Kurt's Tektronix Page

I like tools. Good tools go hand in hand with the desire to build things and fix things. When it comes to building and fixing electronic things, some of the most important tools are the measurement instruments. Measurement instruments let you know what's going on with your circuit. My favorite type of measurement equipment is the oscilloscope. A good oscilloscope provides a lot of insight into what is happening in a circuit. My favorite oscilloscopes are the analog scopes made by Tektronix. I own several Tektronix service manuals which contain complete schematics. Most of what I know about electronics comes from studying Tektronix schematics and trying to understand exactly what the designers did and why they did it.

Many Tek manuals can be downloaded for free from various websites including BAMA. However, I recently was fixing a Tek 464 and could not find the schematic online. Eventually I found it. Here's my copy: Tektronix 464 Service Manual I would like to point out, however, that there is a DC voltage annotation on the high voltage power supply schematic seems questionable. On the left side of page 267, the voltage on C1455 is marked as 0.4 volts, but it seems unlikely to be correct. I recommend ignoring that annotation.

Here is a nice circuit that is used for overload protection at the input of the Tek 7A29 vertical plugin.

This is a 50-ohm input section with DC-1GHz frequency range. The four diodes at the front limit the current that can flow in or out of the amplifier. Since the current sources are set to 17.4ma and the input impedance is about 50 ohms (terminated), the base voltage of the input transistor is prevented from swinging beyond +/- 870mV. The circuit is very simple, yet it does exactly what is called for. It has very low resistance when it is not being overloaded, but becomes an open circuit when the input is too high. The diodes can be small since they will never have much current passing through them. Therefore they do not introduce a lot of parasitic capacitance to ground in normal operation. The circuit behaves a lot like a fuse, but a magic fuse that automatically recovers after the big input is removed.

Tek 7704 power supply
Tek 7704 right side

It is nice to be able to see a waveform in the time and frequency domains together.
I used to work at a chip company and I was spoiled by using an HP 4566B spectrum analyzer. The 7L12 spectrum analyzer plugin for the Tektronix 7000 series is a toy compared to the HP 4566B, but it is still quite useful, particularly when playing with oscillators.

I have a Tektronix 661 scope. The manual for the mainframe is on BAMA. The manuals for the plugins are on Tek Wiki

analog_input -> analog_shift_register -> CPU -> DACs -> X-Y_display

Oct 28, 2007: The problem with the 7D20 is confusing me. It exhibits the same symptom in my 7704 mainframe and in my 7603 mainframe, with the screen being shifted down and compressed and fuzzy. But once in a while (about 1 out of 10 times), when I turn the scope on, the image is perfect and the 7D20 works perfectly, except for the bad EAROM, which doesn't affect normal use. It will continue working perfectly until I turn the scope off. Turning the scope off for a second and then back on puts it in the non-working state. Bumping the machine when it is in the working state will not put it into the non-working state. Bumping the machine when it is in the non-working state will not put it into the working state. Since the signals that are being corrupted are the X-Y voltages that are used to paint the display with trace and on-screen text, I used another scope in X-Y mode to monitor the X and Y voltages in the 7D20 as they go from the X and Y DACs toward the scope mainframe. When the scope is in the corrupted, non-working state, the other scope still shows a nice clear image on its screen. Only the 7D20's host mainframe shows the corrupted image. As I moved the probe points closer and closer to the mainframe connector, I found that the probed signal remained good. Now here is the crazy part. When I probe the signal in the mainframe, where the plugin connector meets the scope's interface board, the signal still looks good! This is really hard to understand how this can be. I confirmed that the vertical signals (A11 and B11) are operating in proper differential mode. v(A11) + v(B11) = approximately zero. v(A11) - v(B11) = 2 * v(A11). So it looks like a good signal is being fed to the scope but the scope is displaying a distorted image. We could easily infer that the mainframe is not working. However, the 7D20 does the same thing in two different mainframes. And both of those mainframes work correctly with other plugins that go into the same connectors.

A guy on the TekScopes mailing list told me how to get rid of the EAROM problem. One just needs to store anything to each of the memory locations and then the contents will be valid and no error will be reported.

