A better elbow

[Martinm210]

While I don't use any myself, I know many with case restrictions that leave them with no choice....

Brought to you by DuckieHo of overclock.net, thanks for lending me the Danger Den Derlin 90 for the test:up:

The DD elbow is about 45% less restrictive than my previously tested nylon 1/2" elbows. I think this is partly due to the slightly larger ID of the high flow barbs and also because of the larger cavity where the water has to turn. Whatever it is it does help.

They are a little pricey at $8 each (http://www.dangerden.com/store/product.php?productid=259&cat=34&page=1), but with this flow rate test result, I'd say it's probably worth it.

http://www.overclock.net/gallery/data/500/Fittingsv1-7.jpg

I will include these as new options in the next estimator release...

[Barb]

At $8 each, I'd say that it's not worth it.
I think it needs to be emphasized that the pressure drops we are dealing with from a 90º Elbow are very small.
The results get pretty dramatic toward the higher flow region of the graph, but those flow rates aren't typical of most systems. Your only going to see 3gpm+ if the ONLY thing in your loop is a 90º Elbow.
Even if these fancy elbows are 45% less restrictive, unless there are a heck of a lot of elbows in your loop, I doubt there would be any tangible improvements.

It would be cheaper to just use a Nylon 5/8" Diameter Elbow. If your using 1/2" tubing it is reasonably easy to squeeze it over the fitting. Even if it is a tight fit, it is likely that the fitting will not need to be removed and remounted often.

If your not using 1/2" ID tubing, then your probably not anal enough to be concerned with the small pressure drop associated with a 90º Elbow.

[Martinm210]

It's true the pressure drop is small, but not that small.

For a ballpark comparison:

2 ea, 90 degree 1/2" nylon elbows = 1ea Swiftech MCW30

or with these other elbows

3 ea, DD 90 Derlin elbows = 1ea Swiftech MCW30

I had the elbow sent to me free of charge to test, so I tested it. My initial assumption was that it would be equally if not more restrictive than a nylon elbow, I was wrong.

That's true about the 5/8" elbows, I have 5/8" fittings everywhere including my MCR320's and T lines, they are the cheaper free flowing option.

But I think there are a number of mods people do to a water cooling system for asthetics. These little elbows do look nice, and at the same time they are better flowing than 1/2" nylon elbows.

And heck, I think I am anal about my water cooling. One elbow or even a handfull isn't going to make much difference, but we're splitting hairs now for any water cooling advances. If I wasn't anal about it, I'd still have my air cooling setup...:D

[Fairydust]

No the prettiest but far cheaper and probably even less restrictive elbows (http://cgi.ebay.com/1-2-90-DEGREE-COPPER-ELBOW-SEE-MORE-IN-MY-STORE_W0QQitemZ110128765918QQihZ001QQcategoryZ6390 0QQrdZ1QQssPageNameZWD1VQQcmdZViewItem), with 2 female ends at 5/8 gives a very snug fit. I guess the best budget solution would be a 5/8 weld in long elbow.

[higgins]

interesting results, thanks for the tests man, your doing a great job.

[ABXG]

I have always wondered no one makes a "curved" elbow piece. It would seem that the water would flow much more smoothly if it didn't come to a straight up 90 degree bend. I tried to find some curved elbows when I was buying parts for my WC'ing, but couldn't find any.

[mcoffey]

that's what I use on my rig, the DD elbows. I wondered if they had less presure drop than the nylon 90's, always thought they did.

thanks for the info. the DD T's are pretty nice also, especialy when used with the stubby and micro stubby fittings. they take up a lot less space, same with the 90's.

thanks again for all the hard work Martin,

andyc

[AndyM]

Interesting findings indeed. I've been wondering, should you want to do some more testing...here's a candidate for another elbow test (Thermochill-suitable elbow when used with a BSPP nipple)

http://i80.photobucket.com/albums/j186/amichalski00/4860K433l.gif

[XS_RoB]

@ Martinm210

I just wanted to say thanks for all of your hard work and let you know that I truly do appreciate you taking the time to always improve and update your spreadsheet!
Thanks and Keep up the Awesome work! :)

RoB

[voigts]

I have been using copper long sweep elbows for years now and they work very well and cannot possibly be very restrictive. See #5520K184 p.187 Mcmaster.com . I solder a short 1/2" copper piece into the female end and then paint it black.

Here is a pic under my previous MB. I still am using this copper elbow with my current Conroe build.

http://i5.photobucket.com/albums/y167/voigts/CustomWoodCaseII/completed/GPUelbow.jpg

[teyber]

Ive got a elbow in mind for a case mod im doing next, it involves some 1" plexi, and my shiny new hitachi drill pres and a new tap :D 8$ each no thank you! they arn't that hard to make, just drill two 1/2inch holes at a 180 degree and a 45 degree angle, then drill just barely liek 1/2" straight down. O well, these look pretty good to me

[Martinm210]

Thanks!

