Looking deeper into the poor airflow observed on my stock of surplus Matrix pneumatic valves, I did a subjective measurement of the airflow on each port, on a scale from 0 to 5, using a 20PSI supply. At zero, the pressure could just barely be felt by my finger over the port, and even then, I could only really feel the pulse of pressure as the valve opened. Five was enough airflow to allow me to make a selection of strange noises by controlling my finger pressure over the port, and would probably be adequate to run the air circuits on the caster.

As I mentioned in the previous post, leaving the valve casing screws a bit loose increased the airflow substantially. A sixth of a turn loose improved the airflow on most ports to well over 5, and the ports that were zero improved to 3 or 4.

The valve body comes apart into four main pieces: The front cover (with the individual air ports), the shutter body, the solenoid body, and the back cover (with the electrical connections and supply/exhaust ports). Four M4×0.7×40mm socket head cap screws (3mm hex key) clamp the covers together holding all the sealing O-rings compressed. An additional two screws hold the back cover to the solenoid body. On disassembly the various O-rings tend to remain in their grooves, but occasionally fall out so you have to watch for that.

I disassembled the valve, numbered the shutters using a permanent marker, and measured the thicknesses of the seals. They are made of NBR rubber, so measuring them is tricky because the measuring tool will compress them.

My first try was to use a pair of calipers, which I would loosen until the shutter fell from their grip. Later I used a micrometer, first tightening it using the “click” knob to get a relatively consistent torque, and again loosening until the shutters fell out. The two measurement methods that involved loosening until the shutter dropped out were very consistent: The shutter seals are about 4mm thick, and the average of the differences between the two measurements was only 0.002mm with a standard deviation of 0.014mm. The difference between the “click” and “drop” measurements with the micrometer averaged 0.19mm with a standard deviation of 0.03mm, so the extra compression from the “click” knob was fairly consistent as well.

I charted the three thickness measurements against the subjective flow value and found a very clear trend:

Besides the clear correlation between thickness and flow, if I take the slope of the trend line (33 flow units/mm) and multiply that by the extra space obtained by loosening the case screws (⅙ turn × 0.7mm thread pitch) the result is 3.85 flow units, which is bang on with my observation that a port with zero flow improved to the 3 or 4 range with the loosely assembled casing.

I also gave a close look at the cross section diagram of the valve from the manufacturer’s literature, on the assumption that it would be to scale. Based on this assumption, there is just a bit less than 4mm of space between the NO and NC ports, so a shutter 4mm thick could plausibly seal both ports preventing substantial airflow from either. From the diagram, the shutter seals should be closer to 3.4mm thick, so they would appear to be about 15% swollen. On the other hand, the metal shutter leaf appears to be about 1mm thick on the cross section, but is actually 1.5mm thick. This could perhaps be another variable the manufacturer changes when building the different models of valves for various pressure ranges, because to some extent a thicker shutter leaf would ultimately get more force from the magnetic field of the solenoid.

So this leads to the question as to what to do to remedy the flow problems. One possibility would be the add a shim between the shutter body and solenoid body but now it appears that to restore like-new flow rates the shim would have to be 0.5mm thick (about 0.020″, or ⅔ of a turn loose on the casing screws) and I’m not sure the O-rings that seal the individual shutter chambers would expand enough to remain sealed.

The other choice is to try to remove from the rubber seals whatever is causing their swelling, using some combination of heat and solvents. In general heat causes rubber to contract, so anything absorbed into the structure of the rubber would tend to be expelled (albeit slowly) by this contraction, like squeezing a sponge. The specifications for the valves list an operating temperature range of -10 to 50°C, and it appears that NBR is generally useable up to 120-150°C depending on the grade, although it is not clear how much hotter it has to get before it starts to break down.

The shutter leaves are iron and probably treated somehow to prevent rust, so whatever I do has to preserve this treatment. The metal appears to be nickel-plated and is showing a bit of corrosion in spots where the plating is worn or lifting. The first thing I might try is prolonged soaking in near-boiling water, with some soap added to carry away any oily materials that exude from the rubber. I could get a slightly higher temperature using salt water, but that would cause rusting of the metal.

The most expensive and bulky aspect of making a system for computer control of the Monotype composition caster is the pneumatic solenoid valves. You need 31 valves, and they typically cost $30-$50 each. They are generally large enough that they need a separate housing with an octopus of hoses running from the valves to the adapter on the air tower cross girt.

I found several surplus air valves that solve both problems. They are compact: up to eight ports can be controlled by a single valve body which is about a 2-inch cube. And being surplus, the ones I have are very inexpensive. As it turns out new ones, though not cheap by any means, at about $25 per port are on the low end of the price range.

