Opening the caseThe first step to attempt the repair is to open the case. It is not an easy task because both covers are glued together. Carefully I applied force to the glued join with a piece of flat metal and with patience I was able to separate both cases. Of course, I made some damage to the plastic cases: They are glued together to be not opened. Once opened, you will notice the PCB is fixed to the front cover by 5 plastic snippets: Two at top and three at bottom. You must to cut them with a cutter to get off the PCB.
When I opened the case, this is what I found:
The meter is based around an AT89C4051 microcontroller running at 3.5795 MHz.. There is a switching power supply to get the supply for the microntroller and an oscillator for the HV (around 400 volts) used by the tube. And, as you can see now, there is no doubt: The tube used is a CБM20-1:
Point C is on a small piece of cooper foil.
B is connected to the rectifier diode, and C is connected to "ground", on the cooper foil.
Point D, is supposed to be also ground, and it's connected to the cooper foil through the track E. It's supposed that E and C are connected by means of the pressure of the transformer over the cooper foil: They are just "in touch".
Over the time, the transformer fixation softens and therefore the contact between C and E starts to be a bad joint, making noise in the HV line, and therefore, producing extra counts.
The solution is easy: Take a small piece of wire and make a jumper between A and D points.
No more counts with movement or vibrations!