By Chris Underwood
The use of quartz crystals in oscillator circuits for improved frequency stability was developed in the 1920s and 30s. Initially they were used mostly by ham radio operators, who ground their own crystals to the required frequency.
Gradually, commercial operators adopted quartz crystal (“xtal”) oscillators and at some point during the 1930s the New Zealand Post Office Radio Section set up a small quartz crystal production laboratory to meet its own requirements.
Each crystal was cut and ground by hand, a laborious and time consuming process but adequate to meet the NZPO’s requirements at the time.
At the outbreak of World War Two, this small laboratory was the only such production facility in the country. With the move of Radio Section to the Wellington East Post Office building, the lab was moved to a room on the second floor overlooking Cambridge Terrace.
Demand for xtals soared to meet wartime demand. The lab not only had to meet NZPO requirements but also that of all branches of the New Zealand services plus those of the American forces stationed here. During peak demand it was necessary to run three shifts a day, seven days a week. The War History of Radio Section records that total wartime production was in excess of 4500 xtal units, each taking approximately four hours labour in grinding.
Early xtals were not always reliable and were prone to many problems such as sudden change of frequency as well as temperature and altitude-related frequency drift. During the war the United States ran an extensive research and development programme rivalling that for the Manhattan Project. This programme determined that Brazilian Quartz was the only “Radio-grade” quartz available, and set up mining operations to secure sufficient quantity of suitable crystals.
As no mass production tools or techniques existed, considerable effort and expense were devoted to the design of early equipment and production techniques.
The book Crystal Clear by Richard J Thompson and published by the IEEE, records the problems faced and the solutions devised.
After the War, it became apparent that the demand for quartz crystals would continue and, if anything, increase with new services such as land mobile and point to point links.
Radio Section Management, after considering the options to update the Lab’s equipment and techniques, decided in 1951 to send a young Engineer, Reg Motion, to the UK for three months to study equipment and methods used by the BPO and other leading commercial manufacturers.
On his return, Reg worked with DSIR staff to develop suitable equipment based on what he’d learned during his time in the UK.
The equipment he’d seen overseas was suited to very large scale production; New Zealand’s requirements were more modest.
Among the equipment designed and built were two precision grinding machines, two lapping machines, two plating machines and a wiring machine. Some of the equipment was built on the 4th floor Radio Workshop and some, such as the wiring machine, by the DSIR.
I’m not sure when the lab moved to Rugby Street, but as this building was constructed in 1960 it was likely around this time.
I first saw the Crystal Lab when I attended my first stage training at the Radio School at Rugby St in the mid 1960s and we were given a tour of the lab. Most of the equipment shown in the following photos was still in use at that time.
One of the things we were shown was a large box of Brazilian Quartz Crystal. Before the crystals could be sliced into blanks they first had to be X-rayed to identify any hidden flaws.
Next came the decision on how to slice up the crystal to get the blanks required.
The “cut” of a crystal blank pre-sets much of the finished xtals characteristics.
When I worked in the Crystal Lab some years later, the box of raw crystals was still in the storeroom although they were no longer being used. They had been replaced by pre-cut blanks from Japan. Made from artificial quartz, these were more consistent and more readily available. Hence I never saw the X-ray machine being used or a raw crystal being cut. However I can remember using beeswax to hold several large blanks together before inserting them in the diamond blade chop saw if more than one blank needed to be cut to size.
We had large 1-inch square blanks for low frequency crystals which sometimes had to be cut down to ¾ in square. Mostly, at the time I was there, we used ½ inch square blanks which only needed to be lapped to frequency. To speed things up we kept a “bank” of blanks of various thicknesses so the nearest thicker blank to that required was chosen. In the foreground of the photo showing Dave and Brian using the grinders the top of what looks to be the diamond-blade “chop saw” is visible. It is probable something like that would have been used for the raw crystals. The rough blanks could then be readily ground to the required dimensions.
The grinding machines were quite complex devices and could cut very accurately. They used a vacuum chuck to hold the quartz blank in place while it was ground to the required thickness.