Anyway, I bought a 519 from a nice guy in St. Louis. Before he sold it, he powered it up and posted pictures of it working. When I got it, I fired it up on a variac. It worked for a couple of minutes and then the high-voltage supply died. I am in the process of troubleshooting it. It is full of black beauty capacitors which are, of course, prime suspects. The transformer looks good, but it's hard to tell. The power oscillator isn't oscillating. It seems unlikely to me that problems on the secondary side of the transformer would cause it to stop oscillating. My guess at the moment is that the cause of the problem is the failure of one of the capacitors in the oscillator. I checked them for small-signal capacitance and low-voltage leakage and they are all fine, except for the expected horrendous drift in capacitance, upwards usually. Still, there are many ways that a capacitor can be bad, while still having the correct capacitance and low DC leakage.

Feb 22, 2008: OK, I got the 519 to work. I replaced all of the black beauty capacitors in the HV power oscillator and it still wasn't oscillating. Finally, even though the 6AU5 tube in the oscillator had a working filament and had reasonable plate current, I decided to steal a 6AU5 from my 545B and try swapping it. Yes! The original 6AU5 was bad. After swapping it, the HV power supply came to life, the neon bulbs lit, and there was a nice trace on the screen. However, the trace didn't respond to the input signal. How could it be? There is no vertical amplifier to go bad. Eventually I determined that my BNC-to-GR adapter was getting a bad connection with the GR plug in the front panel of the 519. When I fed the signal to the center pin of the input connector, it worked. As I looked at the input connector, I noticed that it looks different... Knowledgeable people informed me that there are 125-ohm GR connectors and the 519 uses them, not the standard 50-ohm GR connectors. I had no idea. It makes sense, though. A 50-ohm connector in a 125-ohm system would be a very bad idea. It is great news that the 519 works now. The only thing to do now is to figure out why the triggering is a bit jittery. It does trigger, but there is about 1mm of jitter.

May 16, 2008 I ran across a Nelson-Ross 002 spectrum analyzer plugin for 500-series scopes. Here are some pictures: front, bottom, bottom. I put it in my 547 and fired it up and it does not appear to work. I have used various other spectrum analyzers, so I don't think that the problem is merely my own stupidity. Among other things, it does not generate a sweep signal. It has a wire coming out of the front panel that ends in a banana plug that seems like it should go to the external horizontal input so the spectrum analyzer can use the scope in X-Y mode. I probed that signal with another scope and saw nothing. I'm guessing that there is a ramp generator inside the plugin and that ramp drives a VCO. I'm guessing that the ramp generator is not working. Unfortunately the manual is not on BAMA. Most likely I'll either have to trace the wires and make my own schematic or buy the manual.

May 18, 2008 The same guy who sold me the Nelson-Ross 002 also sold me a Tektronix 555. I was hesitant to power it up because I feared that the electrolytic capacitors would be shorted from years of not being turned on. I used a 0-500V bench power supply to test each of the large electrolytics. This is not nearly a complete test. I just verified that each capacitor could be maintained at the normal operating voltage without any significant DC current. All capacitors were fine. I tried to power up the power supply without it being connected to the scope. That doesn't work. After I plugged in into the scope and powered it up, I was pleased to see that the 555 was in pretty good shape. The lower beam wouldn't do anything, which I traced to a bad black beauty capacitor in its high voltage supply. I replaced it with a 1kV ceramic capacitor from my junk drawer. The lower beam works now. But I smell something hot after the scope runs for a couple of minutes. Each time I smell it, I power down immediately. I will do a high voltage test on all of the capacitors in the lower beam HV supply, including the ceramic cap that I put in there. My current hypothesis is that some capacitor is passing DC and the smell is either coming from the heat dissipated in the capacitor or the heat dissipated in some series-connected resistor that is not designed to have standing current flowing through it. The post-deflection accelerating voltage in a 555 is generated by the lower-beam supply. So the brightness of the upper beam really suffers without this supply working. Interestingly, you can still sort of use the top beam without the lower-beam supply working. It's just dim. Removing V900 (a 6CZ5 beam power tube forming a power oscillator) effectively shuts down the lower-beam supply and the scope runs without the burning smell.

