[Interhouse] Laminar Fountain - 2019

The arches as viewed from the dance platform. You can see where the inner illumination hits the stream breakup right before the arch.

Summary:

This post covers the design and construction of two laminar flow nozzles with LED lighting. The two units have a throw of around 10’ and, with the exception of our outlets, are made from materials that can be easily sourced at home-depot and an aquarium store of your choice.

This was my 2019 interhouse* project, which I co-lead with Jack Caldwell. The overall theme of the party was “bio-luminescent world” so the two nozzles were installed with RGB LEDs over the entry bridge to Blacker’s dance platform. Overall, the fountains seemed to be very well received. Early on in the party they were a popular photo spot, and people were playing with one or both of them pretty much constantly throughout the party (see the video for what that looked like).

Collaborators:

The best part of any interhouse is always getting to work with other moles on a project of mutual interest. I was delighted to work with the following fellow students.

• Jack Caldwell (co-lead)

• Brittany Wylie

• Harrel Dor

• Gracie Suenram

*An annual event which, in Blacker, falls somewhere between a party and a tech-demo.

Outlet:

The outlet of a laminar flow fountain is a bit odd. Since the goal is to minimize turbulence, you actually want to design it with the cleanest, roundest, sharpest inlet edge possible. Then you want the rest of the nozzle designed such that the water doesn’t touch the walls (frequently this results in a sort of inverted cone geometry).

We went through a number of designs before ultimately settling on CNC machined inserts set into 3D printed inserts. The large outlet cap for our nozzle holder takes 12 hours to print, so using a modular design helped us keep the printing time per test down to a much more reasonable 1.5hrs per part. Both the nozzle holders and outlet caps were 3D printed on either the school printers or my own machine. To help with waterproofing, I increased the floor, wall, and scieling thickness to 1.8mm, and upped the infil to 95%. We also found that using 3D infills (any of the ones based around printing cells) worked better than 2D infills, though the difference was not substantial. Some of our rushed 3D prints did leak, but the ones we printed on-spec and on-speed never leaked. The parts were sealed against each other using marine grease for testing (which sort of worked) and using super glue for the final fitup (which also sort of worked, but didn’t require clamps). We used 4-40 threaded inserts for the outlet cap to nozzle holder interface.

Note: Our first quick-swap system used a female threaded pipe cap, and 3d printed threaded inserts. This may or may not have been sufficient, but we moved away from it out of a concern that the threads were disrupting flow and limiting our performance.

The actual nozzles were cut on our Fadal VMC15 out of 1” aluminum stock. The central hole was bored to 5/16 using a 1/4in carbide endmill, and the counterbore was bored to 3/4” with the same endmill to prevent the stream from contacting the walls. On some of the nozzles the outlet hole was then counterbored to reduce the walls to a sharp point, while on others they were left at their full 0.1” length. The water seems to neck sufficiently to avoid the full 0.1” nozzle length and thus both geometries performed equally well. In the hope that it would produce a better edge (which under a microscope it certainly did) we lapped two of our 3 nozzles. However, while this did improve the edge it did not improve performance so it may be that we were limited in other areas.

Note: I am happy with the final performance of our nozzles, and would use design #1 (short bored section followed by a 45 degree chamfer to a much larger bored section) again without reservation. However, we also found that lapped washers performed similarly as long as they were both ID and surface lapped. This may be a more economical option for you if you are trying this project without access to a machine shop. A list of the other things we tried can be found below:

Nozzle Methods Tried:

Each Nozzle was turned from round stock to (very) rough dimension. The lathe-faced side was then placed face down (see above) with a single parallel under the part for reference. The setup was then tightened and the parallel was removed. I found this setup to be sufficient for our purposes, and plan to use it for future insert cuts.

3D printed: This actually produced a great nozzle for almost all of the circumference of the anulus. Unfortunately, we were never able to get rid of the blob when the printer transitioned from the innermost line to the one just outside of it. As a result, performance deteriorated quickly past a 2’ throw*.

Laser Cut: Worked Spectacularly worse than the 3D printed nozzles, it seems the laser cutter introduced a small burr on the bottom side of the cut. After polishing that off it did work better than the 3D printed nozzles and could be promising long term. The process was very fiddly though so we moved on in the hopes of finding a more repeatable option.

