A Philips 667A was the first "antique" valve radio I ever had. Unfortunately, something dropped on it during a removal, breaking the dial glass. I have put up with the broken glass for years before I started searching for a replacement. This quest was complicated by the fact that there are a number of different series of the 667A with different dial glasses. For some time I had no success, but in 2005 I found two 667A's in bad shape with good dial glasses shortly after another. The first one had the right type of dial glass but the lettering had already started to fade. I was not as good as I had hoped so I continued my quest. A few months later I saw an advertisement for a 667A wreck and made the trip to collect it. Though the seller had assured me that he had the right type of dial glass, it turned out to be a different one. But the dial was in such excellent shape that I could not resist buying it. That radio had appparently been stored in a dry environment, because the lettering on the dial had been preserved so well and the chassis had little rust. But somebody had been looting it so it was missing a lot of components.
I really didn't like the idea of changing the looks of my old 667A by mounting the wrong type of dial glass. So I decided to use the slightly faded one after all, leaving me with one dial glass too many on my hands. Gradually, the plan developed to make one good radio by putting together the best parts of both wrecks. But it was going to be a lot of work.
I tested the varnish with a tissue with alcohol and a tissue with thinner, and found out that it was soluble in alcohol, but not in thinner. So it must be shellack, not varnish or nitrocellulose lacquer. If I wanted to restore the original finish as closely as possible, I needed to learn french polish. After a few searches on the 'net I found ample information, the most important message being that french polish can be learned by anyone, but it takes a lot of practice. So I was glad I had the other radio case to experiment on. The first tries were not very good, but after a while I learned and I was confident enough to take on the case I wanted to restore. The result is rather satisfactory. Not perfect, but quite beautiful. The french polish has a deep and warm shine, more appealing than the PU lacquer I used on the AT 660 WK3.
One of the two chassis was rather clean and wasn't too rusty. There was one inconvenience though: all the valve sockets had been removed. Two of these were easy to replace: both the rectifier and the tuning eye are riveted to small brackets that can easily be swapped. But the others had been riveted to the chassis itself with brass rivets. I was able to counterfeit a number of these rivets using 3 mm brass pipe of a type available in a shop that sells stuff for constructing train and aeroplane models.
About half of the electronic components had to be taken from the other chassis. One special part was the first IF transformer. It has a coax cable coming from the top that connects to the top contact (grid) of the IF amplifier valve. But it had been torn off, and this was hard. I carefully dismounted the IF transformer from the other chassis. In order to attach it, I needed a special tool to bend the lugs that hold the IF transformer in place. I made one from an old screwdriver. While I was transferring the components from one chassis to the other, I checked all the resistors and capacitors. I also emptied the wet electrolytics and hid new modern electrolytics in them.
The paper capacitors required quite some attention. They were all very leaky: their resistance varied from 500 kΩ to 2 MΩ. A number of these were in places in the circuit where they would use up too much power or would hamper the radio's functions. But I did not like the idea of just replacing them by modern capacitors. This would be just too ugly. I recently learned about a trick to get the paper capacitor out of the pitch encasement and tried it. The result should be an empty shell, a hollow cylinder made out of pitch. But that didn't work for me. The cylinders crumbled. So I had to think of a different trick to camouflage the replacement capacitors.
I rolled paper cylinders around the capacitors after soldering extension wires to them. Then I filled the ends of the cylinders with melting glue from a glue gun. After this, I melted an amount of pitch in a shallow tin box. I took the capacitors by their wires and slightly touched the surface of the melted pitch, rolling them between my fingers. This way, the paper cylinder was covered with a coat of pitch. After letting the pitch cool down for 10 seconds, I rolled the cylinder between two flat scraps of fibreboard to make them rounder. That also had the side effect of removing the shine and make them look dusty. There was one problem, though: the two ends of the replica "tarball" capacitors were not finished. They had a rough shape, far from the desired conical shape. I did not have a good solution at first. I was going on vacation and after that a lot of other projects came across. So for more than a year, the chassis was left in its brackets standing in a corner on my workbench. In August 2006, I picked up the project again. I took the halfway finished "tarballs" and warmed them up using a blow-dryer, the type that is used to scorch and remove old paint. The pitch became soft again and I could make them rounder and finish the ends using a conical mould. This left me with an odd collection of tar capacitors, as they all had different sizes and they all had become larger than the originals. But if you didn't look too well, they were quite acceptable.
