If I were picking a nice new set of o-rings I would use the Aflas o-ring. http://www.mcmaster.com/#o-rings/=v20df8 It is rated for 400F and a steam/water environment. These are even reasonable in cost. The EPDM will not be happy with the heat but will last longer than Buna-N or Viton, still not a really acceptable service life though..

Just for my edification - what is an acceptable service life?

Those o-ring seals should last for 15-20 years before even thinking about replacement. In most cases they are considered as not being a hidden failure so they are replaced when failed. In other words when they leak. The only reason for leakage in this application is degradation due to the environment which is steam/water and heat. The do not move, they just fill the gap so there is no wear to consider. As Teflon is a trade name and does not accurately define a specific product there are multiple specialty o-rings that are 1 form or another of teflon and they will work.

Name one espresso machine that gets 15 years out of their seals.

AssafL wrote:If only La Mazocco would have bolted the Gicar to the top side of the box (rather than to the bottom), slow leaks would not have the propensity to kill the controller.

What a great idea! Never thought of it. I might look into doing that mod, though I suspect it's not feasible with the current box configuration. As you say below, the mains switch might be in the way.

AssafL wrote:I used conformal coating on the replacement Gicar, but now thinking about it - extending the standoffs would make a great way to keep the "brain" away from water damage. I am not sure there is enough room in there under the mains switch, but I'll look into it next time I take it apart.

Yeah, the mains switch is likely to pose a problem, and I'm not sure there's enough room above the board to raise it much. And I don't think it would solve the problem because water can, and does, wick onto the top of the board via cables that plug in on that side. When my seals leaked I found evidence of water contamination on top of the board.

Suspending the board from the cover is the best solution because water couldn't wick up to the board via the cables and the board would be further from the bottom of the case. Drilling holes in the bottom would be essential for this to work.

Now, an interesting solution might be to turn the brain box upside down and drill holes in the cover. The question is whether there's enough slack in that thick package of cables to twist it 180-degrees. Might require undoing the bundle and re-dressing the cables. Or, it might require extending some of the stiffer cables. Might need longer ribbon cables, too. And then there's the question of whether turning the board upside down would compromise heat dissipation for any of the chips on the (current) top side of the board, especially the CPU and voltage regulators.

Oh, and the connection to the rear cover would probably have to be redone...

I can save you the trouble - when I looked into it - the switch is in the way of bolting the board to the top. That is why I think there may be a few mm available to raise the board but I'll have to take it apart to see (and that is not planned any time soon).

Water on the top of the board -IMHO - are less of an issue at low volume leaks since water will bead and evaporate from the top rather quickly (again - at low leak volumes).

As for turning the box upside down - I think it is a very asymmetrical box with the screws going through the box to the chassis. It serves as one of the bottom covers of the machine and is very irregularly shaped as well.

I highly recommend conformal coating for any GS/3 - make sure you cover any contacts (with masking tape) and do through even coatings (spray from multiple angles). I did 3 layers baked the board between layers in a 50 degree convection oven. Very hydrophobic and water now avoids the board like crazy.

About water wicking through the cables: undersea cables are always filled with a special grease (vile stuff - very difficult to clean) that prevents lateral migration of water in cables. Something like this: http://agrinol.com.ua/en/catalog/smazki ... ol-%D1%96/ Usually it is some synthetic grease filled with PTFE particles that is extremely hydrophobic.

If we source some of this stuff and anoint the cables with this stuff - that should solve much of the issue. (It is nearly Xmas so we should probably anoint our GS/3 - but not with Frankincense or Myrrh).

I hope to be opening my brain box at some point in the next couple of months, so I'll check the clearance.

I'm not so sure water on top of the board is as benign as you think. Depending on water composition, mineral residue left behind by the water could short traces or IC pins on the top of the board. But I do agree that water buildup under the board is much worse.

I'd love to do the conformal coating, but the thought of possibly damaging my board scares the heck out of me!

FWIW, I have two projects that require access to the logic board:

One is to install an optocoupler board I've designed (but not yet built) that translates the AC and DC signals produced by the board so I can monitor status with an Arduino. This will also allow me to replace the heavy cable that carries the motor AC signal to my externally-mounted gear box controller with a safer 5V DC lead. There will also be a relay on the board so I can send it a signal to open the solenoid so I can bypass the 50-second shot limit. I'm using a Slayer-like low-pressure infusion so I can grind finer, and this often requires shot times greater than 50 seconds. There's room for the board where the motor cap goes, if I can get it out of the brain box. My gear pump doesn't use it.

The second project is to replace the heavy manual steam valve with an electronic valve similar to the solenoids used for hot water and autofill. I need to bring AC from the brain box to a button on the front of the machine for that.

