Monday, January 19, 2009

Fountains, fixed

Much happier.  I now have two unit fountains, operating in parallel, driven from a pump, with a filter that takes out most pressure variations.  There is some unsteady flow remaining, which I think is due to bubbles.  I have good reason to believe there will not be bubbles in the production pump system for my fountain, so I think the design is now validated.

As a side effect, everyone involved with the testing is now soaking wet.  Much of the piping is not actually solvent-welded together right now, so there were the occasional blowouts.  Fortunately, we have a blue sky and 75 degree air outside right now, so wet is just fine.  If we had a pool, we'd be swimming (ahem).

Here are two unit fountains running at about 60% of design flow.  Note the smooth flow.  It's splashing a bit when it hits the tub, but that'll go away when it's got more than a half-inch of water to fall into.

  • The streams do not quite reach the same height.  I'm pretty sure this has to do with a small difference in their pointing angle, so I'm pretty sure they'll all line up when properly installed dead vertical.
  • When supplying the fountains with water from the hose, pressure variations from the street supply were making their way to the fountain.  These were probably low-frequency variations, maybe 1 to 3 hertz.
  • I considered a surge tank, but the volume required was too large (over 20 gallons).  Instead, I have a simple open-topped vertical pipe teed into the line from the pump to the fountain.  The pipe only needs to be a few inches taller than the height of the fountain jets.  The production system will have perhaps an extra foot or two.  You can see the pipe in the foreground.
  • I tried placing the tee close to the unit fountain, and that worked well, and then I tried placing it next to the pump, and that worked about as well.  This result indicates the flow instability comes from the pump and street supply, and not from turbulence within the fountain itself.  (Whew!) 
  • The nozzle is a nonlinear resistor, so it was moving the energy from low frequency variations to higher frequency harmonics.
  • The nozzle is nonlinear because it converts pressure into velocity head with almost perfect efficiency.  If you double the pressure, you double the velocity head -- the height to which the water will jump.  I confirmed this by varying the flow through the system and noting that the water always jumped to a few inches lower than the water level in the vertical pipe capacitor.  A doubling of velocity head means an increase by sqrt(2) of velocity, and flow is (sort-of) proportional to velocity.  As a result, flow increases with the square root of pressure.
  • The nozzle diameter is 0.500 inches, but the flow diameter appears to be 0.410 inches.  I didn't expect this.  The reason is that the flow at the knife edge is horizontal, and it turns through a radius of 0.045 inches.  I suspect that at the higher flow velocity in the production system, the stream diameter will actually decrease.
  • As a result of the small stream diameter, I'm going to have to throttle back the pumps.  The fountain was designed to throw 193 gallons per minute 59 inches horizontally with 16 0.500" diameter streams.  Now that the streams are smaller, I'll have to throttle back to 130 gallons per minute, or lower.  I'm not too happy about that.
  • The stream is even more wickedly nonlinear than the nozzle.  The principle problem is that slightly faster flow has over a second to catch up to slightly slower flow.  The stream goes at about 140 inches/second at the nozzle.  So a 1% variation in flow yields a longitudinal variation of over 2 inches by the time impact happens.  If that happens in less than 10 inches (more than 14 hertz), you get blobs large enough to cause the stream to turn into drops.

Sunday, January 11, 2009

Fountain Ballistics

There is a common thread among the questions I get at parties:

"How's the fountain coming?"
"Any news on the fountain?"
"Any progress on the fountain?"

or, from Martha, "Are we ready to can this turkey?"  Sometimes at parties!

As a public service, I thought I'd post how things are going.

First, you should understand that the pool will have an 8' diameter round hot tub stuck into the otherwise straight side.  16 streams of water will issue from jets embedded in the wall of that hot tub, and land in a small circle in the center.  There is a lot of plumbing to make this work, and here is a partial CAD model of it all:

Each of those fat vertical things is a unit fountain, and each contains a diffuser and a flow straightener, and is topped with a head with a nozzle embedded in it.

I've machined the diffusers, cut some flow straighteners, machined the nozzles and heads, and assembled a single unit fountain.  I didn't glue it together.  I supplied it with water from a hose.

Here's how I machined the head.  I started with a 6" PVC end cap, and milled a flat on the end and two shoulders on either side.  These are for aligning the side ports to the nozzle later.  You'll see.

Then I put it on a Bridgeport mill.  Below, you can see that the quill of the machine has been canted 12 degrees over from vertical.  This makes the angle that the water jet will come out on, so that it launches from the wall of the hot tub and lands in the hot tub.  You can also see the parallel on the two studs in front, which fits against the shoulder I cut earlier.  This edge is parallel to the X axis of the mill.  So the mill is canted 12 degrees in the plane of this shoulder, which means I can drill holes in the end cap aligned to this nozzle later.

Here's me working the thing.  It turns out that lubricant makes everything a lot easier.  The Forstner bit was not the best thing to cut this hole with, as I got a lot of heating and chatter.  Occasionally the mill would throw chunks of plastic across the room, so the safety glasses were a must.

