Repairing and damp-proofing a V-Clamp board

A job, long delayed by uncertainty about how to proceed, has finally been taken off my "to do" list: how to restore Voltage clamp boards which have failed. Time to devote to it happened because of Covid, - a minor bonus in these otherwise difficult times.

Early Powerspout turbines had V-clamp boards to keep voltage within limits when the output from the turbine is to an open circuit; the function of the board is to divert power to a dump load and thereby hold the Powerspout's output to a voltage acceptable to whatever device the turbine is connected to: - either an MPPT controlled battery charger, or as in my case, a grid tied inverter.

When manufactured, the boards were covered with a conformal coating intended to keep the underlying electronics free of moisture, but the coating used was hot melt glue and at low temperatures this becomes brittle, separates from the underlying printed circuit board and develops cracks. Through these cracks, moisture could reach the live circuitry beneath and sooner or later this had terminally disastrous results.  

In the early months of my installation in 2013, two boards died in this way; but the years since have been completely trouble free through EcoInnovation having come up with a one-off special, - special in that the board was entirely encased in an epoxy resin block.

So the long delayed job has been to repair the two boards that died, and then to cast them in epoxy, as had been done for the special.  

Here, in pictures and captions, is the process:

An original board showing how the coating of hot melt glue cracks and separates from the underlying pcb, allowing moisture to enter.

Typical tell-tale evidence of underlying pcb damage with blackening visible through the conformal coating and molten glue running out. The white stuff is heat conducting paste applied to the heatsinks on the board where they attached to the bulk head of the turbine; the bulk head, being water cooled on the pelton side, creates a good means of dissipating heat from the power electronics on the board.

Initial view of the damage once glue had been removed; since hot melt glue becomes brittle at low temperature, cooling the pcb in the freezer to -18 deg C allowed the coating to be chipped off, but it is a very time consuming process and care is needed not to accidentally remove any surface mounted electronic components. (addendum 9 Feb 2021: see below for an alternative method.)

Later view of damaged area after further cleaning, showing the glass fibre substrate of the pcb has carbonised and copper tracks have been lost; carbon being a conductor, this all had to be removed using a precision rotary tool (Dremel) and the resulting cavity filled with non-conducting epoxy.

Top-side view of the same area of damage; from some components such as the surge protection thermistor in the centre, the hot melt glue could not be removed, so this device was de-soldered and replaced with new; flakes of hot melt glue are still visible around surface mounted components.

Damaged area after repair; the next stage is a final clean with isopropyl alcohol then casting in epoxy

Topside view after repair with a new surge protection thermistor and adjacent capacitor

This is the 'special' made for me by Andrew Smithies, who designed the board; it was with his advice I arrived at my method of casting; one key piece of advice concerned 'thermal runaway'; this is the intense heat generated by two part epoxy compounds as they cure; the heat was so much when he did it that it melted the mould and let some of the epoxy escape; the resulting void had to be filled later with more epoxy; but epoxy added later to already hard-cured epoxy does not unify into one block, so I was keen to avoid this happening.

Before proceeding to casting the repaired pcb I wanted to check it was fully functional; at this stage it was vulnerable to moisture with the bare soldered components having no protection, so testing had to be done on a dry day with low ambient humidity; the test shows volt meters on the two outputs when the turbine was operating without power being fed to the inverter: - the digital voltmeter reads 381v dc, which is the correct voltage for the 400v version of the V-Clamp board, and the AVO analogue meter is reading the voltage being fed to the dump load. The turbine was then allowed to connect to the grid via its inverter and a test was performed to simulate a 'loss of grid' event to check the circuit board instantaneously diverted power to the dump load; all tests were satisfactory and I felt confident the board could be epoxied.

The secret to constructing a mould which can be used more than once is to use material for the mould to which epoxy does NOT bond, and that means polypropylene; a convenient source is a kitchen cutting board and such is what I used, price £7; for the curved bottom, on my first effort, I used a strip of flat rubber sprayed with silicon lubricant to act as a releasing agent; but the epoxy bonded to it and when the mould was disassembled, the rubber had to be carved off the cured epoxy block; on my second effort I used a non-stick sheet, PTFE coated, marketed as oven shelf  liner; it worked fine but needed to be supported by a backing for which lead sheet, as used in roofing, proved OK. Where pp or PTFE is used, no release agent is necessary, - the cured epoxy separates from the mould very easily.

