What can go wrong? (Preventing catastrophic failure on mains disturbance tester)

[mcinque]

So I ended with this small circuit to see and hear with a SDR any disturbance/noise in 2KHz-2MHz range injected on mains by an electronic light bulb under test.

Now, after simulation and testing with a DMM to see that the output is safe for the instrument (circuit output in normal conditions is 124mVpp while instrument input is MAX 0dBm/225mVRMS, PEAK +10dBm/710mVRMS, that for extra safety will be used with a 10dBm attenuator and a DC block), I'd like to receive some advices and observations about the safety of the SDR/SA instrument connected.

The critical component for the instrument safety connected at the output of R4 is of course C2 since is the primary connection with mains.

I've been teached that that "Y caps are designed to be safe, metallized paper and film caps tend to open circuit IN FAIL MODE."

If C2 fails, since is a Y metallized paper type, it's engineered to go open, so normally all the circuit beyond this component is considered safe.

But since catastrophic failure is always an option and it would fry my SDR/SA, let's consider the worst case:

First of all: it is possible that a 300V metallized paper Y cap go short on failure mode?

If C2 goes short R2 would pull 100mA of current thru earth tripping the 30mA RCD (usually in 35mS). In this amount of time 240VRMS from mains would be present on R3.

EDIT: R6 and R7 would dissipate 10A and 2.451W so they will fry in normal conditions but I think they can keep that power for 35mS. In the worst case, they go open and most of the voltage will go on the ouput of R4, frying all.

Am my thoughts right? What would you modify to have absolute safety of the SDR/SA instrument connected to R4?

[Zero999]

It should be safe. Ideally you should use Y1 capacitors as they have a higher breakdown voltage, than the other types.

Pay attention to the creepage and clearance distances.

A transient voltage suppressor diode or MOV can be used to protect the SDR from any high voltage spikes which get through your filter.

One thing to bear in mind is the circuit will behave differently depending on the polarity of the mains, which could be a problem if the mains connector isn't polarised.

[Kleinstein]

Usually proper class Y caps are very safe: in other circuits if they would fail they could potentially expose mains to the user. So the testing requirements are really high. With Y2 caps sometimes 2 caps in series are used.
Not sure you can get 15 nF though - most class Y caps I have seen are 4.7 nF.

Something like an MOV / sparc gap for protection might be an option. This could be even before the capacitor to prevent any possible spike to reach the capacitor. For the lowest frequency part the RC combination might already be visible.

It might need an input filter, to keep out noise from the grid and offer a defined impedance environment.

For the output signal some kind of over-voltage protection with low capacitance diodes / trans-sorbs or similar are a good idea. In case of a really bad device (e.g. lots of spark on a defect collector motor) the RF amplitude could reach damaging levels otherwise. Some 50 Ohms inputs are not very tolerant to high amplitudes.
Common mode protection towards ground might also be a good idea. There are two grounds: the signal source and PE from mains. They are likely connected at the receiving end (e.g. scope or SA), but might have to be treated separate in this tester part.

[Zero999]

Usually proper class Y caps are very safe: in other circuits if they would fail they could potentially expose mains to the user. So the testing requirements are really high. With Y2 caps sometimes 2 caps in series are used.
Not sure you can get 15 nF though - most class Y caps I have seen are 4.7 nF.

Yes, I believe that's to keep the leakage currents down. 1.5nF capacitors could be used but the lower cut-off will be increased, unless all the resistor values can be increased by a factor of 10.

[mcinque]

A transient voltage suppressor diode or MOV can be used to protect the SDR from any high voltage spikes which get through your filter.

Right.

Pay attention to the creepage and clearance distances.

Yes, I will use proper isolation slots and clearance, thank you for the reminder 

One thing to bear in mind is the circuit will behave differently depending on the polarity of the mains, which could be a problem if the mains connector isn't polarised.

Here main sockets and plugs aren't polarized (bad), so I will manage this issue by connecting the socket in the proper way. I could also use a neon lamp and a 150K resistor between live and earth to show that the connection is done properly before connecting the instrument.

Not sure you can get 15 nF though - most class Y caps I have seen are 4.7 nF.

Found this.

It might need an input filter, to keep out noise from the grid and offer a defined impedance environment.

Yes, I've not included in the schematic but is required, otherwise the measurement would be affected.

For the output signal some kind of over-voltage protection with low capacitance diodes / trans-sorbs or similar are a good idea.
In case of a really bad device (e.g. lots of spark on a defect collector motor) the RF amplitude could reach damaging levels otherwise.
Common mode protection towards ground might also be a good idea.

Right. What do you think about the modified schematics?

[SeanB]

Would not bother with R2, just leave it out, and instead use the divider as the load resistance. As well add another MOV across the mains, and some series inductance to decouple the incoming mains noise ( this will be substantial, unless you are using an isolation transformer  to provide isolation) as this will also be measured.

The neon to show reversed line neutral would work, but there is also a 3 neon circuit that shows almost all forms of socket fault, from reversed line and neutral to loss of a neutral or loss of PE.  Safer, plus more likely to show up a fault.

[mcinque]

As well add another MOV across the mains

This to protect all the circuit from surges and spikes introduced on mains from any capacitive/inductive kickback, right?

and some series inductance to decouple the incoming mains noise ( this will be substantial, unless you are using an isolation transformer  to provide isolation) as this will also be measured.

I planned to use a Schaffner EMI mains filter to keep out the noise. Wouldn't do that job better?

but there is also a 3 neon circuit that shows almost all forms of socket fault, from reversed line and neutral to loss of a neutral or loss of PE.  Safer, plus more likely to show up a fault.

Yes, I've seen it on an RCD tester. I will make and use that type. Thank you for the suggestion!