Drilled Acrylic: Worse than the laser cut acrylic in every way. I would not try this again, a Delrin or HDPE sheet might work better though.

Unprocessed Washer**: This was our first strong performer. Flow remained laminar out to about 4’ - 5’ depending on the washer. Though there were some washers where it broke up immediately, so it’s important to insped each washer before use.

Lapped Washer: Lapping the washers on a flat surface and then ID lapping them significantly improved the consistency of our results. This is where we started to become limited by other factors in system rather than the nozzle. I would personally suggest going with this option unless you have easy access to a CNC mill or CNC lathe. It is however, fairly time intensive because a lot of material had to come off to get flat parts.

CNC Machined: Performed slightly better than the washers, and were much faster to make. We turned the blanks on a lathe out of 1” aluminium rod stock and then processed the actual nozzle geometry on our VMC15. Lapping the inlet surface improved surface quality, but not in a way that seemed to matter.

*all ranges given at roughly 45 degrees of elevation.

** None of our washers were stainless. This was fine for testing, but I’d strongly suggest getting a stainless washer if you want more than about 20 minutes of run time.

INLET:

We designed our inlets to generate a rotating flow in the first chamber of the nozzles. The idea was that this would better spread the water over the whole inside of the nozzle rather than concentrating it in one area. It also allowed us to print our inlets with a small cylindrical body in the center to hold the acrylic rod and LED. The inlets were printed as a single piece and took about 16 hours each. On the whole they worked fairly well, although I think some kind of self-contained tightening system to tension against the outlet caps would have helped during the prototyping phase.

In the interests of full disclosure I should probably note that this is one of the areas where we did the smallest amount of experimenting. We tried axial flow early on, but switched to radial based on the internet consensus and then never really tweaked the design. We got good dividends from the areas where we choose to focus our efforts, but I think the inlet would be a good area for experimentation on future builds.

Flow Straightening:

The flow straightener without optical bypass. I think the rough top screen and straw twist contributed to the poor performance of this unit. With that said, resolving those issues only marginally improved overall laminar throw.

The design we settled on was two pair of 2in disks of aquarium foam spaced by 5.5in and set about 1.5in back from the outlet. We sank, by far, the majority of our prototyping time into the flow straightening system, but I think it remains our biggest area of potential improvement.

In particular, most of the high performance systems we have seen online use straws, or some kind of straw analogue, to artificially lower the renynolds number in the pipe and create laminar flow. We originally started out using straws, and tried a number of different sizes and lengths of straws over the course of our build. Overall, we found that smaller and more consistent lengths, perform better than longer, or less consistent lengths. However, the effect was pretty small (though any straws performed better than no straws).

One possible reason for our poor performance could be our use of a “straw carrier”. These were 3D printed housings that allowed the straws to be packed externally and then inserted into the pipe as a unit. From a fabrication standpoint this was great, but it is possible that the irregular patterns designed into our straw carriers contributed turbulence to the system which then fowled the clean stream from the straws.

The two nozzles we deployed differed slightly in that one used a straw carrier (with straws) to space the two layers of aquarium foam, while the second left that area blank. We so no significant difference in performance between the two units and suspect that the straws were completely canceled out by the foam.

Lighting:

Control:

Our LED control board. It held up well during the part, though the power input connector was damaged during take down.

The electronic for this system were quite simple. An Arduino Nano was used to control two 9W RGB LEDs from Adafruit. The LED brightness was set with a pair of 20 ohm resistors on each channel, and a pair of ULN2803s provided the current capacity required to run each channel at 500ma. For LED connectors, we used some 5 pin molex connectors (because that is what I had) and for main power we used an XT60 connector directly soldered to the board. This system worked well, except that it was short on PWM pins so not all of the channels had brightness control.

The pattern we ran used a fair number of blue/green fades with with the occasional switch over to red/purple. This worked well, though I am certain there are more options to explore.

Block diagram for the LED control board. It should be noted that the nano only has 6 PWM channels so it is necessary to either use some channels as full on / full off, or to use software PWM for some colors.

LED cable wiring diagram, included here for future reference.

It’s hard to find a pinout for the PSM-ISR950EP online, so I’ve drawn this up for future reference. There are a few pins whose purpose I have not yet determined, but I’ve got enough to use the model for stuff like this.

