I’d been in need of a Time Mark Generator for a while, and finally found one that was the right mix of price and form-factor.  With a few exceptions, I’m loathe to spend over $100 on a piece of old test gear, and the only units under that were either RM500 series units which I’ve been avoiding, or the 180 which is huge.  Speaking of the 180, Richard Sears has a nice write-up on both the 180 & 181.

This is clearly an early unit, based on the 3 digit serial number, and the style of the enclosure.  Later models have the familiar textured aluminum side panels and rounded corners, while this bears a resemblance to the earliest of Textronix scopes like the 511, 514, etc.. I think this is also the first piece of tek gear I own thats the ‘wrong’ color.

After replacing the fuse holder, which required some gentle modification of the chassis (hooray drill press), I brought it up on the TU-75 variac, and got a trace.  After being on for 20 minutes or so, the slower markers became unstable and impossible to calibrate, and it would get worse with time.

I checked the power supply and found that the -150V supply was low, at around -139V (or, um, high, absolutely speaking).  The thing is, it would regulate this incorrect voltage like a champ.  Bringing it up on a variac, that -139V was stable from ~80V up to my uncomfortably hot line voltage of 126V.  But, over 20 minutes or so, that would slowly fall to -137V, which is where things would get unstable.

During our semi-annual encounter where he enables my oscilloscope addiction, Kurt suggested I suspect the caps around the feedback loop of the regulator, and he was dead right.  Those .01uF bumblebees were leaky over 50v.  They’re still good below that, so I could probably sell them to some audiot, insisting that the leakage is a part of their distinct warmth and tone.

Lifted one side to test

Their tiny replacements

And here’s the trace on my 453

This is the highest output, 10Mhz.  It’s a frequency multiplier, and I think that modulation is fixable in calibration.

Here’s a pic of with the ‘check count’ switch enabled, showing all of the pulses.  If you notice, there’s 9 small pulses in between every major pulse, not counting the ‘half pulse’ right before the next major pulse.  That means the divide by 10 multi-vibrator is working properly.  This is on my Tek 551 before calibration, which is why it appears to be running fast.

Here’s the 100µS output on my 7D20.  Spot on.

A quick overview of how it works:

At the heart of the unit is crystal controlled oscillator.  Mine has the oven controlled option, which means there’s actually a little container with the crystal, a heater, and a bimetallic thermostat to keep the constant temperature, which minimizes frequency drift.

Here’s a few shots of the crystal oven:

With the outer shell removed

With the top off, revealing the 1Mhz crystal. 

After the oscillator, and the 1Mhz pulse shaper, the signal is applied to a series of cascaded 10:1 frequency dividers, which give the 10µS – 10mS pulse outputs.  In more modern times, you would do this by counting the pulses, letting every 10th through.  These are actually re-triggerable multi-vibrators, with a hold-off time adjusted to only let every 10th pulse through, so it’s not actually counting, it’s waiting.  You can see this when you adjust the trim settings on the front panel.  If it’s incorrectly set, it will hold off for too few or too many pulses, so for example on the 10µS output, you’ll end up with 9µS or 11µS pulse spacing.  Using that ‘check count’ switch, this is easy to observe.

More info and pictures on the wiki

Relatively speaking, the 551 is one of the less useful scopes in my collection, as specially considering it’s size.  With a separate power supply at 42 lbs, and the mainframe 10 lbs over that, it comes in over 100 lbs once you include plug-ins.  It is the least capable of dual beam scopes in the 500 series, having only a single time-base connected to a single set of horizontal deflection plates, and a lone intensity control shared across both beams.
Nevertheless, I’m in the mood to work on a 500 series, and this was the only ‘normal’ one that’s currently the apartment; frankly I’m still too intimidated to work on the 519 or the 661.

Here it is on workbench B (aka the dining room table). On top of the power supply is my TU-75 variac which I was using to check the power supply on the 181 time mark generator (more on that later). Pardon the underwhelming photo quality, proper glamor shots are on the winter todo list.

The vertical section is the same for both beams, and seems to be working well.  The timebase and triggering are are also in good shape, but the timebase is off by 3% – 8%  across almost all of the ranges.  Here’s output from my 181 timemark generator at 1ms.

Parallax induced by the fact that the graticule is forward of the CRT surface makes it difficult to take an accurate picture, but the in the photo above, the left marker is aligned with the left graticule, and the right most marker is at the 2nd to last small tick; making this about 4% off.  Out of the factory, accuracy was within 1%, and I’d like to see about getting that restored.

The manual doesn’t have a section for calibration, it’s a separate document – both available on the Tek wiki, here. To calibrate the sweep, there’s a trim-pot on the back of the ‘Horizontal Display’ control, visible in the photo below, just behind the red test-lead.

Update: What’s written below is mostly incorrect (though I’m leaving it for historical purposes).  I was able to get the timebase into calibration by following the instructions. Before calibrating the sweep, you’re supposed to calibrate the sweep magnification.  This seems completely counter-intuitive, and I never use the mag, so I didn’t bother with this step.  This turned out to be a giant mistake.  Once I did that, I was able to get the sweep calibrated quite easily, with the exception of .1/.2/.5ms, which are running a little slow.  That points to C180A, perhaps I’ll try to replace it another time.  For now, I’m calling this instrument done.

This pot is already as far as it will go, so no luck there.  Turning it the other way increases the error.  What’s interesting is this adjustment doesn’t actually alter the rate of the sweep ramp per se, it changes the amplitude.  There are trimmer caps for some of the faster sweep rates, but there’s nothing in the ramp generator to alter global timing.

The sweep signal is generated in the miller run-up circuit, given a variable DC offset to adjust the horizontal placement of the trace on the display, and then is turned into a differential signal and amplified before being sent to the horizontal deflection plates.
So what’s wrong with mine?  10ms worth of 1ms markers are being displayed in less than 10ms of divisions (1ms per division).  That means by the time the 10th marker goes off, the beam hasn’t yet gotten to the right spot on the screen.
What do we need to do to correct this? The slope of the ramp is determined by both the RC constant of the ramp generator, and the subsequent amplification stage, so it’s possible (within reason) to adjust for a slow ramp in the amplification stage. In the photo above, that’s what I’ve done:  By putting a 500k resistor in parallel with the 100k resistor that’s in series with the trim-pot, I was able to get the sweep into cal.  Is this cheating?  It feels like it.  I haven’t soldered it in yet, I’m just using a substitution box.  At higher sweep speeds, the capacitance of the leads have a significant impact on the waveform.

It’s easy to make erroneous conclusions if you don’t understand the effect the controls have on the measurements, or don’t heed & grok all of the notes on the schematics.  While it’s possible to get the voltages listed on the schematics, if you have to deviate too far from the stipulated settings, or adjust other controls to their extremes, there’s probably something else wrong.  For example, I could get a few measurements within spec if I altered the sweep length beyond normal or listed ranges.  In another case, I measured the length of the sweep ramp at the cathode of V173, and found it to always be about 5-10% too long.  At first I thought this was indicative of the ramp running too slow and I’d found the issue, but then I realized that the sweep length control effects this, and it was only a coincidence.

I’ve gone back and forth between suspecting that the issue is in the ramp generator or the amplifier.  Like I said, the important thing is the slope of the ramp, and that can be adjusted (within reason) in either section.  All of the timing resistors & capacitors are within 1%

Here’s a bunch of the measurements.  It’s extremely challenging to find a 5% error among readings that are 10% off.

This is getting annoying.  I may just solder that damn resistor in place and move on with my life.  More to come.