The epoxy I chose to use was "Water clear" Transparent Epoxy Potting Compound from MG Chemicals; I bought it through RadioSpares; for the size of mould I had made each board needed 1.725 litres of epoxy; conveniently for two boards this meant buying one kit of 2.7 litres (RS no: 181-0370) and two kits, each of 375 mls (RS no: 181-0369); each kit comprises resin and hardener in a ratio of 2:1. It is not wise to attempt a mix of 1.125 litres to fill the mould in one go because this will lead to unmanageable thermal runaway; it could get hot enough to damage components on the board; I mixed 300 mls at a time (200 resin + 100 hardener) at half hourly intervals to stagger the curing time and reduce the temperature rise; it therefore took about 2 hrs to fill the mould; even doing it this way, the temperature still reached about 80 deg C and the 300 ml mixes added last cured more quickly than the earlier mixes because the mould had by then begun to get hot.

You are helped to a good outcome in epoxy potting by being well prepared: graduated beakers for mixing resin and hardener, stirring sticks, cleanliness in the work area, a means of de-aerating the mix before pouring it into the mould, a kitchen timer or clock, warming the resin and hardener in the oven to 45 deg C before you start so their viscosity is as low as possible to facilitate mixing and de-aeration. Thorough mixing is particularly important because any resin not mixed with hardener will not cure; it will lead to a sticky patch in the finished block; I stirred for 5 mins timed on a timer; the type of epoxy I was using will remain workable for up to 1 hour at 20 deg C but for less time if it is warmer; each successive mix of 300 mls will bond to the previously poured batch as long as the cure has not reached the stage of being hard, as estimated by whether a finger nail can indent the surface.

Getting the bubbles out after mixing the resin and hardener is important; it can be done by waiting if the mix is not too stiff, or it can be done by placing the mix in a chamber and reducing the pressure; I chose the latter method and used a vacuum cleaner to suck out from a closed container; it seemed to work and I supposed it worked by making the air bubbles in the mix expand and so rise to the surface more quickly.

There are two stages to the curing process, soft cure and hard cure; you don't want to remove the block from its mould until hard cure has reliably been achieved throughout its mass; this can be ensured by 'post-curing' in an oven at 80 deg C for 4 hours, and this is what I did, waiting until the next day when it had cooled down to disassemble the mould.

The result was better than I had expected: the epoxy had filled all crevices and the mould left clean edges; the original V-Clamp board attached to the Powerspout bulkhead with self tapping screws driven in through the bulkhead from the pelton side but a modification in this epoxied board was to incorporate M5 machine screws into the epoxy, onto which Nyloc nuts fasten to draw the heatsinks tight against the bulkhead. I had expected to have to reduce the bulk of the finished block with a disc sander in order to get it to fit in the confined space it occupies above the Powerspout's bearing housing, but in the event it fitted perfectly by entering it from the bottom right side and then rotating it anticlockwise into its final position. 

Conclusion:
The board depicted was installed in the turbine on 7th October 2020 and has so far worked perfectly; it should, I hope, go on working for years to come.

Repairing and epoxying these two boards has not been without expense: the more severely damaged one was beyond my repair abilities and cost £320 to have done professionally. The other board only cost the price of replacing one MOSFET, - a few pence.

Clear epoxy of the type I used is not the cheapest available but I wanted to be able to inspect the encapsulated components should failure occur, - not that repair would be possible once potted.

For each board the epoxy cost worked out at £232.

I did have the option of ditching the present system in which a V-Clamp board is necessary, either by changing to using a Klampit device and continuing with the SMA inverter I use, or by discarding the SMA inverter and changing to one which can accept 600v dc rather than the 400v dc the SMA is designed for.

Either alternative would have meant disrupting a system which has proved bomb-proof for 7 years, and having several spare SMA inverters and V-Clamp boards, I decided to stick with that way of dealing with voltage capping.

Getting to the point of having these two boards repaired and ready for use has been a long journey; for several years I didn't think it was going to be possible; but having done them, and having learned a good deal in the process, I'm glad of having pursued it.

An alternative method to cooling to remove hot melt glue is to use heat; at 120 deg C the glue becomes liquid and drips off; the circuit board is hung in an oven; 

Stopping water entry

I've been experimenting recently with trying to stop wetness from the pelton side of the turbine creating dampness on the alternator side. There's a seal around the shaft which should prevent water in any quantity getting across but an investigation I've done using bags of silica gel indicates that in spite of the seal about 300mls per month still gets across.

