HP 8648C erratic attenuator

[Oscillosmoke]

First... I'm somewhat of an electronics newbie with access to and some experience setting up and using a variety of test equipment.

So, the problem is an attenuator that is not acting as it should in an 8648C sig gen. I know these are notorious for attenuator issues but I'd like to give a repair a shot for a change. The output is not linear at all! As I moves the amplitude in the upward or downward direction the output jumps all over the place. So I opened it up and connected a known good sig gen to the attenuator input. The output to a know spectrum analyzer and controlled the attenuator (33322-60011) to an HP 11713A attenuator driver - also good. This is a 4 section attenuator that is configured as 10 / 20 / 40 / 60 dB. What I'm getting is 40 / 20 / 40 / 60 dB... and it is very accurately that! So it doesn't seem like a dirty pad or pad/contact issue as I was expecting. But before getting to far in the weeds taking thing further apart, thought I'd ask for advice / thoughts. Thanks in advance!

[m k]

Manual has
A11 Attenuator (8648B/C/D)
The attenuator is a 4-section attenuator (10, 20, 30, and 60 dB sections) that provides 130 dB attenuation in 10 dB steps.
Calibration EEROM on the motherboard contains calibration data specific to this assembly.
It needs to be updated whenever the attenuator is replaced.

Some simple selector problems?

[edpalmer42]

10, 20,30, 60 dB doesn't get you to 130 dB in 10 dB steps.  You can only get to 120 dB.

10,20,40,60, dB can get you to 130 dB in 10 dB steps.

Oscillosmoke, it sounds like either your 10 dB pad is in trouble or the driver is confused.  Yes, it's suspicious that the 10 dB pad is measuring exactly 40 dB, but coincidences do happen.  Is the loss through your 10 dB pad precisely the same as the loss through the 40 dB section?  To the nearest 0.1 dB?  That would suggest that the driver is activating the 40 dB pad when it thinks it's activating the 10 dB pad.  You should be able to power the sections manually to help isolate the problem.

Ed