCR220QP radiator has a fan at 7v as most people run rads like this these days.

Serial: 6.4 LPM, pump adding 16.2W, water temp at exit of pump is 2.2°C, CPU delta-T is 11.5°, radiator C/W is 0.0492
Parallel: 6.05 LPM, pump adding 16.0W, water temp at exit of pump is 2.3°C, CPU delta-T is 11.6°, radiator C/W is 0.0529

In parallel, I approximate the head of each Y connector at 29cm of H20.

Anyone want to do a PQ graph of a common brass 1/2" Y for me? Even a single flow rate value would be SOMETHING!?!?!

Assuming my approximator is correct, it looks basically the same either way you do it. So, plumb it the way that's most convenient, which is typically in serial.

[bobo5195]

Y splitter is something like 0.3*mean velocity^2 of highest speed side (NOT Q)

[billbartuska]

Originally Posted by Cyco-Dude

exactly how will having a higher water velocity in the radiators help? wouldn't it be better to have the water moving slower in the radiators (i.e., spends more time in the radiators) to transfer more heat to the fins? or was that disproved?

Whatever the velocity of the water is , it will spend the same amount of time in the rad. Doubling the velocity also doubles the number of times the water passes through the rad per given time period. But increaced flow will result in increaced efficiency due to turbulance For example see http://overclockers.com/articles1158/

Also note the effect of airflow. Trippling the water flow increaces cooling about 34%, but just doubling the airflow increaces cooling about 38%.....for this rad under the stated conditions....user results may vary.

[redleader]

Originally Posted by Cyco-Dude

exactly how will having a higher water velocity in the radiators help? wouldn't it be better to have the water moving slower in the radiators (i.e., spends more time in the radiators) to transfer more heat to the fins? or was that disproved?

Swiftech actually publishes the data for this radiator. THeres a fairly strong dependenance on higher flow, at least with stronger fans. Google for the page, its quite interesting.

I suppose the answer to this question also depends on the fans used. If you have very low airflow, the difference between parallel and serial for radiator C/W will be vanishingly small, however the overall decrease in flow resistance may be worthwhile.

[Cyco-Dude]

alrighty then, thanks for filling me in guys. i've been outta the loop for awhile

[Long Haired Git]

Originally Posted by bobo5195

Y splitter is something like 0.3*mean velocity^2 of highest speed side (NOT Q)

I got: Head in meters H20 = 0.0088 * Flow in LPM * Flow in LPM.
I did the pressure drop for each barb as a tube of the right ID, the pressure drop for the lip of the entry to the tube, and then the PD over the split, and then the PD for the two exit tubes at half the flow of the original.

Eg: 1/2" Y with 8.46mm ID at 4 LPM entry:

Entry lip: 7.9 PSI
Entry barb: 0.8 PSI
Y section: 4.9 PSI
Two exit barbs at half flow: 0.25 PSI each = 0.5

Hmmm, I add the two exits together, which is a mistake. Nuts. Anyway, adding and converting to mH20 I get 0.143m. I am out by 0.5PSI, but that will be like 3% which is within my margin of error.

I suppose the answer to this question also depends on the fans used. If you have very low airflow, the difference between parallel and serial for radiator C/W will be vanishingly small, however the overall decrease in flow resistance may be worthwhile.

1. My simulator's pretty good. Dropping to 5v fans, parallel is still worse, by about 0.9 degrees C. At 12v this drops to a mere 0.06 degree difference.
2. Current model shows the Y joints result in a larger resistance than the benefit of parallel radiators, so no overall flow rate improvement.

[Kemist]

Long Haired,

If I am not mistaken you made the calculations using two 1/2 wye's. If i do go parallel I will only be using one 1/2 wye and each rads outlet will have a dedicated line to the reservoir. Additionally, i will be using 7/16 tubing which should reduce the lip's effect on flow, correct? Taking the above into account this would probably reduce the difference between serial and parallel making them even more comparable, correct? Thanks for your input guys.

If you did mean only one wye then sorry i misunderstood you. Anyway from your simulator and most peoples comments it seems like it isnt gonna change things by much so ill probably just plumb whichever way is easier.

[Long Haired Git]

Originally Posted by Kemist

If I am not mistaken you made the calculations using two 1/2 wye's. If i do go parallel I will only be using one 1/2 wye and each rads outlet will have a dedicated line to the reservoir.

Will change things, but I have no PQ graph for a reservoir.
I am in the process of blowing up my code in order to enforce tubing ID to barb OD correctness, so will post new model soon.

Additionally, i will be using 7/16 tubing which should reduce the lip's effect on flow, correct?

Ahhh, but due to the smaller ID, more restrictive than 1/2" tubing. I cite lack of PQ graphs and testing, and declared overall it was the same and don't cater for it. You lose on the tubing, win with the lip, and so flow wise you're even. Flexibility wise you're ahead, so you win.

[Kemist]

Originally Posted by Long Haired Git

Ahhh, but due to the smaller ID, more restrictive than 1/2" tubing. I cite lack of PQ graphs and testing, and declared overall it was the same and don't cater for it. You lose on the tubing, win with the lip, and so flow wise you're even. Flexibility wise you're ahead, so you win.

Yeah i didnt really think about the restriction of the tubing. Im just gonna go series, the pressure drop on the mcr220's doesnt look all that high anyway, at least according to the graphs on swiftechs site.

[Long Haired Git]

Kemist: I used that very same data for my calculations. Pump PQ graphs are easily sourced, as are tubing PQ graphs, as are the Swiftech storm PQ graphs, as are the MCR220QP graphs. Just a matter of plugging in the numbers and looking up the CW charts to work out benefits and losses.

[bobo5195]

Discharge into and flow from an infinitely large (for our cases a resviour is infinitely large) body of water the k factor is 1.

Ie dP= K * V^2 where w is the mean velocity.

This is an analytical relationship and is completely geometry independent. A real res where there is flow between teh inlet and outlet is different.

The main advantage of running parallel is not presure drop but increase in fluid dT.

[guandi]

what if you had two pumps, each feeding a seperate rad and then going into a Y peice? would it help if the Y peice was 5/8ID with 1/2id tube?

[Long Haired Git]

My model is flawed in the amount of heat added by the pumps, in that I assume 60% to 80% of heat pump is added to the coolant (depending on flow rate). However, testing at cooling masters has found that this is way too high for some pumps.
Until I fix that, I find that any effort to add in extra pumps results in worse temps due to the extra heat of the pumps.