Diffusion pump on SX100

[sem-geologist]

But how do I know what DC voltages SX vacuum logical board expects for corresponding vacuum? I did reverse engineering scanning of it. (No one would expect anything less from me I guess  ). I fed coaxial cable of secondary gauage with different voltages while watching vacuum values shown in GUI. The setup was simple 20k potentiometer, two 9V batteries (two connected in series, so I could cover with scanning voltages up to 10V), SMB coaxial connector, small breadboard to make some T junctions to attach the voltmeter for V measurement. By changing potentiometer I scanned from 0 to 10V and wrote down corresponding vacuum values. It seems that up to 9V the relation between log(P) and U is pretty linear. Actually in normal circumstances the gauge is engaged with better vacuum so below 9V. Thus that low vacuum range (9V-10V) can be ignored, and then looking for offset and scaling values I was looking that it would cover 0.5-9V range of expected values. The attached spreadsheet contains these recalculations with estimated gain and offset values for DC. tables has two offset values, mathematical and InAmp offset (InAmp - instrumentation amplifier). Mathematically we see no difference between these two equations:

U_sx = U_gauge*G + O

or

U_sx = (U_gauge+O/G)*G

where U_sx is expected voltage, U_gauge is voltage output of gauge, G is gain and O is offset.

However, first equation is not possible to implement in OPAMPS which would saturate at gain, where second is easier to implement. So the offset which will be need to be set by potentiometer is InAmp ref (pin) offset or O/G. Gain is then applied at next stage (the G). Attached spreadsheet shows and presents gains and offests for different kind of gauges.

[sem-geologist]

And there comes a teaser picts (rendering) of ready PCB design:



Hopefully in one week I will have a real prototype. Also I am still waiting for Edwards AIM200 gauge... already for 2 months... Good It is universal and Could buy Curt Lesker active gauge or will test it with WRG gauge.

[Probeman]

We have the turbo on our SX100, but to quote the person I learnt from: "Never trust the gauge!".

Yes, I started to understand such narrative. But is the situation still the same as decades ago and can't we have anything better?

At first, Alcatel Penning gauge (gauge installed on early models of SX100) is indeed hard to trust - there is whole pipeline of things where stuff can go wrong... I dig up through piles of old emails (my late predecessor had printed all communications and emails, which I could look through and learn the whole history of repairs of our SX100), and learned that indeed the problems with vacuum measurements were re-emerging again and again and again... Not only gauge but also cable was changed many times. Investigation of complete construction had revealed major weaknesses of this type of gauge.

1) High Voltage is sent through coaxial cable. That is about 4kV which is constant. If dielectric material weakens somewhere in the cable it start to leak  and generates some current at HV source side. The whole vacuum measuring is based on precise current measurement through shunting resistor at returning path of HV generator, so such additional cable leakage current will sum up with current generated at gauge and worsen the displayed value of vacuum. But, in case of a huge leak in cable there will be no possibility to start penning gauge at all, as there will be significant voltage drop at gauge (leakage in cable will form something like voltage divider). Even if it will be clean gauge - bad cable will hinder its functioning. As for alcatel gauge cables - these are no more produced - all available at second hand shops are of questionable quality! Making new cable (DIY) from scratch is also hindered by requirement of HV reliable cable terminations (tools, expertise), and proper cable availability on the market (There are cables more think, and expensive).

2) Penning gauge, while being in simple construction, can give easily wrong idea about vacuum when not striking. Its lack of current (no strike event) could be compensated with small leak and final reading would look as good properly working gauge value. Experienced user knows that reading is wrong - but only if user tracks vacuum readings closely after the moment the vacuum valve is opened, and tracks the vacuum value response to that event. If vacuum suddenly just goes to 5E-5Pa from very rough vacuum - that is clearly unrealistic. However, if user leaves pumping unattended, and gets back to instrument after 10 min (and cable is leaky a bit), he can see the i.e. 4E-4Pa, even if gauge had not strike. Also Penning gauge needs constant high voltage, that have tendency to deteriorate the weakened isolation material in the cable and progress the problem(s).

