ron soyland

Without doubt, making the vacuum tight wire seals through the pyrex glass is the most difficult part of making a tube. More tubes will fail from seal leakage than from all other problems combined. Thus, I will spend some time here showing exactly how it is done. After several dozen or more practice seals you should be able to relaibly make a seal that will be completely vacuum tight. And have no problems with annealing!


Pyrex (borosilicate glass) is different from the soft glasses in the respect that the expansion coefficient with heat is drastically lower. The glass doesn't change size much when it is heated. This very fact is why pyrex is so desirable to use. It has a very low tendency to crack when it is heated or cooled.
There is a serious downside to this though. There is only one metal (common metal) that has a coefficient of expansion close enough to pyrex to successfully make a seal that is vacuum tight. Many wires will seal through pyrex for liquids but not vacuum tight for gases.
Tungsten metal has a coefficient of expansion that is within 10% of pyrex. Thus, by using a trick, a vacuum tight seal can be made. This trick involves the use of tungsten oxide as an interface between the pyrex and the tungsten.
Other metals, such as Kovar, are available for sealing to some types of borosilicate glasses, but these are special glasses that are not commonly available. You would have to buy the glass directly from the manufacturer, which would mean a full case minimum. (a lot of glass, a lot of money!) Kovar will not successfully seal to common boro 7740 lab glass (pyrex and equivalents) that is the most commonly available borosilicate glass. The tungsten seal is the only successful seal for this glass.

The tungsten oxide inherently seals to the tungsten, by forming a compound. It also will go into solution with the pyrex glass. The result is a seal to tungsten on one side, a layer of oxide, and a seal to pyrex on the other side. There are strict requirements for this process to work:

(These requirements are for a simple pinch seal in 7740 borosilicate glass: pyrex or its equivalents.)

 First, the maximum size wire you can get away with this seal is about 20 mils (1/2 mm) diameter wire. Try with larger diameter wires and the seal may fail at any time due to stresses. You might be successful when you do the seal but an hour later, a day later,a week later, the seal may crack and be ruined. This doesn't happen with wires below 20 mils diameter. The smaller the wire used for the seal, the more likely it is to succeed. The current handling capacity of the wire is the determining factor of the wire size, and to a certain degree handling complications for smaller diameters also limits the smaller sizes.

Second, the wire must be polished mirror smooth. A rough as-drawn wire will not seal because the small longitudinal grooves from the drawing process will conduct gas past the seal. Most tungsten wire is easily buffed to the required polish in a few minutes using fine emery cloth. (400 grit)

Third, the oxide coating on the tungsten must be of the correct thickness. If the coating is too thick, it will crack. If it is too thin, it all will mix with the glass and there will be no oxide left to bond to the tungsten.

Fourth, the temperature of the glass when the seal is made must be just right. Too cold and the glass will not completely form tightly around the wire, leaving a gap for air to leak through. Too hot, and the oxide will break down and release oxygen gas making bubbles along the joint which can make passages for air leaks.

A fifth somewhat more rare occurance is longitudinal laminations in the tungsten wire. When the wire is drawn down, any imperfection in the original wire gets pulled through the die with tremendous force. This can cause it to laminate and form an invisible crack that runs longitudinally along the wire for a few millimeters. If this crack just happens to fall where you make the seal, you can have a perfect seal to the tungsten but the crack will pass gas along under the seal! The wire used for the seal must be free of the longitudinal cracks. Welding wire (TIG wire) is not tested for these cracks so you must be aware when preparing the wire for any indications of a crack. A sure sign is when you bend it sharply to break it, the break is not clean. It splits the wire instead of cracking it cleanly in half. In this case, throw that piece away and move down the wire a few millimeters and try again. The cracks are never longer than a few millimeters so it is easy to get by them.
Annealed tungsten wire (light bulb filaments) are inspected for cracks so you can be sure they do not have any. It is important to use NEW light bulbs for the filaments, since in used bulbs the filaments will be so brittle they will shatter if you try to straighten them. Auto headlamps have heavy tungsten filaments, but can be expensive. Simply buy the wire from a supplier.


The cheapest source of tungsten wire is welding wire that is used with TIG welders. (tungsten-inert gas) This wire is available in sizes from about 10 mils to over 100 mils. This type of wire is not annealed. Thus it is stiff and cannot be easily formed. It can be cleanly broken by gripping it with a pliers and sharply bending it over. This wire in the size used for tube feedthroughs is about 70 cents a foot. (A 100 foot roll at $73 lasted me over 5 years even with the large number of tubes I have made! (almost 1000 tubes now!)
Annealed tungsten wire is available from a company called HYPERTRITON which sells on ebay. The price for the size used in tube feedthroughs is about $3.80 a foot, which is much higher than welding wire. The wire is annealed, so it bends to form easily. It is cut with a good side cutters. The annealed tungsten wire doesn't have any longitudinal cracks to worry about, and it is already polished to a mirror finish. One foot will make about 40 seals so the cost per seal is about 10 cents each. This is not a problem. I have several rolls of welding wire left or I would buy the annealed wire myself.

 The current carrying capacity of the wire is what determines the size you need. For most small vacuum tubes, the filament will be the highest current that will flow. Thus, the wire should be chosen to carry that load safely. The rating for the wire (in a glass seal) is less than the rating in free air, since the wire must not get hot enough to risk cracking the seal. For 1 amp current, 15mil tungsten wire is satisfactory. I have used 15mil wire (.39mm) to carry 3 amps of current with no failures, so the ratings are generous.

