Sunday, May 31, 2009

Fountain test

Belle, non?  (1/80 sec exposure)


Non. (1/4000 sec exposure)


Here it is with Martha for a sense of scale:

First, what went right?
  • The geometry is right.  In this spreadsheet, I calculated the height of the fountain (69.4 inches, was actually 65.5 inches), and how far it would throw the water, and where the jets would come down into the hot tub.  Although some of the jets land about 5 inches off where I expected, and two jets collide in midair just before they hit the water, the geometry is about as good as can be expected, and fulfills my goals, which were:
    • It should be possible to walk between the rising jets without being hit by them.
    • It should be possible to sit in the hot tub without being hit by the jets.
    • It should be possible for a child to stand in the middle of the hot tub and have the jets come down all around, without actually hitting the child.
  • All the jets rise to the same altitude, within about half an inch or less, which means the balanced binary tree distribution system with the shorted end terminals worked.
  • The pump-side pressure stack does not overflow.
  • The pumps don't cavitate.  They are incredibly quiet.  You cannot hear them unless you walk over to the pump vault and stand on top of it.  Once the lid is installed on the pump vault, I doubt you will hear the pumps even when you are on top of it.
However, the jets are not laminar.  Gloppy blobs of water fall into the water and make a dull roar instead of the quiet sizzle that I had wanted.  There is enough splashing from the jets entering the water that you wouldn't want that a few inches from your face.  The kids love it, of course, because it's loud, fast, and wet, but it's not so great for the adults.

I'm pretty bummed.  What happened?

The individual jets were flow tested in January, and this is what they looked like then:

As you can see, the jets were smooth, and landed smoothly and quietly, back in January.  Now Martha jokes that if we move the back yard table to the farthest corner of the yard, we can still have a nice conversation.

I'm not entirely sure why there is a difference.  Here are possibilities, ranked by my guess of most to least likely.
  • In January, the heads had no lateral ports in them.  In this latest trial, there are two ports in each head, connecting each to the heads on either side.  These ports keep the pressure even across all the jets, which makes them shoot to the same altitude.  But these ports may also be causing the water to tumble slightly as it passes the edges, and that turbulence may be causing the breakup that I'm seeing.
  • In January, the heads were surrounded by nothing, and so small amounts of water on top of the heads ran down the sides, away from the jets.  Now, the heads sit inside recesses in the gunite.  Each head has a small pool of water in it that terminates at the jet.  The water in this pool greatly disturbs the jet during startup, but it gets cleared in two or three seconds and then I don't think there is any more water recirculating through that pool and into the jet.
  • In the January trial, I had an open-topped pressure stack between the pump and the jets.  In the production version, I have a stack after the pump, but it's not quite the same.  In this one, the water from the pumps goes to a Tee.  In one direction, the water heads for the jets, and in the other direction, the water heads for the stack.  It's possible this alternate arrangement works less well.
  • This test has a closed air volume right before the jet, which was intended to be an additional flow smoothing device.  The January arrangement used open-topped pressure stacks either right before each flow straightener, or right after the pump, and both worked well.  The closed air volume is known not to work as well (since the pressure changes more with a small surge in water).  Also, since the current arrangement has two capacitors with an inductor between, it's possible that there is oscillating pressure being stored between the two capacitors.
  • The nozzle holes are 0.590 inches, rather than the 0.500 inch holes that I tested in January.  This makes the jet diameters about 0.500 inches, which is necessary for all 200 gallons/minute to flow.  As a result of the larger jet and the larger jet velocity, the flow through the flow straightener is perhaps twice as fast as it was in January.  It would be great if this were the problem, since I can reduce the flow later when I have that plumbing finished.
  • In January, the pump had air in the lint basket bowl, and the pump could be heard continually injesting air.  Now the pumps have no air in their lint basket bowls.  I would expect this to make things better now, but I thought I'd list it because it is a difference.
I also have two unexpected observations which may be a clue to a solution if I can figure it out:


The ports in the sides of the fountain heads are connected via riser pipes to a plenum that is fed from a pipe that will ordinarily lead to a blocked valve.  This valve is used when the fountain is off to backflush the flow straighteners.  However, that plumbing is not yet finished, and so the pipe currently leads to many other pipes that are currently filled with air.  There is also a hose bib and a pressure gauge connected to those pipes (this is how we did the pressure test).  I have calculated that the static pressure at the top of those fountain heads is about 3 psi above ambient, and so I expected the plenum to be pressurized at 3 psi.

But that's not what the gauge says.  The gauge shows zero pressure (I don't have any gauges that show negative pressures).  If I open the hose bib while the fountain is running, then cover the opening with my finger, I feel a little pull.  It's very feeble, but it's there.  WTF?

The flow in the head is moving at 1.6 inches/second, and I calculate a dynamic pressure of 0.85 Pascals, or 0.00012 psi.  That isn't diddly compared to 3 psi pushing out.

[Update: Mystery retired: it turns out that the pipe connected to the top manifold is capped off right now, and those other pipes are just not connected to the fountain.  I can't explain why I was thinking that there was a small pull of air, but it certainly wasn't measureable.]

The second unexpected thing happened the first time I started up two of the three fountain pumps.  All three pumps are in parallel.  I had difficulty taking the lid off the third pump's lint basket bowl, so I had left that bowl filled with air, and just started the other two (which were properly filled with water).  I expected the first two pumps to push water backwards through the third pump, flushing the air into the intakes of the first two, where it would be blown into the fountain or otherwise ejected from the system.

Nope.  There was no noticeable flow through that third pump.  Later, I pulled that lid off and removed the air.  When I ran just two pumps again, the third pump did have flow going backwards, and in fact the impeller was turning backwards at perhaps half the RPM of the two powered pumps.

It's not clear to me how the air can block a >3 psi pressure drop.  The total drop from the top of the pump to the bottom of any associated piping is perhaps two feet, which would account for a 1 psi drop block, but not 3.

I suspect that the solution to these mysteries, especially the first, will tell me something about the fountain behavior.