Power:

The power supply for this project was a PSM-ISR950EP server power supply pulled from E-Waste several years ago. After a bit of sleuthing I was able to get it running, and found that it was way more than adequate to run our two 9w LEDs. Putting an XT60 connector on a 5v rail did feel a bit weird, but I think it was better than using alligator clips or some bandanna plugs (the other two medium-current connectors I have on hand).

We ended up placing the power supply behind the fountains, outside of the flooded area, and then running the 5v and ground wires over to the fountains in a pair. Putting the power cables in the water was not an ideal choice, but certainly much preferable to having 110AC suspended over the water.

Light Pipes:

We made the decision early on to place our LEDs outside of the housings and move the light into the nozzle with a light pipe. This saved us from having to worry about waterproofing the LEDs, and also gave us a bit more flexibility on our timeline. We used 1/2in by 10in acrylic rods as our light pipes. These were inexpensive, and seem to work reasonably well for transmitting the light once it has entered the rod.

The one major issue we ran into was coupling. The LEDs we choose have a domed geometry that creates a very braud distribution of light. This means that with flat ends we were really only getting 30% - 40% of our light into the tubes. This was especially clear comparing the 300 Lumen LEDs to a 150 Lumen flashlight. The Flashlight produced a much brighter stream despite a lower total power output. After reading through some industry documentation on light pipes (and talking to an alumn who works in optics) we ended up drilling a small hemispherical hole in the inlet of each rod and then epoxying the LEDs in place. This increased overall performance slightly, and made the rods easier to handle and install (since they were now single integrated units. We also wrapped each rod in Teflon tape in the hopes of improving transmission, but it is unclear if this had an impact.

The as-of-party performance was pretty solid, with a nice lit up section at the peak of each arch. I am happy with the results, and think that other improvements (like special effects or more quantity) would produce greater dividends for future builds. With that said, a switch to real fiber-optics, or maybe better pre-tube focusing would likely produce a dramatic increase in transmission.

A note on resistors: The electronics for this project took place on pretty short notice, so I ended up using 1/4w resistors for all of my current limiting. With two 20ohm resistors in parallel sinking 500ma this works out to about 1.25w per resistor. I strongly recommend against doing this. However, I did run several tests and even running full brightness we were fine for the ~20hr design life of the board. If you are making this project with more time to spare please just purchase correctly rated resistors.

Doming the input and then using epoxy helped us capture more of the LED’s light flux. The paper we read suggested we should get a 80% or so improvement. In practice I think it was more like 20%-30%, but still substantial. (Also human vision isn’t great for scaling brightness differences so it’s hard to know for sure)

Sustained testing (10hr/day for 8 days) resulted in no performance drop from submerging the LEDs. There was some electrolysis going on at the contacts, but there did not seem to be any significant contact degradation as a result (over our relatively short test). I would not suggest using this kind of setup for a long term installation, but for something that only needs to last a few hours or days it could be a good option.

Fluids:

Water for our fountains was provided by two pool drainage pumps: a 1/4hp pump which has been in the house for many years, and a 1/3hp pump purchased for this project. The fountains were set up inside a flooded basin which provided the water-return for the fountains, and was a substantial interhouse project in it’s own right (our thanks to the flooding team for letting us set up in their installation).

A rough sketch of the water flow through our system. It should be noted that the line running to the second nozzle was about twice as long as the line to the first (hence the need for an equalizing clamp).

Water straight out of the pumps has a fairly jittery pressure head, we did not measure the pressure, but I would guess the jitter was around 1khz. Left unattended, this turned into lower frequency pressure variation in the nozzle and degraded fountain output. To resolve this issue, we installed an 8’ section of vertical 4” pipe, capped at both ends, as a low pass filter. Both pump inlets were positioned at the base of the filter, with the two fountain outlets, and the pressure relief valve were positioned about 8” higher up the tube. This did a good job of filtering vibrations from a single pump during testing. However, it was less effective (though still sufficient) with the two pumps both running. It seems as though it may do a better job of catching the higher frequency variations from the 1/4hp pump than from the 1/3hp pump.

