Tek Model 464 repair

[bd139]

Good advice. I usually spend as much as I paid for the scope on the bloody things.

[GeorgeOfTheJungle]

[Ero-Shan]

Have you got an IR camera? If so, turn it on and have a look at each and every tantalum first. If not, have a look at each and every tantalum first :-)

It seems to me that tantalum capacitors are not held in the best esteem around here. 
I actually didn't have a single bad tant in many years. Shorted, leaky and lost-their-capacity electrolytics, yes.

When I resurrected that scope, I was actually thinking: Tek + not working = Tant. My approach to troubleshooting has always been: Don't assume too much, measure. Be systematic. Since all main supply rails are fine, there's certainly no shorted tantalum cap there. And there are no tants in the vicinity of the HV generator, either.

The tip with the IR cam is a good one (and I do currently have such a camera). On a day like today however, everything is hot.

[GeorgeOfTheJungle]

Mind you, I've only fixed one 465, and had no luck there with the "tantalums trick" either. But I've managed to fix quite a few MacMinis with this method: it's always the same tantalum (!) and you can see it very clearly with the Flir, as soon as you attempt to turn it on, it lights up very clearly on the screen.

[Ero-Shan]

That's a classic.

Wonder if it helped piling up all my components on a desk. 

Hmm interesting one. This is typical tek though.

If it’s not working in situ it’ll be in the feedback loop from the HV. On a couple of occasions, both 465b’s, I have broken that loop and hooked it to a bench supply to see if it had any kind of regulation. Turned out to be a resistor in one of the divider chains was out of spec so it stopped the whole oscillator dead.

Driving the feedback at the divider (R1525 C/D) should allow assessing the regulator. I've considered it healthy so far, as the missing HV lead to a positive bias voltage for the switching transistor (where the schematic says -3.8 volts). But I might have missed something obvious. Wouldn't be the first time. (Pulling the transformer out and still having to put it back in may be my well-deserved punishment, then. If it was so.)

“Working” is a very narrow margin between two posts of failure on these HT oscillators.

Edit: meant to say I like your HV probe. I bought a big Tenma “anal probe” out of fear in the end. I think yours is excellent for such low volume repair jobs. Maybe I need to cure the HV fear

My fear involving these 'weak' high voltages is more of inadvertently damaging things with them, or secondary damage caused by an unexpected jolt.

[GeorgeOfTheJungle]

My home made HV probe is just a long string of (IDK)MΩ resistors in series inside a heat shrink tube, and works well, helped me get the 465b back to life This 465b:

[David Hess]

Tektronix used more than 3 different 2N3055s and graded some of them.  If you use the wrong one, then the high voltage oscillator may either not start or suffer from spurious oscillation.

The 151-0140-00 is the 0.3MHz low hfe variation.  The closest modern part is the 2N3772G.

My Siemens 2N3055 seems to work all right. If they use a selected component in the high precision circuitry of a HP 3458A I'm all with them, but depending on it in a simple dc/dc converter doesn't get my blessings.

Saturation of the transformer is suppose to control the oscillation frequency but my understanding is that there are issues with LC resonance in the transformer which complicate things so the circuit can be picky about the output transistor gain and bandwidth.  Tektronix used this design for decades and finally updated it in the 455/465M oscilloscopes which use a fast high gain ring emitter type of transistor although I am not clear about what they changed to allow this without problems.

Checking the low voltages at the error amplifier is usually enough to tell what the inverter is trying to do and what the condition of the high voltage side is.

Sure. It tells me there is no high voltage. I still don't know why, however. I don't know how much my body's capacitive load was (and probe to fingers), but did it amount to more than, say 20 µA, of beam current?

Remember I isolated pins 8 and 9 already while it was still built in, and it looked totally overloaded. So I took a closer look at CR1512, C1512, CR1514 and C1514, but they all appear to be perfectly healthy. Judging from my external hookup, I'd say the transformer is also good, but with much less confidence.

The most common failure is a short in the high voltage multiplier.  Next is a shorted high voltage decoupling capacitor.  Unfortunately a low voltage test may not reveal either of these.

The high voltage multiplier can be tested simply by disconnecting either its power or ground inputs.  The CRT will work without its PDA (post deflection acceleration) voltage applied although it will be dim, fuzzy, and have about half of its deflection sensitivity.

I always get suspicious if a secondary side short does not blow the primary side fuse and one of the things I would definitely do while diagnosing an apparent short in the high voltage inverter is replace the fuse with or add a constant current limit.

[Ero-Shan]

The most common failure is a short in the high voltage multiplier.  Next is a shorted high voltage decoupling capacitor.  Unfortunately a low voltage test may not reveal either of these.

The high voltage multiplier can be tested simply by disconnecting either its power or ground inputs.  The CRT will work without its PDA (post deflection acceleration) voltage applied although it will be dim, fuzzy, and have about half of its deflection sensitivity.