Once the blanks were cut to size and slightly above the calculated thickness they were taken to the lapping room where they were lapped to the finish thickness. I last did this many years ago and most of the fine detail I can no longer remember. However some low frequency xtals had to be lapped in a special way to make one surface convex. They were known as PC (Planar Convex) lapped xtals, and we made “Deep PC” and “Shallow PC” versions of them.
The machine used to do this was similar to a Potter’s Wheel and had three interchangeable lapping wheels, one for each type of PC xtal and one for doing bevelled-edge xtals. The machine in the photo is similar to the one we used and gives an idea of what’s involved. It’s a messy business until you get the hang of it.
The wheels we used were much smaller than in the photo, about 150mm (6”) in diameter, and the frame of our machine was round, allowing you to sit close to it.
You had a choice of two polishing compounds: fine and medium. They came in the form of a powder which you had to mix with a “carrier” (we used kerosene). The trick was to get the mix to the correct consistency. If you got it wrong it would spray everywhere when you put it on the wheel.
You then put a dollop of the mix on the wheel then turned the machine on. Next you had to hold the blank between the nail of your thumb and your forefinger then carefully position it just above the side of the spinning wheel then snap it into contact with just the right pressure of your finger and hold it there.
Snapping the blank into place was definitely an acquired skill. Too hard and you broke or chipped the blank. Too lightly and the wheel would snatch it and hurl it across the room.
I did this job for a while and my skin became saturated with kerosene. You couldn’t wash it away and I’d go home on the train stinking of kerosene.
You regularly had to lift the blank to check progress and the best way to do this was to slide the crystal up the concave side of the wheel and catch it again with your thumb nail when the edge of the blank overhung the edge of the wheel. As you had to do this many times for each xtal you got very good at it.
Bevels were another finish used occasionally. For these you would use the flat wheel and hold the blank between finger and thumb at about 45% to the surface of the wheel as you ground each of the four edges of the blank to a V shape. Fortunately the blanks were usually larger and thicker, so were easy to handle, but you had to be careful to keep the V shape symmetrical on each edge.
One trick always played on “newbies” was the boss would suddenly walk in and give you what looked to be a very large thick piece of quartz crystal and say it was an urgent job:
“The 50Hz crystal controlling the National Main Grid frequency has shattered and they need a replacement urgently!”
I didn’t fall for it, but some did. The “crystal blank” was actually a piece of 1” thick glass, about 4” square.
Ordinary xtal blanks were lapped on small machines like the one in the photo above being used by Dave Burger. The blank is held in a holder between the two wheels and both faces (from memory) were ground simultaneously. The blank had to be cleaned and checked regularly as things progressed fairly quickly.
In the photo above, Dave is measuring the resonant frequency of a blank. The blank is inserted between two pads connected to an oscillator. From memory, the activity of the blank could be checked on the meter and the frequency on the NZPO Radio workshop-built frequency counter mounted above the oscillator.
Once the blank was lapped to specification it was washed and ready to be clamp-mounted or plated.
In the early days of the lab each finished xtal would be placed in a crystal holder similar to the ones below, clamped in place, tested, labelled, and that would be it.
By the time I joined the Crystal Lab, clamp-mount xtals were no longer produced there. Instead, plated xtals were being made, as they were more stable over time and generally more robust.
In the photo below of Brian Fanning using the wiring machine, there is another important piece of equipment used for making plated xtals. It is the small item looking a bit like a paper hole punch in the bottom left hand corner. In using non-clamp-type mounting systems some method of firmly holding the crystal blank in place yet allowing it to vibrate has to be used. The system we used was wire mounting. To prepare the blank for wire mounting small spots of a silver-based paste were placed in the corners of the crystal blank. This small device was used for that purpose. First the “spotting plunger”, a small metal rod about 1/8 inch diameter was dipped into a small dish of coating paste by pressing down on the handle being careful not to get too much, nor too little, paste on the end of the plunger. The dish was then replaced by the crystal blank and four spots on larger blanks and three or two on smaller ones were deposited onto the blank.
When enough blanks had been prepared they were placed in a small electric furnace. The furnace had to be pre-heated and run at an exact temperature. There was little tolerance – just a degree or two. Too much heat would ruin the crystal blank and not enough heat meant the paste would not bond properly to the surface of the blank. The furnace was on a timer. I don’t remember the time now but it was relatively short – just a few minutes. Having your “rocks” in the oven was a stressful time!