Thus I came up with initiative, to make a replacement card which would interface modern gauges to SX100 (giving-up alcatel penning). That would be not jumping one tier in technology (like SXFive's used Agilent IMG) but two tiers. The next technological tier, IMG (inversed magnetron gauge), is more advantageous compared to Penning, but still requires the high voltage cabling and its correct termination and custom made current sensing circuits are difficult to do properly. Interfacing stock controlers of gauges is also hard as it would require changing firmware and finding some serial port on Cameca SX100 vacuum logical board. (That is rather completely out of equation).
Inside SX100 vacuum supply box, there is small board interfacing Alcatel gauge. It produces HV supply, measures the return current and sends the calculated vacuum as log10(P) as DC signal from 0 to 10V (10V roughting, 0V -ultimate vac) through coaxial cable to Vacuum logical board.

Thus naturally my attention got stolen by active gauges, which is higher tier than normal IMG. Active gauges work with low voltage cables, and produce all required HV for integrated IMG inside the gauges integrated electronics and send out only low voltage 0-10V signals (see the pattern?). Such gauges are also flexible as they can be powered from 14.5 to 40V, and such supply is already present in the Vacuum supply box. Also it looks that few different vendors have very similar connection based on 8C8P (or FCC68, or well known RJ45 ethernet cable/sockets), with same basic pin configuration: at least I found out few such gauges from Edwards, Curt J. Lesker and Inficon with such connections.

But what are advantages?
1) No HV cable. Basically simple Ethernet cable will do the job, as only 2W +15V is required to power such gauge, and signal is 0 to +10V. Replacement cable can be easily made or bought. (but it should not age at all, differently to high voltage cables).
2) These active gauges have builtin microcontrollers and LED(s) which show(s) status of gauge (i.e. if strike was successful) - the meaning of vacuum reading is clear. Situation then strike failed, and we have low current and very low vacuum reading is easily to catch. Also Active Gauges reduce high voltage to minimum after strike - this means less contamination of gauge - longer exploitation time between need of cleaning.
3) Edwards gauges (i.e. AIM200) has special multi-strike geometries which guaranties striking even in dirty environments.

How it will work?
My idea is to translate the voltage from such modern gauge to corresponding expected DC voltage by logical vacuum board of SX100. Basically my design is few OPAMPS for DC offset, scale and clip (values out of expected range). Two potentiometers will allow to easily recalibrate board to different kind of gauges from different vendor. It would be possible even to use WRG from Edwards (albeit pirani range would not be seen). The idea is to have seamless replacement (no firmware or software changes).

So how many people would be interested in such design?

I have schematics alreadynearly finished (see the attachement), and now I am at stage of PCB design.
I won't produce and sell the boards, the EU laws and handicaps are too enormous for me (Basically RoHS3 forbids me to make and sell reliable electronics). But I could share the design, gerber files, bill of materials and notes for proper assemblage at site overcoming RoHS hindrances. PCB when having these is possible to order in most of the world. Buying parts and assembling them on PCB should be not too difficult, board is designed to be hand solder-able (no surface mounted parts, everything "though hole" parts). Albeit design could be updated with surface mounted parts if someone would decide to produce and sell larger batch of such boards.

Our instrument engineer made this comment regarding the above comment from SG:

The penning gauge on the SX100 is pretty robust. It's easy to clean.

We don't have much trouble except that it does cause GV3 errors (vacuum system shuts down) whenever it decides to arc. Ash buildup makes it  more prone to arcing, but even when no ash is present, small arcs can  trigger a GV3 error if the specimen chamber pressure is higher than usual (pinhole in spectro window), which offsets the DC signal the arc voltage spike is riding on, tripping the window comparator. Such narrow spikes are supposed to be filtered out by a capacitor, but  even adding more parallel capacitance does not seem to make the filter effective in stopping such spikes. That fact makes me think the arc duration is rather long, negating the effectiveness of the filter.

