Making a HV transformer for the Tektronix 310 scope


Recently I was given a Tektronix 310 oscilloscope, which was missing several parts. Among them the HV transformer. Here is a quick photo sequence about making that transformer.

At the start I didn't even know how that transformer looked, let alone its design. Several people provided information, which combined with the scope's manual to allow designing a replacement.

The original ferrite core is a quite short and fat double E core with octogonal center leg. In place of that one, I chose the power supply transformer from a junked TV set dating to the early 1990s, which has a round center leg of roughly the same cross section as the original core, but allows a much longer coil assembly. I put this transformer in the oven to soften the glue holding it together. Then I could easily separate the core halves, and then unwind it, recovering the intact bobbin.


Okay, let's start. I wound the primary first,  using #33 wire,  154 turns  for the main winding and 77 turns for the feedback section, in three layers, leaving a generous unused margin on each side, and using painter's masking tape for insulation and to make up the sides. This photo shows the completed primary.




Then I wound the two high voltage secondaries, in what I call "thick layers", which are a type of scrambled winding, but instead of building it up evenly from top to bottom, it's  built up evenly from one side to the other. That requires winding on a conical face of the winding, which takes some care. But such a winding is very quick to do, is space-saving because it doesn't need any insulation within it, and it has much lower capacitance than a layered winding. I have made many such thick layer windings with good results, but making one for close to 2kV with only 20mm length, like here, is critical at best.

The HV secondaries have 1269 and 1346 turns of #39 wire.

Here you can see the first secondary completed. It looks like a plain scrambled winding , but all cold end turns are on the left, and all hot end ones on the right.




I brough out the cold end connections by stranded wires, to avoid having high voltage between the pins of the bobbin, which are a tad close together for 2kV. Here you can see how those connections are made. The hook-shaped ends of the stranded wires prevent damage to the fragile #39 wire if the stranded wire is twisted and pulled.

On the right side you can see the bobbin pins having some relatively thick wires connected. These are the filament windings, one turn of  bifiliar #30 for each. A single strand of #27 would have been as good, but more bulky - and I didn't have any solderable #27 at hand!




This shows the core and the completed coil assembly. The core center leg is 13mm diameter. The external core dimensions are 40x45x13mm. The air gap is 0.88mm wide, which is pretty large for this transformer. but still acceptable. It will result in a slightly higher than nominal oscillation frequency.




The new transformer installed on the 310's HV supply board. I fashioned a simple bracked from aluminium sheet for mounting the transformer. It looks a bit strange with the bobbin pins poking into the air, but if it works, so be it!




The problem is that it didn't work. At least not for long! After a total of about two minutes working nicely, oscillating at roughly 35kHz and producing what seemed to be a normal HV output, it started sizzling, and the HV broke down. That means arcing inside the transformer.

I tore down and unwound the secondaries, searching for the cause of the problem. And I found it in the Hv winding I had wound first: While winding the thick layer, one turn had slid away from its righful place, and hed been pushed further and further away by the new turns being wound over it. Finally it ended up several millimeters down the coil, in contact with wire turns that carried a very different potential. At the place where the underlying insulation sheet pressed this turn up against the others, thus slightly compressing the enamel, it broke down and a short between turns resulted. Right in the middle of this photo is the failure point.



And if you couldn't see it, here is a close-up, clearly showing the burned enamel. Don't worry about the shiny place left to it, that's just a reflection of the lamp. 




I could have tried to just wind the secondaries again, with the same thick layer technique, but with more caution to avoid turns slipping out of place, and hopefully more luck. Instead I decided to try a conventional layered winding. Each secondary would take 8 layers. Interlayer insulation would be provided by a single layer of masking tape.

This photo was made in the middle of the winding process, showing such an insulation layer with a small overlap, a 3mm wide strip of tape on each side and the wire coming out between the tapes, ready to start winding the next layer. 




And this shos the layer complete. It's not perfect, specially at the start and end, but decent. The loose piece of tape on top is used to keep the wire from unwinding while I shoot the photo.




