Tektronix 2213A scope - Exploding power supplies!

[magnehh]

This one has been bothering me for weeks on end.

I have an old Tektronix 2213A scope that I'm trying to repair. The second time I powered it on after having had it sit in my storage room for a long time, the fuse blew. I didn't have any 1A slow-blow handy, so I just put in a 2A one. Bad idea! The fuse blew, again, and with it an unknown amount of circuitry in the power supply.

Here is the service manual for the Tek 2215A, which has an identical power supply to the 2213A: http://www.ko4bb.com/manuals/129.241.229.7/Tek_2215A_Service.pdf
Scrolling to page 180 (as seen by the PDF reader), we find the power supply schematic.

This guy has provided us with some troubleshooting tips: http://www.hakanh.com/dl/docs/troubleshooting_tips_on_2200_ps.pdf

CR907 was shorted and badly burned, so I swapped it out with a new one. Put a new 1A slow-blow fuse in, it blew. Then I checked all the other components he mention. For some reason, R909 had way higher resistance than 39 ohms. The color code said 39 ohms, the meter said many kiloohms no matter how hard I rubbed the probes against its legs.That got swapped with a new one. I swapped out U930 for good measure, and either CR908 or CR920. Unfortunately, I can't remember which. It was an avalance diode I believe.

Turns out, the reason the fuse blew now was because rectifier diodes CR902 and CR904 were shorted. So I went and switched all four rectifier diodes with new 1N4007's. No life, not even a blown fuse.

I desoldered C925 and found that it looked bulged on the bottom. So I went and picked up an ELNA cap with the right specs. Still no life.

There is more than 300 volts across C906, so everything before that is likely OK. When I apply 43VDC to test points TP940 and TP950 with Q9070 disconnected, the scope works exactly as it should. So the problem must be between the diode bridge and the test points. I was not able to power U930 and check the waveforms tonight. I will also take Q9070 for a spin on my Tektronix 577 curve tracer sometime soon.

Originally, Q9070 was an MTP6N55, which even had a hole punched through it when it exploded, but I couldn't get it anywhere. So I bought an IRF820 in the hope that it would work. This could be the problem. I won't know until I put it on the curve tracer. I have still to check the windings on T906. If any one of these is not the answer, I just don't know what else could be.

Do you have any suggestions? The circuit took a pretty good load when I ran it with the 2A fuse. I suspect that's when the rectifier diodes died.

At least I learned that there's a good reason one should not use a heavier fuse than specified.

[rf+tech]

Okay, I'll bite. Start by measuring resistance/continuity from the bridge rectifier (junction of CR901 and CR903) to directly on the drain pin of mosfet Q9070. We want to rule out a vaporized track somewhere between the two points.

Next, check resistance from directly on the source of Q9070 to the negative end of C906. This checks the current sense resistor R907 and should be 0.2 Ohms plus residual resistance in your meter probes. Again, we're looking to rule out a vaporized copper track.

Post back what you find and we'll work from there.

RF+ Tech

[magnehh]

Thanks for the tips! Glad I got you hooked. 

OK, so I did as you said. The values were measured with an HP 3456A in 4 wire ohms mode. These are the results:

• From bridge rectifier to Drain of Q9070 = 2.025 ohm
• From pin 1 to pin 2 on T906 = 2.020 ohm
• From negative on C906 to Source on Q9070 = initially 512 ohm, goes down slowly
• From negative on C925 to Source on Q9070 (which I think you meant) = 0.198 ohm
• From negative on C925 to TP940 = 2 milliohm
• All windings on T906 are about 2 ohms or less
• One of my newly bought IRF820 (Q9070) was dead, the other well alive. As seen with ~5-10V D-S, ~4-7V G-S on curve tracer. The dead one was acting like a funky resistor regardless of Gate voltage.
• There's little to no visual indication of broken tracks, except for the pads of CR907, which I messed up trying to solder in a new diode. But the immediate tracks measure OK.

EDIT: Some additional debug info.

