For this article to make sense, it is necessary to learn about my CA-2-B15 here, and also the CA-1-B15 here.
While the CA-2 evaporator blockage problem had been completely solved, it became aparrent after running for many months that there were problems with the float valve, either sticking shut, or not seating properly. Given the success of the capillary tube conversion performed on the CA-1, it was clearly obvious the same needed to be done for the CA-2.
I also took the opportunity to do something I'd been giving a lot of thought to; working on the cooling unit while it was still fully charged.

Float valve problems.
Ever since I had restored the CA, there were intermittent problems with the float valve. I had hoped that it was the same substance that had caused the evaporator blockage causing the problem, and that it would dissolve in the same way. The fault would make itself evident in a number of ways. The most obvious was a sticking valve where the frost line would suddenly start dropping, eventually causing all the frost to melt. Without my compressor timer, the motor would then run continuously. Another sign of a bad valve was an excess of refrigerant flow. One could clearly hear the refrigerant rushing through the system, even to the point of the float rattling. This meant the evaporator temperature didn't get as low as it should. Usually, the float would stick around 5am, when the room was at its coldest, and the pressure of the methyl formate at its lowest. I'd arrive in the kitchen in the morning to find a defrosting evaporator.
Persistent running did in fact seem to clear out all the obstruction, and I had it running very well for a month. Cycling times were excellent and it almost seemed ready to start using to store food....but alas, one morning the valve stuck again. This went on a few more times and it was becoming evident that the float could block at any time, making the reliability unacceptable.

Replacing the evaporator bolts.
It had transpired that the nuts, bolts, and washers I'd been sold for the evaporator were actually zinc plated, and not stainless steel. After almost a year, signs of corrosion were evident. With the cooling unit out of the cabinet ready for the capillary conversion, the opportunity was taken to replace the hardware with stainless steel. I was hoping to be able to do this without having to remove the evaporator and top plate again. Indeed, not only was it possible, but was done in only a few hours. Carefully separating the top plate and getting an open ended spanner into the gap allowed the nuts to be loosened. They, and the washers were carefully hooked out. Installation of the new ones was just a reverse of this procedure.

The screwdriver handle gave enough clearance to access the nuts.

After a year, this was the condition of the zinc plated hardware. Stainless steel eliminates this problem. The original screws were aluminium.

Making the capillary assembly.
Having already been through the procedure with the CA-1 six months before, it was plain sailing this time round. As before, the filter drier was prepared by drilling a hole and extracting the pellets. It is not known what effect methyl formate might have, with the possibility of them dissolving and blocking the capillary. Additionally, methyl formate flowing past them could cause abrasion, breaking down into smaller particles, again causing blockage. Incidentally, the latter can be problematic with the modern refrigerants these filter driers are meant to be used with.
Once the pellets were removed, a brass screw was put into the hole and silver soldered over. Thus, we were left with a simple filter having a coarse and fine strainer.

Filter drier with pellets removed, and 1/4" tube with 1/2" reducer attached.

Because I had planned to do the installation with the cooling unit charged, I thought it best to make the capillary unit up as one piece first. Then only two connections need to be soldered onto the cooling unit. As with the CA-1, I used 29" of .026" capillary tubing. This is of course soldered into the end of the filter that has the fine screen. To the other end was soldered a length of 1/4" tubing, suitably bent, and terminated in a 1/2" to 1/4" reducer for connection to the Everdur float seat tube.
Flow through the assembly was checked with nitrogen, in case the capillary tube was blocked during soldering.

Preparations for the conversion and working on the unit fully charged.
Methyl formate has a very convenient property in that it's a liquid at atmospheric pressure, and it boils at 32 degrees C. This means it should be possible to open up the cooling unit on a cool day and not have to extract the methyl formate inside. I had given a lot of thought about how to go about this over many months, so this was the opportunity to try it out.
Everything went to plan...

