"Out of the Box" solutions for diode switching noise

[hylaphone]

Hi folks. Somehow many of my projects end up in small enclosures, and while I've learned to design for quiet grounding and lead dress, the dreaded diode switching noise often makes an appearance.

There have been a few very informative threads on the subject here, I've been through them and tried most every solution offered - snubbing, slurring, twisting, swearing, you name it.

On a recent project giving me fits, I relocated the bridge rectifier outside of the chassis as an experiment, leaving the rest of the power supply inside the box. This quieted things right down, but obviously is no solution.

Got me wondering about shielding - in the myriad of modern semiconductors, is there such thing as a rectifier in a shielded package?
Or, if I were to experiment with mu metal or something, which composition might be most effective?

Lastly, I've only been using full wave bridge configuration. Theoretically, would switching noise be more or less prominent in half wave? Or even voltage doubler circuits?

[Gingertube]

One of those treat the disease and not the symptom things.
Best not to generate the diode switching noise in the first place, or at least limit the diode switching noise.
Best options are of-course the most expensive - use Silicon Carbide Schottky diodes or incorporate a tube rectifier in a hybrid arrangement.
Next best option is to use Ultrafast Soft Recovery diodes, the UF4007, UF5408 etc. these generate about 1/5th of the switching noise of 1N4007, 1N5408.

Most power supply filters are reasonable about removing the switching noise from their supply line but do nothing about the switching noise coupling back through the power tranny to other windings. This is where snubbers can help, put a snubber across the AC winding before the rectifier.

Cheers,
Ian

[R.G.]

G.T. is right - before you can stop something, you have to figure out what you're stopping.

Rectifier noise is a problem with RF transients being radiated or conducted into your audio circuits. It's a form of radio interference.

It's quite important to be sure that you are not causing "rectifier noise" by your wiring. You're making the right noises about this, but the connection of wires from transformer and rectifiers to the first filter caps is critical, as improper connection of these wires can and will cause noise otherwise indistinguishable from picking up the RF pulses of diode turnoff.

The source of diode switchoff noise is usually ringing in the parasitic capacitances and inductances of the wires connected to the diodes. At RF, these are pretty high-Q resonators and the sudden turn-off of diode conduction when the diode's stored charge is swept out of the conduction region in the diode makes the diode slam off so fast that the inductance of the wires makes the wires ring with the parasitic capacitances. The wires are self-antenna-ed, so bursts of RF radiate and conduct down the wires until the energy stored in the wire inductance is dissipated.

There are three ways to get rid of RF noise: (1) don't generate it in the first place, (2) make the circuit immune to it, and (3) shield it out if you can't do 1 or 2.

G.T.'s note is right on for 1; change rectifiers so less RF energy is generated. Fancy rectifiers can do this to some extent by not slamming off. Less unwanted RF is always good.

It may not be enough. You're clever enough to recognize that you're stepping on your own toes by compressing things into smaller and smaller boxes. This reminds me of the old story where the guy says "Doc, it really hurts when I do ...this..." and the doctor replies "Well, don't do that." Tubes are high frequency amplifying devices with RF gain. Eventually, if you keep stuffing them right up next to RF sources, they're going to pick it up.

So you go to 2. You can make the circuit less prone to pickup by doing good RF practice on the tubes, with RF bypasses on the B+ and other voltage supplies, by using grid stoppers, ferrite beads on the grids, and low value, low inductance resistors shunted to low-impedance planar grounds on both sides of the resistors in the signal path, especially at the input tubes. You can also tune the tubes' frequency response to be very low above, say, 100kHz by tinkering small caps from plate and grid to ground.

You started with 3. Shielding the rectifiers may or may not work. A shield doesn't so much eat up RF energy as force it to go elsewhere. And it's not the rectifiers that are transmitting that grak to you, it's the wires connected to them. So a shield on the rectifiers will only force MORE of the RF down the wires. Whether this is better or worse depends on whether conducted or radiated noise is worse in your circuit. What you want to do is to either not generate as much RF, or eat up more of it right out at the source. Putting a shield around an RF source is much like putting a glass bottle over a leaking faucet. The water direction may be changed, but the total water coming out will be the same. The water's got to get out somehow. For RF, we want to convert it into heat inside the shield. If the shield is just an RF-impervious barrier, the RF will leak out the holes or wires.

Snubbers are designed to eat the RF close to a noise generator. Snubbers are our friends, even with silico-nano-hyper-Schottky-silk rectifiers. They take tuning though, because they have to DE-tune the wiring parasitics. Small caps and low value carbon comp (YEAH! We found something that carbon comp is really, really good at!) resistors can help here.

Last stop is 3, shielding. The best and cheapest shielding is distance. RF is an inverse-square-law thing. Double the distance, quarter the energy. Half the distance, FOUR TIMES the energy. This is the one you're cutting yourself on. Less effective is shields that lead to RF traps to dissipate energy as heat, less effective still is dissipative shielding. Least effective is reflective shielding.

[hylaphone]

Much to respond to here, thanks guys.

I have seen switching noise coupled to other windings before, in this case I'm using a separate filament transformer so no problem there. And no sign of spikes on the B+ rail.

This morning I mounted the rectifier to a long phenolic strip, with the terminals carefully insulated. Moving it around the chassis I was able to listen to the buzz and observe the spikes on the scope.
Of course, the quietest spot was furthest from the signal path. So this suggests radiated noise. I will relocate the bridge, and reorient that transfer for shortest twisted run of secondaries.
Then move on to RG's fine suggestions for RF if need be.

I've found UF diodes only slightly quieter than their 1N counterparts, but more responsive to snubbering for whatever reason. I will save my pennies to try out some SiC Schottkys..

Will play around with this and report back. Thanks again ---

[trobbins]

I'm using a separate filament transformer so no problem there. And no sign of spikes on the B+ rail.
This morning I mounted the rectifier to a long phenolic strip, with the terminals carefully insulated. Moving it around the chassis I was able to listen to the buzz and observe the spikes on the scope.
Of course, the quietest spot was furthest from the signal path. So this suggests radiated noise. I will relocate the bridge, and reorient that transfer for shortest twisted run of secondaries.

Does this mean you can't locate the rectifier diodes and first filter cap close to the transformer, and so keep all interconnecting wiring very very short and twisted and not placed in close proximity to signal circuitry? Just the leads between transformer and diodes (whether twisted leads are used or not, and whether 1N or UF or special schottkys are used) will capacitively couple hum and leakage inductance related dV/dt noise to circuitry, so extending those leads and moving them around an amp is very likely to show up as noise in amp circuitry.