May 19, 2008 The lower beam HV problem remains, but in the meantime, I noticed 2.7 volts of ripple on the 500 volt supply. The problem is that C760, a dual 40uF capacitor, has lost most of its capacitance. I left the capacitor in place, but electrically disconnected it, and mounted two 50uF electrolytics on a little piece of plexiglass using hot glue and screwed the plexiglass to the chassis of the 555 power supply using an existing screw. Now the ripple is below 3mV. There was time-domain blur before and now it is gone. I can't remember if the blur got worse from left to right across the screen. Maybe the ripple on the 500 volt supply was affecting the triggering somehow, which would cause uniform blur across the screen, or maybe it was affecting the Miller runup circuit, which would cause increasing blur from left to right. The ripple was full-wave rectified 60Hz power line. It is interesting to think about the manifestation of a "modulating" signal on the sweep rate.

By the way, I have a Tek 155-0147-02 chip from a 1470 NTSC generator. If anybody needs it, it's yours for the cost of shipping, or zero if you're in NYC.

May 20, 2008 I played with the 555 for a little while this morning. The lower-beam supply acts strange. For one thing, the HV pot has no effect on the cathode voltage. The upper-beam supply behaves normally. However, the lower-beam cathode voltage is in the right neighborhood, around -1380V. And the post-deflection acceleration voltage is around +6000V, which seems reasonable. But there's no lower-beam trace. I verified that the lower-beam filament is good, and I swapped the horizontal plugins. My guess is that there is a problem with blanking. The burning smell has decreased or even stopped. Maybe something in the blanking circuitry was burning and has finally burnt out. Of course there could be a fault in the elaborate distributed vertical or horizontal amplifiers, but if that were the case, I'd expect to see some vague patterns on the screen, not just darkness. So my current hypothesis is that the grid-to-cathode voltage is holding back the electrons. This will be verified with my multimeter, floated to the cathode voltage. This is one reason why I prefer my battery-powered multimeter to my bench multimeter.

May 21, 2008: The 555 HV supply is better now, but still not proper. The HV ADJ control still doesn't have any effect. But at least both beams are working now. It's a bit hard to believe, but it seems that V962, the 5642 high-voltage rectifier tube for the lower-beam cathode failed during the few days that I've had the scope. It must be caused by me but I don't see how. My guess is that C931, C943, C944, or C948 was intermittently passing a bunch of DC current, and this was overloading the little tube. The filament glows and it doesn't look gassy, but it fails to rectify. I borrowed a 5642 from another working scope (I dislike doing this.) and installed it in place of the bad V962, and the cathode voltage came back. But I don't want to kill this good 5642 tube. I will figure out a safe way to continue debugging without endangering the rectifier.

By the way, note what happens if V962, V914, or any of the associated circuitry is not working. The HV supply will be out of regulation and will run at full-throttle.

I replaced some more bad components and got things closer to how they should be. Still, the CRT cathode voltage is a bit weak. Here are some notes about the feedback that controls the HV supply.

V914 is a 12AU7 dual triode used as a feedback amplifier for the HV supply. The grid of V914B is compared with the cathode of V914B, which is -150V. A voltage difference near zero will result in about 200uA of plate current, which will pull the grid of V914A down to about -14V, which will cause 3mA of plate current and will put the plate at +260V. This is the second grid voltage for V900, the 6CZ5 power tube that forms the HV power oscillator with transformer T901. So, for the same reason that one makes the "virtual ground" assumption when analyzing op-amp circuits, the analysis (and troubleshooting) of the 555 HV supply can start at the grid of V914B. If the system is operating in the proper closed-loop mode, we expect to see -150V on the grid of V914B. This is a fork in the road. If there isn't a constant -150 volts on the grid of V914B, we have to ask whether the HV oscillator is oscillating at all, or trying to start. Relevant issues will be whether the HV transformer is good, whether the oscillator tube is good, whether the tube is getting the right DC voltages to start oscillating, and whether there are bad passive components, particularly capacitors in the circuit. In the case of the HV supply of my 555, the grid of V914B is a constant -150 volts. So the question is why CRT voltage is wrong while the feedback system is in a steady state equilibrium. The design of the HV supply uses something like a 9:1 voltage divider on the CRT cathode voltage to get the grid voltage for V914B. If the resistances in the divider drift, the divider might be 10:1 or 8:1, and one would expect to see 10*(-150V) or 8*(-150V) on the CRT cathode.

