LED driver PCB draws less power when scope probe connected to it

[ocset]

We have some offline , 150W LED drivers. These don’t comprise a mains filter because they are switched linear regulator based. 
The driver PCB  sits on an earthed heatsink, but earth isn’t taken to  any pads on the driver PCB.

Today a production staff member brought me a driver PCB and complained that it was drawing 74W instead of the usual 71W. (This is a dimmed down version).

Anyway, I therefore cranked up the AC power source  in the Test Dept (with its isolated output) and supplied the problem board with isolated 240VAC. I noticed indeed that the board was drawing 74W. I  therefore attached the scope probe to the reference voltage on the driver board  that controls the amount of LED current (this was across a  ceramic capacitor and the connection was made via a twisted pair flying wire) and then switched the product back on. –With the scope probe connected, the  board only drew 66W. I   then took the scope probe  off and the board again drew 74W.

I  then  manually added some 1n5  Y capacitors from the live and neutral input traces of the board, and connected the other lead of these caps to an earth wire. The problem did not improve.
Next I am thinking of adding some ferrite beads to live and neutral. Also a common mode choke. Can you think of any more  solutions? Maybe I need to use an isolated differential probe instead.

(As an aside, when I unhooked only  the  ground clip of the scope probe from the flying wire, [leaving the main  scope probe  clip connected] the board then drew 363W permanently!....However, this problem was indeed solved by adding the Y capacitors as described above)

[DaJMasta]

Have you tried bridging the points you measured and ground with either a small cap or a 1M/10M resistor or both?  Maybe there's some instability or resonant mode that the loading of the scope probe altered and got rid of, improving the efficiency of the system.

I assume the device measuring the power consumption has sufficiently high resolution and accuracy that this isn't in the noise of the measurement.

[jeduffy]

Sounds like there's some capacitive coupling between your isolated supply and earth ground, have you tried connecting the isolated supply ground to earth ground?  Aside of that, if adding capacitors didn't work, have you tried adding 1/10M resistors across the capacitor?

[ogden]

I  therefore attached the scope probe to the reference voltage on the driver board  that controls the amount of LED current (this was across a  ceramic capacitor and the connection was made via a twisted pair flying wire) and then switched the product back on. –With the scope probe connected, the  board only drew 66W. I   then took the scope probe  off and the board again drew 74W.

It could be so that your scope DC-loaded that voltage, especially in case of 1:1 (1Meg) probe. [Edit] could be AC-loading as well. What is impedance of (reference) voltage source you are  measuring? Any chance to see that part of the circuit? Did you check results with just scope ground connected to test point?

Maybe there's some instability or resonant mode that the loading of the scope probe altered and got rid of, improving the efficiency of the system.

That shall be noticeable as driver output waveform change. In case scope has more than one channel - it is not that hard to test.

[ocset]

http://www.ti.com/lit/ds/symlink/uc3854.pdf

Good point, the cap i measured across has a 1meg resistor into it....i should have checked at the output of the opamp buffer which has  its  input from the cap voltage.

Regarding the output of the led driver increasing to 363W (a large overpower) when the AC source was supplying it, (but not when supplied by the mains isolation transformer) it is bizarre that this happened, and more bizarre that it stopped happening when i connected Y capacitors from Mains Live and neutral inputs to earth.

Sounds like there's some capacitive coupling between your isolated supply and earth ground, have you tried connecting the isolated supply ground to earth ground?

Thanks,  i assume you mean when the scope isnt connected to the circuit, since as you know, scope ground clip is connected to earth, so i wouldnt be able to connect the output ground (ie neutral output) of the AC source to earth whilst the ground clip was connected in the circuit.