Another HT Diode Rectifier Question.

[Leo_Gnardo]

Thought I'd pitch in a little caution about a snubber cap across the AC windings in a bridge. Rivera does this, and I've had to replace too many of them - they generally fail spectacularly when they go. Select a cap well over rated in voltage to avoid failure. The Riveras have a 1000V rated cap. I replace with an orange drop rated @ 1600V. Haven't seen a 1600V OD go bad yet in this application.

[R.G.]

It is mentioned in Bob Cordell's excellent book, "Designing Audio Power Amplifiers". He suggests splitting the filter capacitance in half with a low value wirewound resistor in between. He says that even 0.22 ohms works wonders in a solid-state power amp. Of course the idea originally came from tube amps where we would use two filter caps separated by a choke.

Well, there you go. I'll have to go find and read Cordell's book.

Back on the original topic, the idea of protecting a resistor with a fuse makes me uneasy.

Don't be silly. Anything worth doing is worth overdoing. I would extend that and then protect the fuse protecting the resistor with a power transistor.

Adding a bypass cap directly across a secondary HT winding will effectively achieve a damped RC snubber, due to the considerable winding resistance that is normally in situ - the damping may not be too bad.

Yep, like the old adage says, if you have to pick luck or skill, pick luck.

The harmonic structure of the current step waveform when the diode is being commutated off by the falling secondary voltage waveform is a fundamental aspect of rectifier noise that is often overlooked when diode noise is discussed. Most conversions about diode noise focus just on the final sub-microsecond of reverse recovery current turn-around and snap off (the highest dI/dt sections), and seem to forget about the major section of commutation with the ramp down in diode forward current from full conduction to zero current.

Yeah; the thing is that there aren't very fast edges on that one. Diodes in rectifiers turn on on the front edge of the incoming sine wave when the secondary AC wave rises more than on diode drop above the run-down DC on the filter cap it's driving. The current then runs up at a rate driven by the slope of the incoming AC at that instant and the various resistances that limit it. It's one of those imponderables that varies with load because the point on the incoming AC wave where the diode starts to turn on depends on how far down the first cap has run, and so it turns on faster with higher loading.

But still, it's visibly a slope, not a step.

The slope of the AC input to the rectifiers constantly decreases as it heads for the peak value of the sine (assuming it hasn't been too badly distorted by outside forces and the incoming AC is a sine), and the first filter cap constantly charges, so the incoming current through the diodes ramps up quickly to the peak of the charging pulse, then tapers off at a rather slower rate as the filter cap charges, then tapers to zero.

An infinitely fast diode would simply stop there as the AC wave decreases below the peak and the remaining voltage across the ESR, wiring resistances and diode resistances get reduced. Again, this is a slope, not a step. The harmonic content is there, but it's not the multi-hundredth harmonics that lead to RF issues. That's not to say it's not in the audio range, if that can be coupled to the circuit. Just not much into RF.

However, the penchant for normal rectifier diodes to stay conducting even when the voltage reverses leads to the current actually reversing, and as the AC power waveform goes down, the diode lets current increase in the leakage inductances and parasitic wiring inductances in the reverse direction until charges are swept out, and then it bangs off. That *IS* a step current change, and it starts with high frequency content to spare, being composed of the turn off slope of the junction at the end of its sweep-out time, and the instant flip in voltage from the loaded-up inductances. There is a lot of RF in that one. It's not as big as the main pulse, granted, but the edge is much, much faster and hence easier to couple.

Adding a series resistance with the winding is different to adding it after the first filter capacitor, for a number of issues.

Yep, very. At one point I worked on deliberately making transformer winding resistance and leakage inductance bigger as part of a filtering/smoothing setup for a controlled-thyristor rectifier setup, regulating the output voltage by delaying the turn-on time of the thyristor to get regulation. The effect of impedance before the rectifiers is quite different than after.

Not being a hi-fier per se, I prefer to put such resistance in series with the secondary when using a ss diode so as to effectively return the diode commutation towards a valve style, and use relatively small main filter capacitance, so pn diode reverse recovery is a much more benign secondary issue.

You really should try the UF series of diodes. They really do a good job of suppressing the RF squarks. And you can then use a bigger main filter and get away with much less ripple. This used to be impractical but now the fast, soft recovery diodes are becoming much cheaper.

[trobbins]

The slope of the AC input to the rectifiers constantly decreases as it heads for the peak value of the sine (assuming it hasn't been too badly distorted by outside forces and the incoming AC is a sine), and the first filter cap constantly charges, so the incoming current through the diodes ramps up quickly to the peak of the charging pulse, then tapers off at a rather slower rate as the filter cap charges, then tapers to zero.

I was just looking at how flat-topped the waveform was the other day at someones house (and how a bidirectional inverter was handling it) - definitely a modern day staple and hopefully getting better with modern technology (but one step back each time a hifi amp gets connected!).

