power supply input stage: surge protection, EMI filtering, how?

[Jenalee K.]

Could someone please explain how to design a decent input stage for a
power supply?

Let me explain: I have two wires coming in from somewhere with about
24VDC between them. These two wires go to an 8W DC/DC converter (Traco
TEN8-2411) which gives me 5V at its output terminals. I now have to
take care of three things (industrial device):

- overvoltage and reverse voltage protection: I've put a 6A 100V diode
(P600B) and a transil P6KE33A in parallel with the power input through
big fat pcb traces. In case of reverse voltage I assume the PSU's
current protection will kick in before my diode and/or pcb starts
burning.

- EMI filtering. According to the datasheet the DC/DC converter meets
EN55022 class A and FCC class A specs. How does this relate to CE
requirements? Do I have to add anything? I've thought about a common
mode choke with 100n caps between its terminals at both sides and 10n
1kV caps to earth. (as seen somewhere)

- surge protection. How do I handle that? I've seen varistors to earth
but I've also seen pcbs with burned varistors to earth. I found in the
archives that varistors are notorious for starting fires. How do I
choose a varistor? Do I want a varistor?

Thanks,
Jenalee K.

 

[Terry Given]

Could someone please explain how to design a decent input stage for a
power supply?

Let me explain: I have two wires coming in from somewhere with about
24VDC between them. These two wires go to an 8W DC/DC converter (Traco
TEN8-2411) which gives me 5V at its output terminals. I now have to
take care of three things (industrial device):

- overvoltage and reverse voltage protection: I've put a 6A 100V diode
(P600B) and a transil P6KE33A in parallel with the power input through
big fat pcb traces. In case of reverse voltage I assume the PSU's
current protection will kick in before my diode and/or pcb starts
burning.

If you can handle the losses, place the diode in series with the input.

- EMI filtering. According to the datasheet the DC/DC converter meets
EN55022 class A and FCC class A specs. How does this relate to CE
requirements? Do I have to add anything? I've thought about a common
mode choke with 100n caps between its terminals at both sides and 10n
1kV caps to earth. (as seen somewhere)

I'd take a large grain of salt with that datasheet statement.

you need to know the input current shape (IOW spectra). then choose a
suitable input cap (most dc-dc bricks have little or no input
capacitance) to turn the wobbly I into a slightly wobbly V. Then bung in
a suitable series (DM) L to turn the slightly wobbly V into an almost
wobble-free I, that meets the EMI spec

(rough guide: I_slightly_wobbly*50R = Vemi_limit)

The DM L can help the transorb deal with a voltage spike, if the
transorb is across the DC-DC rather than the input.

and some form of CM choke too.

EMI filtering is all about parasitics. Mark Nave's book is worth reading
(power line filtering for switch-mode power supplies)

- surge protection. How do I handle that? I've seen varistors to earth
but I've also seen pcbs with burned varistors to earth. I found in the
archives that varistors are notorious for starting fires. How do I
choose a varistor? Do I want a varistor?

Thanks,
Jenalee K.

personally, I dont like MOVs, they have a built-in failure mechanism. I
use transorbs instead. regardless of what you use, either device can
(and will) fail short-circuit, so you need to deal with that too.

HTH

Cheers
Terry

 

[w_tom]

To "design a decent input stage for a power supply", first define
different types of problems. For example, brownouts, blackouts, noise,
short circuits, and surges. What will happen if that 24 volts drops
too low or blackouts repeatedly cycle. Are those problems solved in
the DC-DC converter or in 24 VDC supply? For example, when voltage
drops too low, the converter might disconnect (shutdown) making low
line voltage irrelevant. Repeated power cycling may cause the
converter to lockout - another internal protection.

Noise would be made irrelevant by that DC-DC converter or solved by a
line filter. Not just incoming noise to the converter, but also so
that the converter does not radiate noise. Such filters include a
choke on each line that is inductively coupled to its peer. Learn more
from line filter manufacturers such as those listed in Digikey, Allied
Electronics, Mouser, Newark Electronics, etc.

What happens when the 24 VDC supply is shorted? Will it safely move
to foldback current limiting as has been standard for generations?

What happens when converter output is shorted. Again, does converter
also protect itself - or must you?

Consider differential mode and longitudinal mode transients. The
transil as described would shunt a differential mode transient - ie
excessive voltage between both 24 volts wires. However, if transient
power is too high, then that transil may fail shorted. A line fuse
would disconnect a shorting transil from 24 VDC source.

Longitudinal mode transients come down either or both volt wires to
seek earth ground. These are best shunted back at or before a 24 volt
source via MOV (varistor) or transil/tranzsorb from each wire to
earthing. Not just chassis ground - earthing. Longitudinal mode
cannot be shunted from one wire to the other. A longitudinal mode
transient may be 300 volts on both wires; and zero volts across the
in-parallel 6KE33A transil.

Protection starts by defining both types of transients AND by
designing for when that protector device might fail catastrophically.
Also learn what is already inside the 24 volt supply since longitudinal
mode transients would be better solved on other side of that supply and
since it should make short circuits irrelevant.

Again, each protector device should include some type of protection
(ie thermal fuse) so that a short circuit does not create other
dangers. Check Littelfuse for their TMOS series that includes a
disconnect function and application notes for designing.

Using 10nf 1Kv caps to chassis should be of type so that a failure
causes an open circuit - not short circuit.