Equalising the pressure.
First thing to do is get the cooling unit up to atmospheric pressure so that no air flows in, and no methyl formate flows out, when the unit is finally opened. As described elsewhere, the CA unit is in a vacuum in its dormant state. Only the high side works in a positive pressure, and that's only when the machine is working hard.
So, a connection to the nitrogen bottle was made, and the unit filled slowly until the pressure was just above 0psig. At this point the nitrogen bottle was disconnected and the charge valve slowly opened to equalise the pressure to the atmosphere. This needs to be done slowly so as not to stir up the oil and methyl formate causing any to be blown out.
Nitrogen is used rather than simply opening the charge valve to the atmosphere. This is because we don't want moisture entering the system, and also because nitrogen being inert will not support combustion of the methyl formate - important given the proximity of the blow torch during the work.

The cooling unit is brought up to atmospheric pressure with nitrogen. This reduces the chances of the methyl formate igniting, and prevents moisture entering.

Cutting the float seat.
With the cooling unit on its side so any liquid would remain away from the area to be worked on, the float seat was cut off using a tube cutter and junior hacksaw.
It was immediately obvious what the problem had been all along. The seat assembly was coated in blue granules of what looked like copper sulphate. Apart from that, the seat had a strange bleached appearance.

Soon as I saw this, it was the answer to the unreliable float valve.

The seat material looks somewhat bleached. My assumption is the blue granules are copper sulphate.

Obviously some sort of reaction had been going on. What, I'll never know as it would predate me owning this fridge. Quite probably it was these blue granules that had blocked the evaporator.
As I'd hoped, I could not smell any methyl formate vapour and the severed Everdur tube was dry.
Holding the flame from a stove lighter near the opening didn't cause any effect, so it seemed I'd be able to solder on the 1/2" reducer without any problems. And, so it turned out to be.

Evaporator connection.
Next was to desolder the 3/16" tube running to the high side of the evaporator. Then, the capillary was inserted in a few inches and soldered. The unit was now sealed again, and without losing any methyl formate.

All connected and sealed. The unit is now ready for purging the nitrogen.

Purging the cooling unit.
The astute reader may wonder now, if the unit is at atmospheric pressure, how is the vacuum restored? Normally, one simply connects a vacuum pump to do this, but as there's methyl formate inside, which we want to keep, then we can't take that approach.
Simply, the unit is purged as if it has NCG's.
To remove all the nitrogen took about an hour and a half, following the GE purge instructions of one minute open, and three minutes closed. It was clear as to the progress, for in the same way as an NCG purge is done, the top of the condenser warmed first, then the lower half, and then the float chamber.
By this time the evaporator was frosting, and the whole exercise was a proven success.
Unlike with the presence of a needle and seat, NCG's do not cause the rattling sound in the same way when the capillary conversion is done. There is after all no valve for the NCG's to force shut. What happens when NCG's build up in a capillary converted CA unit is that the frost line drops first. The rattling begins after, when the high side pressure has become higher than normal. This is one of the advantages of the capillary tube conversion - the frequency of NCG purging is reduced.

Tidying up.
As with the CA-1, I had to make a bracket to secure the filter drier. It would be just too easy for something to be pushed to the back of the evaporator bending the 1/4" tube.
I used a plastic cable clamp screwed to the shelf. Silver paint was applied to all the new tubing, and a new 1/2" grommet placed over the Everdur tube where it exits the cabinet top. This was sealed with Silastic to prevent moisture entry. The cooling unit was ready for installation, but I took the opportunity to do some cabinet work first.

The filter drier is secured to the shelf and all bare copper painted. A new grommet seals the float valve tube.

The cabinet.
With the cooling unit out, I decided to do some touching up with the cabinet - things I had overlooked in my haste to get everything together and working the first time round, a year ago. The inside of the cabinet liner and the door needed a tiny bit of rust clean up and touching up. It was also an oppurtunity to have a closer look at the insulation, which was all dry and in good condition.  One thing I wasn't happy about was the No-ox-id cloth seal. While various holes and tears can be patched, the problem is that it catches on the cooling unit when lowered in. The slightly sticky surface doesn't help here. With the success of the duct tape replacement I tried with the CA-1, it was obvious the same would be appropriate for the CA-2. It make a big difference, forming better to the inside of the cabinet, and staying in position. The top lowered in very easily.