I think I can preclude that. Thad had occurred to me as a likely cause, also. But disconnecting the HV multiplier and the cathode rectifier, leaving both pins 8 and 9 open, did not change anything.

I always get suspicious if a secondary side short does not blow the primary side fuse and one of the things I would definitely do while diagnosing an apparent short in the high voltage inverter is replace the fuse with or add a constant current limit.

That is a very good suggestion.

[Ero-Shan]

A quick update. I have now replaced my 2N3055 with Tek's original in the HV generator, and it works even better. Allows me to draw some 1 mm sparks from pin 8. Reminds me of the glory days when we drew several centimeters from the 25 kV of color TVs.

I think it is time to check the regulator. Maybe I shouldn't have ruled it out so quickly.

[Ero-Shan]

The regulator appears to be working just like it is supposed to, from what I understand.

So next I try to run the oscillator closer to its original environment. The only windings of the transformer used are primary and feedback (and an AC voltmeter at heater). The regulator is still driven by a PSU.
With expectations running high, I switch on the power. No oscillation. I vary the input to the regulator, which has a profound effect on the collector current of the driver: It's either in the vicinity of 0.45 amps or around 72 mA. But it does not oscillate.
Next I pull out Q1484. No change. I use the exact same 47 nF capacitor I used on the breadboard, a 5.6 k\$\Omega\$ resistor and drive it by my PSU, exactly like I did yesterday. The collector current stays around 72 mA (didn't measure explicitly yesterday, but the PSU showed 0.08 A), but it does not oscillate.
The remaining difference is the 22 volts supply, the inductor (I had only 56 µH), the 47 µF 'lytic. Added my 47 µF from the breadboard, just to make sure (I had had measured Tek's cap and it had shown good). No change.

This is ridiculous. The circuit is now exactly the same as yesterday, where I had rock solid oscillations. I better call it a day. 



With the original fan removed, I also had to create some extra wind - at the current temperatures it seemed very necessary to me. One more hindrance.

[David Hess]

If the collector current is low with no output then there has to be something wrong on the primary side; the error amplifier should be pushing the 2N3055 hard producing a high collector current.

[Ero-Shan]

If the collector current is low with no output then there has to be something wrong on the primary side; the error amplifier should be pushing the 2N3055 hard producing a high collector current.

It is. I just squinted when I saw it and drove the error amp down. Then I took it out of the loop completely and set the bias with the PSU through the 5.6 kΩ resistor. To no avail. It cannot be persuaded to oscillate any more. Tomorrow I will go back to square one and try it again on the breadboard - with the scope a few feet away.
There must be a reason for what I'm seeing. I just can't seem to find it.

[David Hess]

I hate to say it but it is not unknown for the high voltage inverter transformers in these to fail.  But your earlier test indicated that it was good.  Still, be careful about overdriving the transformer and damaging it.

[floobydust]

I say don't run the transformer without a load, unless you want to make a Tesla coil. It's too hard on the old insulation. A junkbox CCFL lamp makes a reasonable 1kV (bi-dir) zener plus a ballast resistor, for a dummy load.

We might be oversimplifying the DC-DC as a one-transistor oscillator. Would Tek really use that here?
How Tek tuned this resonant to get sine verses a noisy flyback converter, I'm not sure.

For this circuit, there might be AC (ripple) as part of the feedback signal which helps the oscillator.
So not a strictly DC error voltage going into Q1484. Any open capacitor like C1455, C1472, C1494 could cause problems too.

I'm just rambling because this blaming the transformer probably leads to a sad ending for the scope.

[David Hess]

We might be oversimplifying the DC-DC as a one-transistor oscillator. Would Tek really use that here?
How Tek tuned this resonant to get sine verses a noisy flyback converter, I'm not sure.

Tektronix used this high voltage inverter design from the beginning in their tube oscilloscopes up through the 400 series portables and the 7000 series mainframes that had linear power supplies.  Some of the 7000 series used a more well known two transistor push-pull variation.

For this circuit, there might be AC (ripple) as part of the feedback signal which helps the oscillator.
So not a strictly DC error voltage going into Q1484. Any open capacitor like C1455, C1472, C1494 could cause problems too.

AC feedback for sustaining the oscillation is exclusively through the transformer and not the error amplifier.  The oscillator operates at the resonate frequency of the transformer.

[floobydust]

I remember seeing a Tek service manual showing a shorted-turn on the HV DC-DC transformer to set the resonant frequency. I'll dig for that model number. You think that is the case here? I've never understood shorted turns, they always baffle me.

