HP5890 TCD signal going to 0 after pushing start bottom

[CJCheng]

I'm using an old (~25 years old) HP5890 GC/TCD for methane gas test. Usually I wait about 30 mins for letting the signal baseline to be stable (signal will be about 50-30).
Recently, when I push the "start" bottom on the panel to start a run, the signal will go down to zero and bounce back to 30 after the run. I tried several times with calibration gas, the results are repeatable. But according to my colleague, the numbers are much lower than before.
My questions are:
(1) Is the TCD unstable? Can I use these number for quantify gas concentration?
(2) I have tried changed columns (HP-PLOTQ and CarbonPLOT), changed analytical methods, checked gas leak (N2 is the carrier gas), baked TCD. What should I try?

Thank you all in advance!

[chromatographer1]

Have you tried to manually zero the detector BEFORE you start the run?

best wishes,

Rod

[CJCheng]

Yes, I tried to adjusted the base line to zero before start the run, then the signal will go down to negative values, then back to zero after the run.

[Russ]

Sorry if this is very obvious. Did your signal accedentally get changed to use the column compensation? If you press SIGNAL 1 (or 2) does it read A (or B) or does it include - COMP 1 (or 2)?