Anritsu MG3633A Repair #1

Wait, what's your output level again?

The missing link

Turning on this unit first time revealed a red "Uncal" light and the error message saying: "No Ref Signal".
Hmmmm ... Just take a quick look at the back side of the unit. Yes, the U-shaped BNC bridge is missing, leaving me staring at two open BNC ports, labelled "Ref In" and "Ref Out".
U shaped ref link U shaped ref link

So no surprise this unit is complaining here. I've connected Ref out and Ref in with a short piece of BNC cable, the unit beeps and still complains, WTF? Check the Ref out with a scope, there's clearly a 10MHz signal of rather ugly square shape available. So is there something broken within the OCXO, leaving me with a bad waveform on its output? I've connected for now the Ref in to the CMU200's Ref out, the unit beeps again and is fine with that. So I pushed this issue away for later investigation.

What's that Attenuator stuff, anyway?

Having an apparently working unit for now, I've tried all kinds of settings. It quickly turned out the output level setting behaves in a rather random way. The output level correlates in no way but random to the setting. One changes the output level by 1dB, a nice and beefy "klack" inside the unit, and the output amplitude jumped by some 10dB up or down. The noise gives a clear hint to mechanical contacts within the attenuator, digging in the service manual reveals: The attenuator has a limited life and is supposed to be replaced after some millions of cycles. There's even a built-in counter for this purpose. Stickers on the unit indicate: This service has been done before on most of the units.
Anritsu MG3633A Attenuator module
This is the attenuator module with its control board.

Nah, buying a new attenuator is clearly out of scope for this repair, even if it should still be available. A last resort would be to salvage the attenuator from the completely failed unit. But this attenuator might also be worn out, or I might decide to repair that unit at some point and then it would miss the attenuator. There must be some other way to deal with it.

So lets go all the way down to fix this attenuator. Turns out a piece of art inside. I felt like unscrewing a thousand screws to find my way inside.
Anritsu MG3633A Attenuator module Anritsu MG3633A Attenuator module Anritsu MG3633A Attenuator module

Now, we've discovered some nicely engineered gold spring contacts
Anritsu MG3633A Attenuator module Anritsu MG3633A Attenuator module

Remembering that good old stuff of how to clean relay contacts: Draw a solvent soaked strip of paper through the closed contact. Carefully push your finger on the magnet pieces to close the contact to be cleaned.
Cleaning contacts

Do this really carefully to not bend the delicate strips and don't lose the small copper pieces used as an additional shielding for some of the attenuation resistors.
Put it all together again and repeat for the second box. Now restore the whole unit, screwing another thousand of screws to get it all in one piece again.

Again for the Reference

So after having re-assembled the unit, I turned it on again. But now, I've put the short BNC bridge cable in place before I turned it on, et voila: It doesn't complain anymore about the missing reference. So I have no idea what caused the unit to complain at the first check, but now it's gone (so I hope). Watching the output frequency, the oven appears quite a bit drifty while heating up, but finally stabilizes to a very small deviation from the CMU200's reference. Using the CMU200's Ref out to synchronize to unit the frequency is spot on, just as one would expect.
Playing with all the settings, especially the output level, reveals no further issues for now. Let's call it a day, first unit fixed. Including a screwed up adventure, not even one screw was left over after finishing the job, WTF?

Back ... und ein Zaehlpixel hab ich auch :-)