Putting a datalogger on this signal would answer this question. What I've done to stop these nuisance arcs from causing GV3 errors for the time being is to defeat the window comparator so that the CPU logic does not see these arc spikes. This is not ideal, but we rest easy knowing that the  analog pressure reading remains active, and the system can do a shut down if that reading goes high.

     The author of the post does have a point saying we need to have a replacement for the stock gauge since it is now obsolete and out of production.

Steve

I also note a post from Jeff Streger from Rave Scientific on LinkedIn noting this new product from Edwards. The APG200 Piriani gauge:

https://www.edwardsvacuum.com/content/dam/brands/edwards-vacuum/edwards-website-assets/scientific-vacuum/documents/datasheets/3601-0681-01-Active-Pirani-Gauge-APG200-Product-Datasheet.pdf

[sem-geologist]

Pirani is of not much use in our case, as 1) pirani gauges are useful at atmospheric - low vacuum ranges 2) there is low vaccum gauges already, tightly integrated with Cameca SX100 vacuum logic. It is not possible to use low vacuum range of pirani without modifying logical interface of SX100 vacuum; 3) separate cold cathode and pirani (or other low vac type) are more robust. I mentioned WRG (pirani coupled with cold cathode inside single package https://www.idealvac.com/Edwards-WRG-Wide-Range-Gauge/pl/7-33-291), as sometimes those are off-shelf available at vacuum shops (applied more widely), where cold cathode type gauges needs to be waited for production as it is produced in small batches, unfortunately. What I would like to get ultimately is Edwards AIM200:
https://www.edwardsvacuum.com/content/dam/brands/edwards-vacuum/edwards-website-assets/scientific-vacuum/documents/datasheets/3601-0753-01-AIM200-Active-Inverted-Magnetron-Gauge-Product-Datasheet-LR.pdf
If they fail to deliver that till I get the functioning interface board, I am going then with this:
https://www.lesker.com/newweb/gauges/kjlc-cold-cathode.cfm
Eventually shop could lend me temporary the WRG (but I need documentation to be sure it is safe to connect), so I would not be so upset and would wait for AIM200 a bit more. WRG is only a temporary option, and not due to pirani, but due to widely available off-shelf cold cathode gauge.

As for arcing, new generation of gauges prevents this by reducing the high voltage after started - that significantly reduces risk of arching. Albeit we are unlucky with that (had not observed this), In our case we rather have lowered voltage at gauge as our cable is partly leaking.

And mine project aims to be as much as robust and to give possibility to make reliable replacement.
1. All parts are THT, for these reasons:
  a) the solder joint inspection can be done with bare eye, and is quite straight-forward
  b) easiest way to mitigate RoHS enforced tin whisker and solder embrittlement problems (design uses sockets for DIP8 and DIP14 components, not because replacability, but because then IC components can be NiAuPd plated. In case directly soldering such parts it risks in forming cracking when using PbSn solder (have risk of embrittlement and cracking after some time like few years); Where other passive components (i.e. with pure Sn coating) can be dipped into PbSn solder for whisker protection, doing that with Integrated circuit components (in case of pure Sn plated) is not wise as can be damaged by thermal shock. The Not-green components (with PbSn coating) aimed at space and military are restrictively expensive and are sold in packs, thus (over-)plating passive components and using "green" NiAuPd plated ICs is the optimal solution.
  c) Can be easily soldered by less experienced.
2. Aims to work with wide range of gauges (currently to my knowledge at least three vendors, and different models). Post-pandemic market fluctuations had shown how not-wise it is to rely on a single supplier of components.
3. Is not tied to single supplier of components, has alternative integrated circuit models and suppliers for circuit to work near identically.
4. No software or firmware changes - just drop-in replacement - The software or firmware will not notice it is different gauge than Alcatel.
5. You probably had noticed Open Hardware mark on schematics and board? yes, after testing the first prototype I will release it as Open Hardware (probably on github, or other repository) - that means anyone will be able to tailor and keep the design up-to-date with best market-available gauges (OH - is like open software, just for hardware). Anyone can modify and tailor for his needs, order the parts and assemble at-site.