At some point I should include a few of those other options. I particularly like the long copper sweep:up:

[migueld]

Comparisons are indeed interesting, according to Martin's estimator:

1 Fuzion GFX (GPU) = 15 elbows :)

A 2 elbow figure looks pretty harmless in that context.

Ah yea, comparisons comparisons...

[migueld]

I have always wondered no one makes a "curved" elbow piece. It would seem that the water would flow much more smoothly if it didn't come to a straight up 90 degree bend. I tried to find some curved elbows when I was buying parts for my WC'ing, but couldn't find any.

There is :)

http://www.mcmaster.com/catalog/113/gfx/large/5670kp4l.gif

It's item 5670K18 for 3/8" tubing at www.McMaster.com, costs $3.37 for stainless steel which is pretty nice. Unfortunately for 1/2" lovers it's double the price :shrug:

[CyberDruid]

There is :)

http://www.mcmaster.com/catalog/113/gfx/large/5670kp4l.gif

It's item 5670K18 for 3/8" tubing at www.McMaster.com, costs $3.37 for stainless steel which is pretty nice. Unfortunately for 1/2" lovers it's double the price :shrug:

Very nice--I like that piece.

Martin I really appreciate your hard work and constant updates and upgrades to the spreadsheet. An invaluable tool and a great asset to the Forum.

The only thing that puts me off of those DD parts is the feeling of being RIPPED for something so simple that uses so little material... a 1" cube of delrin is like a dollar tops...mill and drill in production what another dollar...then tack on $6 because OMG it's for a PC... gimmee a break.

I think the copper sweeps for a quarter apiece are a much more attractive option.

Good work man! we can bicker about fittings all week...but the work you are doing stands apart from the bickering :up:

[alpha0ne]

I have been using copper long sweep elbows for years now and they work very well and cannot possibly be very restrictive. See #5520K184 p.187 Mcmaster.com . I solder a short 1/2" copper piece into the female end and then paint it black.

Here is a pic under my previous MB. I still am using this copper elbow with my current Conroe build.

I too have been using exactly the same and painting black

Easily the cheapest and least restrictive option available :up:

[Martinm210]

Thanks all!
I'll add some of those others after I finish up some thermal testing I have planned...Fuzion GFX vs. MCW60.

BTW...Version 1.7 of the estimator is up with this new and controversial elbow ready to roll!

Elbows are cheap, as the adding of new components dyes down, I'll have to follow up with a few more elbow options to satisfy my curiosity. Obviously those long sweep types are going to be next to nothing in terms of restriction, but I bet even the tight copper elbow does really well because of the thin wall and radiused change in direction.

[Barb]

Martinm210, I took a look at your Excel Watercooling Flow Rate estimator, very nice work indeed.
I just have a question.
How did you come to the 4' per elbow estimation?
I did an analysis a while back an came to the conclusion that one 1/2" elbow is roughly the equivalent of one foot of tubing. I realize that the theory will likely differ from reality, but 4 feet seems like a lot.

I looked at how you calculated the pressure drop for the tubing and I compared it to what I got. Your results are a little more than double what I got. Which I suppose is believable when adjusted to reflect the real world. But I am still surprised that the theory is so far off.
I haven't done any real world testing, so if you conducted your experiments correctly, I'm inclined to believe the real world results. But I would like to hear your comments on why you think this discrepancy exists. Is it the friction factor? Is the tubing typically used in our systems really that much rougher than the "smooth" tubing often used in calculations?
Also, I know that the Darcy equation tends to fall apart near the transitional region between turbulent and laminar flow. I wonder if the theory gets better at predicting pressure drops at higher flow rates. comments?
Whew....

[Martinm210]

Martinm210, I took a look at your Excel Watercooling Flow Rate estimator, very nice work indeed.
I just have a question.
How did you come to the 4' per elbow estimation?
I did an analysis a while back an came to the conclusion that one 1/2" elbow is roughly the equivalent of one foot of tubing. I realize that the theory will likely differ from reality, but 4 feet seems like a lot.

I looked at how you calculated the pressure drop for the tubing and I compared it to what I got. Your results are a little more than double what I got. Which I suppose is believable when adjusted to reflect the real world. But I am still surprised that the theory is so far off.
I haven't done any real world testing, so if you conducted your experiments correctly, I'm inclined to believe the real world results. But I would like to hear your comments on why you think this discrepancy exists. Is it the friction factor? Is the tubing typically used in our systems really that much rougher than the "smooth" tubing often used in calculations?
Also, I know that the Darcy equation tends to fall apart near the transitional region between turbulent and laminar flow. I wonder if the theory gets better at predicting pressure drops at higher flow rates. comments?
Whew....

Thanks!