Indeed, drift in the resistors in the voltage divider was responsible for the error. R950, an 820K resistor, had drifted to 920K. I replaced this, but it didn't completely fix the problem. R952, a 1Meg pot, had drifted to 1.26Meg. I shunted the pot with a 5Meg resistance to correct this. R956, a 2Meg pot, had drifted to 2.5Meg. I shunted it with 10Meg to correct it. The result of this sort of shunting is not quite the same as if the pot were correct, but if the wiper current is very low, the difference will be small. After making these three changes, the HV ADJ control works correctly and I set the CRT cathode to -1350 volts, as it should be.

May 22, 2008: I traded a DEC PDP11/23 for a non-working Tektronix 310A. The 310A would blow the fuse at power-up. I suspected shorted filter capacitors in the power supply, but this was not the problem. The problem was a bad 6AQ5 tube in the HV power supply. I happened to have a 6005 in a box of old tubes, and it works well in place of the 6AQ5. The CRT cathode voltage is correct now, without any adjustment necessary. Now the 310A works perfectly except for the calibrator which puts out a low-amplitude, high-frequency waveform instead of a low-frequency square wave.

June 1, 2008: I fixed the calibrator in the 310A. It was a bad tube. The circuit calls for a 12AU7 and the closest I had was a 12AT7. It works fine.

June 2, 2008: I never was able to get my Tektronix 661 sampling scope to work. (Update: It works.) Recently, I bought another 661, not working and missing its timing plug-in. My hope is that from the two non-working 661's, a single working 661 can be built. Just today, I put the new 661 on my bench to take a look. It is certainly dirty. So far, I have confirmed that the main power supply works correctly, the cathode voltage is good (no PDA), and the CRT works. There appears to be a problem with triggering, timing, and/or blanking. The screen is completely dark in what should be normal operation.
June 3, 2008: I got the "new" 661 to work. There was a bad tube in the 4S1 sampling plug-in. Debugging the 4S1 is not easy, but fortunately nothing was wrong with any of the exotic components. Here are some photos of the 4S1 schematics. I feel that the best way to debug the 4S1 is to leave it in the 661 and lay the 661 on its side with the bottom cover removed. That way you can easily probe the inputs and outputs of each module on the 4S1's interconnection network. The manual provides examples of what the signals are supposed to look like. I have a 5T1A and a 5T3. The 5T3 works only at high sweep rates. The 5T1A doesn't work at all. I'm glad I have a working configuration to start from. Next I'd like to get the 5T1A working. I'm not intimidated by a half-dozen tunnel diodes.

June 4, 2008: The Tektronix 661 is really a nice machine. It appears to have been designed in 1961. The scope takes two plug-ins, one for the time base ("timing") and the other for the vertical inputs ("sampling"). It is a high speed sampling scope, so it is somewhat more complex than the average scope. The external styling and internal construction resemble the 500-series scopes like the 545 and 547. The mainframe provides regulated DC voltages to the plug-ins, and provides them with a low-speed X-Y display. Unlike most Tek scopes, there is nothing tricky about the CRT. Basically the scope works as follows. The signal comes into the sampling plug-in. There is a trigger pick-off that sends some of the input signal to the timing plug-in through a coaxial cable that is part of the mainframe. The timing plug-in has triggering circuitry and sends a signal back to the sampling plug-in, telling it when to sample. The sampling plug-in does some signal conditioning to the output of the sampler and sends this to the Y amp of the display. The timing plug-in sends a slow ramp signal to the X amp of the display.

The mainframe is actually quite a bit simpler than a 545 or 547. Failures, therefore, tend to be in the plug-ins. Tek made extensive use of fancy diodes to do the high-speed work. These fancy diodes seem to be the most common cause of a 661 not working. At first, it seems like the sampling plug-in and the timing plug-in are in a very intimate embrace, and it seems like it would be hard to determine which one is at fault. However, this is not the case. As I mentioned before, probing the sampling plug-in from the bottom gives you access to many of its internal signals while it is running. It is possible to probe most of the timing plug-in's signals by removing the side cover of the 661. So, for example you can determine whether the sampling plug-in is sending the timing plug-in a signal. If it is, you can check whether the output tunnel diode in the timing plug-in is firing. If not, you have restricted the problem to the front end of the timing plug-in. If so, you can check the signal that the timing plug-in is sending back to the sampling plug-in.