[sem-geologist]

Gauge Upgrade Complete! The gauge replacement to AIM200 Edwards gauge shows that chamber vacuum is OK (The vac was off for two weeks, and after start up yesterday currently reached 2E-4Pa, but seems still dropping). Good that I patiently waited and had not did any stupid things with diffusion pump.

There, let me give you few teasers from this new system.
One from 5 PCB's manufactured in JLCPCB (china). (I have 4 spare laying around, as minimal board number to fill the PCB standard blank PCB board is 5). The daughter board is connected through "mouse bites" with main board, so that manufacturer of PCB would treat is as single, and it is easy deatach:


Old and new gauge side by side:


... and their connectors, please note that Active gauge works with low voltage supply and signals thus RJ45 does the job perfectly:


The original and upgrade boards side by side. Color red so it would be very clear it is not original part.


Upgrade board sitting in the Vacuum supply box. It is drop-in upgrade:

[sem-geologist]

Getting back on the topic of this thread (The Alcatel Crystal Diffusion pump in some of SX100 models).

I have all answers which I asked initially, and more, as I was forced to go through all of this.
Below is applicable only to earlier models of SX100 equipped with Alcatel Crystal 102 diffusion pump (pump painted in orange), for more detailed maintenance sequence instructions AND properties of Balzers PDI100‐W diffusion pump installed in later models of SX100, the info is there:
https://smf.probesoftware.com/index.php?topic=1698.msg13025#msg13025

TLDR;
For Alcatel Crystal 102 diffusion pump main maintenance parameters:

• Fill/drain O-rings replacement: 2x FFKM o-rings of size AS-568 (IDxCS: 9.19x2.62mm)
  (In my case, I used Kalrez 6375 FFKM, as required size Oring with that type of rubber chemistry could been sent from the nearby shop the next day); I do not recommend using Viton (FKM; the pump originally used 9.2x2.8 mm Viton green O-rings; CS of 2.62mm Kalrez type is standard size (AS-568) for inner diameter of 9.19 mm, the smaller cross section compensates with higher thermal expansion; in case Kalrez would have 2.8mm - it could deform with thermal expansion and could lead to failure).
• Amount of diffusion pump oil: 100mL
• oil wet section off dipstick of fill port plug, if DP has nominal 100mL oil filled: 10mm
• Type of oil: definitely works with 704 or 705 type silicon oils, probably would work with Santovac 5 or other similar oil.

A long/short story about vacuum trouble with 25 year old Cameca SX100. Few months ago there was planned power line of building maintenance/checkup (planned power interruptions). Our SX100 is not covered with UPS (as I already mentioned before). As such maintenance work would interrupt the power in whole building few times for few seconds, I decided to be at the safe side and to minimize the risk I completely shutdown SX100 the evening before the planned maintenance work. So the next day, afternoon, after power checkups finished, I tried to turn on the SX100 back, and... Diffusion pump could not reach the roughing vacuum required to start it up. After excluding all other possibilities, the diffusion pump was the only place where the leak could had occur.

I found that fill and drain port O-rings had hardened in 25 years (Original Viton O-rings) and so after complete cool-down with loss of elastomer compression during many years spent at high temperature, it introduced a leak. While pump was hot, the hardened O-rings were still compressing toward surfaces it seal, due to additional compression force of thermal expansion of O-rings. But as Diffusion pump cooled down, the contracted hardened ring would leak, and as diffusion pump would no more be possible to turn on (too bad roughing vacuum), it would never heat up back and would not seal the leak. I could just measure dimensions of old O-rings, order the new Viton O-rings replacement, replace an move on...
But instead I wanted to achieve a more long term fix, and so I took some time to read and learned some fascinating things about O-rings. Viton(R) or in general FKM type rubber (flourinated elastomer), while is taken for granted as one of the best elastomer rubbers, it suffers in heated environments and even if it is used below maximum rated temperature, due to very long exploitation time (years) can harden-in and loose totally the elastic properties loosing its sealing capabilities.

