advertisement


Snubber capacitors at main rectifier

13mh13

Pi-s Artist
I've encountered a few different ideas on how a snubber at the PSU rectifier (i.e., mains transformer/diode area) should be implemented.
At its simplest, the two or four rectifier diodes should have a small-value poly/film cap across each of them, as in this Arcam Alpha 5 CD player:
Arcam_apha_CD_5_PSU.jpg
.


Alternatively, Peufeu suggests a somewhat different option:
" The snubber is not placed around each diode, but rather across the transformer secondary, which is IMHO better. "
supply_outboard.gif


I think there are other snubber options which may also work (improve PSU performance).
Any thoughts/suggestions?
Certainly, the first approach is the easiest to implement.
Note: I think I have implemented both methods in past projects. Alas, I did not take careful notes; so, your input would help!

Thanks!
 
Yeah, that's the now-classic reference. And it seems everyone and their DIY uncle implement snubbers AFTER the main PSU (i.e, at the regulator or at opamps).
I'm more curious about implementation at the mains rectifier.
What do you think about that Arcam topology? Note different subber cap values (across diodes) for various regulator voltage: 1nF (for 5v) and 2x47nF (for 15v).
 
The series resistor is important to absorb the RF ringing energy. 33R to 100R is a good starting value
 
There was a very good article in one of the linear audio magazines, cant rememer which right know.
 
I replaced smoothing caps on my XTC Pre 1 yesterday - the rectifiers in this (crappy small bridge for a 12v supply), separate diodes for the +/- 30v supplies all have caps across each diode junction, but no resistors - worth adding? For the small cost I'd normally just replace the diodes with better ones.

Richard
 
I am not keen on caps across the junctions - just feeds crap through when the diode should be isolating.

The input RC snubber damps resonances in the transformer; when the diode dumps out its charge packet the main problem is exciting these resonances. With the input snubber, you no longer need to provide a path for injecting high frequencies, but instead absorb the energy at source.

As davidrsb noted, you want this damping to work broadband, so use a reasonably low inductance cap on short leads, or better still one of the ready made RC snubbers that Maplin sell.
 
"Better" diodes can make things worse by turning off more abruptly. I recently had a Schottky ringing strongly at 50MHz with about 5cm of track. It was radiating all over the place. Just using a capacitor reduces the ringing frequency and can even make the amplitude and duration worse with higher Q. The resistor part is what stops the ringing
 
The phenomenon of noisy diodes is called "shot noise" and was discovered by Walter Schottky who was researching noise in vacuum-tube diodes, but it applies to solid state. In silicon it has to do with quantized electron migration across the PN junction. Starts in the hundreds of KHz and moves up from there.
 
I am not keen on caps across the junctions - just feeds crap through when the diode should be isolating.

The input RC snubber damps resonances in the transformer; when the diode dumps out its charge packet the main problem is exciting these resonances. With the input snubber, you no longer need to provide a path for injecting high frequencies, but instead absorb the energy at source.

As davidrsb noted, you want this damping to work broadband, so use a reasonably low inductance cap on short leads, or better still one of the ready made RC snubbers that Maplin sell.

Correct in every point
 
"Better" diodes can make things worse by turning off more abruptly. I recently had a Schottky ringing strongly at 50MHz with about 5cm of track. It was radiating all over the place. Just using a capacitor reduces the ringing frequency and can even make the amplitude and duration worse with higher Q. The resistor part is what stops the ringing

The resistor damps the ringing by converting the noise into heat. This must be a non-inductive, so carbon composite is better than spiral-cut metal film. Resistor size is a tradeoff. The smaller the resistor the more power it routinely passes, but the faster the damping. The bigger the resistor the lower the power it routinely passes, but the longer the damping takes.

The capacitor prevents the QRR, which is the charge held internally to the diode, from creating a high-frequency spike in the power supply. (Charge across a junction.) Because the QRR is rapidly dumped, it creates high-frequency noise that passes through the filter caps.

Yes, High Q makes it worse. Low Q caps (ceramic or mica) are commonly used for snubbers. Short leads always.

Schottky diodes are better than regular silicon (no PN junction, so QRR is vastly lower), but different types are better than others.
 
I think you mean value not physical size. Suitable values are in the range ~22R to 220R.
SMD is even better than carbon, lead length is critical at VHF frequencies
 
I think you mean value not physical size. Suitable values are in the range ~22R to 220R.
SMD is even better than carbon, lead length is critical at VHF frequencies

Yes, of course. Thought that was clear. Larger in terms of value, not physical size.

Lead length also matters for HiFi, because the Qrr dumped by a diode is at UHF/VHF frequencies. (So is shot noise.) This acts a carrier wave which is then AM modulated by mains frequency so it's buzz all around if any wires are near the diode to act as antennas. Transformer LC noise is much lower frequency, BTW.

SMD resistors (usually tin oxide on ceramic) are great for switching power supplies where space matters and frequencies are very high, but not very practical to piggyback onto an existing rectifier diode (or a tube socket).

Adding RC snubbers to existing circuitry, where they were omitted out of ignorance of bean counting, reduces noise and is an easy tweak.
 


advertisement


Back
Top