Assuming all went well, the next process was to use a dentistry tool to carefully burnish the oxide coating off the spots leaving a bright clean surface. Before these could tarnish, the blanks were next placed one at a time into the wiring machine Brian is using in the photo below.
The wiring machine was an ingenious device, developed by Reg Motion and staff of the DSIR. It was built by the DSIR.
The centre part of the machine held blank slides on precision tracks. The fine mounting wire was fed from either side through a positioning guide. The wire was fed out so that a small length (about 2mm) protruded from the inner end. A small device could then be activated to bend the protruding end over into a right angle hook. The wire was then fed further through until just past the crystal blank, the blank was moved to position one of the spots under the wire hook, the wire was then pulled back fractionally to bring the wire in contact with the spot. Next the centre assembly was slid back and forth so the wire and spot passed under the two hot air jets, quickly at first but more slowly as the crystal blank warmed. At the right moment thin silver solder was touched to the spot and wire forming a soldered joint. The wire was then cut, leaving a short wire tail attached to the crystal blank. This was done in turn for each spot until all had been “wired”. It was a job requiring total concentration.
With the blank now wired the next step was to fit it to its base. This was done by soldering the wires to support wire connected to the base’s pins. While I was there several different bases were used, depending on the final physical form the xtal was to be. The largest (NZPO type D) was mounted in a manner similar to an octal valve, another similar to a miniature 7 pin valve (NZPO type E). The rest were various sizes of two pin mounts (NZPO types F and the smaller type J). There were also others I can’t remember which were rarely used.
Next the wired and mounted xtals were placed in an ultrasonic cleaner containing a weak solution of Teepol and water. After a period of at least ten minutes the xtals were taken out and washed under running water. Each xtal was then carefully blown dry with compressed air. The xtals were then ready to be plated to frequency.
The process we used was known as evaporative plating, in which the mounted crystal blank was placed in a socket in a small vacuum chamber fitted with a viewing window. Metal masks were next placed close on each side of the blank. Cut-outs in the masks, looking very similar to the outline of an old fashioned cast iron frying pan (see photo below), allowed the plating material, normally silver but on occasion gold, to form on the exposed surface of the blank. The cut-outs were arranged so that the handle of each “frying pan” covered different spots with a wire attached.
While the NZPO Crystal Lab used evaporative plating, most commercial manufacturers at the time, such as Rakon, used sputter plating. Sputter plating was relatively cheap and quick but produced a technically inferior product. In sputter plating you prepared a bank of pre-plated xtal blanks then using a process that sprayed (sputtered) silver to one side of a blank to bring it on frequency. This causes unequal loading on the xtal as it vibrates, due to one side being heavier than the other. Also, if the sputtering is not done properly it can fall off causing the xtal to go off frequency. When the NZPO Crystal Lab closed, Rakon purchased the two plating machines. I think the Government may have been lobbied regarding a Government Department competing with private enterprise.
In the photo below, technician Graeme Thompson is placing small hairpin-shaped lengths of silver wire on a heating element. One piece of silver wire was placed on each side of the crystal blank adjacent to each frying pan. Next, the cover of the vacuum chamber was put in place and the vacuum pump started to pump the chamber down. The chamber was fitted with a spark gap, and to check the pressure a button was pushed to create an arc which could be viewed through the small window.
The arc lasted as long as you kept your finger on the discharge button. From memory the voltage used was around 25kV and it looked like forked lightning or a Van de Graaff generator arc. It was initially quite intense which is why you only gave it just a quick burst, but as the air in the chamber thinned it became more blue-coloured and progressively much weaker until the arc could no longer be sustained as there were not enough air molecules to support ionisation.
At that point the two heaters in the chamber were turned on. In the low pressure the wire hooks would start to evaporate at low temperatures and a mist of silver steam would flow through the “frying pans” in the mask and condense on the cooler quartz blanks.