A common cause of failures of sampling plug-ins is the sampling diodes. They are small and have low capacitance, so they are vulnerable to overload and ESD. Their forward voltage should be around 400mV. Blown diodes may show higher voltage on a diode tester. Also, there are Nuvistors in the sampling plug-in. Nuvistors sometimes fail. Fortunately, many of the components can be swapped back and forth. So if you have one working channel, you can swap parts to figure out what's wrong with the second channel. Keeping notes is recommended. If all of your sampling diodes are dead, don't worry, there are modern replacements. Dead tunnel diodes are a more serious problem.

June 12th, 2008: I ordered some HP Schottky diodes to replace the blown GaAs sampling diodes in my 4S1 and other blown samplers that I have. They come in pairs, and HP says that each pair is well matched. I am a bit unclear on the matching criteria for sampling bridges. It is clear that the left half and right half should be matched, but it isn't obvious to me why top-to-bottom matching is required.

I scanned some manuals. They are on my manuals page.

June 17, 2008: The Schottky diodes for the 4S1/661 haven't come yet, so I've been playing with other things. I've been trying to get my HP-85 computer to communicate with my Tektronix 7912AD via GPIB. Let's see, I'm not sure whether the 7912AD's GPIB works. I'm also not sure whether the HP-85's GPIB works. And I don't know anything about GPIB. I am able to get the 7912AD to switch back and forth from local to remote mode via GPIB, but when I try to read data from it, nothing happens, and my I/O procedure times out. The HP-85 works by itself, and the 7912AD works in TV mode. Most likely the problems is that I don't know anything about GPIB.

June 28, 2008: I am fortunate enough to own a Tektronix 519. Here is a photo of the delay line in my Tektronix 519. For the purpose of comparison, here is a photo of an automobile supercharger.

Oct 24, 2008: I've been using my 547 lately as my main bench scope. From the time I bought it until today, it had a blanking problem with the A timebase. The cause of the problem was a bad PNP germanium transistor, a 2N2207. I replaced it with a 2N3906 and it works fine now. Other than that, my 547 is in pretty good shape. One of the triggers is a bit fussy. I'll see what I can do about that.

Oct 26, 2008: I calibrated the triggers and sweeps and my 547 is really working well now. The cause of the fussy trigger was actually a dirty switch. I used contact cleaner on all of the switches and worked them back and forth. After that, the trigger adjustments helped a bit more. Finally, I used my time mark generator to calibrate the sweeps. They were running a bit fast on some sweep rates.

Dec 6, 2008: I started a wiki for classic Tektronix equipment.

Oct 8, 2009: Today, I fixed a Telequipment D54. The scope was making arcing noises and the brightness control wasn't doing anything. The 100V zener diode in the CRT cathode supply was shorted. I replaced it and the scope works. The construction of the D54 is not friendly to a person making repairs. Removing the PCB requires removing controls from the front panel. An alternative is to cut the wires of the old component and to solder the new component to the stubs that remain. The D54 uses an unregulated power supply. They play some games with voltage-controlled amplifiers to keep the gain stable across supply voltage variations. It doesn't really work. When I play with my variac, I see significant changes on the screen. That doesn't happen with, for example, my 547.

Oct 9, 2009: Due to lack of participation, I consider my Tektronix wiki a failure. I will move back to a non-wiki webpage.
Update (April 2010): Things have gotten more active on the Tektronix Wiki so I will continue to post to that.

Jan 27, 2010: Recently I have been fixing up a Tektronix 567. The original electrolytic capacitors in the power supply failed due to excessive ESR. I replaced them with new capacitors. This affected the low voltage supplies the most. Some ripple in the pre-regulator capacitors is normal. The regulators handle this variation at their input without allowing it to affect the output voltage. But there is a limit, and certainly if the voltage at the input of the regulator dips down to (or below) the assigned output voltage, there is nothing the regulator can do. That is what was happening in the 567 before I replaced the capacitors.