These photos tell the story of the sequence of steps I've taken:

1. Limescale deposits on the shaft indicated that a considerable amount of water enters the top-hat labyrinth chamber

2. A V-ring seal (purchased here) was mounted on the shaft; the seal turns with the shaft and its lip seals against the plastic face of the top-hat, with the idea of preventing water tracking alongside the shaft

3. Inspection after 3 weeks running showed the seal had badly scored the plastic face of the top-hat, presumably from softening of the plastic by the heat of frictional contact, - I must have applied it too tight to the face.

4. A friend who is skilled on his metal lathe kindly turned a stainless steel cap to fit over the plastic top-hat so the seal rubbed on metal; the cap is held on only by being a tight fit.

5. Suspecting that water might also enter the top-hat via its drain hole, a deflector was devised to shield the hole from the upward direction of water leaving the pelton from the lower jet.

6. So the complete arrangement as it is at the moment looks like this:

Only time will tell if it makes any difference.  The early signs are that the silica gel bags do seem to be taking up less water but I'm yet to be convinced this is a genuine observation.

Whilst I was working through these stages of development, EcoInnovation have come up with a slightly different approach:

Theirs is a neater solution but care will be needed not to apply the V ring seal too tightly against the face of the top-hat.  The seal only needs to just touch. After observing the scoring illustrated above, a new top hat with the seal just touching ran for 3 weeks with not even a mark being caused.  A smear of grease is also a good idea.

Yet more on moisture

Dealing with the problem of moisture in the electrical side of my Powerspout is proving more difficult than I'd anticipated.  A year ago I started putting bags of silica gel in there in the hope I could create a warm, dry atmosphere. I also blocked off the ventilation louvres to stop moist outside air exchanging with the dry inside air.  But it hasn't worked out as well as I'd hoped.

Initially the strategy seemed promising; relative humidity in the compartment could be got down to 10% (ambient outside being in the 90's) and the lack of ventilation made it warm but not too hot. The warmth did have the disadvantage of reducing electrical output slightly but the reduction was only by 1 watt per degree Centigrade rise and I came to accept this as a price worth paying.

What has become evident in the longer term is that the silica gel doesn't keep the absorbency it had when first purchased; it cannot be re-charged to be as good as when new; I have been re-charging it by putting the bags of beads in a fan oven at 120℃ for 2 hours, but even prolonging the time to 3 or 4 hours doesn't seem to drive the last 7-8 mls of water out of each 100g bag; in consequence, when put back in the SmartDrive housing they only last for 3 weeks before needing to be re-charged again. 

More recently I've tried a new tactic: if a compartment is COMPLETELY sealed it ought to be possible to get the relative humidity down to a low level where it will stay; but the sealing does need to be thorough: ALL possibility of moist outside air exchanging with dry inside air needs to be removed;  to achieve this for the electrical side means blocking two openings in addition to the louvred ports which I have already blocked: the drain hole on the floor of the compartment and the notch in the bearing block. The notch is there to allow any water getting past the seal on the shaft to dribble down, and so not come close to the outer shaft bearing.

But though plausible in theory, this tactic doesn't seem to work in practice; admittedly I haven't blocked the notch in the bearing block because it's tricky to do, but plugging the drain hole seems not to have done anything to extend the period before the silica gel needs replenishing. This is disappointing. The drain hole gives direct communication with very moist air and spray from beneath the pelton; blocking it ought to have produced a benefit.  That it hasn't leaves me scratching my head for the next bright idea...

Update on two previous matters

Two subjects to report back on in this diary entry: variable speed running and moisture.

1. Variable speed running

Since the last entry in which I floated the concept of variable head / variable speed operation, there hasn't been much opportunity to test out the idea. A spell of rainy weather has meant that instead of decreasing flow, I have needed to put in bigger nozzles to harvest as much as possible from the unseasonably generous flows available. I'm still aiming for that elusive total for the year of 4 MWh.