(As silver was deposited over everything inside the chamber, you had to keep wiping it clean so that you could see the arc and also so the oscillator would not be shorted out. Every month or so the buildup up on the non-critical parts got so thick you could peel it off in great wafers of silver or gold. After we had been “Gold plating” for a run of xtals for a big customer, one of the guys peeled the gold flake out and melted it on a block of carbon using a gas torch. He ended up with a pea sized chunk of gold but the effort was so great and time-consumming no one else ever bothered.)
The xtal’s resonant frequency could then be measured as the socket the xtal was plugged into was connected to an oscillator. Once the desired frequency was reached the heaters were turned off, stopping further evaporation of the silver wire.
Pressure then was permitted to build back up in the chamber and the cover removed. Oscillation at the specified frequency was checked and capacitance noted. If within specification, the xtal was ready to have its cover fitted and labelled.
At the time I was there, this meant applying a transfer and hand writing on the label the frequency and other identifying information. Sometime after I left an engraver was purchased and xtals were engraved with the frequency and other data.
Glass envelope crystals
The lab also made other types of xtal but these were not that common in later years. These xtals required special finishing with vacuum sealed glass mounts.
We used a small spot welder to weld extender wires from the glass bases. We then built up the mechanical structure of the mount by sliding mica wafers onto the two wires and spot welding them in place before mounting the blanks to them so D’s and E’s were quite labour intensive but only about half the complexity of the XT xtal. The NZPO was about the last place to produce those and we used to get orders from all over the world. Our main local customer was NZ Railways who used them in their trackside signalling gear. Brian Fanning was a master at making these and on occasion would make several XT xtals on the same long strip of quartz that just fitted length wise into a D type mount. I understand these were used in some type of Xtal Filter working at around 100kHz. It would take him several days to cut the quartz then make the XT xtals. I never saw anyone else make these. I hate to think what they cost to make and I bet we didn’t charge enough for them.
Fitting a crystal into a glass envelope was a skilled process. In the photo below, the technician is sealing a Type D xtal. Covers can be seen washed, ready for use and warming in the middle left hand side of the photo.
The base was mounted in a holder in the centre of a ring of gas burners. Then a glass cover was taken from the rack and placed in position. The heat from the burners melted the glass of the base and the cover together; this was a critical make-or-break point. If the glass was not heated evenly, stress cracks developed which meant the xtal had to be stripped down and re-wired to a new base. Sudden drafts, undetected flawed glass components, and under-heating the junction all contributed to cracks developing.
In the Type D a hollow glass tube protruded from the base. This tube was connected to a vacuum pump to suck the air out of the xtal. I can’t remember how it was determined that the vacuum inside the xtal was correct but once achieved the glass tube was melted close to the base to seal the xtal. This was also a skilled judgment call, as once the stem was sealed, the excess stem had to be pulled away and broken off without breaking the seal.
(Peter Mulhare recently told me that they used a “wand” charged up to about 25kV which was held alongside the xtal being pumped down and the air inside the xtal would glow from static discharge, a bit like the St Elmo’s Fire you see around aircraft wings when you fly through highly charged clouds. When the static discharge glow could no longer be seen then the xtal was evacuated enough. I never saw Brian Fanning using this idea, but I have vague memories of the vacuum pump used running slower and slower and finally stopping and that was when the xtal was considered finished and ready for sealing.)
The Type E was done in a similar way, except the hollow tube formed part of the glass cover hence the small blob of glass on the top of this type.
The Type D then had to be fitted to a Bakelite octal valve base with the two active wires from the crystal being soldered to the appropriate pins. The last job for both types was labelling and this was done using white paint to print the frequency and other data onto the glass of the xtal. Sadly, as with the markings on radio valves, this paint was not particularly durable and over time often rubbed off.
- When I worked in the Crystal Lab, Brian Fanning was the 2IC and taught me most of what I know about making xtals.
- All this happened many years ago and I’ve probably remembered some things wrongly or not at all.
- The Type D xtal shown on these pages was made by me personally except for sealing the glass which Brian did. It is a 1MHz 34pf Shallow PC xtal. I was allowed to make one as a training exercise.
- Some of the xtals pictured were not made by the NZPO but are used to illustrate the type.