• When powering from mains, there is a steady 1.065V from TP940 to TP950.
• The attached picture shows the waveform from Gate to Source on Q9070.
• On pin 5 of U930 there is the same waveform. There should have been a ramp.
• The waveform on pin 10 is nearly identical to that on pin 5. Here we should have had nice pulses as per page 177 in the service manual.

This whole waveform issue could potentially be caused by my "poor man's differential probe" setup. In that case, the interference comes from within the scope.

[oldway]

If

CR907 was shorted and badly burned

, you must also replace C907 1µF.

[rf+tech]

• From negative on C925 to Source on Q9070 (which I think you meant) = 0.198 ohm

Yes, C925 is what I meant.

EDIT: Some additional debug info.

• When powering from mains, there is a steady 1.065V from TP940 to TP950.

This may indicate a problem in the crowbar circuit. Should the voltage between TP940 and TP950 momentarily exceed the zener voltage, Q935 will get triggered. Check resistance across VR935 to see if it has failed shorted. Then apply mains and check crowbar SCR Q935 to see if it is getting triggered and clamping the supply between TP940 and TP950.

Check the pre-regulator output to U930 pin 12 against TP940 reference point. The TL594 datasheet states Vcc can range from as low as 7 Volts to as high as 40 Volts. I would expect to see something like +12 to +18 Volts.

And yes, as oldway states, C907 may have failed shorted and caused CR907 to fail.

RF+ Tech

[magnehh]

It looks like Q935 and VR935 are OK. Also, the scope powered up fine when I applied 43VDC limited to 2A to TP940 and TP950.
From the scope, it looks like the crowbar circuit triggers. After that, there's a small (<0.1V) 50Hz or integer multiple of 50Hz oscillation on the SCR Gate.

Between U930 pin 12 and TP940 there's 0V! VR925 and Q928 seems OK. I think I'll bring Q930 home tonight to test on the curve tracer. Between its pins 1 and 3 there is ~0V forward voltage drop either way. Resistance between the pins 53 ohms either way.

All the resistors around U930 seem to be working fine.

EDIT:
I lifted one leg of C907 and meas'd it. A bunch of resistance and nearly spot on 1uF. Seems to be as alive as anything.

[magnehh]

I think I'll bring Q930 home tonight to test on the curve tracer. Between its pins 1 and 3 there is ~0V forward voltage drop either way. Resistance between the pins 53 ohms either way.

The results are in, and I can tell you Q930 looks pretty much like a resistor.

EDIT: I've got some PNP's with -40V C-E breakdown voltage, as opposed to -60V in the original transistor. Can't see any huge warning signs. Will try it in the morning.
New PNP: http://www.farnell.com/datasheets/1841929.pdf
Original PNP: http://www.onsemi.com/pub_link/Collateral/2N2907A-D.PDF

[magnehh]

OK, making progress! After replacing Q930 with a slightly different transistor, the scope now powers up.

But now, there's a nasty squealing sound. I reckon about 1kHz - 3kHz. The frequency increases steadily. I only let it sit there for ~10 seconds, so I don't know whether it'll disappear or if the scope will blow up again. At least the fuse didn't blow yet.

Any tips? Perhaps I need to change some resistors to match the new transistor?

[rf+tech]

Bias and feedback resistors are used to ensure transistors behave in a predictable and repeatable manner. Bias and feedback resistors remove/minimize variations from one device to the next. If there is a resistor that is out of tolerance, then it should be replaced with one of the same value, to ensure the circuit operates as the designer intended.

The differences in gain between the original 2N2907 transistor and the replacement 2N4403 for Q930 are insignificant and effectively absorbed.

However, the lower breakdown rating of the 2N4403 is a concern and may fail unexpectedly. Yes, it "works" but with reduced operating margin. A PN2907 is the plastic TO-92 equivalent of the metal TO-18 2N2907.

It would be a good idea to measure the voltage now present on pin 12 of U930. And measure the current drawn from the mains. Check waveforms 37 through 44 and apply the finger tip test to see what parts may be heating up (especially Q9070) following the 10 seconds of operation. Sticking ones fingers in a live circuit is to be avoided, do this only after removing mains.