Don't forget that the diode conducts load current up until the time that commutation starts (rather than tapering to zero).

You really should try the UF series of diodes. They really do a good job of suppressing the RF squarks. And you can then use a bigger main filter and get away with much less ripple.

Ahh, but the mains ripple is partly where the tone comes from for guitarist use - the antithesis of a hifi requirement :-)

And just to keep on the good side of the moderator - if you have a choice of 3W resistor types you may well find a type with one-time transient power dissipation peak capability of 10 to 100x the continuous rating (I use PRO2 metal films and they have excellent peak ratings even though they aren't specifically aimed at pulse applications http://www.vishay.com/docs/28729/28729.pdf ).

[Leo_Gnardo]

Ahh, but the mains ripple is partly where the tone comes from for guitarist use - the antithesis of a hifi requirement :-)

If you're playing gut-bucket blues yes mains ripple may be part of the sound. Lots of guitarists would just as soon do without it. "When I play single note leads an octave up the G B and E strings, it sounds like someone is gargling along, out of tune." Clean jazz players are my toughest customers. Their amps get the hi fi treatment. Courses for horses. (yes I know thats reverse to the usual - on porpoise).

Meanwhile it's good to see the UF rectifiers getting a good mention in this thread.

[Steve Conner]

I've heard that nowadays a solid-state power amp without some sort of snubber on a standard recovery rectifier will fail EMC testing. Probably on conducted emissions. I don't know what the EMC situation is when ultrafast diodes are used, or with tube amps that tend to run smaller diodes with much lower currents.

Diodes generate RF noise by acting as fast opening switches. They don't turn off immediately the current passes through zero. They hang on for a little while as the current reverses, then suddenly snap off, interrupting the reverse current that built up in the meantime. It is this high rate of change of current that generates the RF.

It follows that to get best performance from a RC snubber, it must be connected as close as possible to the diode it's snubbing, with minimal loop area enclosed. When the diode turns off, the reverse current just switches from the diode to the snubber instead of kicking the whole circuit into ringing.

Morgan Jones showed a design with one RC network for each bridge diode, connected directly across their respective diodes. This is the ultimate cost-no-object approach. I recommend a cheaper version, with a single RC across the bridge's AC input terminals. I believe the filter capacitor across the DC output terminals completes the loop for the recovery transient current.

[R.G.]

I recommend a cheaper version, with a single RC across the bridge's AC input terminals. I believe the filter capacitor across the DC output terminals completes the loop for the recovery transient current.

... if the filter cap across the DC output terminals has a low enough ESR and ESL. Getting to there may need some ceramic disks across the main DC filter caps.

Actually, I buy the 400-800V 25A bridge rectifiers when I find them in the electronics surplus places for cheap. Likewise 1kV ceramic 0.01 and 0.001 caps and small-value resistors. The bolt-down-bridge makes stringing an R-C between terminals easy, especially if you do it before you bolt it in and hook wires to it.

But I do like the fast, soft-recovery diodes, too. If you don't have time or equipment to mess with tuning a snubber, these are great.

[bob p]

And just to keep on the good side of the moderator - if you have a choice of 3W resistor types you may well find a type with one-time transient power dissipation peak capability of 10 to 100x the continuous rating (I use PRO2 metal films and they have excellent peak ratings even though they aren't specifically aimed at pulse applications http://www.vishay.com/docs/28729/28729.pdf ).

I've been stocking with PR01 and PR02 for the last few years. One very nice feature of the PR01 is that these MF resistors have a PCB footprint that fits in the space that used to be occupied by a 1/4 watt resistor. And the PR02 allow you to fit a 2W resistor in a space tht used to be occupied by a 1/2 watt.

I am happy to hear someone else mention that PSU ripple can be an important part of the sound. IMO too many people try to completely eliminate it, which can lead to very sterile sounding amps.

[bob p]

... if the filter cap across the DC output terminals has a low enough ESR and ESL. Getting to there may need some ceramic disks across the main DC filter caps.

This is an example of why cap selection is so important. Getting low ESR and ESL really does matter, and if you're willing to spend time pouring over a stack of rather boring cap data sheets, you'll find that there are some amazing differences in these values from one manufacturer to another. I recently spent 3 days pouring over cap data sheets while I was in the planning stage for the rebuild/restoration of a pair of a PL700 (and a Carver C-500 that descended from them), just so I could find the best caps available for the job. After all, this was a HiFi application where noise and hum reduction were important.

Needless to say, the caps that you want to buy for long life, high ripple current, low ESR and low ESL are the usual suspects from the premier cap companies. These screw-type caps are becoming harder and harder to find in-stock as the modern market begins to move away from them. They are either very expensive in single unit quantities or you have to place a large minimum order to get a decent price, so bring money. You absolutely do not want the grade of caps that we commonly see people selling for the MI tube amp applications. Their ESR and ESL specs are just horrible. The good news is that if you're willing to create a terra-cotta army, then an array of modern-day snap-in caps will bring your really close to the ESR and ESL specs of some of the premier screw-type capacitors.