I used Darcy's equation to estimate pressure drop in tubing for the first few iterations of the estimator and I found that Darcy's was in error by over 20% with real world tests, here is my old Darcy curves if you're interested:
http://www.overclock.net/gallery/data/500/Tubing_Friction_Loss.JPG
And these were the real world checks that I did on 1.0:
http://img455.imageshack.us/img455/9877/v10verificationteststu2.jpg
It was clear to me, that I had error in the tubing curves, which wasn't signficiant overall, but error regardless.
Eventually, I ran pressure drop curves for each section of tubing and sure enough, very much like my real world error results...the tubing is more restrictive than Darcy's produces. Particularly the 1/2" and 3/8" curves:
http://www.overclock.net/gallery/data/500/Tubing-.PNG

My only explanation is model scale error. Anything that is reduced in size will have some variable affect the scale model a little differently than real world tests. Friction factors if using something like Hazen Williams also apply to pipes of large size...smooth PVC to an 8" pipe is alot different than smooth on a 1/2" hose.

My initial 4' per elbow was an old test I ran that was fairly crude and simply based on measuring flow rate on a pump and length of tubing. I added in an elbow and starting snipping off 1' increments of tubing until the flow rates matched again. Here is that original test, it was one of my first flow rate tests, and I've since learned I had some error built into the test due to static pressure supplied by my reservoir and I didn't have a very good reservoir to clearly see water marks. Now when I pressure/flow rate test, I measure at the outlest side and fix the outlet elevation so I can eliminate that testing error. I also have a section of tubing with a barb that very clearly indicates flow level. Regardless is was a good approximation and based on testing, not theory:
http://www.overclock.net/gallery/data/500/Elbowsarebad.jpg

If you look at fitting equivalents in most engineering manuals you will find a more common length value for an elbow is 1.5'. This however is based more on a PVC type elbow where the ID is the same as the pipe. In a hose connection, it's typically a step down from 1/2" ID to 3/8" ID...that's the main difference here. You not only have a smaller/tighter restriction...you also get losses from the step down and up in ID sizes.

So currently, all of estimator curves are based on real world pressure drop tests. The largest error resides in the fact that a large portion of the curves come from a wide variety of sources and testbeds. I suspect most people that create the curves don't correct out the drop in the testing equipment because it's so small, but it's likely in general the error would be such that the estimator gives you slightly lower results than actual because of this. I've recently been correcting all my own curves by subtracting out the testing pressure drop (Particularly for these small pressure drop elements like elbows and tubing). Anyhow...sorry to be so winded....that's what I've found.

[ABXG]

Comparisons are indeed interesting, according to Martin's estimator:

1 Fuzion GFX (GPU) = 15 elbows :)

A 2 elbow figure looks pretty harmless in that context.

Ah yea, comparisons comparisons...

Are you serious? Wow, I never would have bought that block if I new it was that bad (or am I just a noob and there isn't anything better?). I bought it because it would fit with the 8800GTX UNISINK (I didn't want to deal with the RAM sinsks that evidently don't stick to the card very well). Oh well, maybe I will replace the thing sometime in the future if I get the money :mad:

[Martinm210]

Are you serious? Wow, I never would have bought that block if I new it was that bad (or am I just a noob and there isn't anything better?). I bought it because it would fit with the 8800GTX UNISINK (I didn't want to deal with the RAM sinsks that evidently don't stick to the card very well). Oh well, maybe I will replace the thing sometime in the future if I get the money :mad:

It does take second place behind the MP05 with nozzle 3 as being the most restictive block in the estimator based on D-teks published curve.
http://www.overclock.net/gallery/data/500/V17Componenets.jpg
But I have a copy of the block here and plan to test it myself to confirm, and then later put it up against my MCW for thermal testing. My guess is that it does well thermally, but it's restrictive so it works best in GPU only loop setups where you're not impacting other blocks. But as long as you have over about 1 GPM, I wouldn't go replacing it.

Out of curiosity what kind of temps are you getting with yours?

[ABXG]

It does take second place behind the MP05 with nozzle 3 as being the most restictive block in the estimator based on D-teks published curve.
http://www.overclock.net/gallery/data/500/V17Componenets.jpg
But I have a copy of the block here and plan to test it myself to confirm, and then later put it up against my MCW for thermal testing. My guess is that it does well thermally, but it's restrictive so it works best in GPU only loop setups where you're not impacting other blocks. But as long as you have over about 1 GPM, I wouldn't go replacing it.

Out of curiosity what kind of temps are you getting with yours?

My 8800 is in RMA limbo ( being shipped back to me on Monday) right now so I can't say. But with other parts in my sig my CPU idles very close to ambient and loads about 15C warmer. I am going to wait for the GTX before I get my final OC in place.

Getting back to (off?)topic :rofl: I can't wait till the card gets back and I can see what kinds of temps I get.

[weescott]

I used copper pipe from Homebase a few years ago. A real PITA to fit but looks awesome.

http://server5.uploadit.org/files/weescott-pipes_inside_sma.jpg