I am working on repairing a power supply board of a LCD TV (200W power), more specifically, the PFC section (active power factor correction). See schematic (adapted from here):

Power supply partial schematic

I've replaced some defective components and now the power supply seems to be working OK (the voltages at their output seem OK, and some waveforms measured on the oscilloscope look good).

I still have no confidence to connect the power supply with its output connected to the TV, so at the moment I'm doing bench tests. The PFC has 2 MOSFETS in parallel (see diagram), but I'm currently testing the power supply (no load at the output) with only one MOSFET on the board.

I am analyzing some waveforms triggered at the moment the PFC is turned on (it charges the capacitors CP815/816 from 300V to the target 380V, so it have some work to do).

What's bothering me is the current spikes in the SOURCE terminal of the QP801S MOSFET (measured through the voltage on the RP820 (0.08 ohm) with the oscilloscope, using a very short lead).

switching current waveform

See the detail of a negative current spike (the greater that occurs):

detail of current spike

My question: Is this negative current peak (shortly after MOSFET turn off) normal? Note that it reaches -67A (-5.36V over 0.08 ohm) with a duration of ~50ns. [EDIT 2: figure and paragraph edited to show the greater negative spike that occurs].

Note that the positive portion of the current (while MOSFET is ON) is a normal ramp, charging the inductor with up to 4.75 A (0.38V over 0.08 ohm)).

EDIT 1: probe setup used to get the measurements, with short wire connected to probe GND (Agilent N2862B, 1:10 passive probe):

my probe

EDIT 2: See the figure below for a longer time capture (the scope was set into Acquire=Peak Detect mode, to ensure that the peaks are captured. The scope's A/D converter works at the maximum rate in this way).

It is possible to observe that the amplitude of the negative peaks follow the waveform of the input voltage of the PFC (60 Hz mains rectified with full bridge => 120 Hz half wave, T = 8.3 ms). May be this provides some hint about these spikes.

enter image description here

  • 2
    \$\begingroup\$ Before I believe that 50ns waveform is the true voltage on the shunt, I would need to see a picture of your probe setup. In this case it may be advantageous to solder 50 Ohm cable directly to the shunt and run it back to the scope on 50 Ohm cable. Set the scope input to 50 Ohms. The 50 Ohms in parallel with 80 mOhms will have negligible effect on shunt impedance, but this setup should reduce any ringing that arises purely due to test setup. \$\endgroup\$
    – user57037
    Mar 7, 2019 at 1:57
  • \$\begingroup\$ You say things seem ok, but readings of MOSFET current spikes seem high. Like mkeith mentioned you need to treat this as a RF circuit. Keep scope ground wire as short as possible. Negative spikes on a ground rail can be a sign of scope ground loop or back-EMF. I doubt you have a circuit problem. At these power levels tiny problems can cause big bangs. With strong current pulses sometimes you need to use differential probes. \$\endgroup\$
    – user105652
    Mar 7, 2019 at 2:56
  • \$\begingroup\$ @mkeith please see the picture in my last edit. \$\endgroup\$
    – Berk7871
    Mar 7, 2019 at 3:38
  • \$\begingroup\$ That is good technique. But I still am skeptical. Do you have a piece of coaxial cable you can sacrifice? Solder center conductor to one side and shield to the other side of the shunt. Put the scope on 50 Ohm input impedance and check the signal again. \$\endgroup\$
    – user57037
    Mar 7, 2019 at 3:43
  • \$\begingroup\$ The excerpt you are showing is not revealing some potential debug opportunities. I would check if DP807 is OK. This diode should be clamping any negative excursion on the CS signal to a diode drop below GND to protect UP801S. This diode is a ultrafast rectifier (Trr of 1.6ns) and should be doing, well, something to limit that spike. If the diode is intact and working I would start to suspect a measurement artifact. The waveform you showed does appear to be CCM mode so I don't think the controller is bad. \$\endgroup\$ Mar 7, 2019 at 13:37


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.