Conformal coating is pretty easy - just make sure you spray in overlapping lines across the board. Make sure to cover all connectors before you do. Being rather impatient and wanting to do 3 layers - I baked my board in a 50 C oven... Conformal coating is pretty compliant so it should cause undue stress on any of the SMDs.

I don't think water on the top is benign. I just think it evaporates. Also, if my memory serves me right I think the bottom of the board has more of the small signal and more SMD components. I think the area surrounding the connectors is more or less bare. Perhaps I am wrong.

I do know the board has a serial port which I assume was meant to connect the Maestro lines of Espresso controllers to coin devices. Perhaps you can use that to get the necessary signals out? Also, the board is pretty isolated (transformer based power supply, and SSR outputs) and the only ground ref is the fill sensor. Is ground loops or ground reference the reason you want to isolate the signals using Optocouplers?

As for the 50 second time delay - if you need it as part of the volumetric system - you'll need something that will continue to count pulses to control the relays; otherwise, you can use a timer. Also, you need to control both the pump as well as keep the 3 way open.

I've tried this ant it works - setting the GS3 preinfusion timer can extend the 50 second time limit to 55 seconds with the 3way open or 5seconds open, 5 closed, and 50 open. That maintains the GICAR volumetric system operational.

AssafL wrote:I do know the board has a serial port which I assume was meant to connect the Maestro lines of Espresso controllers to coin devices. Perhaps you can use that to get the necessary signals out? Also, the board is pretty isolated (transformer based power supply, and SSR outputs) and the only ground ref is the fill sensor. Is ground loops or ground reference the reason you want to isolate the signals using Optocouplers?

Neither. There are other reasons for using optocouplers. First, the only way to get to most of the low voltage signals is to solder leads to traces on the board, and I don't want to modify the board. It's much easier to tap the AC signals at the terminal block. I can do that with optocouplers I found that convert AC to 5VDC that I supply externally. Second, if I directly tap the DC signals on the board, a short in my circuitry or the cable connecting it to the machine might blow a hard-to-replace semiconductor on the board. It's safer for the logic board to use optocouplers to tap the signals. The board can't be damaged even if one of my AC optocouplers shorts -- that would just blow the fuse on the logic board.

Well, that's almost true. One of the signals I'm planning on tapping is the flow meter, which is DC. I'm using a standard optocoupler for that, which will provide isolation from my circuitry and the cabling to it, but the source of that signal will be vulnerable to a short in the optocoupler.

AssafL wrote:As for the 50 second time delay - if you need it as part of the volumetric system - you'll need something that will continue to count pulses to control the relays; otherwise, you can use a timer. Also, you need to control both the pump as well as keep the 3 way open.

I'm not using the volumetric system. The plan is to completely bypass the GS/3's CPU during the brew cycle.

I'm going to use an external "brew switch" to toggle between normal operation of the machine (i.e., using the brew button on the GS/3 front panel) and "untimed mode". In untimed mode, a relay on my interface board will be activated. The relay will switch AC to the 3-way and will interrupt AC to the autofill solenoid (the CPU suppresses autofill during the brew cycle, so I need to do that in untimed mode.) The switch will also activate the pump.

As it turns out, I'm already controlling the pump externally. I outboarded the stock pump years ago, then about a year ago I replaced it with a gear pump and associated electronics for pressure profiling. At the moment, I'm tapping the AC motor signal to trigger a relay that activates my gear pump control circuit. I'll remove the motor AC tap when I install the above-mentioned board, which will convert the motor AC signal to 5VDC. My pump control circuit is already setup for the conversion -- just have to move a jumper on the board to activate the pump when the 5VDC signal comes in from my interface board.

This is the manual version, After I get it working, the plan is to add an Arduino controller so the brew mode can be selected from a touch panel, I can experiment with various profiling techniques and will be able to store and recall pressure profiles. I may try to add a scale so the shot can be cut when the target weight is reached. That also requires taking control of the 3-way and the pump.

AssafL wrote:I've tried this ant it works - setting the GS3 preinfusion timer can extend the 50 second time limit to 55 seconds with the 3way open or 5seconds open, 5 closed, and 50 open. That maintains the GICAR volumetric system operational.

I discovered that as well. I'm using one of the volumetric buttons to get a 55 minute brew cycle. But it's not long enough. When I'm not doing slow preinfusion, I don't use the volumetric system at all. I prefer cutting the shot by weight and appearance.

You are separating the "Brew" cycle and the GIcar controller: One of the things that would be concerning is since the two operate independently is that the autofill valve may decide to fill the steam boiler while the pump is operational - causing a pressure drop on the puck. Perhaps you can cut power to the autofill valve?

You are very close to completely replacing the Gicar - just add control for the autofill and tea water solenoids - and 2 PID loops (and a steam water level A2D). (Just my tuppence - can add a lot of functions once we are in Arduino land...).

Assaf