Point is OP might not be able to test his transformer here. Looking for crack in the core or a bigger airgap if the clip/adhesive moved, but an L meter on a shorted-turn transformer where we have no number as a reference, kind of a guessing game if a winding is at fault.
Could change gain with a different 2N3055, the part appears to be vanilla but antique specs of fT=0.3MHz

[David Hess]

I remember seeing a Tek service manual showing a shorted-turn on the HV DC-DC transformer to set the resonant frequency. I'll dig for that model number. You think that is the case here? I've never understood shorted turns, they always baffle me.

As far as I know the 400 series did not use any shorted turns on the high voltage transformer unlike the 500 (starting with the 535 or 545?) and 600 series tube oscilloscopes.

There are a few discussions about the purpose of those shorted turns in the TekScopes@groups.io archives.  One reason was to balance the flux through the E-cores that had asymmetrical windings for supplying heater power to the high voltage rectifiers.

[floobydust]

Ah yes, I found it- "flux balance winding" in the TM500 series (i.e. SC-502, SC-504). I'm mentioning it I thought it might be in a 464 as well, as the filament was later changed to DC power.
I think the only method left is to connect a signal generator to a winding and look at the waveforms.

Sphere has a couple 120-0909-01 transformers USD$65

[Ero-Shan]

Ah yes, I found it- "flux balance winding" in the TM500 series (i.e. SC-502, SC-504). I'm mentioning it I thought it might be in a 464 as well, as the filament was later changed to DC power.
I think the only method left is to connect a signal generator to a winding and look at the waveforms.

Sphere has a couple 120-0909-01 transformers USD$65

I'm not going to spend that amount of money on this scope that I never liked all that much. I have enough working scopes. And not just scopes. It's generally "either I can fix it cheap (mostly with components I already have) or forget it." There are exceptions, of course, like my HP3458A. Getting gear repaired is fun and satisfactory, and I'm willing to invest considerable time and energy.
But thanks anyway. If 464s all of a sudden become admired as the coolest scopes ever built, I might reconsider.

Had to change my plans for yesterday, soldering iron stayed cool (i.e. around 30°C these days). Before going back to square one, I will try one other thing: power the oscillator from my lab supply with 22 volts. I've even got new 12 kV rectifiers (2.89 a piece is OK, and they can be used for other scopes as well).

[Ero-Shan]

So I tried my interim step, not hoping much, not getting anything.

Back to square one it was. Everything exactly as on Friday. Except that the [expletive deleted] wouldn't oscillate any more. The collector current corresponds nicely with the bias voltage, showing a DC current gain of about 70. Have I somehow killed the transformer?

[David Hess]

Since you had the transformer out, I would try testing it with a function generator.

[floobydust]

I think if you ran it stuck at 0.45A for long time, it might have overheated the primary winding.

If you're sure the circuit has changed- you have the phasing correct still? Are you running it on the bench.
The feedback winding is biased at a negative voltage, so a lone 5k6 could be the problem for startup.

Transformers have several resonant frequencies and I'm surprised there's nothing to stop the leakage inductance from dominating.

[Ero-Shan]

I think if you ran it stuck at 0.45A for long time, it might have overheated the primary winding.

If you're sure the circuit has changed- you have the phasing correct still? Are you running it on the bench.
The feedback winding is biased at a negative voltage, so a lone 5k6 could be the problem for startup.

Transformers have several resonant frequencies and I'm surprised there's nothing to stop the leakage inductance from dominating.

The 0.45 A lasted all but a few seconds, then I pulled back.

I've checked the phasing several times, as that might have been the most likely blunder.
The 5.6 kΩ resistor's bias comes from a PSU, so I can easily modify it. Last week the damn thing started with robust oscillations at around 3 volts bias. Now even 6 volts don't do anything apart from increasing the DC current.

I will definitely check with a function generator also.

[Ero-Shan]

I have now driven the HV transformer with a signal generator. The resonance was at 44 kHz, and the ratio of all voltages was correct. Except the high voltage ones, which I couldn't measure - even my home-brew probe was too much of a load.

I also reversed the polarity of the feedback winding in the breadboard circuit, just to make sure. As expected, it didn't change anything. Guess I'll try with a different 2N3055 once more. All components seem to be OK, it just doesn't work.

[floobydust]

How frustrating, I would be tempted to increase gain, try a higher hFE transistor or Darlington for the experiment. Or a few 10's pF C-B if leakage inductance is a problem. I would just mess around and try things to get it to start.

Grasping at straws, I thought the 2N3055 might see avalanche if the oscillator ran into a heavy load.
I recall older power BJT's (but epitaxial vs homotaxial) could take more abuse, spikes way over the VCEO 60V rating.

In your breadboard circuit, C1483 0.05uF I think important for crisp switching and it looks like Q1484 runs about 1mA plus R1483, for bias, that looks like a couple mA.

The only other straw I have is the secondary load might be part of the resonant tuning.