Why would someone use 10nF Capacitors in parallel with each diode of a FWR?

[jakeisprobably]

Why would someone use 10nF Capacitors in parallel with each diode of a FWR? I came across this inside a small wall wort plug and was curious about the reason. The diodes are simple 1N4007's and the capacitors are all ceramic in parallel with each diode. Then these feed into the typical electrolytic cap for smoothing.

[Benta]

I've seen this before.
The turn-off of the rectifiers creates current spikes, and to get the supply through EMC testing, it can be necessary.

[Fraser]

As already stated, the diodes create noise that the capacitors counter. Such is often recommended for producing a low noise power supply.

Fraser

[David Hess]

Standard recovery rectifiers can behave like step recovery diodes in power line frequency applications causing significant impulse noise repeating at the power line frequency.  One easy solution is to shunt each diode with a small capacitor although I generally find that 220pF is plenty for a 1N400x rectifier.  I have also seen EMI beads placed in series with each diode to solve this.

Whether diodes do this varies by manufacturer and even by lot so another solution is to qualify and test the diodes.  The problem does not seem to occur with fast recovery rectifiers.

[Kleinstein]

One sees this quite often in older audio circuits. The extra capacitance seen by the transformer is not bad as the magnetizing current is a inductive load and the current through extra capacitors is opposite phase. So there is no problem even if the caps a larger. I have seen 100 nF too. Already only 2 caps can help.

It is unusual to find it in a wall wart - these are usually build to cost.

[MrAl]

Hi,

If i remember right it also helps with the reverse voltage distribution across each diode as the circuit operates.  To see the effect, you could do a simulation with say 0.01uf caps, then decrease one to 0.001uf or remove one completely.  You might also be able to see an effect with just one cap across one diode.
The idea is to simulate a larger difference in equivalent parallel capacitance between diodes and see how the reverse voltage builds across all four of the diodes.  By tacking caps that are significantly larger than the intrinsic capacitance across each diode helps to ensure better capacitance matching and thus better voltage handling.  This is probably more significant when using 200v rated diodes on a 120vac system (for example) where the peak is closer to the rating of the diodes.  If the circuit acts as a voltage doubler for even a tiny fraction of a cycle it may blow the diode.

It would be interesting to get reader feedback on this issue to try to see if they went to using 400v diodes now or are still using 200v diodes and the four capacitors.  Maybe we could form a consensus.
Readers would have to submit their diode part numbers and capacitor values if any, and of course operating line voltage.

[Seekonk]

I was always of the opinion they did this back when diodes had quality issues and this increased the spike immunity a long time ago. No one back then was worried about an EMI issue. Certainly it has gone out of fashion.

[T3sl4co1l]

Even soft-recovery diodes exhibit recovery; the mere fact that recovery exists, guarantees some harmonic generation.

Or, well... let me take an even further step back...

The mere fact that diodes exist, guarantees harmonic generation.

Whether that's a problem, depends on what harmonics are a problem.

FYI, the usual mechanism of diode noise, is reverse recovery causing transients in the circuit around the diode.  This can be dampened by placing an R+C across the diode, where C > diode CJO, and R ~= sqrt(Lstray / CJO).  If a very large C is used, R doesn't need to be very large, so you often see 10nF caps and that's that.  (The diodes still cause ringing, but the amplitude is small, and it's shifted to a lower frequency.)

Any effect that resembles reverse recovery still has the same result, so merely using fast recovery diodes, or schottky diodes even, isn't necessarily good enough.  (Schottky diodes have so much capacitance, at Vr = 0, that it looks like reverse recovery anyway.  The difference is, it's conservative rather than lossy, so you get the energy back the next time you forward-bias the diode.  On the other hand, if that energy is simply dissipated as switching loss, it's gone for good, so you can actually incur higher losses by using schottky diodes in a circuit not suited to them.)

In extreme cases, even just the simple fact that the diode rectifies -- even if you had an ideal one, gives rise to harmonics.  If the surrounding circuit is easily excited, then one might add some damping across the diode, to tone it down.  Here again, an R+C will help.

If the surrounding circuit is largely inductive, it will ring against the diode capacitance, in which case a ferrite bead can be used to make that capacitance lossy.  This doesn't work so well if the surrounding circuit has its own capacitance and such.

Tim

[David Hess]

Any effect that resembles reverse recovery still has the same result, so merely using fast recovery diodes, or schottky diodes even, isn't necessarily good enough.

But some diodes even of the same type but from different lots and manufacturers are much worse than others and some standard recovery rectifiers seem to be the worst when used for line frequency rectification.  I have yet to see the problem show up when replacing 1N4001 series diodes with 1N4933 or 1N4942 series fast recovery rectifiers or schottky rectifiers. 

I get the feeling from studying the differences in various Tektronix schematics over time that at some point they became aware of the problem and initially added capacitor shunts and then later qualified the diodes for low recovery noise.

[T3sl4co1l]

The most common problem will be 1N400x having 2-3us recovery time, making them 10-20 times more prone to problems than fast recovery diodes will be.

If that recovery is soft, turn-off (the falling edge part) may be slower than the entire recovery phase of a fast diode, which wouldn't be too bad, even with the longer overall recovery time -- time in which extra energy is stored up.  Whereas if it's sharp, it's only going to be worse.

Downside to slow diodes, of course: you have no guarantee of recovery time, or softness.  And in practice, they seem to group into a couple types: lightly doped, diffused, PIN, whatever.  Only way to tell is to test it.

Tim