Two pumps config - serial or paralell?

[Puzzdre]

I need your help on this guys,

wanna build new rig, having two pumps (maxi jet 1000), one rad, plexy res, #rotor style block on cpu and nb, flow indicator, all 1/2 tubing.

What I had in mind is res as a highest point in system, followed by pump or pumps. I wanted to hook the pumps to the res in paralell, than feed the rad - cpu block - nb - flow indicator - res.

BUT...

Is the paralell pump setup way to go? I meant to connect pump outlets through the Y, than feed the rad.

OR...

One pump sucking from the res, feeding rad, than second pump from rad to blocks to the res.

Ofcourse, I can hook up pumps one after another right behind the res, but that would be a little tricky (but doable) in tower I want to mod, so I want to put them side by side in paralell setup, or serial but not one after another.

Any comments/infos/even flame appriciated!

[airspirit]

The non-scientific answer would be serial. In a parallel configuration you risk backflow and possibly pump damage if one pump is pumping harder than the other. If my memory serves me right you'll get better results through parallel, but the risks outweigh the benifits. Essentially in series you'll be keeping the flow rate in GPH of the stronger pump, but adding the head rating of the two, so if you have two pumps that pump 250 GPH at 5ft head, you'll be pumping 250 GPH at 10ft head in the new config (its been a while, so I may be wrong). There was a big discussion on this a few months back, so you might want to search. The other nice benifit of series is that if one blows up, you wont lose flow since the other will still be working ... in parallel most of your flow will flow back through the parallel loop and the rest of your system will be nearly stagnant leading to flaming CPU death.

Anybody else have better input?

[airspirit]

Oh, and you want to keep the pumps in front of the res rather than behind (relative to flow direction). The res will absolutely destroy your flow if your pumps are pumping into it. If you put them both after the res you'll get much better system flow since centifugal pumps are more efficient pushing than pulling. That is something I know for certain.

[Puzzdre]

Thx airspirit,
I didn't express myself right way, when I said pumps after the res, I meant pumps inlets to the res, and outlets toward rad. Sorry, I just have the pic (in my head) of it how I want to mount it in case, so I made a mistake not clarifying that.

Yes, I did the search before I started the thread, good thing I found morphling1's thread about several pumps in series, but couldn't find much on pumps in paralell.

Oh, and your'e right, paralell pumps = same head as one, double flow, serial pumps = double head, single flow.

And for those cheapo' pumps, I didn't even think of possible failure of one, and consequences of that.

Thx!

[decodeddiesel]

Anyone have any conclusive data as to the long term effects running 2 identical pumps in serial? Will it cause long term pump damage? I am not looking for speculation, just some hard data.

[myv65]

This is not that difficult a situation. You can think of pumps as devices that add pressure to a system. The amount of pressure they add varies with the flow rate produced. If you put two identical pumps in series, the maximum possible pressure you can develop essentially doubles while the maximum possible flow rate remains the same as for a single pump. If you put two identical pumps in parallel, the maximum possible head remains constant while the maximum possible flow rate doubles.

In real-life systems, you always have some flow resistance. This means that a pump will never produce its peak flow (which only exists at zero head). In a typical system, you'll normally not see much performance difference between the two options. In restrictive systems you may see a nice boost from running in series. In rare instances, the difference may be dramatic. Heck, in open systems with enough vertical lift, you routinely see multiple pumps in series (lift stations). I'm now talking more on the industrial side of things.

For simplicity's sake, series piping is easiest. The impellers in these pumps are open enough that if a pump fails, you'll still generate enough flow to protect a system. The "dead" pump becomes just another resistance in the system at that point. In parallel systems, you really ought to have check valves on the pump outlets otherwise a dead pump will allow a short circuit and effectively kill the flow going through your radiator and block. This setup handles a failed pump better because the failed pump doesn't restrict the flow of the working pump. Check valves are not zero resistance either, though, so also impact overall flow a little.

Airspirit's comment about working better pushing than sucking only really applies if the suction pressure drops to the point of cavitation. So long as cavitation is avoided, a pump merely adds head to a system irrespective of its location within that system. Ultimately most of the energy consumed by the pump shows up as thermal energy in the fluid, so keeping all pumps ahead of the radiator offers some slight benefit. This point pales versus air temperature fed to a radiator. If putting the pumps both ahead of the radiator means using warmer air for the radiator fan, don't do it. Keeping the coolest possible air feeding the radiator takes priority.

Here's a link to a file explaining the stuff in a little more detail. Head on down to page 7 for more info on series vs parallel operation.

[g.l.amour]

serial seems to be the best option. everything else aside, doubling the head seems to be the best bet, performance wise. both pumps will be closer to their ideal flowrate (u know, like the eheims starting to rattle if impeller is not teflonned to the shaft, in case of too much backpressure).