What has interested me is that I have come to read that variable head / variable speed operation isn't new and is by no means confined to the small scale of a Powerspout: pumped storage installations of several megawatts in size in Switzerland, France, India and elsewhere are being built to work this way, or in some cases converted to this way of operating when older, conventional synchronous pump-turbine units are refurbished. Apparently, the new way of operating confers efficiency gains on both the pumped and generating cycles. More can be read about it in Hydroworld magazine here and here.

2. Moisture

The silica gel bags I placed in the SmartDrive enclosure 6 months ago have lately been seeming to come to the limit of their ability to take up moisture:  relative humidity in the enclosure has risen to 51%.  This is still below ambient RH which is usually about 70-90% but the figure is significantly up from the usual 10-20% which was evident when the bags were fresh.

Weighing the bags before and after recharging them in the oven (120℃ for 2 hours) showed they had taken up 245 mls of water during the six months.  Certainly the bulkhead has always been completely dry on the few occasions when I have looked inside and I conclude that the use of silica gel has been a good thing. On the basis that moisture and electricity are never good bed-fellows, I plan to continue using silica gel and re-charge the bags every 6 months.

At the time of removing the silica gel bags, I took the opportunity of experimenting with packing out the rotor to see what effect it had on operating voltage and shaft rpm, - hence the tachometer and rubber O-ring visible in the picture above.  The experiment needs to be repeated at different flow/power conditions before a conclusion can be drawn and if there's anything useful to feed back, it'll be the subject of another update.

More on moisture

Eleven weeks ago, I wrote about Managing moisture. It described my attempt to create a warm and dry atmosphere in the electrical side of the turbine by packing in a kilo of silica gel beads and blocking the ventilation ports. The aim was to improve insulation resistance and reduce corrosion. What's happened since?

At the time I sealed the compartment, generation was only 300W, so temperature inside did not immediately rise very much.  But as winter generation picked up (it has been > 780W since 1st December), the temperature rose to 18-20℃ above outside temperature, reaching nearly 40℃ on warm days. There was nothing worrying about this. But at about this time I began to recognise that the warmer the SmartDrive operated, the lower was its output, about 6 watts less per 5℃ temperature rise (see last diary entry and also here).  After thinking about it and deciding that maximal efficiency was my main aim, I re-established ventilation by putting back the louvred ports in two of their three locations. Ever since, the temperature inside has been lower at between 19 and 24℃.

Restoring ventilation predictably made relative humidity inside rise. When sealed, it was as low as 10%, but unsealed it rose to 20 - 38% because more air exchange occurred with humid, outside air.  The inside humidity now fluctuates in the 20-38% range, tracking the humidity of outside air, and the temperature in the compartment. I have been assuming the silica beads are still taking up moisture as the inside RH is well below the outside value. Later on I'll take the bags out and weigh them to see just how much water they've taken up.

Today, I've been measuring the temperature of the water powering the turbine.  It was 8℃.

It's the temperature of the bulkhead that I really want to know and I'm assuming it will be the same as the water. The temperature of the bulkhead determines when moisture will condense on its surface and this measurement was the only factor I'd never got round to measuring in my quest to predict the conditions which will create condensation.

In the diary entry "Managing moisture" I gave a link to a dew point calculator.  This is a handy tool for working out at what point condensation will occur, handy because the computation is complicated: there are three variables to consider: the relative humidity of ambient air, the temperature of the air, and the temperature of the surface on which condensation is to occur. 

Convenient as the calculator is, a graph gives an alternative way, a more visually predictive way, of appreciating how relative humidity and ambient temperature can be manipulated to avoid condensation. These two variables, temperature and humidity, are the only ones which can be manipulated, since the temperature of the bulkhead obviously cannot be changed.

To understand the graph requires an understanding of humidity: humidity is the amount of water carried as invisible water vapour in a mass of air; the amount of water vapour carried will depend on how moist the air happens to be: deserts are dry, rain forests are wet; but the maximum amount of water air can carry in either region depends on the temperature of the air: the warmer the air, the more water; the cooler the air, the less water.  When a body of air at a certain temperature and carrying a certain amount of water (as invisible vapour) comes into contact with a surface which is colder, the layer of air immediately above the surface is cooled; in being cooled it finds it can no longer carry as much water vapour as the warmer air further away from the surface, and if it cools below a certain point, water comes out of being in a vapour phase and condenses on the cold surface as liquid, first as misting, later coalescing to droplets.