Though not a perfect analogue, it may be helpful to think of this system as an RLC circuit. With the tube having many of the properties of a capacitor (the water is pressurized against a “spring” of air, giving the tube storage additional storage capacity with pressure), while the lines behave a bit like resistors (in that they reduce pressure flow) and inducters (in that there are momentum effects from the flowing water). This was not a major focus area for us this time around, but I look forward to more fully exploring this system in the future.

The overall water flow through the nozzles was tuned using a ball valve connected directly to the low pass filter. This let us bleed off pressure and helped keep the arch-lengths to something reasonable (such that the flow would remain mostly laminar). To equalize throw lengths between the two nozzles a clamp was placed on the tubing running to the closer nozzle, and then tightened until the pressure drop along both nozzle supply lines was roughly equal. This worked well, but required about 10 minutes of hand tuning to get right.

Budget:

Our original budget for this project was 275$, of which we spent about 315$ (this was not a budget we were expected to adhere to). Our largest single cost was the new pump, with much of the rest going to prototyping and brass fittings. The control electronics and power supply are not included because I either had them already (power supply) or purchased them separately (arduino). Not including prototyping costs, we ended up spending about 25$ per nozzle in final parts.

The most significant areas for cost reduction are likely the brass fittings (4$ at home depot, 1.5$ online) and the vinyl tubing which is cheaper in bulk, or which could be replaced with hose at lower cost.

IMPROVEMENTS for next time:

Use a collar or other quick-connect mechanism for holding the two ends on. This would make it much easier to prototype. Something based around 1/4-20 cap screws would be good because you could then use an impact driver to set it up and take down.

Real fiber optic is not that much more expensive than the acrylic for small numbers of fountains, and seems likely to work much better. For fountains where light is a priority I would absolutely make the switch. On the other hand, for fountains where bulk is important or where other effects are the focus then the current design works well.

Resources:

• Similar Project Link: https://www.youtube.com/watch?v=o5L6W0YoAd4&t=603s

• Similar Project Link: https://www.youtube.com/watch?v=HBeQkX0WzCo

• Similar Project Link: https://www.youtube.com/watch?v=SjOZmby5K8c

• Similar Project Link: http://laminarfountain.blogspot.com/2014/03/water-works.html

• Light Pipe Reference: http://www-eng.lbl.gov/~shuman/XENON/REFERENCES&OTHER_MISC/Lightpipe%20design.pdf

• Project Documents: [click me]

Testing Pictures:

Testing maximum flow rates with one of our earlier nozzles.

Unintentionally siphoning off the water test bucket through the nozzle. After this we started placing the nozzle above the level of the water, but it was fun to see this in action.

Some early testing with our 6” nozzles. You can see that the nozzle was hooked directly up to the pump with no filtering.

The pipe-clamps let us change the internals relatively quickly, without a great deal of fuss. Overall I liked the design, though I’d probably build some threaded rods into any new future nozzles to serve the same purpose.

Appendix A:

I have been getting some questions about where exactly to hook up the enable jumper for the supply, so I’ve added this picture to clarify. The connection should be from the top right pin to the bottom left pin as shown above. You can use a plain jumper or switch depending on your preference, but do note that they pads can be a touch delicate so you want more strain relief (and insulation) then is shown in the test setup above. As always, don’t hesitate to reach out (use the contact form under “about”) if you have any questions.

[Interhouse] - Stairs 2018

Overview: 

The bow stairs prior to painting. Photo credit: Ethan Jaszewski

Blacker's 2018 interhouse featured 5 staircases, all cut and assembled fresh for this year's platform. We made 48” staircases running from the lower platform to the upper platform, and one 53.75” staircase running from the ground to the lower platform. All five staircase units were structurally self-contained units, bolted onto the main platform. The 53.75" module was constructed relatively early on to aid with general platform construction and allow viewing platform access during PFW. The four 48" units, although nominally freestanding, were assembled in place so they could be adjusted to the real gap between the upper and lower platforms. 

Our overall assembly time ran about 16 student-hours per unit, with efficiency peaking at three people for fabrication and four per staircase for assembly. Discounting costs associated with elevated construction, each unit had a real-world cost of 100$, with an expected amortized cost of 40$ and would have cost 150$ using all-new wood. More details about construction time and budgeting can be found below. 