[Puzzdre]

Thx for the input guys!

Just to clear the things a little bit more, here's how I see it in my head:

[Puzzdre]

I've read the (part of) .pdf on the paralell vs serial pumps and it cleared a little bit more what was said here.

I still have this dilemma: if I put the setup like in pic, I can easily make it a looker, but also want to make it a good performer. I can think several ways of making the kinda flip valve in Y piece to ensure that if one pump fails, the valve closes that outlet, but it seems like too much hassle (but also makes it more fun ). I believe its manageable with a little added restriction to the setup.

If I go serial, I will drop the res, no need for it. The res is here just to make the two pumps in paralell have most equal supply of water on the inlets.

Also, there is MBM capability of shutting off the system if overheated, so if one pump fails, I know that the temps will rise rapidly, but still it's not matter of seconds but minutes, so the MBM should be capable of switching it off (I tried it, switched off my pump and watched the temps rise).

Dunno why I'm stuck to this paralell setup, must be it reminds me of those speed boats with two Z drive props, I always admired the beauty (and the horse power too )

I'm not in a hurry, so still, any info/thoughts/hard data welcome!!!

Thx again to all of you guys!

[Puzzdre]

Just some more clarification...

The fan behind the pumps will blow the air out of the case, cooling the pumps a little, so they will (I hope so) add less to the water...

The pumps will ofcourse be in the same level, side to side, not like I draw it on the pic (that was only for adding some crappy 3rd dimension)...

If I build the res, the two pump inlets will be separated by piece of plexy in some (1/3 I think) length to prevent them "stealing" the water on the inlet holes from eachother...

Flow indicator is optional...I like it, but will not insist on it, except if I manage to make it flow METER, that's another story...

Got no more ideas for now...got headache...should I go to sleep or light just one more cigarette and stay a little bit more here...hmmm, the question is now...

[#Rotor]

leme ask this first, how restrictive did your blocks come out?..... you probably was surprised at the amount of flow you got....right.


I would split the paths to each block from the radiator, have them completely separate right back to the reservoir, and then add the pumps in each of those lines,

make sure the line between the reservoir, and the radiator can handle the increased flow. That would also be where the flow meter would want to be.

as per usual, make sure the pumps are in the lowest part of the loop, so as to prevent airlock.

[Puzzdre]

Hi #Rotor,

thx for the tips!

I would really like to answer your question, but I can't cos I haven't made the blocks yet. Not sure yet should I go with 3 mm or 4 mm drill bit. I counted the number of holes in a row on some of your's just to get the idea, but I'm still stuck with 10 mm thick pieces of copper only and don't know if this would be enuff if I go 7 mm deep, leaving 3 mm base. Also, must go and buy a good drill bit, and dremel cutting wheels.

But I do believe that the blocks will have a good amount of flow...

Second 'problem' are the barbs. I cannot just go and buy them (they don't exist in that form here:shrug: ), so must go to the machine shop to make them for me, or just (ugly!!!) solder the 1/2 ID copper pipe on the blocks.

make sure the line between the reservoir, and the radiator can handle the increased flow. That would also be where the flow meter would want to be.

I can make those lines 16 mm ID, but I thought of putting the flow meter as a last component in the whole system, after all the components that would restrict the flow would be the way to go.
Ofcourse, I can always try both, right behind the pumps, and on the end of the loop, just before water entering the reservoir. But I must first make the flow meter first.

Oh, yeah, and I think I'm going with 4 mm bit for start...I'll draw the grid today, if nothing comes in my way...

Btw, I cannot buy goop of any kind here, so I thought to go with liquid metal epoxy in thin layer between the plates. It's solid to the 120 deg. Celsius, and resistive to chemicals we use in wc...

I like to do lots of planning, cos' when I start making things, there'll be still lots of things I didn't thought of comming in a way, and still lots of rearanging for the fan sake...

Thx again!
Cheers

[Volenti]

I'm using 2 small 700L/H 1m head pumps in series with my very restrictive direct die water cooling, I feel quite comfortable knowing that I have the redundant pump setup. (pretty critical with direct die, no saftey margin)

[g.l.amour]

euhm puzzdre, how are u making your blox. i'm interested because u seem to live in a small place too, and if u can pull it off in an apartment, then so can i...

[Puzzdre]

Thx for infos guys!

g.l.amour: A few months ago, I bought a drill press for somewhat 55 euro equivalent, it's cheapo' but it works fine for me. I usually keep it in basement (every apartement here has it's small portion of basement area, so I can keep my biggie stuff there), and when I want to work with it, I simply bring it to my balcony or kitchen. It's weight is somewhat 20 kg, so it's not a problem. I fixed mine onto one large piece of MDF to give it more stable base. For some time I kept it in my grandma's house, but, when I need it, I bring it here into apartement.