The IRF820 is a poor choice for Q9070 as the datasheet shows Rds_on to be twice that of the original part. An IRF740 would be better suited,

What part number was chosen to replace CR907? The DSR3400X is an ultra-fast 20 ns recovery type. If this was replaced with a 1N4007, then there's your problem.

With device substitution parametrics, the best rule of thumb is to match everything as close as possible. Factors of 1.5 or 2 in the wrong direction should be avoided, when the objective is to ensure post-repair performance to be equal to the original.

RF+ Tech

[magnehh]

By the way, the squealing does not start immediately upon power-on. There's about a ~1 second delay.

• Between TP940 and pin 12 on U930 there's 16V.
• I don't have any tools to measure mains current. Although in a pinch, I could probably make some fixture for it. But it doesn't draw more than max 1A anyway. (mains fuse) Oops, spoke to soon, see last point...
• The diode used to replace CR908 or CR920 was a 1N914.
• The diode used to replace CR907 was the original type, a BYD73G. I'm not sure where you got DSR3400X from?
• I was going to measure waveform 37, but then the fuse blew. I heard a snap before it went, which I assume was the fuse popping. I did use an isolation transformer, so something else than my probe ground lead must have overloaded the fuse.

I guess I just have to unsolder components and test them now... 

[oldway]

Probably T906 (120-1439-00) with internal short between turns or layers.
Sadly, replacement is very hard to find.

[rf+tech]

The parts list shows A1CR907 to be DSR3400X. The parts list and the schematic are my only points of reference. The BYD73G is fine in this application.

The snap heard before the fuse blew would be the familiar sound of a component splitting its case open. The fuse itself makes little audible noise when it fails, and there is some small time delay between the "event" and "flash", as shown by the datasheet curves of overload versus time to open.

A good visual inspection is in order.
Check Q9070 to see if the IRF820 has failed.
Check CR907
Disconnect P9070, replace the fuse, apply mains and check for Vcc on U930 pin 12.
If Vcc is missing, check Q930.
Check CR908
Check Q908

My previous recommendation of an IRF740 is off-base, the Vds breakdown is much too low for reliable operation with 240 Vac mains. I did not have the proper datasheet for the MTP6N55. The manual shows waveform 40 swinging to 175 Volts peak, this would be correct for 120 Vac mains. Since your mains is 240 Vac, and your measurements indicated over 300 Volts on C906, waveform 40 will be closer to 350 Volts peak.

If you get the scope working again briefly, check waveform 40 first thing followed by waveform 41, 39 and 38 in order.

Since the original failure resulted in CR907 shorted, capacitor C907 remains suspect - even though it measures 1 µF. The snap heard may have been an internal arc-over.

RF+ Tech

[magnehh]

I will do these checks tomorrow. But I brought Q9070 home with me tonight, and it is indeed broken, but the package looks fine. The attached picture shows the I-V curve. Of course, these curves should have been your typical MOSFET Id-Vds curves.

Does the value of C907 have to be 1uF exactly? I think I have some old thin film caps laying around, but not sure of the values.

[oldway]

By the way, the squealing does not start immediately upon power-on. There's about a ~1 second delay.

This is typical of a fault T906.
In reality, T906 is an inductor.
When there is a short circuit between turns or between layers, the inductance value becomes very low and the peak current in the MOSFET is very high.

In such a situation the tl594 is current limiting and may oscillate.

This can also happen if there is an overload or short circuit in the 42V but this has already been checked since magnehh had energized C940 in 42V and the oscilloscope had worked normally.

[magnehh]

I ran a check on the coils in T906 prior to this, and they measured about 2 ohms each. What should they be ideally?

And, is T906 identical between Tek 2213 and 2213A? I also have a non-working 2213 laying around.

[oldway]

No way to measure such a fault with an ohmeter, you must use a rlc meter and compare the inductance with a good working T906.
I don't know if the power supplies of 2213 and 2213A are the same.
If they are, yes, T906 are the same, as for 2235 and 2236.
There are two models : 120-1439-00 and 120-1439-01.
The last one seems to run a little cooler.

