How to fix PS conducted interference triggers RCD?

[Grant]

Hi there,

A friend bought a Taiwanese made switchmode power supply that
delivers 0 - 24 V at several amps. Problem is that it triggers
his RCD (Residual Current Device, Core Balance Relay) in normal
use. He asked me to take a look, since the nuisance tripping is
major annoyance.

Seems the RCD is rectifying the 40kHz ripple from the PS to produce
what it thinks is a fault current.

Unlike PC power supplies I've opened up, this one solidly grounds
the output and does not have the usual large clearance between
mains side and load side on the PCD, that I'd expect to see in
switching power supplies. Other PS units I've seen usually have
a large gap with an obvious break between mains side and load side.

I made up a power lead with a 1.0 Ohm, 3W resistor (6 x 1R5,
1/2W, 1%, with some bypass diodes for fault current) in the
ground lead to measure earth current from the PS. It's bad.

The RCD triggers at around 50mA p-p of 40kHz noise from the PS.

I've been asked if I can stop the nuisance RCD tripping.

Because the DC side is grounded, isolating the DC side from
mains earth is non-trivial as I can't just cut the earth trace
and isolate the DC, as that trace is needed for safety. Need
to isolate the DC side then add tiny caps to ground to stop
RF interference.

Got me beat why this PS doesn't have the usual clearances?

I put photos of the waveform, measurement setup, RCD and the
offending PS PCB up on http://grrr.id.au/ps-earth-current/
for the curious. With summary of this problem.


What I'm seeking is an easier way to stop the PS dumping that
40kHz switching noise current down the ground lead? I have no
circuit for the things.

This is second one I've looked at, the first one triggers the
RCD too, so it's not a unit fault, it appears to be a design
fault.

In the worst case, we may need to dump this brand PS and get a
better ones for the task, I don't like chucking stuff if it can
be made to work.

The PS claims a 'C tick' approval mark, I didn't think lack of
trace clearance and mains separation like I see here was allowed
any more? But I haven't worked (professionally) with consumer
gear since the '70s.

Thanks,
Grant.

 

[Tim Williams]

Board top view?

I'm guessing it has neither X nor Y-type capacitors on the input? The way it's supposed to be laid out, Y-types RF-ground the circuit side, and a common mode choke limits ground loop current. X type caps provide differential filtering, bypassing line and circuit sides; leakage inductance in the CM choke provides a crude lowpass, but this is sometimes enhanced with a series choke of 10-100uH to provide more differential filtering.

The maximum value for Y-type caps is somewhere around 4700pF; more and the 60Hz leakage can cause problems (564kohm at 60Hz, or 0.4mA at 240V, enough for a minor shock hazard I guess). However, this sounds substantially better than the ~18mA you measured.

Tim

 

[Grant]

Board top view?

Yup, there now, with commentary

"
Those three resistors standing up behind the fuse are part of
the output current shunt resistor, another three resistors are
further back.

Next trace to left is mains neutral, next trace to the left of
that is mains active, going to the fuse's right hand terminal.
Next trace goes under the fuse is mains earth!

There's none of the usual large separation of secondary side and
mains side traces.

Left side of fuse goes offboard to power switch, back on board
to the right of the yellow X2 cap onto the tab also marked switch.

Behind the green core common-mode filter there's bridge rectifier
just out of sight. It feeds the two 330/200 electrolytics connected
in series in foreground with the classic voltage doubler option
tabs east and northeast of the left electrolytic cap. A couple of
150k 1/2W resistors balance the caps.

On the left side just behind the mains entry, and to the left of
the right hand side electrolytic capacitor you see the two small
blue Y caps.

In particular, note how the Y cap on left goes from neutral to
earth, which is also the DC output negative line.

The blue cap to the left rear of the green cored CM filter is
centre-tap series capacitor for half bridge drive to the power
transformer. Half bridge power transistors are K3469, controlled
by a signal transformer. No sign of opto, but there's a second
signal transformer likely used for feedback.