After helping rescue last year's debacle, my big focus this year was on establishing efficient fabrication practices, and using good documentation to parallelize construction. Although there's still a lot to work on next year, I was successful in cutting our assembly time by about 60% per staircase, while reducing the total unit cost by 30%. Having full mechanical drawings let students with a passing interest take over fabrication of a single component, and made it easier for more engaged students to get up to speed on the overall design. 

My role: 

I ran the project, managed the budget, procured the materials, created the cad, wrote the documentation, and oversaw the assembly. However, this is a strange project for me in that I did relatively little of the physical cutting and assembling of wood. Having a complete drawing packet meant that I could outsource most of the parts fabrication and focus on cutting stringers and teaching people how to use our tools. 

General Fabrication:

With 5 staircases to produce, we focused on trying to produce as many of the parts as possible in batches. This was largely successful for the stringers, support posts, treads, and kick-plates. However, we found that irregularities in the platforms themselves necessitated custom fitting for the railings.

Stringers: 

Stringers are always the most difficult part of a staircase to source and fabricate. Since we had fairly non-standard stair heights on this year’s interhouse I decided to fabricate the stringers from 2x12s instead of purchasing them ready-made. The appropriate mechanical drawing can be found at the end of each design packet (below). The generating cad is parametric in nature (along with the entire stairs assembly) and so easily accommodated our two different stair heights. That file can also be found below.

We experimented with two different types of stringer fabrication:

1. To begin with, 5 high quality stringers were fabricated as follows: a reference edge was cut onto each of several 2x12s. Then the critical points for that stringer were marked according to the drawings, and then connected to form the stair outline. The outline was then cut-out with a circular saw, placed against measured and clamped straight edges.

2. Once 5 high quality stringers had been fabricated, we picked the best 3 and used them as drawing templates for the remaining stringers. This removed the longest step (measuring and marking), and allowed us to quickly produce the bulk of our stringers. These stringers were cut either with clamped guides, or free-hand depending on the confidence and skill of the saw operator.

In both cases, we found it necessary to finish each cut with a hand saw, so as to ensure clean corners. This added about 5 minutes per stringer, but provided a 10% increase in strength relative to over-cutting with the circular saw. Likewise, we found that teams of 2 people were most efficient for both types of fabrication.

Reflection: Stringers cut free-hand using drawn templates were substantially less consistent than those produced directly from measurements. However, they were structurally equivalent and proved largely sufficient in practice. The 70% reduction in speed was more than worth having to throw out 2 wasted blanks.

Assembly: 

Assembly went very well this year overall. I was able to train both our work-frosh to manage assembly groups. This fread me up to do the safety checks and help out where needed. We were generally able to keep two groups working at a time, and were largely limited by our clamp supply more than people or materials. The general process we used ran as follows: 

• Place, clamp, and bolt vertical alignment posts. Where possible these were bolted directly to the main platform supports with 1/2" bolts, but in some cases we used an additional 4x4 to act as an intermediary so the stair supports and platform supports could be placed corner to corner.

• Clamp both outer stringers and drill the stringer to post bolt holes. Note: In the future I would recommend creating a drilling jig for this process. It was difficult for some participants to remember that consistent drilling is important for re-use.

• Place the central stringer, and connect it with treads at the top and bottom of the stairs.

• Continue up the stairs from bottom to top, placing treads. We used three 2x4 treads per stair, with all three pressed towards the outside of each stair. This left no gaps between treads, but a significant gap between the kickplate and treads. It was determined that this was better for students wearing heals.

• Once the treads were all placed, they were then screwed down. We found this to be most efficient with two people (one per side).

• This process was then repeated for the kickplates, placing them with the gap towards the top of the kickplate.

• Finally, railings and railing supports were added. We cut these parts beforehand according to the CADs, but ended up needing to make some slight adjustments to account for the placement of the dance platforms.

Recommendation: Purchase an additional six 18" fast acting clamps before next year's interhouse season begins. 

Documentation: 

• Mechanical Drawings (Short Stairs)

• Mechanical Drawings (Tall Stairs)

• CADs

[Interhouse] Stairs - 2017

Story:

I have always wanted to work on the interhouse stairs, and when stairs team ran into personnel issues (through no fault of their own) I got a wonderful chance to put together our front staircase on the night of the final safety inspection. The experience itself was rather surreal -  the only other person avaliable to help had never worked on stairs either, but ultimatly succesful.  