This inductance operates at high temperature and I put a fan in my 2235.
It's easy, just look at how 2236 is mounted because it has a fan.

[magnehh]

Thanks for sticking with me here! Here's what I've got.

• Q9070 failed, as previously stated.
• CR907 is OK.
• CR908 is OK.
• CR920 is OK.
• Q908 seems to be OK, testing the pn jct. with multimeter, but I replaced it with the same as Q930. I'll bring the old one home and test on the curve tracer anyway.
• I'll have to wait until I get new fuses from DigiKey before I can try powering it on again. Unfortunately, I don't have any sort of current limited AC power supply.
• I replaced C907 with a similar cap from a Tek 2213 (also broken). So there's no guarantee it's OK, but we'll have to see.
• I desoldered T906 from the for-parts 2213, and measured the windings with an LCR meter. Then I measured T906 of the scope I'm trying to fix, in-circuit. They measured almost exactly the same. Between pins 9 and 10 there's very little inductance, but I believe that is to be expected.
• Q930's junctions seems to be OK, measured with multimeter.

I'll order a Q9070 that is rated for 400V or more from DigiKey. If I can find it, I'll order a new C907 as well. And about 15 fuses!

[rf+tech]

Just one comment:

I'll order a Q9070 that is rated for 400V or more from DigiKey

Using a 400 Volt breakdown part operating from 350 Volts leaves very little derating margin. The original MTP6N55 has a 550 Volt breakdown rating and a healthy margin. Reliability is proportional to derating margin. See if Digikey has the IRFBC40PBF-ND in stock close to you.

The incandescent-lamp-in-series-with-the-mains trick may be of some benefit. In fact, lowering the mains voltage to 120 Vac with a variac or step-down transformer, might reveal something. The TL594 will respond by widening the pulse width to the mosfet as the drain voltage will be half.

This is certainly an intriguing problem.

RF+ Tech

[magnehh]

I'm ordering 3 of those FETs now + 10 fuses. Never know what else might be wrong.

My buddies are asking me if I'm really still trying to fix that old scope. But you know, I can't give up now. Too much time has gone into it to do that.

I'll try the light bulb trick if I can find one. Those bulbs were outlawed here in Norway two or three years ago.

Lucky me found two more fuses. My isolation transformer can be wired to step down the voltage by 30-50V, so I'll try that today. The old Q908 seems to be cold dead. I'll salvage an IRF820 from my other 2213 and hope that the lower mains won't kill it right away.

[magnehh]

It works! Or it did for a couple of minutes anyway.

This is what I did:

• Replaced the fuse and applied about 200VAC mains.
• With Q9070 disconnected, I measured waveforms 37 and 38, see attached pictures.
• Turned off and connected Q9070
• Turned scope back on, it works!
• Then I think my scope probe moved a little and shorted Gate and Drain on Q9070, which killed the fuse.

So I guess I'll have it working when I get new fuses and that MOSFET you recommended. Maybe C907 was bad all along, I guess we'll never know. Q908 could have been the fault for the last few tries, but I'm pretty sure it was OK at one stage.

Anyway, I'll keep you posted on how things are working out, when I receive the replacement parts. Thanks again guys! You've been a great help.

EDIT: I'll probably replace Q908, CR908 and R909 for good measure. I'll also check CR907 and R907. Then I'll check waveforms 37 and 38 before I connect the new Q9070.

[rf+tech]

Congratulations magnehh, on a job well done. 

I've had my share of pr bing err  s, it's just part of the learning curve.

RF+ Tech

[magnehh]

So here's the result after using it a bit the last two weeks: It works beautifully!

I did all the stuff I said I'd do, and nothing flashed, smelled, got excessively hot or otherwise didn't work. So no problems!

To all those who find this thread and also have a broken scope, know that the power supply of the Tek 2213A and 2215A are more or less identical. I had a really hard time getting hold of a Tek 2213A service manual, but the one for 2215A is readily available for free.

Thanks again!