So things seemed to go fine. Using the replicas I completed the radio circuitry. Then I carefully tested the radio without valves, using a variac. First I tested the isolation of the mains transformer. Then I inserted the rectifier valve. All voltages were OK, except there was a positive voltage after the coupling capacitor to the output valve. The replica capacitor I had used, appeared to be leaky. I was surprised to find this and expected I had made a mistake. But when I unsoldered it and tested it, indeed the capacitor turned out to be leaky. It was a polystyrene capacitor. I assume I have warmed it up a bit too much in the process of making the tar mould.
I replaced it with a mylar capacitor and went on. Uncamouflaged, alas. Now it was time to insert the EBL1 output valve. I had five used ones in my valve box, so I was confident at least one of them would be OK. It turned out, however, that four of them were dead and did not have any emission left, at leat not at the normal heater voltage. The fifth one did seem to work, but after a few minutes I saw the plate voltage collapsing from 260 V to 190. It turned out that its grid voltage had gone up from -6 V to 0 V. I feared my new mylar capacitor had broken down. But the valve had the same problems when I inserted in a test circuit. So it was gassy or it suffered from grid emission and it was not usable any more.
I scratched my head. Now you have five EBL1's, now you have none. Fortunately, I could buy two "new" Tronal EBL1's through Ebay. The valves arrived and I tested them. They were both stable and worked excellently. After this, I added the other valves one by one, and my radio worked! The band switch was cracking badly cracking which was fixed by cleaning its contacts.
There was another thing to be replaced. The tuning cord drives a large bakelite pulley and gear that drives another gear that is coupled to the tuning capacitor. The mounting point for one of the springs that stretch the tuning cords had broken off the bakelite pulley and gear, so I wanted to replace the pulley and gear with the one from the other chassis. But that was nearly impossible because the pivot on which it rotates, is riveted and soldered to a metal plate, the "driving plate". I decided to replace the whole driving plate sub-assembly containing of two gears (including the bakelite pulley and gear). There was something odd about the way the driving plate was fixed to the chassis. Obviously, you could adjust the way it was driving the tuning capacitor, resulting in some uneven, non-linear movement. In the manual, I found out how the driving plate should be adjusted to correct the movement of the pointer.
But first, I completed the mechanism and strung the driving cord from the pulley at the shaft of the tuning knob to the larger bakelite pulley. I first used the original cord. It looked a bit dried out and indeed, it broke when I tried it. So I replaced it with a shiny white new one. Then the tuning knob worked ok.
After that, I wanted to restore the string that moves the pointer on the tuning dial. One of the dials had a thin steel cable, the other one had a solid bronze wire. The steel cable was untwining and a loose strand of thin steel wire penetrated my skin when I touched it. Ouch.
For the alignment of the oscillator and the tuned circuits at the input, the set point that Philips prescribes is the point where the tuning capacitor is rotated over 15° of its total range of 180°. This corresponds to a frequency of 1442 kHz. Philips specifies the usage of a “15° gauge” for this purpose. This is a tool that you slide over the stopping pin of the tuning capacitor so it cannot turn to 0° but stops at 15°. By carefully measuring the dimensions of the capacitor of the chassis to be discarded, and doing some trigonometry, I was able to calculate the required extra clearance from the stopping pin. Then I made a 15° gauge out of a strip of aluminium with a hole in it, that I carefully filed to size. Using this tool, the alignment of the tuning capacitor went well.
The most awkward part was the aligment of the pointer movement. As I mentioned, the driving plate of the tuning assembly must be positioned such that the pointer indicates the right frequency on three points specified in the service manual. Three M4 screws fix the driving plate to three studs. The holes in the driving plate are oversized, so that it can be moved. By slightly turning and shifting the driving plate, the movement of the pointer can be influenced. It took me a lot of iterations before I finally got it right, but it is really quite precise now.
Then I looked at the result and listened to the music. Great.