After replacing them, I could look at the finer details of the situation. I found one dead tunnel diode in the 3T77 sampling sweep. They symptom was that it would never really trigger on the input signal. Depending on the position of the trigger level control, it would either not generate trigger pulses or it would free-run. For debugging sampling scopes, I like to look at signals that pass back and forth between the plug-ins. And I like to start with an external trigger signal since it is simpler than internal triggering. The 567 is working well now except for one small problem with the 6R1A digital unit.

Feb 1, 2010: I bought a Tektronix 454A on Ebay. It was sold as not working. It arrived, indeed, not working. The main problem was that the A trigger was not triggering. It would free-run when in the AUTO mode, and not trigger when in the NORMAL mode. I found abnormal DC voltages on the terminals of Q623, which is a JFET source-follower at the input of the trigger circuit. Current was leaking out from the gate of the JFET, a bad sign. I removed Q823 (a similar JFET) from the B trigger board and inserted it in the Q623 socket. It worked. With Q823 missing, the scope still works, including the B sweep. Q823 is only needed for the "triggerable after delay" mode, not the "sweep after delay" mode. I will find a JFET to replace Q823. Q623 is in a vulnerable position. When the trigger input is set to EXT, if a big pulse (e.g., electrostatic discharge) is applied to the external trigger input, it goes more or less directly into Q623, killing it.

Another problem with my 454A was that the horizontal position knob was pushed in, cracking the knob and breaking open the potentiometer. To fix this I had to remove the structural aluminum piece on the side of the scope. The rear panel of the scope must be loose to remove the structural piece. I was able to push the potentiometer back together and bend its metal retaining tabs back in place. It works fine now.

Feb 1, 2010 (continued): I replaced Q823 with a 2N3819 FET. It works well. The pinout of the 2N3819 is different from the pinout of the original 2N4416. After changing the transistors, I recalibrated the DC levels of the triggers and checked them for stability at low input levels.

Feb 3, 2010: Nose-to-nose measurements are common for high-speed sampling scopes. Today I wanted to measure the risetime of 454A that I recently got, which is specified to be about 2.4ns. Lacking a subnanosecond risetime pulse generator, the first thought that came to mind was to use the kickout pulse from one of my sampling scopes. The 567 was convenient, so I connected the output of the 3S1 sampling vertical plug-in to the input of the 454A. I drove the 3T77A sampling sweep horizontal plug-in with an external trigger signal from my HP pulse generator (7ns minimum risetime). The sampling pulses were visible on the 454A, but I was disappointed by how low in amplitude they were. They had a more or less gaussian shape, with a 10% to 90% time of about 2ns. Later, I thought about it and it makes perfect sense. The tall skinny sampling kickout pulses become lower and wider due to the bandwidth limitation of the 454A. This is, in fact, exactly what one should expect. And the 454A is behaving as it should.

Feb 4, 2010: I found that the amplitude of the kickout pulse from the 3S1 sampler is strongly dependent on the mV/div setting on the 3S1. By choosing a setting that gives a strong kickout pulse, I was able to get a better trace of the impulse response of the 454A. This photo was taken with the 454A set to 2ns/div (0.02us/div and 10x horizontal magnification).


June 4, 2010: I bought a 564. It had a problem with HV arcing (photo 1, photo 2) The scope was very dirty when I got it. It was arcing from one position in the porcelain strip to a the silver cup of the adjacent position, and then to the next position to complete the path. Corrosion or etching of the porcelain strip can be seen in the photos. One end of the arc path is the junction of a resistor with a wire. I lifted that junction off the porcelain strip and the arcing stopped. I tried to clean the porcelain strip but it is permanently discolored from the arcing and I don't trust it not to leak current at high voltage. For now, it seems fine to let the junction float. It is mechanically stable because of the stiffness of the resistor and the wire with which it joins.

It is an interesting 564. It has been modified with the addition of a module labeled "MODIFIED FOR SPECIAL LINC DATA TERMINAL BOX". It has four relays added. These relays are in sockets. The relays are controlled by signals brought in through an added connector on the rear panel of the 564. The relays appear to provide remote control of the storage circuit.