Accidents do happen. When I was three quarters through, the wire snapped and broke. Not wanting to undo the whole winding and make it anew, I cut out all the wire that could have stretched, then soldered the wire together, and embedded it between several layers of tape, as shown here. Such a patch needs much more insulation than what goes between layers of wire, because the solder joint doesn't have enamel around it, and does have sharp points, so that it's much more prone to corona effect than the undamaged wire.
For the same reason, I used a relatively thick blob of solder, so the sharp points and edges would be minimized.

This is also a good time to mention that all those things that add bulk to the winding, such as overlap of insulation layers, embedding joints, wires running across the winding sense, etc, are ALWAYS made at those sections of the coil that will end up outside the core! So the final coil assembly is somewhat oval rather than perfectly round, with the shortest diameter at the place that goes into the core.  




In this version of the transformer I started winding one HV secondary at the cold end, completing it in the middle of the eight layer. I made the single turn filament winding right there, then applied two layers of masking tape, then wound the single turn filament winding of the other secondary, and then started the second HV winding, still in that same layer. All high voltage stuff is embedded inside the secondary windings, and comes out at the bobbin pins. Instead the cold ends come out as wires, and both the outside layer, and the bottom layer close to the primary, are at low voltage.
  
The winding is now complete. The two cold end wires are poking out on the left, ready to be soldered to stranded wire pigtails.




Version #2 of the transformer is ready. Externally, the only thing that changed is the color of the wire pigtail insulation. Not that I would love pink, but I didn't like that the blue wire insulation melts so fast while soldering!




And here it comes: The result!




But there is still some work to do. The new transformer has too much capacitance. As a result it warms up quickly, and the HV regulation circuit then gets out of range, the HV starts to drop, and the trace gets weak. When cold, it operates just at the limit. So I will have to do a version #3 of this transformer. I think it will be OK to wind the two secondaries using half the bobbin with each, so that each secondary uses 16 layers of half the width, resulting in one quarter as much capacitance as they have now. Another option would be a bifiliar wind, both secondaries at the same time, as was apparently done in the original transformer. That would result in the same reduction of total shunting capacitance, but with lots of capacitance between the two HV windings, which is probably acceptable. But given the added difficulty in making bifiliar windings, I think I will make side-by-side windings.

Also the scope still needs several other repairs, but nothing too terrible.



Transformer #3:  

Finally I wound it!  I reduced the turns number to raise the volts per turn to 1.3 rms. So, what I wound this time is:

Primary:  115 turns for the main winding, plus 57 turns for the feedback winding, #33 wire, in two layers.
Secondaries: 975 turns for the cathode supply, 1050 turns for the grid bias, #39 wire. I wound these side-by-side, in 13 layers each.

The insulation between layers was two turns of masking tape, which is 0.08mm thick. This was done to reduce interlayer capacitance as much as possible.
The filament windings are 1 turn of bifiliar #30 wires, for each filament.

This photo shows the completed grid supply secondary, wound over the left side of the primary.



And now the cathode supply secondary is ready too. 




The two filament windings are wound on top of the cathode secondary, because it has one layer less than the other winding, so there is more room. This photo shows the grid supply filament winding, and the hot side connection of the grid secondary. The other filament winding is faintly visible through the tape.
Note air spaces between both secondaries, and between the cathode secondary and the wires coming out of the windings. Corona is less likely to form around wires separated by air, then if they have lots of dielectric material in between, and some air around the wires!


Then the whole winding assembly was wrapped in NMN insulation material:



This transformer is working fine, with the voltages in the oscillator section being quite normal. There is just one thing I would change: The grid secondary turned out to deliver about 3% too much voltage, so that the intensity control needs to be turned pretty far up for the trace to start appearing. So, if you want to copy this transformer, I would suggest just 1020 turns for the grid secondary, instead of the 1050 turns I used.

I was too lazy to take out the transformer again to remove those 30 turns, even if it would be quite easy. Instead I added a dropping resistor to bring the bias voltage to the correct value.

The air gap of the core is now 0.4mm. I sanded down the core side legs, using a belt sander, to achieve this value.

I think this pretty much completes the job!

Back to the homo ludens electronicus page.