There is however other elastomers designed for hot environments - perfluorinated elastomers (FFKM), i.e. DuPont's Kalrez(R). The only caveat is that Kalrez has twice a thermal expansion to Viton, thus in hot environment that needs to be taken into account, when choosing Kalrez replacement inplace of Viton.

As I had to open drain port, I also had to drain oil. Less than 80mL had flown out. I could not find out what kind of oil was there - no notes what was it filled with - for sure it was not filled with Santovac 5, as it had quite a water like viscosity. It could be petrol oil or silicon oil or some other nasty thing. Raman spectrum analysis showed very mixed data, there were some benzene and some other "smoking-gun" signatures pointing to cracking products of petrol oil (or cracking of silicon oil somehow). Definite answer came from liquid density measurement showing it most probably was 704 type silicon oil. As only 80mL from required 100mL had dripped out, I decided to disassemble pump for cleaning. It was absolutely nastily dirty that it took me few days to completely clean it up (using acetone and iso-propanol alcohol only).

As for filling back with oil I had few possibilities: go for Santovac 5, or go with silicon 704 substitutes, or 705 substitutes. Santovac 5 availability had lots to wish for, lead time to be ordered was few weeks.. however I found out local 705 substitute, German manufacturer which could send me oil just right away. Another consideration was if thermo-regulation of the pump should not be tweaked in case switching from previous 704 type into santovac 5, which require higher temperature. 705 type is closer to 704 and could work without any thermal pump tweaks (and consider that having no manual on the pump), so I had chosen 705 as more safe option. The mentioned German diffusion pump oil is Indomet PENTA.

I had more troubles waiting me - the chiller failure... which postponed diffusion pump integrity testing, and so after about 1 month after initial pump failure I could finally turn it on.

There are few tricks and steps I had taken:
1.Cold dry test. After cleaning and assembling the pump (including the the new fill/drain O-rings replaced), but without refilling the pump with oil, I started up SX100 vacuum system with pump power cable still detached, so that SX100 Vacuum firmware could not energize the heater if roughing vacuum is good enough to start the diffusion pump. After pumping for a while and seeing that threshold can be reached without a problem in very short time, I also did a "vacuum halt test for a night", to check how tight the vacuum can be kept inside the diffusion pump with roughing pump cut off from the diffusion pump. It was a sharp contrast (practically vacuum stayed the same) compared to same test with previously leaky old O-rings.
2. I Filled the oil through backing vacuum pipe side (there it is attached to Vacuum buffer tank – the 90 degree elbow is easy  to detach, as it is kept in place by two KF type clamps) instead of using hte dedicated fill port. Generaly, Diffusion pumps can be filled much easier through vacuum backing connection side, which on SX100 is in much more accessible position, than the fill port of the pump. Additionally I could put the glassware upside down on the pipe opening, after dripping the majority of oil and let the very last drops of oil to drip into the port. I am convinced it is easier to get it precisely 100mL that way. Otherwise, if using fill port, it would definitely require a syringe, there would be oil leftovers in beaker and syringe, and beaker should have some little excess of oil then (i.e. 102mL).
3. After waiting few (10-15) minutes for oil to level out in the pump I took the measurement of fill drain dipstick for future reference. (10mm oil-wet end of the dipstick)
4. Then I had assembled back the pump, but still had not plugged the power to the pump heater. I restarted the SX100 vacuum and let the cold degassing of fresh filled oil - in my reasoning, the oil in the shipped bottle clearly contained bubbles with entrapped air - I think it is safer to let the majority of air to degas in cold, instead in the hot diffusion oil (less possibilities to crack oil).
5. After some degassing for few hours, I stopped the vacuum, plugged back the power to the pump, and restarted the vacuum system letting it go full startup sequence. (of course in my case as chamber was at atmosphere pressure for whole 1 month, I had preventive kept Ion pump off, so it would not be prematurely too early switched on and consumed.)
6. After weekend of pumping, I had switched vacuum off, turned the breaker for ion pump power back on, restarted with full sequence and lived ever happily.

