Thanks all for your comments.
The clock has an analog part, the receiver, with a very small ferrite
antenna, and a large black wax blob on the PCB, with many traces
reaching it. Usually this is the low cost "packaging" of an IC, which
probably is a sort of simplified microcontroller.
But it turned out that Andy and Johan were right. I did what I should
have done since the beginning, and what I usually do when building
an oscillator to test its stability : the "blow test".
I blowed for 20 seconds on the exposed components of the clock,
and the result was a change of more than one microsecond....
so the 1pps signal is not derived directly from each pulse of DCF-77,
but from the internal oscillator, corrected every now and then with
the received signal...
I think I can use it as a high sensitivity thermometer, to measure
fractional degree changes in the temperature of my shack :-)
73 Alberto I2PHD