Output is rectified by TO247 U30D20C with snubbers, feeding
2200/35 with 200 Ohm 2W across it, through a filter toroid to
another 2200/35. Secondary power centre tap goes to the right
hand side of the vertical current sense resistors mentioned
above.

There's more control circuitry in a tinplate shielded box not
in frame, control signal in front panel is a triple(!) gang pot,
two sections control voltage while a third section in opposite
sense controls max current.

Output adjustable 0 - 24V and output current depends on voltage
and loading.
"
Hope that helps?

I'm guessing it has neither X nor Y-type capacitors on the input?

Got them, just doesn't follow the common layouts I'm used to seeing

The way it's supposed to be laid out, Y-types RF-ground the circuit side, and a common mode choke limits ground loop current. X type caps provide differential filtering, bypassing line and circuit sides; leakage inductance in the CM choke provides a crude lowpass, but this is sometimes enhanced with a series choke of 10-100uH to provide more differential filtering.

Yeah, that I know X2 cap is 100n, Y2 rated caps are 2200pF.

No extra series chokes, there's a couple more Y caps on other
side of CM choke, snubber cap 4700pF plus 5W resistor across
power transformer primary, that's about it for the mains side.

The maximum value for Y-type caps is somewhere around 4700pF; more and the 60Hz leakage can cause problems (564kohm at 60Hz, or 0.4mA at 240V, enough for a minor shock hazard I guess). However, this sounds substantially better than the ~18mA you measured.

Tim

As stated in the text, that signal on DPO is twice current reading
for voltage shown, as I'm feeding 50R into coax with 50 Ohm
termination on input (divide by two from one Ohm shunt). So you're
seeing about 40mA, trip point is about 50mA (+/- 2 1/2 divs at
20mA/div).

Oh, and the PCB reference designators show component values, so
those four Y caps are all marked CF222, the two 150Ks across mains
filter electros: R150K, I see an R39, the colours are orange white
black gold. This is one very strange layout!

I'll put the thing back together and measure the ground noise
with high load (200W of lamps) connected and post results later,
might need to remove the diodes -- I put the diodes in, in case
of fault. After all, the idea of cutting the earth lead is not
appealing, I need to check for DPO ground interference to the
measurement too. That the RCD triggers on this noise is real.

Mains is 240V, 50Hz in the land downunder.

BTW: to the poster suggesting line filter, I've yet to meet a line
filter that does anything for interference conducted along the
ground lead

Grant.

 

[Grant]

Yup, there now, with commentary

Now added new measurements, without that confusing x2 factor

http://grrr.id.au/ps-earth-current/#new_measurements

DPO now shows earth 1R shunt rms and p-p voltage, as well as PS'
DC output voltage.

Grant.

 

[Hammy]

In particular, note how the Y cap on left goes from neutral to
earth, which is also the DC output negative line.

Are you saying that safety earth ground is used for secondary
refrence---> return currents?

Strange I didnt think any current other then small leakage or fault
current was allowed into earth.

I know its not uncommon to have earth ground on the secondary but its
for fault detection. Primary to secondary insulation failure not for
secondary return currents.

If that is the case wouldnt this cause your earth ground which is
ideally at 0V to bounce all over the place? Given transformer DCR and
trace R and the high peak currents bounceing around the secondary.
..

 

[Grant]

(snip)


Just to humour me (note spelling - also in LDU) trying running ALL the
supply (A/N/E) though a single ferrite toroid. If it makes a
difference, I'll happily explain the rationale, otherwise I'll STFU.

Thanks, I didn't think of running all 3 leads through toroid, will
try it. I see your rationale

All I saw was line filter == X and Y caps and CM filter for A & N,
nothing for E, therefore nothing for my problem? After all, I am
asking here for suggestions, even if I need to be hit with cluebat
at times.

Cheers,
Grant.

