Tektronix 7603 / 7623 / 7A18 repair

or strange effects with simple fixes

The Incident

Within the last half year, I've acquired these to have some decent analog scopes in my lab:
7623 and 7603 showing sine wave

A Tektronix 7623A and a 7603, both of them equipped with the usual 7A18 and 7B53 plugins.

At first glance, they were working fine, apart from some noisy pots and switches. One of the 7A18 has the push-out knob for variable V/div broken, so nothing uncommon or unexpected.

Wiggly waves and dancing readout

First, I noticed something strange on the 7623A: The displayed waveform is distorted (non-linear sawtooth) at the left hand side of the screen:
distorted wave left

If one turns on the readout, this starts to look even more weird:
distorted wave left with readout

In case the videos (should do on firefox, I don't care for other browsers) are working for you, here's the wiggling and dancing in motion at different speeds:

This effect usually appears if the scope was turned off for a few days and is then turned on again. After some warmup time, it disappears and everything starts behaving totally normal.

Having no idea what is the root cause for this behaviour, I started investigating. Using ice spray on several components of the horizontal amplifier, nothing really changed. I did some interesting experiments like swapping the wires leading to the horizontal deflection plates, this results in a mirrored picture:
distorted wave right

Another one was using a second scope in X/Y mode to display the horizontal and vertical amplifier output voltages - the result is interesting: The 7623A still shows the wiggly waves, while the second scope shows a stable image. This ruled out the horizontal amplifier as the source of the defect. In the meanwhile I discovered a shaky beam finder switch which I suspected to be guilty. Some contact cleaner took care of this, but wasn't the cause for my problem.

Next guess was some broken spot weldings inside the CRT, which would have been unfixable then. In my imagination, the broken spot welding gained contact with warmup of the inner workings of the CRT, matching my experience. So I took the 7623A on hold for a while.

Non-square square waves

In the meantime, the 7603 showed up on my workbench. Some checking revealed grossy decompensated square wave response on all inputs (only one shown here):
square wave at 5mV/div   grossly decompensated square wave at 10mV/div   grossly decompensated higher frequency square wave at 10mV/div

The decompensated signal edges were visible on all V/div settings, the 5mV/div range looked less decompensated than the others. My first guess would be the vertical amplifier since this affected both of the 7A18 plugins that came with the 7603 mainframe. Swapping the plugins with the 7623A (which had its own set of 7A18) ruled this out quickly - all of the 7A18 inputs from the 7603 were decompensated.

Trying to adjust the X2, X4, X10 and X100 attenuators of one input wasn't successful, their adjustment range wasn't wide enough to get square edges and a flat top wave form as required by the adjustment procedure. I wondered which kind of defect would cause all of the inputs to be decompensated this much.

Next, I checked the input impedance of a "bad" 7A18 in comparison to a "good" 7A18. I was able to get meaningful results by removing the plugin from the mainframe and placing it on a cardboard box to isolate it. The LCR meter showed some not so good results for the 5mV/div setting:
isolated bad 7A18 set to 5mV/div   input impedance of isolated bad 7A18 set to 5mV/div

Way less than the expected 1M resistance and about 30pF instead of the specified 20pF.

Things get somewhat better with increasing V/div settings, here the 10mV/div which puts the X2 attenuator into the signal path:
isolated bad 7A18 set to 10mV/div   input impedance of isolated bad 7A18 set to 10mV/div

The good 7A18 plugin showed results that were nearer to the specification. For the remaining differences, I decided to put the blame on the non-ideal setup for the LCR meter, still having large stray capacitance to ground. The LCR meter doesn't show anything meaningful if the plugin is plugged into the scope.
isolated good 7A18 set to 5mV/div   input impedance of isolated good 7A18 set to 5mV/div
isolated good 7A18 set to 10mV/div   input impedance of isolated good 7A18 set to 10mV/div

Aged or damaged capacitors

Checking the components inside the plugin revealed these capacitors (their nominal value is 1.8pF), placed in paralled with the first amplifier stage (a JFET voltage follower):
bad capacitors  impedance of bad capacitor

The LCR meter shows them having a really bad "D" and capacitance reading way above their nominal value. This matches the too high input capacitance reading, since these capacitors are effectively wired in parallel to the input connector if the 5mV/div range is set.

