Noise manifestation in rectifiers?

[Chris55000]

Hi!

I was reading with great interest the thread that dealt with silicon diode/thermionic–valve rectifier combinations in Marshall amplifiers, and the point was made about RF rattle from silicon rectifiers.

Nearly all consumer, T & M and industrial gear using standard linear PSUs I've seen has a capacitor of about 10-50n across each diode to suppress it, but can anyone tell me:–

1) What were the audible manifestations of this, and can it occur eleswhere in linear PSUs?

2) Did metal or contact–cooled rectifiers suffer from this or were good ones very quiet in operation?

I'm well aware gaseous mercury–vapour rectifiers can cause horrendous noise problems, even when otherwise functioning apparently normally, what did this sound like?

Chris Williams

[trobbins]

Whether or not you can hear something that is related back to the rectifier would, imho, depend completely on the particular equipment and how well it was made.

As I see it, a cap across a diode could be for three reasons - one is that the equipment needs the cap to pass EMI compliance levels (ie. > 150kHz noise emissions either by conduction or radiation) - not that other means could not have been used, but that the cap addition was the simplest way to get a pass with the compliance test. The second is that the choice of transformer and diode and layout has meant the manufacturer has experienced a noise problem and used the caps to subdue it. The third is that the manufacturer or designer is just copying from other equipment without have any good reason or idea about what they are doing, and is just hoping that using the caps causes no issue.

Some transformers have substantial leakage inductance, and even with fast diodes, there is a splay of noise each time the diodes turn-off, that eeks in to audio stages through poor layout. A properly determined snubber across the offending winding is the preferred means to suppress that leakage inductance issue, but that can be a step too far for some, so they find solace with using a cap across each diode.

[stevehertz]

I once made a HT power supply unit for use with vintage battery radio sets. Using the PS produced lots of noise hash. At the time I wasn't really sure what was going on so I approached a wily old design engineer at my place of work armed with a diagram of my rather basic bridge rectifier circuit, plus smoothing of course. He pondered it for a minute or so then told me to fit a cap across each diode in the bridge. I went away and did so and hey presto the noise hash was gone.

[Radio Wrangler]

Several things all come together.

Noisy rectifiers make RF noise, not AF noise. So badly chosen diodes or a lack of suppression won't necessarily upset an audio amplifier, but it will interfere with any radios in the vicinity and put up the general noise floor of hash. They're just antisocial.

Modern products have to meet EMC requirements before they can be legally sold and soon after testing became routine, people started discovering surprising levels of emissions coming from even circuits that didn't contain any RF stuff. This got traced to the rectifier diodes and action had to be taken to get a pass mark.

Wiley designers had been fitting suppressor capacitors across transformer secondaries and across rectifier diodes for years since the rise of semiconductor rectifiers. Reasons were twofold. Filtering is a two-way thing and the capacitors filtered spiky transients coming from elsewhere on the mains network and reduced the incidence of rectifier failure. This made a big difference in the BY100 era. Secondly, the filter capacitors across diodes in a radio receiver stopped the PSU interfering with its own set.

The EMC regulations don't care about causing interference with your own stuff, that's your own problem, pal, and is assumed to be incentive enough. They are about interfering with your neighbours. Rectifier and SMPS noise using the mains lead as an antenna is one of the main problems.

Rectifiers have improved over the years. The development of types with controlled avalanche breakdown characteristics lost the fragility of the early types and boosted reliability significantly.

Faster acting diodes have also reduced a nasty habit of early diode designs. The higher voltage parts in the 1N4000 series were designed as PIN diodes with an intrinsic layer in the sandwich. This boosts their voltage rating dramatically, but it also makes them hellishly slow. The slow turn-off is the problem. Even as a 50Hz rectifier, they are too slow. Appreciable reverse current has started to flow before they turn off. There is energy stored in the diode, and in the inductance of circuitry around them. They turn off with a regeneratively-accelerated snap and it is this pulse which gives harmonics into tens of MHz.

Peppering capacitors around will help, but not eliminate the problem. Using a fast diode is a better start, and with modest filtering gives a better result.

It is disappointing just how many people automatically reach for a 1N4007. They're the electronic equivalent of lead-based paint and asbestos ceiling tiles. They were once the standard and everyone used them without knowing of their drawbacks. At least the diodes don't cause health problems, but radio users (us!) would be better off if the things were banned.

That said, the 1N4007 does make a useful RF switch and a fair varactor diode. Just don't use them as rectifiers.

Metal rectifiers and selenium rectifiers aren't a problem.

David

[stevehertz]

So instead of reaching for a 1N4007 for use as a rectifier, what should be used? What's the best or a good alternative that does not share its highlighted problems?

[ionburn]

What about the faster 1n4007's being supplied?

[turretslug]

Ba159.

[trobbins]

Even with a faster UF4007 (almost an equivalent to the 1N4007), there is still switch-off step current change - just not as bad as when the step also includes a reverse conduction blip. Although onerous, if you have a noticeable problem from your transformer then I'd recommend applying a 'quasi-modo' style C-RC snubber across the winding. However you may be able to avoid problems by careful layout of current loops and grounding, as that is one easy way that rectifier noise couples in to nearby sensitive circuitry.

[daviddeakin]

Originally Posted by Chris55000

1) What were the audible manifestations of this,

Imagine a puny pantry transmitter inside your chassis, producing brief bursts of 100kHz, twice every mains cycle. You can probably guess the audible manifestations: very tiny spikes coupled into the audio. Not really a problem for a Marshall, but perhaps less palatable in a high quality amp.

[stuarth]

I'd seen a reason for the capacitors across the diodes as a cure for modulation hum, the theory being that the RF coupling to "ground", ie the mains supply, went high impedance when the diodes were off as the mains cycle when through zero, and this reduced the effective RF signal input. The capacitor, according to this theory, provided a low impedance path at all times.

That was a long time ago, and it might have been slow switching diodes causing the problem, not modulation hum at all.

Stuart

[Refugee]

I once had a very old and cheaply made cassette recorder that used to buzz if the chassis was not earthed either directly or to a large ground plane such as a bed frame.
There were no capacitors on the rectifier.
There was obviously no buzz on batteries.

[Skywave]

Originally Posted by stevehertz

So instead of reaching for a 1N4007 for use as a rectifier, what should be used? What's the best or a good alternative that does not share its highlighted problems?

Suggestion.
BYV96E:
PIV = 1,000 v.; Typical continuous fwd. current = 1.5 A.; Typical reverse recovery time = 300 ns.
1N4007:
PIV = 1,000 v.; Typical forward current = 1 A. Typical reverse recovery time = 30 µs.

Al.

[TrevorG3VLF]

Or UF4007 1kV 1A 75ns.

I bought a fistful of BYV96E at a radio rally for very little and was lucky to spot that some 30V zeners had infiltrated.