[sem-geologist]

... As only 80mL from required 100mL had dripped out, I decided to disassemble pump for cleaning. It was absolutely nastily dirty...

That above (question how so big 20% loss of diffusion pump oil could happen), and emergency fire situation in the building a month ago made me to look back into importance of backup battery which I mentioned previously in the beginning of this thread: https://smf.probesoftware.com/index.php?msg=11923

As this diffusion pump repair had teach me a lot about the diffusion pumps, It changed my perception and recognition of the backup battery in completely different light. It is a crucial safety feature for SX100 equipped with diffusion pump (for turbo pump equipped SX100 it is not crucial or even unnecessary) - it covers most of failure cases which UPS and backup generator does not cover, and recent fire hazard event and complete cutoff of power to the building for 3 days had teached me the backup battery for Diffusion Pump inlet valve is not a "maybe", but a "must" – unless cleaning the column from deposits of diffusion oil vapor is seen as a very nice pleasant afternoon activity.

I was partially wrong in my initial reasoning - safety valve is not to protect diffusion pump oil from residual gases originating in chamber – but its protection is in opposite direction - it is to protect clean sample chamber from back-streaming of diffusion pump oil in case of abrupt power cutoff to EPMA. The reliance explicitly only on UPS/generator can be dangerous, as UPS and power generators covers only and only event of general external power blackout. In some of countries in particularly with high capacity UPS installations, UPS is prone to remote switch off at critical circumstances. In example our Polish laws and safety standards (and I believe many other European standards and laws) require to have remote switch off of UPS, which is situated near entrance of the building, and in case of any fire emergency situation those UPS power brake switches are triggered without any questions asked or any hesitations at the moment firefighters enter into the building during any fire emergency event. UPS and generators can cover only black out from the city - but that is not a single weak point for chain of power-failures leading to most important - power loss of vacuum system which is equipped with diffusion pump.

The side-lined power loss also strongly depends from how lab power topology is designed, Comonly UPS power output covers also peripherals of EPMA (chiller, air compressor) and so is roated through some distribution electric wall box. Failing peripherals , i.e. a water chiller then has very high chance to trigger RCD (in between the UPS and EPMA) cutting also power for EPMA - I had seen just few months ago this kind of event, where cooling fan motor in the a chiller, after 3 years of constant work, just developed reduced resistance between Phase Neutral and Ground and triped the RCD which triped all outputs from distribution power box (including EPMA). Looking further just into EPMA desing itself, SX100 contains quite a complex electric power system with different safety devices including RCD covering whole vacuum system, any failure in any vacuum component, which would trigger RCD, will send whole vacuum system into state of no power.