June 5, 2010: A few weeks ago, I found an early 535 being discarded. It is the type with the chassis that slides out from the case, which is brown. Before powering up the scope, I did a few things. First, the scope was filthy so I washed the entire scope with warm water from a hose. After letting it dry for two weeks, I remounted the fan, which was originally mounted on rubber anti-vibration blocks, which predictably had become brittle and disintegrated. Next, there was a HV rectifier tube that had a broken plate lead. Based on the appearance of the inside suface of the metal box that encloses the HV circuit, it seems like the plate had been arcing for a while before it eventually weakened the wire enough for it to break. I would guess that this is the failure that ended the scope's previous phase of service. I reconstructed the plate wire by soldering a thin wire to the 1mm stub remaining of the 5642's plate wire. Next problem, there were arcs forming between two of the secondary leads on the HV transformer, T801 and the core of the transformer. These cracks in the insulation/potting can be seen in this photo. The blue lines drawn on this photo show the path of the arcing. I cleaned the area of the arcing with contact cleaner and a toothbrush and then dripped hot candle wax on the area. This completely stopped the arcing.

The -150V supply was quite far off, -142. I adjusted it to -150. If the scope was calibrated while the supply was wrong, then setting it to the correct voltage will mess everything up. But I doubt that anybody would calibrate a scope without checking the -150 supply. A couple of volts off is not a problem, but 8 volts is too much. After that, I checked the other supplies. The +225V supply was +195V. It it was in regulation, and not noisy, just the DC voltage was wrong. Here is the schematic of the +225V section of the power supply circuit, annotated with what I think was happening. R772 and R771 are not too far from their specified resistance values. The 12AX7 has normal emission and no grid shorts. The problem was DC leakage through C770. Let I_leak refer to the DC leakage current through C770. By my calculations the output voltage of the +225V section of the 535 power supply is 225 - 5.8*10^6 * I_leak. So the the output voltage drops by 5.8 volts for every microampere of I_leak. Based on this, it seems like C770 (an original bumblebee capacitor) in my 535 was leaking 5.17uA with 195V across it. I replaced C770 with a good capacitor of the same ratings and now the output voltage of the +225V power supply measures +227V, close enough. The reason it is 227 and not 225 is that R772, a 333K 1% resistor, has drifted up a bit. I will measure the DC leakage of the original C770 on a voltage source set to 195V.

Update, June 5, 2010: I measured the leakage of the original C770. It leaks 1.5uA with 195V across it. Although 1.5uA is a totally unacceptable amount of leakage for a 0.01uF capacitor, I was expecting more like 5uA. I have no explanation for the discrepancy at the moment. In any case, C770 is clearly bad.

June 6, 2010: I found that the leakage current through the bumblebee C770 of the 535 is extremely temperature sensitive. At 195V, the leakage varies from 1.5uA at cool room temperature to 20uA when warmed so the body of the capacitor is warm to the touch. I tested a 0.047uF 600V Orange Drop (polypropylene film) capacitor at 195V. The leakage is 100nA cool, 200nA warm. Normalizing for the difference in capacitance, at room temperature the bumblebee's leakage is 70x worse than the Orange Drop's. Warm, this becomes 465x.

The 535 works now. All supply rails meet their voltage and ripple specs. I was seeing some strange erratic nonlinearity in the sweep, but it went away after I used the scope for a while. I think it was a dirty contact in one of the switches, most likely the horizontal mode switch. The triggers for both sweeps work, but are very tempermental, only triggering correctly in a narrow range of stability settings.

July 10, 2010: I bought a 230 Digital Unit. It was not working at all. All of the DC voltages at the test points on the regulator board were close to 0V. One of the low voltage electrolytics has failed (high ESR). I connected another capacitor in parallel with the failed capacitor using clip leads. That fixed the ripple on the post-rectifier capacitor, but the outputs of all of the regulated voltages were still all dead. There are interdependencies between the regulator subcircuits. For example the -50V output is used by the error amplifier for the +50V regulator. And the +50V output is used by the -50V regulator.

There is a 1-ohm sense resistor used by the +50V regulator for overcurrent protection. This resistor had failed, open-circuit. I replaced it and then the whole power supply came to life, pretty close to the specified voltages for all outputs, and less than 2mVrms ripple on all of them. But the 230 is still not working in other ways. Some Nixies don't light up at all and others have multiple digits lit apparently simultaneously. None of the three limit lights illuminates.

main page