What the graph shows is a family of curves. Each curve describes the ambient humidity and ambient temperature at which condensation will occur on a surface having a given temperature. Several curves are needed because each shows the humidity / temperature relationship for a different surface temperature. In my graph, curves are shown for surface temperatures of 0, 4, 8 and 12℃. 

The measurements I took today recorded the temperature of the bulkhead as 8℃, the temperature in the compartment as 23℃, and the humidity in the compartment as 28%. The red arrow in the graph has its point at the place for the two temperature measurements: where today's compartment temperature 23℃ intersects with the 8℃ curve.

At this point it can be read on the Ambient Relative Humidity (RH) axis that dew (condensation) will form if ambient relative humidity is 38%.  It will of course form too if the RH is at any figure higher than that. Since the ambient RH inside the compartment today was only 28%, it follows that condensation would not have been possible under today's conditions.

We can use the graph in an alternative way to see at what ambient temperature condensation will occur when the RH is 28% and the bulkhead temperature is 8℃: it will be 29℃ (as best as can be discerned from this crude graph).  Now 29℃ is a temperature which might easily be reached if ventilation was blocked off so we might conclude that blocking off ventilation would be a bad idea from the condensation point of view (it will certainly be bad from a power generation viewpoint); but if ventilation was blocked off, ambient RH can be expected to be lower, maybe only 20%, perhaps less, perhaps as low as 10%, and then it can be seen by extrapolating the green 8℃ line to the right, that ambient temperature could be allowed to rise to perhaps 45℃ before condensation would occur, but only so long as the bulkhead temperature is at 8℃.  And if the bulkhead was colder, say 4℃, we would have to jump to the 4℃ curve which gives a whole new set of temperature / humidity relationships.

You can begin to appreciate what a complex business condensation is, and yet it is a very precise and predictable subject ... if only one knows the values of the variables. 

What a fun occupation it is measuring things and using the data to better understand the world about us!  If it prevents a V-Clamp board failing from poor insulation resistance caused by dampness, it will be useful as well as fun.  Here's to hoping!

Managing moisture

It has been a busy few days with the Powerspout.  With autumn well on the way, the weather has been warm and wet.  The flow available to the turbine has been gradually increasing so I have been able to change from the reduced core stator to the 42 pole, full core one. This will now see the turbine set up for its winter period of peak output.

Whilst doing the change of stator, I also changed the bearing block.  I had to deliver on a promise to Michael Lawley that once winter flows arrived, I would run another trial of ceramic bearings.  The first trial was done in June and the bearings lasted just 4 weeks.

This second trial started yesterday, - and finished yesterday ! Just 8 hours !  Inspection of the failed bearings showed much the same as the first trial: pitted balls causing locking up of the race, but this time there was also abrasion of the PTFE spacer ring, leaving particles of PTFE everywhere.  I think this has to be the end of the road for the ceramic bearing dream !

In the course of this repeated delving into the turbine, I have progressed the idea of trying to run the Smart Drive compartment as dry as possible.  This I've done by stopping the ingress of moist outside air by blocking off the ventilation ports and de-humidifying the captive air inside with bags of silica gel.

The idea is to create a warm, dry environment for the electrical side which will inhibit corrosion and promote insulation.

It seems easily possible to obtain very low levels of relative humidity.  Ten bags of silica gel, each of 100g, brings the humidity down to just 10% within an hour of closing the housing, and this is with an outside ambient humidity of 98%.

The rise in temperature is to about 6 ℃ above ambient when the power output to grid is 300 W. At the reading showing this morning, 22.4 ℃, I am happy with this, although as power output increases in coming weeks, and with it greater heat output from the alternator, the rise in temperature will have to be watched.

To touch briefly on the theory of humidity and its relationship to dew point, if the relative humidity inside the housing can be kept at this 10% level and the temperature in the housing does not rise above 36 ℃, it will completely prevent any condensation (which is dew) forming on the bulkhead, even when the temperature of the bulkhead gets down to 0 ℃.  Since it is condensation forming on the bulkhead and then dribbling down over electrical components which probably causes most of the issues with insulation breakdown, this would be a significant advance.

The unknown will be how often the silica gel bags will need 're-charging', and the answer to that will be how often the compartment is opened allowing an air change with moist outside air.  I'll be keeping my inspections there to a minimum.

For those interested, 25 silica gel bags of 100 g cost £18.59 and a useful dew point calculator can be found here.