LIVE from POOLSIDE, it's GHETTO TESTING!

[Gooserider]

AKA Test like BillA (NOT!) on $20 or less!

Way back when I first started on ProCooling, there was a survey about 'how much would you spend to test your stuff?' or something like that. I think I responded on the order of $20 max.

I then started building blocks, and since my rig was a bit more complex than average (dual processor, lots of hot SCSI drives, etc.) was also wondering how best to plumb it.

Since I was making two different sets of blocks, namely CPU blocks that were using large tube ID's, low restrictions, and high flow rates; and Drive blocks that used small tubes, were fairly restrictive, and had low flow rates, I planned to make at least two seperate parallel cooling loops, one high flow, one low flow. But with 2 CPU blocks, 4 drive blocks, a flow switch, etc, I thought there might be advantages to splitting the flow further.

There was a certain amount of discussion about this, involving factors like:
When running serially, how much hotter is the water going to the second block?
Is it better to have more total flow if each block sees half of it, or less total flow with both blocks seeing all of it? (I ignored the drive blocks, since they would be restrictive enough that I didn't expect them to significantly change anything)

In the end I decided it was necessary to do some testing to find out. My theory was that if the two CPU blocks in parallel gave twice the flow of the two in series, then parallel was better. Also that more flow in general was better. I also realized that precision testing was neither needed, nor possible on a minimum budget.

Fortuneately I already had available the largest peice of necessary test equipment - an in-ground gunnite swimming pool! (A most useful peice of test equipment, and you can swim in it as well...) ;D Swimming pools are good since they offer an effectively unlimited supply of water, and you don't have to worry all that much about spills. However I suppose that one could make do with the kitchen sink or a bathtub if desperate and unable to talk parents / Significant others into building a pool for testing....

I also learned about Manometers and PQ curves, and the infamous 'Bucket test' and decided to combine the two. I knew that if one knew the head pressure of the pump, it should be possible to determine the flow volume from the pump PQ curve. This should be confirmed by timing how long it took to fill a 5 gallon bucket.

My Iwaki MD20RT pump has a maximum head of 14' so I went with a 15' manometer.

I first built a manometer with a number of lengths of 1/4" plastic tube zip tied to a 15' long 1 X 4 (another advantage of swimming pools, they are usually outdoors where you don't have to worry about ceiling heights.) I fastened the manometer to the side of a step ladder to hold it up and enable reading the higher numbers.

My initial plan was to measure the resistance of each block, plot a bunch of PQ curves, and then calculate the optimum combination and confirm it by testing.

To do this I fed the pump into a T-fitting with a manifold outlet on the side arm. The through arm went through a ball valve and into a manifold made with 1.5" PVC pipe, and outlets of different tube sizes, also containing a manifold outlet. The ball valve was used to set flow resistances, and to shut off the flow when changing setups.

Problems included the supprisingly high flow resistance of the bare manifold and the difficulty of interpolating values off the Iwaki PQ curve chart (I was unable to get a table of values from Iwaki, so I had to deal with a fuzzy blowup of the graph from their website...) I also didn't have enoug blocks completed when I started to do much testing.

However I was able to confirm the theory of reading flows off the manometer via the PQ chart, as the bucket test results matched consistently.

But I realized this was alot of extra work for data that I really wouldn't have a use for - I wanted as much flow as I could get, and so why not test for that?

My phase two setup consisted of connecting the T fitting directly to the pump, and getting rid of the manifold. This left me with one manifold connection which I used to monitor the head of the system configuration under test. The ball valve was left in to shut off the flow during setup changes.

I ran a hose from the ball valve to the radiator input. On the radiator output I have a 4 outlet manifold that I constructed from copper pipe with machined copper fittings soldered into it, in the fittings are the female quick disconnects that I will be using to allow easy servicing of the system.

I then created a variety of plumbing combinations consisting of the male QD fitting, lengths of hose, a block (or blocks) and a male / female QD set. Aside from the hose length and material (for testing I used short peices of cheap stuff, will get fancier in the system itself) this reproduced the hardware that I'll be using in the system except for the rez and pump inlet section. Since I was sucking up about 8" during the test, but the res will be gravity feeding into the pump in the system, I don't expect this to cause significant differences.

I hooked up different configurations, and got a manometer reading for each one, along with bucket test timing.

 SETUP DESCRIPTION      'Head  PQ GPM  Bkt time Bkt GPM
2 CPU blx in series       ?      ?       1'36"   3.12
CPU Block A only          11'8"  4       1'29"   3.33
CPU Block B only          11'8"  4       1'30"   3.33
2 CPU BLX in parallel      8'8"  5.5     1'01"   4.92
2CPU Par, 4Drv + flow ser  8'4"  5.5     1'04"   4.7
2CPU Par, 2drv + 2drv&flo  8'0"  5.5     0'57"   5.26

Note, manifold reading error on first test, probably should have been ~12'
'Head - head pressure in feet and inches
PQ GPM - Gallons per minute per PQ chart
Bkt Time - time to fill 5 gallon bucket, minutes'seconds"
Bkt GPM - calculated GPM rate - (5/time(in seconds))x 60

It was not practical to do bucket testing, but I did also try testing with just male QD fittings (no blocks)
in the manifold to see what manifold readings I would get.

QD fittings used,               'Head 
1 x 1/2"                         10'6"
2 x 1/2"                          7'6"
2 x 1/2" + 1 x 1/4"               6'7"
2 x 1/2" + 2 x 1/4"               6'4"

I consider these results to be within my limits of accuracy, and adequate to the task.

I conclude that I would get about 2.5GPM through each CPU block in parallel vs. just a little over 3 running them in series. I think the overall flow increase is worth the slight loss in flow through each individual block.

The drive blocks are more restrictive than I had planned (I may replace them later) but I think they will flow sufficiently to cool the drives I'll be using.

I am finished testing for the forseeable future, as I need to close the pool this weekend (bad part about living in New England) and if all goes well, will have the system running with this hardware before I re-open the pool in the spring.

Gooserider