The valves I have are Matrix Mechatronics BX758DE2A324 units with date codes indicating they were made in 1998. The manufacturer offers valves in this series with several options in terms of coil voltage, air pressure range, Normally Open/Normally Closed operation, wiring options, etc. This particular model is designed for 8 bar (120PSI) of pressure, Normally Open (NO) operation, and 24 volt coils with diodes to damp inductive kickback wired for a common negative terminal. Because the air ports on the Monotype spend most of their time at atmospheric pressure, NC operation would be preferable because there would be reduced power consumption and less heat produced. However I can make the supply manifold to suit either NC or NO valves, and holding NO valves closed most of the time as opposed to opening NC ones occasionally would just be a minor change in the software that drives the valves.

This cross section shows two of the individual port valves with the ports themselves on the right, the upper one in its normal “closed” position and the lower one in the energized “open” position. O-rings are marked in red. Air paths are marked in green.

The mechanical structure of the valves seems to be the same for all these options, and it wasn’t really clear to me what difference they actually made. In particular, it seemed to me that the only distinction between NO and NC operation would be which port was connected to the compressed air supply and which one would vent to the atmosphere.

I hooked up compressed air to the central NC port with no wiring connected and found that most of the valves leaked air into the individual ports. Energizing one of the coils would increase the flow but this increased flow was not always phenomenal. This proved to be the case for all the valves I tried.

Thinking there might be more to the NC/NO distinction than just connections, I attached the compressed air to the side (NO) port instead and found that with all eight coils energized, there was no leakage at all. My theory now is that the NC and NO versions (and the 8 bar and 4 bar versions) differ in the strength of the spring that holds the shutter closed. The total force on the shutter is a combination of the spring force, the magnetic attraction from the coil when energized, and the difference in air pressure between the shutter chamber and the sealed port. For best performance, the spring force would be customized based on what these pressure differences would be. On the NO versions this spring is weak because when the shutter is in the normal position, the supplied air pressure will be pressing it closed. When I try using it in NC setup, the air pressure is trying to push the shutter open and the spring is not strong enough to resist this, so it leaks. Used in NO setup, the solenoid coils are strong enough to hold the shutter against the air pressure, so they seal properly.

By this time I had removed one of the driver integrated circuits from the circuit boards that came with the valves, and built a manual 8-channel driver on a prototyping board so I could easily operate each port individually.

Although operating in NO mode gave good sealing when the airflow to the port should be off (i.e. with the coil energized), airflow with the coil off varied from almost nil to strong-but-not-as-much-as-I-expected depending on which port I tested. This was similar to what I observed when trying to use NC mode.

Taking the valves apart turned out to be fairly easy, as long are you’re careful not to drop any parts. By taking apart the valve and assembling it with sections rotated a quarter turn I could change which coil and which shutter was used with which port. I found that the good/bad flow seemed to follow the shutters themselves as I moved things around. By chance once I failed to tighten the casing screw completely and was blown away by the increased air flow. I put all the parts back to their original positions, and found that there was excellent airflow on all the ports if the casing screws were left backed off about half a turn from fully tight.

My theory is that the rubber button which forms the seal on each shutter has become swollen (for instance, from absorbing oil from the air supply, or perhaps just from age) so it in fact almost seals both the NO and NC ports at the same time. Leaving the casing loosely assembled increased the space between these two ports allowing for better air flow.

Swollen shutter seal (blue) blocks both NO and NC ports.

Loosely assembling the casing moves the NC and NO ports apart a bit so the air can pass through the port that is supposed to be open. The O-ring around the shutter chamber (red) expands enough that it still seals.

Based on the model number, the seals on the shutters (as well as all the O-rings) are NBR (Nitrile rubber also known as Buna-N). This type of rubber is subject to damage by ketones, aldehydes, and esters, but not by regular oils or glycols.

According to the North American distributor, internal parts are not available for rebuilding these valves, so it looks like I’ll either have to fix them myself or buy new valves.

Fixing them myself might involve somehow shrinking the seals on the shutters a bit, or somehow machining off a bit of the rubber, or adding a shim to the valve assembly to permanently hold the shutter chambers a bit wider.

I might also be able to strengthen the springs by lengthening the tiny push rods that transmit the force from the spring to the shutter.

Before doing any of this I need to be able to measure the shutter seal thickness, first to see if there is indeed a correlation between thicker seals and poor air flow, and second to be able to determine if my attempts to shrink them are working. Measuring these seals will be tricky because they are soft; any force applied in measuring will compress them and make them seem smaller than they are.

Similarly I want to measure the actual spring force to see if it is indeed plausible that the air pressure is strong enough to blow the NC valve open, and to determine whether just adding a spacer to compress the spring more would suffice.