Actually, I buy the 400-800V 25A bridge rectifiers when I find them in the electronics surplus places for cheap. Likewise 1kV ceramic 0.01 and 0.001 caps and small-value resistors. The bolt-down-bridge makes stringing an R-C between terminals easy, especially if you do it before you bolt it in and hook wires to it.

These amps also use the high-voltage, high-current bridge packages. During the restoration process I decided that I needed to scrap the amps' original PSU design and update them to 21st Century technology. Part of this included designing the new PSU boards to use 1kV ceramic 0.01 and 0.001 caps, as you mentioned. It's good to know that my thoughts were on the right track.

[bob p]

Thought I'd pitch in a little caution about a snubber cap across the AC windings in a bridge. Rivera does this, and I've had to replace too many of them - they generally fail spectacularly when they go. Select a cap well over rated in voltage to avoid failure. The Riveras have a 1000V rated cap. I replace with an orange drop rated @ 1600V. Haven't seen a 1600V OD go bad yet in this application.

Yikes. I read with great interest your comments that the 1kV rated caps are failing. This makes me wonder if the application would benefit from surge protection devices on the primary side.

[Leo_Gnardo]

I read with great interest your comments that the 1kV rated caps are failing. This makes me wonder if the application would benefit from surge protection devices on the primary side.

There's a nearby store that's a Rivera dealer, and any problems get sent to me. Really it's only a couple amps a year that have this failure. So far the Riveras seem to be one of the least troublesome amps over the past 15-20 years or so. I'm sure any amp would benefit from an appropriate surge device. And if Rivera would just install a 1600 or 2000V rated cap here it wouldn't hurt. Would be a good idea to keep a couple proper replacements handy for anyone who handles Riveras.

Thanks for the like!

[Steve Conner]

Well, was that 1000V rating DC or AC? Typically a 1kV DC rated plastic film cap can only stand about 350V RMS AC on a continuous basis, not 700V RMS as you might think. See Table 1 here for an example. http://www.rubycon.co.jp/kr/catalog/...onFilm_Eng.pdf

[trobbins]

By 1600V, do you mean 1600VDC (ie. 500VAC) rating ?

X and Y rated mains emi suppression caps are also very appropriate for across a secondary winding as long as the AC voltage rating is ok (eg. easy to get caps up to 275VAC) - and are easy to find.

I'm also keen to find any references that specifically investigate the diode reverse-recovery impact on a mains transformer, rectifier, capacitor input style configurtion. Every reference I've looked at either just does simulation, or sketches, or relates to operation or test circuits for switchmode operation.

[Steve Conner]

What do you mean by "impact"?

[trobbins]

Sorry, by impact I meant the relative effect in a variety of circuit setups.

In a typical reverse recovery test setup, there is a constant forward current level followed by a set dI/dt ramp down through zero. A range of those test conditions (initial current levels, and dI/dt rates) are sometimes presented to indicate their relative effect on the actual reverse recovery peak current amplitude, and the recovery character of the current waveform. But I'd be second-guessing that those current levels and dI/dt rates are consistent with switchmode type circuit situations, but may not be representative of mains rectifier circuit situations.

A range of mains rectifier circuit setups could be reasonably found in amps - where important parameters can be quite different, such as secondary winding leakage inductance and effective resistance and voltage, load resistance, filter capacitance, and diode on voltage (ss to valve).

One could muse that if leakage inductance of the PT windings were made a lot smaller, as per the effort put in to OT design, then that would benefit a hi-fi amp - maybe there are examples of that in the hi-fi market where cost is little detergent in the pursuit of perfection.

[Steve Conner]

The only effect of diode recovery in an old-fashioned line frequency rectifier is that it may fail EMC if you don't snub it, and it'll couple little spikes into your audio path if you get careless with loop areas, grounding and stray inductances. These may possibly be heard as a very high-pitched tinny buzz. Seemingly they can also be picked up by an AM radio held near the circuit, though I've never tried it personally.

The di/dt in a line frequency rectifier is orders of magnitude lower than in a switchmode application, so the reverse recovery transients are that much smaller. The energy loss is insignificant, unlike in a SMPS where a standard recovery diode can be blown to smithereens in milliseconds.

Leakage inductance in PT windings helps to smooth out current pulses, increase rectifier conduction angle and low-pass filter RFI. I think a good toroidal PT already has less leakage inductance than you would like from this point of view. I've experimented using toroidal PTs as OTs and interstage transformers, and measured HF bandwidths in the hundreds of kHz.

On the other hand, leakage inductance also impacts regulation, a bad thing from the audiophile point of view. And increased leakage inductance tends to mean worse stray flux, a bad thing in any audio application.