Tektronix 491 spectrum analyzer problem...

Hello there is a problem with my Tektronix 491 spectrum analyzer. Some times it works ok, but other times the horizontal movement of the dot on the screen moves like crazy, backwards and forwards. Some times the problem is corrected when I hit the analyzer with my hand, some other times not. I suspect it it a poor contact problem.
But which part should I check first? The sweep circuit, the sweep generator and timing circuit, the blanking circuit, the vertical amplifier, the dispersion or what?
What do you think?

Check the grounding / earthing of all the modules and all cable shielding

You can often find such intermittent faults by doing some localized tapping with an insulated tool, like a thin wooden stick or plastic knitting needle. Open the unit and while powered on, gently tap around on the modules you may expect to cause such a problem. Once I have located the problem area or module, I will push against individual components to see if I can get it to act on my command. Sometimes, I will pull out the complete module or pcb and inspect it carefully under my magnifier for dry joints. Don't be surprised if you eventually find the fault where you never expected it. :)

From your report, the problem seems to be related to sweep generator / horizontal amplifier. A contact problem of the time base switch could be a simple explanation. As you probably have seen from the block diagram in the instrument manual, there's a sawtooth output jack between sweep generator and output amplifier, so it should be easy to narrow down the problem.

Thank you all,
I forgot to mention that apart from the "crazy" point on the screen sometimes the line width on the screen goes narrower for about 1-2 divisions than it should.

---------- Post added at 18:25 ---------- Previous post was at 17:05 ----------

That is absolute right!
I connected the sweep output to an oscilloscope, and at the time where there was a "crazy moving" dot on the analyzer screen there was not sawtooth signal on the oscilloscope. When the whole thing worked after a wile, a nice sawtooth signal appeared on the oscilloscope.

I will check the sweep generator / horizontal amplifier modules for lose wire connections or loose transistor connections, as the analyzer has it's transistors clipped in sockets and not soldered. Maybe it could be a dust problem as well on the transistor sockets...

Your test showed, that the problem is not in the horizontal amplifier. The purpose of the suggested test is exactly to check this point.

Yes of course, there may be some problem with the sweeper. I have checked for bad earthings. I have also checked for bad pin contacts on the PSU board. This was the only board that hadn't be checked for these bad contacts, all the others have been checked some time ago and corrected. I have found 4-5 very loose pins on the PSU board and corrected them.
Now it seems the sweeping is ok, both in amplitude and in timing.
But I would operate the analyzer for a few days before I say for sure...