 

[Grant]

Are you saying that safety earth ground is used for secondary
refrence---> return currents?

Yup, madness in there

Strange I didnt think any current other then small leakage or fault
current was allowed into earth.

Same here.

I know its not uncommon to have earth ground on the secondary but its
for fault detection. Primary to secondary insulation failure not for
secondary return currents.

What I have seen is earthed screens (electrostatic screen?) in
power transformers, single turns of foil (non-shorting) between
primaries and secondaries, connected to mains (safety) earth.

If that is the case wouldnt this cause your earth ground which is
ideally at 0V to bounce all over the place? Given transformer DCR and
trace R and the high peak currents bounceing around the secondary.

That's exactly what I think is happening, seeing high injected
current to earth, which is also DC output negative. There's no
earth to the transformer indicating shielded windings.

The RCD is tripping at about 12mA rms, as reported by the Tek
TDS3034, dunno if that is true rms, ought to be, for an instrument
with AU$13k replacement value But there may be some vagueness
in the measurement as I don't have a diff amp input module for the
DPO, and I'm wary of isolating the mains ground to it. After all,
we got safety earth fitted for a reason.


In general, the product seems a new design (over-designed, too many
internal connectors for a tight production design), by somebody
who's not read the book on modern mains rules or layout techniques
very well (example, the ref. des. -> part value, not a part
reference).


Contrast this power supply with others I often see (at similar
power levels) that has multiple CM chokes, several X caps,
usual Y caps, TVS + PTC components, and an obvious divide between
mains and output sides. Might add some examples of what I consider
decent layouts for the 250W power level. Most I see are switching
battery chargers, 6A to 10A into 12 cell SLA or VRLA batteries.


The problem power supply is sold by a .au & .nz nationwide
electronics chain (Jaycar), you'd think they'd follow the rules?

I dunno which authority over here to ask if this does meet the
C-tick mark it claims. All stuff sold here is supposed to have
one, on the one hand, but is ignored by some store chains on the
other, seemingly because they can claim to be simple importers
and sellers? But I doubt Jaycar can claim that.


It doesn't meet the rules as I remember them from decades ago,
particularly the physical isolation barrier between mains and
secondary.

But then, I once return a PC PSU under warranty when its active
PFC board went Bang! --> tracks too close together for 240V.
Oddly enough, the warranty replacement was a different model by
same manufacturer, indicating a design flaw?

Grant.

 

[Grant]

(snip)


Just to humour me (note spelling - also in LDU) trying running ALL the
supply (A/N/E) though a single ferrite toroid. If it makes a
difference, I'll happily explain the rationale, otherwise I'll STFU.

No help But, there's more!

Added another filter after the toroid and now the power supply runs
the 200W lamp load without tripping the RCD ) Definitely see a
reduced earth current waveform, it's an avenue worth exploring a bit
more.

I could manage five turns of A/N/E through a couple of 1 inch OD
by half inch joined to make 1" by 1" toroid. Did only the slightest
improvement on its own, but adding the filter that has X, 2 Y caps
and another CM toroid did the trick. Nothing I can fit inside the
PS box though.

The ferrite used is from mains filters, not switcher toroids, I had
some larger ones in a bag but cannot find them. I've got some other
large toroids from switcher power supplies to try as well.

This result means I got to go through the other testing done with
toroids just in the ground lead (as is quite common in this size
power supply), to see if the extra filter helps there too, be a
bit smaller. Five turns of just the earth lead through a small
toroid helped a bit, I want to see if the extra filter helps there.


Will add this info and photos to the web page soon. Tell me your
name if you'd like a credit (my email addr above is valid)

I'm thinking of changing the power supply DC filter by removing
the first cap connected to the main diodes, a choke input filter
would reduce peak current through the diodes and is a far more
common configuration for these higher current converters.

Assuming, of course, that the thing wont blow up or go haywire
with the different feedback dynamics.

Thanks,
Grant.