I have a prototype board which I'm currently testing the signals of its zero-cross circuit. This is the schematic of the PSU, the power supplies of the low voltage part of my circuit (12V, 5V, 3V3) are galvanically isolated from the high voltage part.
This is the schematic of the zero-cross circuit:
The circuit works well, the output signal is not perfect, not a perfect square, but its waveform is very suitable to control a TRIAC (this part is working fine). The output signal waveform I consider as 'ok', even when input supply is 100Vac.
Here are some scope screenshots of some points of the zero-cross circuit.
Note that all the waveforms below are in respect to GNDP, and the negative probe of the scope is connected exactly at the emitter of the transistor.
(1) Here is the cathode of diode D8.
(2) Here is the cathode of diode D8, with zoom next to ground point.
(3) VBE, when it is all working correct.
(4) Some days, at some moments, a glitch occur on the circuit. Today I was able to catch the problem with the scope, and here is the output signal when the glitch occur (the fast falls from 1 to 0):
(5) Of course this problem comes from a glitch on the input, and I was able to catch the glitch on the input also, here is the VBE signal when the glitches were occuring. Check the short pulse from 0 to 1 at almost the mid of the level 0 period. I realized that when the glitch occurs is because the amplitude of this undesired pulse reached a voltage sufficient to VBE to conduce.
I already modified D8 from 1N4007 SMD/SMA to S1M diode, but it makes no difference. I can't short D8, if I do this there will be much power dissipated at de 39K resistors. And I do know that the zero-cross circuit is consuming around 1mA from 6VP, and this do increase a little the temperature of the IC LNK626, but I consider this is not a problem.
What can be done to avoid such undesired pulses? Or maybe, how to reduce these noises? For me, the noise seems to be the switching of the PSU at around 100KHz. Regards.
EDIT(1): new tests.
I placed a 100nF capacitor across R19 and the VBE signal remains the same, it makes no difference.
Below is shown the waveforms of AC2 vs GNDP. In all these measures, the negative probe of the scope was connected exactly at the emitter pin of the transistor Q5.
(2) Zoom, closer to the reference
(3) Zoom on X
(4) Zoom, even closer to the reference. Noise observed.
(5) Noise observed at a different moment. Maybe occured when the output of the PSU demanded more power.
Is there any way to correct this?