New duct tape top seal eliminates all the problems with the original No-ox-id cloth.

The test!
As with the CA-1, I found it not necessary to run the heater for very long to get a quiet start. It was just really nice to see the cooling unit start cycling and not have to worry about the float valve blocking. The cabinet cooled down without any problems and it seemed all was well at last. The frost line was as it should be. At this point I should say that the cooling unit was running with what I presumed to be 3lbs of methyl formate. Apart from the initial 2.75lbs put in when I first had the unit running, six months later I added another 4oz (.25 lb) in attempt to get the float to work more reliably. Not that it did, it just worsened the cycling time instead.
Given that I'd found a charge of about 90% to be right for the CA-1, I'd wondered about the charge in the CA-2. Again, it seemed that an overcharge is not evident with the frost line, all it does is worsen the cycling times. And, these I was not happy about - not as good as when the float valve did work.
The obvious thing to do would be to remove some methyl formate, so about 75% remained, and then slowly add more to get the optimum level.

Methyl Formate Extraction and optimising the charge.
Not just for the purposes of optimising refrigerant level, but also should one want to evacuate the system for extensive work, I came up with a method of recovering the methyl formate.
The cost, at about $130 per litre, is too high to allow the liquid to be wasted. A lot of the ideas used came from jhigdon on the forum.
What I came up with was a connection that could be screwed to the charge valve, and empty into a remote bottle.

Connection to the charge valve was by a flared piece of 1/4" tubing to which was connected 8 feet of capillary tube. The charge valve adaptor was used to connect the 1/4" tube to the charge valve.

As there would be vapour present, as well as the liquid, it was necessary to condense the vapour in order to recover everything. So, a long capillary tube, about 8 feet of it,  was used to connect the bottle.

Capillary is inserted into empty methyl formate bottle for collection. The bottle sits in an iced water bath to futher increase condensation. The tin is actually that which the bottle came in.

By the time the vapour got to the end, some of it would have condensed. However, the most dramatic improvement was to run the capillary through an ice bath, and to also place the bottle in an ice bath.

Passing the capillary through an ice bath improved results considerably.

The liquid flow was continuous, and I could not smell any methyl formate. It is a slow process taking about three hours to get .63 lbs out. But, it does seem to be very efficient, and is easy to implement. Of course, the evaporator has to be warmed with a pan of hot water to get a positive pressure on the high side.

The complete extraction setup. Note the saucepan of hot water in the evaporator to ensure positive high side pressure.

After three hours, .63 lbs had been collected.

With .63 lbs out, and a charge of 86%, the frost line hadn't gone down, and the cycling was still poor, with short off times. So I removed more.

Right side header tank with 86% charge. Run times were still poor.

Left side header tank with 86% charge.

With 76% of the official charge, the frost line had finally dropped, but was now too low. The right header tank had no frost, and the frost above the shelf was not evenly spread across the evaporator.
Now it was a matter of adding very small quantities to bring it back, while keeping a close eye on the cycling, optimising for long off times. The run times had stabilised around 2.5mins, but the off times were still barely getting to 9 mins.
Putting in .04 lb quantities, and observing the effect over a day got the off times to nearly 12 mins. That's what I was looking for! Charge was now supposedly now 2.48 lbs, which is 90.2% of the original charge.
It is very important that the cycling times are observed when the room temperature is the same as the last reading - the fridge is very much influenced by room temperature, and this can mislead one into thinking the cycling time is worse, when all that has happened is the room is much warmer. It is also necessary to wait a couple of days to observe the effect, as the system takes a while to stabilise after the change of operating conditions.

With such an improvement, I wondered if adding just a little more would get another minute for the off time. So, I added .03 lb. Alas, it had the reverse effect, with off times now around 9 mins for a 70F room temperature, approaching 11 mins for 60F.  It does seem that the charge is actually critical within one percent. However, I was still not satisfied with the cycling time. It had been better with the float valve.