Volenti: On die cooling is critical, so redundant water supply is almost a must. I'm glad you succeeded. Did you experience any damage to the serial driven pumps, mentioned by decodediesel?

/edit/ This drill press is not 'too' noisy, so the neghbours won't be bothered, there's more noise when you vaccuum clean...

[g.l.amour]

thx for the info !!

but my ratio tells me to just buy the damn block i want, it most be alot of trouble getting drill bits - pieces of copper, etc, prolly will end up way more expensive than normal.

c ya

[Volenti]

Volenti: On die cooling is critical, so redundant water supply is almost a must. I'm glad you succeeded. Did you experience any damage to the serial driven pumps, mentioned by decodediesel?

I havn't noticed any problem with these pumps, however when I was doing some rather extreme testing with 3 large pumps in a series-paralell setup I had this happen;


had 2 2000-2800L/H pumps in paralell pushing into the back of a 4000L/H pump that was trying push water through a 4mm jet (pressure test) I guess it failed the test...

[Puzzdre]

prolly will end up way more expensive than normal.

LOL!!! This is ALWAYS the case, at least in my experience...but you got the satisfaction of making it yourself, and when u want to make another, even different one, you'll won't have to buy the tools you already have...but there's always another tool...

[MadDogMe]

If you're worried about the flow sensor being too restrictive you can always run it after an equal Y (5/8~14mm have 12mm inner diametre for less restriction) then rejoin with another Y, that way not all the flow will be going through the sensor,enough to turn it, but not so it'll restrict flowrate...

[#Rotor]

10mm thick plates are good. do not go too deep. I'd say about 4mm on each side should give you awesome flow-rate and yet you will have turbulence and surface area to spare.

depending on the size of your block, but 4mm diameter is a good overall size. Leave at least 3mm on the sides, for sealing the block up. and before you seal it... lap the surfaces, so they fit snugly on one another. then just a minute amount of sealant, in between, and you would be surprised at the bond being formed between the plates....

of coarse, make sure you do not have any oil or dirt on the surfaces....

[Puzzdre]

@Volenti: Woah, I saw that pic on another thread or on OCAU, that sure don't look pretty!!!


@MadDogMe: Thx for the tip! I'm not sure yet about the flow sensor cos' I'll have to think to fit large barbs onto it (the one I made had 9 mm ID barbs). If I don't succeed, I might just do what you suggested!

@#Rotor: Man, thank you for giving all the secrets of the trade!!! I really appriciate it, it will save me lots of trouble and experimenting!
Thank you!

I had enough free time this afternoon to draw the grid and print it on self adhesive paper, now I just have to find the workshop with the circulary metal saw thingie to cut my piece of copper to the size I need. Bottom plate 50 x 50 mm, top plate 80 x 50 mm (that's because I have one piece 80 mm wide and 400 mm long).
A note to myself: check that no capacitors are in the way on the mobo for 80 mm wide block!

I cutted some pieces of another 50 mm wide copper bar, but handsawing didn't look to pretty, more to it, the saw always managed to run not along the line I draw but little right or left...
Cutting cannot cost me much, and it will save me lots of sanding the edges to be perpendicular to the flat sides...

Thx again!

[g.l.amour]

damn puzzdre, u really make me want to make my blox also. not that they would be any good, but it sounds like sooo much fun!!!!

a swiftech sort of block, shouldn't be that hard to make.

[Puzzdre]

damn puzzdre, u really make me want to make my blox also

C'mon m8!!! Somehow I knew that you'll get hooked up on this, LOL!!!

not that they would be any good, but it sounds like sooo much fun!!!!

IMO, any block you made yourself will be good enough for time being. I made my first (somehow glad that I couldn't buy one already made then), it looks like this:

[Puzzdre]

(can anybody enlighten me how to post a pic and then keep the writing, thx )

It works, its kinda ugly, the sides are not *exactly* perpendicular to each other, and so, and so....but I made it, it works, no leaks, etc... It's good to go forward, so now I learned more and I'm making the wc rig mk2...

Thats what I'm saying, anything you make will be good enough. It's to keep YOU satisfied.

To me, this is fun. Enjoyment. It's not to be the best, its to be part of it. Not going for breakthrough, just participating.

Go for it!!!

[g.l.amour]

problem is that the more specialized it gets , the harder it is to get your stuff. the drill bits u need for that tool are prolly not found at the local diy store, nor the copper bars. no need to hurry, i will start to look out for some tools, but it would have to be a good second hand deal...


as always, u make me consider/do stuff, that i would normally hesitate to do. u evil puzzdre