In any case of power failure of EPMA vacuum system (including fire hazard with remote UPS power cutoff, chiller failure, roughing vacuum pump engine coil isolation weakening, or weakening of isolation in coils of electrovalves, or even some random dust built-up and current leak through it on any of Vacuum PCB's with 230V on it triggering RCD) will lead to situation where previously fully operational hot diffusion pump will go through time window where oil curtains gradually stop being formed, and still hot oil vapor will escape into sample chamber through the stayed opened diffusion pump inlet valve. In case of power loss such oil vapor backstreaming can be amplified by pressure built-up at diffusion pump backing vacuum buffer side, as primary pump and outlet valves are off, for one minute still working diffusion pump (still hot oil) will keep building pressure. As soon the diffusion pump ultrasonic oil curtains start to fail (due to decreasing energy of oil and decreasing oil vapor speed from nozzle) the higher pressure of gas at backing side will very efficiently push oil vapor from diffusion pump upward into the sample chamber. It is possible that time window of vapor transport to chamber can be longer as with power loss chiller water circulation is also off - thus chances of partial trapping oil vapor on its way upward by cold diffusion pump walls is even more reduced, as diffusion pump walls also stop being cooled. The dangerous time window is probably 1-3 minutes long, just after 1-2 minutes counting from power loss (the oil curtains will stay in motion for a minute after power cutoff to diffusion pump; this is in contrast to my initial assumption that danger occurs only about 15 minutes after the power loss). The danger of oil vapor back-streaming into chamber is real and can happen in just few minutes if inlet valve to diffusion pump is not shut.

Basically any failure occurring after UPS can cut off the SX100 vacuum system power (UPS, and generators are unable to cover those cases) - and in any of such cases the inlet valve stays open, as H-bridge used to move motor forward or backward to open/close the valve is without the power. In our case as that battery was not replaced for nearly 2 decades, every power loss of EPMA was cumulatively getting rid of diffusion oil from pump into the chamber - that explains: how 20% of oil was lost from the diffusion pump, and why a few years ago, the maintenance full disassemble of column revealed such enormous filthy dirt on all the apertures. That contamination practically killed also BSE detector, where its replacement had cost us tens of thousands of euros/dollars.

The battery replacement.

This is where backup battery comes into for prevention. It is connected with diffusion pump inlet valve motor only in one way - only to energize the movement of motor in one direction - that is to close the valve if open. It is activated utilizing the NC contacts of few relays, which closes the backup battery circuit in case there is no external power (If there is power then relays disconnect powering of motor from battery and switch it to use power from H-bridge instead). Originally battery was 12V NiCd rechargeable battery pack (inside I found 10x 1.2V NiCd cells spot-welded into a pack) of 100mAh. I had seen a few issues against replacing it with similar chemistry pack: 1. recharging the pack with no balancing would lead to premature aging of battery (it would rather not survive for 10 years, but more close to 5-6). 2. I could not find any packs of 100mAh; Commercially available 12V NiCd packs starts at much higher capacities (i.e. 2000mAh). The trickle charging circuit present on the vacuum power distribution board on SX100 would be a bit too weak in case of huge package (i.e. 2000mAh).

Thus I looked to alternatives and decided to replace it with smallest industrial Sealed Lead Battery (800mAh) I could get, and add some voltage limiter to limit charging voltage down to 13.6V - the most appropriate float charge voltage for these kind of sealed lead batteries. The circuit is just adapter to interface JST battery connection, tab type (FASTON) connectors for connecting the wires for charging from SX100 (the wires with FASTON connectors previously used to connect to NiCd battery), 2 Zener diodes of 6.8V, which connected in series would limit voltage from 14.4V (actually from 15V) down to 13.6V, and socket for a 1A slow blow fuse. Fuse is there to protect from eventual shorting of winding in motor (or motor stall for some unexpected reason) or shorting anywhere on the circuit from battery to the motor - 800mAh 12V VRLA battery if shorted would be enough to pose some fire hazard (13V * 3A = 39W; 12V * 3A = 36W), taking into account the quite thin wires used for cabling from the distribution board to the valve motor (originally use 100mAh NiCd had no such power, maybe up to 1A in short pulse, for sure not more than 0.5A continously, where VRLA can give about 3A in continous mode).

If anyone would be interested I could send the kicad project and/or gerber files, although it is such a trivial circuit (only 4 nets) that I made it on a bench with help of small modelling dremel scraping cooper from single layer PCB laminate and drilling the holes for THT components.

I had tested sudden power cut off - everything works correctly. Such battery should cover protection for about 10 years.

[sem-geologist]

I just add the picture with more down-to-earth implementation.