50 cycle operation.
A chance discussion on the forum made me realise that my compressor is operating at about 1425 rpm, whereas in the U.S, it's about 1750 rpm. While I'd always been aware of this, what I hadn't thought about was reduced compressor capacity. It's reduced by about 17%. Of course, the float valve accomodates this automatically, which is why the monitor tops are sold for 50 or 60 cycle operation with the same compressor. However, as I'd changed to a capillary tube, it suggested that the not quite optimum cycling times were a result of the lower capacity. A capillary tube only works properly with one set of conditions. And so began a lot of intensive research into the subject.
First thing was to observe what happened when I ran the compressor on 60c/s. To do this, I used an Omron variable speed drive.This is simply an inverter with variable frequency output. There was a dramatic increase in the frost level when I tried it. This confirmed that mains frequency really does make a difference. It also suggested that the capillary tube was probably not optimum. I also found that the start relay calibrated for 50c/s doesn't work properly on 60c/s. So, I had to start it in 50c/s each time then wind up the frequency once started.

Capillary tube experiments begin.
Now I had to learn all about sizing of capillary tubes. One would initially think that for less capacity that the tube should be shortened to provide less restriction. And so I tried using a 38" length of .0315" ID capillary. As before, I worked on the unit fully charged and brought up to atmospheric pressure with nitrogen. There were no problems doing this. I did however get to see what methyl formate does in the presence of flame. It burns with a green/yellow flame. This occured while removing the old tube from the filter drier. There was just enough positive pressure inside, presumably from the room warming up, to get the vapour out. There was no explosion or anything like that - it burned just like a candle. I simply blew it out and kept soldering.
While there was no problem getting the required temperature with the lesser restriction, the cycling times were never better than about 9 minutes off time. Adjusting the charge level didn't greatly improve things.

Longer tube.
There had been suggestions that the tube should be longer for less capacity, and looking at compressor catalogs seemed to verify this. It sounded illogical at first, until I started thinking in terms of pressure rather than rate of flow. For the evaporator to cool, there has to be a pressure difference between the high and low sides. The greater the difference, the more it cools.
Imagine the compressor is reduced in capacity. Now, for the same capillary tube, the pressure drop across it won't be as much. So, the evaporator doesn't cool as much as it did before. Shortening the tube is the worst thing to do, for now the pressure drop across it will be even less, and the pressure drop between high and low sides of the evaporator will be less.
In order to restore the correct pressure difference between high and low sides, the capillary has to provide more restriction to create a greater pressure drop.
As to how much longer it should be, that would have to be found by experiment. While there are formulas available, they are extremely complicated, and any of the internet based calculators do not accomodate low pressure refrigerants. This is not surprising as in the modern world, refrigerants like R11 and R123 are used with float valves because of the size of the systems. However, I did get a starting point by looking at recommended capillary tube sizes in various catalogs.
For a reduction in compressor capacity of around 17%, the cap tube was about 1.4 - 1.5 times the length, depending on refrigerant.
Time to try a longer tube, and this time, I used 44" of .026" ID capillary. This gave the best results. I did find the charge level had to be higher; about 2.7 lbs. The performance was now as good as the float valve.
It is interesting to note that despite the low pressure of methyl formate, the capillary tube sizing is not vastly different to that used for R12 or SO2. Literature suggested that capillary tube length is critical, but it's important to note that again this was for high pressure refrigerants. My experiments revealed that with low pressure refrigerants, it is not so critical.
All three of the tubes I tried did work, and the difference between them was not huge. If one was to use .028" ID tube as a reference, a change in 10" of length alters the off time by about 2 minutes.

Update: Improved Methyl Formate Recovery Unit.

This is an improvement on the original method I used. Here, a car fridge keeps the bottle and ice bath cold throughout the extraction procedure. Two freezer blocks also help with condensation in the bottle. Importantly, a rubber stopper has been added to the bottle to prevent moisture absorption. Because it is now sealed, it is necessary from time to time to release the pressure build up. Extraction is more rapid with this new set up.