Documentation:

To hopefully make he build process a bit faster next year, included below, is my solidworks model of the staircase we actually ended up building. As of now, the only significant design change I would make, would be to use three stringers in place of the two stringers and the 4x4 supports. Although the latter option is stronger, and marginally cheaper using pre-made stringers, it is also much slower to build and significantly less elegant than the three-stringer build.  

Model Files

Purchasing Note: 

The home depot in Pasadena has expressed a willingness to loan us stringers for tracing as long as we also purchase the wood we use in the final project from them. Since we buy all of our wood from them regardless, that seems worth pursuing. 

3d printed model: 

One of the great things about having a CAD of interhouse is that it becomes easy to print 3D models of parts of the build. This is especially useful for projects like stairs that are a bit harder for new folks to visualize. With that in mind, I decided to print out the individual parts and glue them together to make a 1/20th scale model of last years staircase. 

Build photos can be found below. Each stage was glued using CA glue with no kicker (to keep it clear), and then allowed to dry for 12 - 18 hours before the next glue stage. All parts are 1/20th scale. 

[Interhouse] River - 2017

My guess is that if you are reading this then Blacker has decided to do a river for interhouse and you were foolish enough to volunteer (It's actually an easy project aside from the paperwork). This page outlines how we did the river for Blacker's 2017 "Japan Through Time" interhouse, but the design is modular, so you should be able to use it for whatever setup you want.  For the pump system and waterfall I have included notes from our postmortom design review with things to watch out for or possible improvments that could be made to the system. 

River Sections:

The river itself is composed out of 8ft sections (see right). Which can be built before hand and then raised to height and connected when needed. These units are intended to be single-interhouse and so emphasize cost over durability. That said, they should be good for a few short uses. 

The assembly procedure we used for each 8ft section runs as follows:

• Cut the two 6" lengths, two 4' lengths, two 22.5" lengths and the 18" by 8' plywood section.

• Lay out two 2.x3s 18" apart (from outer edge to outer edge) and check that they are parallel (if bowed, place so the center bulges).

• Put a continuous line of glue down each board (this acts as a last ditch sealant).

• Place the plywood on the two boards and fix it in place with a screw at each corner. As far as we can tell it is safe not to drill pilot holes if the house is out of bits.

• Add a screw and washer every 6-10 inches down both of the two 2x3s.

• Measure 33 inches from opposite corners of the structure and secure the two 6" sections with two screws. It is important to drill pilot holes.

• Connect one 4' sections to the main structure 36" inches in from the ends using a single screw. It is important to drill pilot holes.

At this point the structures can be moved into place, lifted, and then the legs can be fixed in place using an additional screw at their pivote points and two screws where they cross eachother. I find it also helps to add two braces connecting the sides. Once the structures are in place, make sure to use hot glue (or simiilar) to seal any remaining cracks. This is best done last since moving them will open up new leaks.

Note: During our post-party design review we determined that the seal between plywood and the 2x3s would be improved by the presence of a 1x3 underneath providing additional support. This board should be pre-bent so that it can be used to compress the plywood against the 2x3s. This should help prevent the small leaks that sprung up on 3 of the 4 sections we deployed. This design can be found in the V2 version of the uploaded documentation. 

Unit Cost: 15$

Unit Time: 30 min

Materials: 

• [4.5] 2x3 soft wood.

• [40] 1.5" Deck-Mate screws (go torques).

• [5] 1cm diameter sticks of hot glue

• [1/12] of a medium sized bottle of Elmers wood glue.

• [1] 18in by 8ft section of thin plywood or press-board.

Documentation: 

• PDF V1

• PDF V2

• CAD ZIP

Water Reciculation

Water for the river is pulled from the catch basin, down a length of 3" pipe, and into the pump. It is then pushed up to the waterfall, where it runs out a series of holes drilled in the bottom of the pipe. It then falls down the waterfall, entering the river. The remaining water is then piped back to the collection basin to be re-circulated. Of particular note, the pump sits at the bottom of the waterfall structure and below the level of the water in/out takes. This allows the pump to be primed by poring water down the intake until it begins to overflow back into the basin. 