I didn't have 1.8pF capacitors quickly available, so I decided to roll my own - just to check if the decompensation goes away with the replacement of these capacitors:
twisted wire capacitors   impedance of twisted wire capacitor

They aren't particular good at "D", but they did the job. So I ordered some fresh 1.8pF ceramic capacitors and replaced them all.
impedance of new capacitor

Having replaced these capacitors, another issue arose:
The input capacitance of these plugins is "normalized", say adjusted to a specified value (20pF here). This is quite handy, since you can swap your usual X10 scope probes in between the inputs and scopes of the same kind without having to compensate them each time. The service manual describes the adjustment procedure, but requires a little peace of gear, called the "Input RC Normalizer" which I don't have. Luckily, there's just a 1Meg resistor and an adjustable capacitor inside, so I've rolled my own. The normalizer must be adjusted itself, that wasn't a big deal, since I have a good 7A18 plugin from the 7623A to adjust the capacitance of the normalizer to show a perfect square wave. The normalizer simply acts as a compensated 2:1 probe.
homebrew input normalizer

One more problem showed up:
The input capacitance adjustment isn't accessible if the plugin resides inside the mainframe. You can remove the right hand side enclosure from the scope and also from the plugin to gain access to the attenuator adjustments, but not to the input capacity trimmer cap. Tek recommends the extender board, which I don't have. Now this simple adapter comes in handy:
power supply and signal output for plugins

Using this one and some lab power supplies to get the required supply voltages:
lab power supplies connected to get +/- 50V, +/-15V, 5V

At the right hand side is a black box supplying a fixed +/- 15V. Then connected in series is the powerbox supply set to 35V on each channel, connected in series to the +/- 15V. This results in the required +/-50V. The third output of the powerbox supplies the fixed +5V. There's also a +130V supply voltage on the plugin connector, but wasn't required here.

Now I was able to operate the plugin outside the mainframe. Connecting the signal output to another scope, I was able to watch the square wave while adjusting the input capacitance using the normalizer.
access to input capacitance adjustment   view plugin output signal on another scope

The plugins signal output operates into 50Ohm, so the TDS scope is set to 50Ohm terminated inputs. The output voltage of the plugin is around 20mV/div into 50Ohm. The input capacitance adjustment is inside the leftmost shielded box, labelled C100. Next to it, the adjustments for the input attenuators. Most right, just in between the white ceramic soldering terminals, near the green tantalums, you may spot the small 1.8pF capacitors (the new ones are small brown disc type).

Having adjusted the input capacitance, I put the plugin back into the scope mainframe (with side panel removed) and then adjusted all the input attenuators of each channel. The result:
good square wave at 1V/div

A nice and clear square wave on the 7603.

A fix for the dancing waves

After having read some of the Tektronix "Concept Series" books, especially the one on cathode ray tubes, I found a simple fix for the "dancing waves" on my 7623A:
In some chapter, internal static charges affecting the deflection were mentioned, and a way to dissipate them quickly: Turn the intensity knob fully clockwise, and shift the trace outside the screen. In my particular case, shifting the trace right at full intensity does the job.

Siblings in peaceful coexistence

Finally, I've got two useable 76x3 mainframes in my lab:
7623 and 7603 showing square wave

Attenuator compensation

At the EEVBlog forum someone found out a shortcut method of adjusting the attenuators:

As promised I'll describe the easy procedure I found for adjusting the attenuators quickly and easily. Let me first explain how these attenuator assemblies work and then the procedure.

Attached is a typical attenuator adjustment. This one is from the 7A16A and with some variations they are basically the same on many Tek plug-ins and scopes. The 5mV range does NOT go through the attenuator blocks and it's a direct link from input to the first stage of the vertical amplifier. Typically that cap (C134) does NOT need to be adjusted unless it's been burnt out or someone monkeyed with it. The 10mV and .1V ranges are in the same path between input and first vertical stage. Same with 20mV and .2V ranges, 50mV and .5V ranges. The 1V range and above on a separate range. If you follow the adjustment table sequentially as shown I guarantee you'll spend a frustrating hour or more trying to get the 1KHz square to look decent across all ranges. And you'll still have some waveforms that look like crap. But I figured out the secret and I'll bet the gray beards at the factory used this method and didn't want anyone on the outside to know.

Let's take the 10mV adjustment as an example. An as I mentioned the .1V range is in the same path. So whatever adjustments you make to 10mV range it will affect the .1V range. And this is where it will drive you nuts because if you get one to look good the other doesn't. Back to the 10mV adjustment. The table says simply to ?check?. That's a load of garbage. Adjust C114 for square corner but do NOT touch C115. Switch to .1V and adjust waveform for 6 divisions. It will look terrible. Adjust C115 flat top for best waveform. Do NOT adjust C114. Go back to 10mV and it may or may not need a minor tweak of C114. Go back to .1V and check and you are DONE with that range. The waveform should be perfect. Follow the same procedure with 20mV/.2V ranges and 50mV/.5V ranges. Adjust the square corner cap ONLY on the mV range and the flat top cap ONLY on the V range. The 1V range you can adjust both and it should be OK. That's it. 10 minutes and the compensation is done.

I applied this procedure to the 7A15A plug-in which I had to rip apart and remove the attenuator blocks because I had disassemble the switch assembly to clean noisy contacts. So the compensation was way off upon reassembly. In 10 minutes back to perfect. Touched up a military 7A15 with same results. And I went back to the 7A16A that I adjusted the other day but I wasn't happy with the results. It's now perfect. This post is getting a little long so I'll end it here. Later I'll post some waveforms to prove that I'm not giving you guys a line of bullshit.

Screenshot from EEVBlog forum

Table of Contents ... und ein Zaehlpixel hab ich auch :-)