One characteristic of welding wire is that it is "as drawn" with no surface polishing at all. Thus, if you use welding wire you must polish the wire. To do this, cut about 2 feet of the wire and grip one end in a bench vise and the other in a visegrips pliers. First, use 300 grit emery to polish the wire by pinching the wire in a fold of the emery and sliding it forcefully along the wire rotating it to get all sides. Then do the same with some 400 grit. This only takes a few minutes to prepare two feet of wire, which is a lot of seals.
If you purchase the annealed wire you don't have to do this. (recommended)
This size of wire may be a bit large if you are going to make a multi-lead pinch with over 8 wires. Drop down to 10 mil (,25mm) tungsten and you can get more leads per seal. The current carrying capacity of 10mil tungsten is right at 1 amp, which will be satisfactory for tubes using a 2mil (1/20mm) tungsten wire for the filament.
If you use the annealed tungsten wire, you can use the tungsten wire itself inside the tube for element support since you can bend it around quite easily. But at almost 4 bucks a foot that will get expensive quick, since you will use as much as half a foot per tube in some cases. Also, welding to tungsten wire requires a buffer metal, which is not convenient in the tube support construction. I recommend welding nickel or stainless steel wires to the tungsten so the tungsten is used only for the seal itself.
Thus, make the seal wire itself about 1/4 inch long and weld a copper wire to one end for the external connecting wire, and a nickel wire to the other end for the internal tube element support. Use 20mil or so copper wire and about the same nickel. This gives good flexibility for the copper and good stiffness for the nickel. See the page on the pincher welder to see how to make the welds to the tungsten.


The tungsten as purchased has air trapped in the surface of the metal. If the seal is made without driving out this air, bubbles will form which can form a path for a leak. It is important to do the preparation immediately before making the pinch. Do not do it ten minutes earlier, do it just before you are heating the glass to pinch it. This is important to get a perfect seal.
To prepare the wire, have it supported right ready to insert into the glass where the pinch is to be made.
Use the small pinpoint tip (1mm flame) and carefully swipe the flame across the tungsten wire (or wires if you are making a multi lead seal) heating each wire to white hot for a fraction of a second. BE CAREFUL NOT TO LEAVE THE TORCH ON THE WIRE LONG ENOUGH TO MELT OFF THE WELDED WIRES! The tungsten will not melt in the flame but if you heat it too hot, the copper wire especially will melt and drop off. Play the flame back and forth over the wires, heating each one to white hot, until they have a black coating, AND THEN STOP. If you get a white coating or a yellow coating, you have done it too long. The oxide coating should be a dark gray-black in color. The wire is now ready to make the pinch. Move it into the glass right to the position where the pinch will be made.


NOTE: During all phases of the pinch, be sure the mashers do not get red hot. Glass will stick to them! Let them cool somewhat if they get red. If the glass sticks to the mashers, immediately remove the torch and let the joint cool slightly while moving the mashers slowly back and forth side to side. They will pop loose when the glass cools. Do not try to pull them away while the glass is molten.

The pinch is best done in three stages. First, the internal nickel lead wires are pinched to hold the wires exactly in position. Second, the actual seal pinch is made. Third, the pinch to hold the exit lead wires is done.

The pinch is the most critical part of the seal. It is best to use a "crossfire" torch that has a jet on each side of the glass to heat both sides at the same time.
Use the mashers to clamp the wires right at the edge of the pinch so you will not burn them off when you heat the glass. First, heat just the edge of the glass, the part that will grip the nickel wires. Get the glass to the point that it is starting to sag and quickly use the mashers to pinch the glass down onto the wires. Note that this is not the seal pinch. This is to pinch the glass down onto the wires to hold them in place.
Now, the wires are in position inside the glass. If there are any wires that are touching or out of position, carefully heat the glass and use a small screwdriver to pry the wires into position.
The electrode ready to make the pinch. The top wire is .028 nickel, which is used for the internal wire to support the tube elements. The center is .016 tungsten, which will be the actual seal. The lower wire is .024 copper, which is the external connecting wire. Note the thorough penetration of the welds so they will not pop loose while the pinch is being made. The total length of the tungsten seal part is not particularly important. It can be anything from about 2mm to 10mm or so. The 4 mm length gives a good chance of having a vacuum tight seal while not making the overall length of the seal so long as to be messy. If the tungsten is made very short, there is a greater chance of a leak if the pinch is not perfect. A slightly longer seal area makes it more likely that part of the seal will be hermetic and make the seal a success. There are some situations where a very short seal length is necessary, like through the side envelope of a tube. It is best to pre-bead the tungsten in this case.
To pre-bead the tungsten, the electrode is prepared as in the picture. The tungsten is flared to remove the air and form the oxide layer. Then, a short piece of capillary glass is placed over the seal area. The torch is then used to seal the bead onto the wire starting at one end and carefully working along so that the air is excluded.
The second photo is of a seal bead ready to be used in a vacuum tube. Note the perfect dark gray, almost black color of the tungsten seal with no signs of bubbles at all. This is a perfect hermetic seal. The seal was made using the electrode wire in the other photo. When making the bead, it should be about .125 (3mm) dia. Be sure the bead completely captures each end wire weld in the glass.
 If you are going to use the bead in a pinch, the total diameter can be made smaller, .06-08 in. (2mm) dia. However, it is not usually necessary to go to the trouble of making beads when making a pinch. Just do the seal in the pinch!
Note that seal beads like this can be prepared ahead of time, since the tungsten is completely covered with glass and will not tarnish.
A pyrex flare flattened to make it easier to do the pinch on a bunch of wires.
Two tungsten seals in pyrex. Note the perfect dark gray color of the seal itself. The bubbles on the nickel end wires are not usually a problem, but can make outgassing the tube take longer. Heating the nickel weld to red hot for a few seconds before making the pinch reduces the bubbles significantly. So I forgot.