High pressure hand pump

September 2016

As I progressed with plans for the nitrous oxide tank, I realised that I needed a means to safely pressure test this and other home-made pressure vessels to 1500psi. So, I made this hand pump…

Safety first

Well… maybe not first…

Why 1500psi? Well, nitrous oxide self-pressurises at about 750psi. This means that if you fill a pressure vessel (a sealed tank) half-full with liquid nitrous oxide, the liquid will boil off into gas in the remaining space. The pressure that the gas will reach is dependent on temperature, but is about 750psi.

It is good practice (or so the internet tells me) to pressure test vessels to twice the working pressure. Or, twice the pressure you expect it to withstand under normal conditions. 2 x 750psi = 1500psi.

I have bottles of various welding gasses that are 3000psi or thereabouts. I could use this inert welding gas to pressure test by tank, but that is really not a good idea…

Gas is compressible. Liquid much much less so. If I were to pressure test with a gas, and the vessel failed during the test, the gas would eject explosively and likely carry some particles of the broken tank with it (this is shrapnel to you and me and anyone standing nearby). This is because it would take some time for the high pressure in the tank to return to atmospheric pressure, because it has a lot of stored energy (like a spring). Although there would not be an actual explosion, the results would be very similar.

If I tested using a liquid (such a water), and the vessel failed… there would be a spurt of water but the pressure in the tank would almost instantly return to atmospheric pressure. This is because the liquid has stored very little energy (this is why hydraulics work like they do, because the hydraulic fluid does not compress very much under load).

Gas is like a spring, with the potential to store a lot of energy. Liquid is like a solid block, it may compress under load but does not store energy like a spring does.

Trial and errors

Initially I tried using pressure washers that claimed to output 1600psi, but even with the pressure switch disabled they all crapped out at 400psi. Very disappointing. (I was inspired to try them after watching some Colin Furze videos on hydro-forming).

Next I looked at commercially available pumps, powered by compressed air or electric motors. I got this idea after watching a video on hydroforming pipes by those crazy Danes at Copenhagen Suborbitals (I am a sponsor, you should be too!). However, they were quite expensive and had performance that I didn’t really need.

In the end I came up with the approach of a simple pump: a piston mounted in between two one-way valves. The idea is not new (or mine), here’s a pic that explains how it works:

Check valves

The first thing I needed to make and test was one-way or ‘check’ valves. These valves do as their name suggests: only permit flow in one direction.

Because I don’t have a lot of spare time, and because I had never made anything like this before and was likely to cock it up, I designed and made a pair of matching valves, which could be screwed into a body that housed the piston (yet to be designed). If I could make the check-valves then I was pretty sure this project would work.

I found BSP standard threads to be very useful for hyraulic fittings, and tooled up with a set of 1/8 and 1/4 BSP taps and dies. The male ends of the valves were threaded to 1/4 BSP. The insides were tapped with 1/4 BSP to accept the threaded inserts.

(Shown disassembled for clarity) The round thing is a stainless steel ball. The browny thing next to it is supposed to be a spring. The terracotta slug on the other side of the ball is an o-ring. The aluminium body is the large item. The idea is that the spring holds the ball against the o-ring, and if there is pressure from the right side of the pic (against the direction of flow) then this just forces the ball harder against the o-ring, preventing flow in that direction.

When some pressure is applied with the direction of flow, this pushes the ball away from the o-ring and compresses the spring a little. The fluid then flows around the ball, and through the spring. I don’t think this is a very efficient design, but I don’t need high flow rates.

The other objects in the pics are brass inserts. There are two for the top one, and one on the bottom one. The job of the inserts is to hold everything in, including applying a little pressure on the spring so that the ball sits against the o-ring.

Above is a pic of the finished item (the top one in the two diagrams above). You can see how the brass insert threads into the body, and you can also see the ball in the middle. The brass insert is a little roughed-up from the screwdriver I used to screw it in place. It took some fettling to get the right preload on the spring.

With the two check valves made, I tested them by joining them with a 1/4 BSP adapter, and then attaching one of the valves to a garden hose (1/2 BSP thread). I made some minor adjustments to the preload on the springs, but it all seemed to work OK.

But, would it hold up to 1500psi? Would the o-rings just tear apart?

Here they are installed in the pump body. Flow is left to right. The valve on the left has a 1/2 BSP garden hose adapter attached to it. The valve on the right has a (manual) pressure-relief valve, and then a 1/4 BSP thread to attach to the length of pressure washer hose that led to the pressure vessel. The pressure relief valve is a variation on the ‘ball and o-ring’ design of the check-valves – the threaded actuator just presses the ball into the o-ring.

Piston & body

The body and piston are relatively simple and straightforward compared to the check valves. I made the body out of aluminium, the piston is 304 stainless (scavanged from my 3D printer).

I set the body up in the lathe, and drilled and reamed the 8mm bore. At this stage I was planning to machine o-ring grooves into the piston as this seemed natural. However the surface finish of the reamed bore had some scratches, and I felt that an o-ring sliding up and down under high pressure would surely tear or burst.

I spent a while travelling down the dead-end of trying to lap the bore: I made some D-bit laps (based on some very useful info I found here) but they did not seem to improve the surface finish of the bore. Not for the first time, the project seemed to be at a dead end.

However, I am nothing if not bloody-minded, and spent a couple of weeks searching for a solution to my problem. I finally stumbled on a single-sentence reply in a forum which made me go “ahhhhhhh”… “O-rings on the piston will jam the pump, put the o-ring in the bore.”

The pump body went back in the lathe, and I machined the two o-ring grooves in the cutaway diagram below. The piston just displaces volume, the job of sealing the volume is just as well managed at the top of the body.

I designed the handle-to-piston mechanism in SolidWorks, I did not use textbooks or mathematical equations… I just kept changing it until it gave me the stroke I needed. The link pins are brass, the connecting rods are 2mm stainless sheet.

There is a plug with o-ring that closes the bottom of the piston bore. The pump body bolts to the aluminium bottom plate, which is turn is bolted to the steel frame (the tube in fetching Yamaha Tenere blue).

The eagle-eyed reader will notice that Captain Cockup was not yet done with this project, as the link pins are all about 4mm too short… but they work so they are still there.

Testing & first use

To test my newly machined nitrous tank, I hooked the garden hose to the inlet of the pump, and the attached a high pressure hose (pressure washer hose) to the outlet of the pump, and to the bottom of the tank. The top cap of the tank has a pressure gauge.

The top cap of the tank has a bleed valve, which is open in the pic above (hence the stream of water). I filled the tank with water but just letting the hosde pressure run through the pump, bled and tank of air and prepared to start pumping…

I was ready for failure in a variety of ways: endless pumping on the pump, tank failure, failure of the pump… I was not prepared for complete success. Four pumps on the handle and the needle of the gauge scooted neatly around to 1500psi, and stayed there.

I learned a great deal in this project, especially about o-rings and BSP threads. Lots of fun and the thing actually worked. Result.