I cut the diffuser on a ShopBot at the Sawdust Shop, out of 1/4" PVC sheet.  The ShopBot is a CNC router.  You program the computer to move the router (they call it a spindle because it goes up to 25000 RPM) in 3 axes.  Programming isn't too hard -- you use a ShopBot program to convert line drawings (AutoCAD type, I made them in SolidWorks) into tool paths.  We wrote some text programs to step and repeat the pattern once we had it properly tweaked.

The table is covered in chips because we didn't run the vacuum.  The owner didn't want to mix PVC chips into his sawdust since he uses the sawdust for something.

Result: just after the hose is turned on, it works great.  Flow is smooth almost all the way to the end, certainly good enough to be acceptable.  Here it is just after startup.  Understand that the unit you are looking at will be one of 16 buried in the wall of the hot tub, and the streams will issue from holes cut into the sides of the tiles that cover the top of that wall.  So you won't be looking at ugly PVC.

Up to something close to working pressure, and it's holding together pretty well:

However, within a matter of seconds, the flow becomes less smooth, breaks up, and soon I have a gloppy mess at the impact point.

Here's a closeup of the top:

I'm not happy.

Another (minor) problem: I've got a small leak between the nozzle and the endcap, right where the shoulder is thinnest because I've got the flat spot.  This is eventually going to be buried in concrete, which will greatly slow the already trivial leakage, but I'm not thrilled about this either.  This problem I probably won't fix.

My guess is that I'm getting variations in the flow rate.  The fountain is extremely sensitive to these variations -- a 1% change in flow (from 193 to 194.98 gallons/minute) makes a 2.1% change in vertical height and a 2.0% change in horizontal throw.  2% doesn't seem like so much but it's over an inch, for a stream which is around a half-inch in diameter.  I think the variations I was seeing were at least an inch of throw, at frequencies from 3 to 10 hertz.

There are two possible reasons I wasn't seeing the variations at first:
  1. Because the hose wasn't fully up to pressure.  The hose is not linearly elastic.  When I first turn the water on, the hose fills with water, but it doesn't develop any pressure until it sees back pressure from the nozzle.  At partial flow through the nozzle, the back pressure is low enough that the hose is quite stretchy, and so it acts as a surge tank and filters flow variations.  At full back-pressure (probably just 2-3 psi), the hose has developed it's fully round shape and is considerably stiffer, losing the filter function.
  2. If there is air in the fountain, it will act as a surge filter.  If I could figure out a way to insert a tennis-ball-sized flaccid bladder filled with air into the fountain, I'd probably have this problem licked, but I'd need something guaranteed not to leak in 30 years immersion.  I don't know what that is.
Obviously I need to figure out exactly what is going on, and how to fix it.  I'll post some analysis later.

Wednesday, January 07, 2009

Tooth Fairy Traffic

Anya just had a tooth pulled by the dentist.  She put it under her pillow but the Tooth Fairy seems to have not made it last night.  A few notes:
  • The Tooth Fairy came on the fourth night after she lost her last tooth.  Anya's theory is that the day before she lost that tooth, a little boy lost one of his, but he was sick, and the Tooth Fairy caught a bug when she touched the tooth.
  • I believe Anya is lying, eyes closed, in tense anticipation of the Tooth Fairy.  She was watching Martha and I as we climbed the stairs last night at 1:30AM, and she was wide awake and ready to go at 7:00AM when I got her up this morning.
  • Little girls, of which I have 3, have 20 baby teeth each.  Even discounting the molars, which might come out much later, I can forsee a time when the Tooth Fairy will be making more than one visit per month.  With multiple little girls engaging in outright deceit in their attempts to catch the poor fairy, I forsee delays and uncertainty.
I'm not sure if the full-body tension I'm feeling right now is because I'm worried that my contractor won't be able to lay out my circuit board properly, or because the damn house creaks even though it's only six years old.  Yeesh!

Thursday, January 01, 2009


If you're out there, send me a better email address.

5 Ways to Die During Reentry

If you haven't already seen it, the Columbia Crew Survival Investigation Report.

During reentry, there is a 10 minute long window of maximum heating.  They almost made it through all 10 minutes.  Right at the end they lost their hydraulics.  Makes me wonder if they could have flown the orbiter at a funny incoming angle to spare the load on the left wing.  Maybe they wouldn't have gotten Columbia onto the ground, but if it had broken up five minutes later things might have gone a bit better.

There were 40 seconds after loss of control during which the Columbia pitched up into something like a flat spin, and the folks inside tried to get their hydraulic systems back.

After that, they had a depressurization that took less than 17 seconds and probably, hopefully knocked everyone unconscious.  Nobody dropped their visors (which would let their suits handle pressurization).  Apparently they were all in "fix the vehicle" mode and not in "survival as long as possible" mode.

After that the cabin seperated from the rest of the vehicle, the crew's shoulder and other restraints mostly didn't work, and they got thrashed to death: fatal trauma to their heads from the insides of their helmets.  Owww.

From my reading, had they dropped their visors, gone to suit oxygen, and braced, several of the crew could have made it through both depressurization and cabin separation.

But then the cabin blew apart and they were in their suits in a mach 15 airstream.  I didn't actually read this anywhere, but it sounds like most of the suits came off before they hit the ground.

Side note for camera geeks: notice how crappy the home video shots of the breakup look.  Then look at the Apache Helicopter shots of the same thing, especially when it zooms in.  That chopper has some nice telescopes!