By far the hardest part of this project was getting the pump system to play nice without giving party-goers an impromptu shower. The pump is significantly more powerful than we had anticipated, and left to it's own devices, was more than happy to project the waterfall a good 3-4ft beyond the catch basin that feeds the river. To lower the pressure, we ended up having to add another stretch of pipe returning unused water to the collection basin. This is clearly not the ideal solution, it would have been nice to pump that water into the river itself instead, but it did work and kept safety happy. 

One thing to keep in mind is that the pressure on fittings after the pump seems to be much greater than on fittings before it. It is well worth clamping or gluing any right-angle fittings or the like on the water return run. On a similar note, it is worth watching the system for a bit after starting it up, because temperature effects from the cold water will change the fit of any press-fit fittings during the first 5 or so minutes of operation. 

If you end up using one of the houses's large sump pumps it is worth thinking about how you are planning to prime the pump before building the system. In our case we found that placing the pump below the water intake level, and then not gluing one of the fittings so the intake can be trivially turned up and used to fill the lowest foot of the system was sufficient. However, what works will depend on how your system is constructed. The important thing is that you have a way to flush bubbles out of the path from the intake to the pump. 

Notes: 

• 3" pipe is much more expensive than 2". Make sure to do flow tests to determine which you need before purchasing your piping (3" is probably overkill).

• Placing "spoilers" or some other mechanism to slow water passing over the holes should help increase flow.

• Make sure the pump is below both the intake and the outtake to the system. That will make it much easier to prime.

• Any fitting which changes the direction of water such that the flow pushes against it should be glued or clamped in pace.

• Having at least one flexible fitting in each straight run to reduce the risk of cracking is well worth the extra cost.

Water Fall

After a bit of back and forth, we decided to start the river off at a waterfall and end with a catch basin (vs flat with recirculating water). That helped us hide the pump mechanism, and gave us an impressive backdrop to the left for folks entering the North Gate.  Ultimately, we ended up building a platform, and then using a single contiguous piece of thin plywood to form the upper basin for the river. That worked reasonably well, but I'm not totally happy with the solution we came up with, so rather than post a detailed build log, I have included a few lessons/ideas that we picked up along the way below.  In a similar vein, you will also find - below - notes from our post-project design review detailing how we think the the waterfall could be done best, and with as little pain as possible in the future. 

The general design that we settled, more as the result of time than anything else, was an 8' by 8' by 4' frame made out of 4x4s and braced with 2x4s. We then used 2x4s of varying heights to bend a single sheet of 1/4" plywood so that it would direct water over the intended edge. The water itself was provided by a 3" pipe with holes drilled in the bottom. The waterfall overflowed into a 2' by 8' by 1' catch basin and from there into the river proper. 

Notes: (Design Review)

• Check the waterfall distance/behavior before building a catch basin for it, could be it needs to be larger than you expect.

• Screw a line of 2x4s 3 '- 4' below the top of the platform where they can be used as scaffolding (a built in ladder is really nice as well).

• Using an upper basin with the water spilling over a lip will help create a nice even sheet of water.

• Creating the upper basin and structure desperately may make it easier to maintain the basin and keep it rigid.

• Make sure the Calk you use is rated for water (oops) and that you finish early enough to let it dry for 2 days before the party.

• Building a ladder and some scaffolding 2x4s to allow easy access to the entire structure (would recommend a single line running 4' below the top) would make maintenance and assembly much easier.

Notes: (During Construction)

• Calk not hot glue!

• The top of a tall structure like the water fall will flex enough to break wood to wood calk bonds when picked up and moved.

• Black plastic can be pained, although it stops being waterproof if you use staples.

• Plywood bends way way better when wet (soak for 2-4 hours in warm water).

• Surface tension means that getting a uniform sheet of water from a flat surface is hard. If you need even distribution it can help to build a small lip and then cut channels through it.

• Clear plastic drop cloth from home depot turns actually clear under water (whenever possible better than black plastic).

• You can soak thin plywood sheets and then bend them to get continuous water tight surfaces as needed.

• Water will stick to drop-cloth and adhere it to nearby surfaces, so if you have leaks you need to direct towards more productive destinations than the bricks, that can be a good way.

• Most leaks don't really show up until the wood and plastic have had a chance to saturate. That takes around 20 minutes, but once that's take place they don't seem to get much worse until the wood starts to warp (by which time interhouse is hopefully over).

Pictures: 

Because pictures are never a bad thing.