VALVE RECTIFIER "BACK UP DIODES"

[ViperDoc]

I was reading Blencoe about rectifier diodes, and he makes a calculation about the 1N4007 only being suited for up to a 290-0-290 HT secondary load, but also references the common series solution in solid state rectifier circuits, like two or three 1N4007 diodes in series on each HTS output. I would think the same calculation would apply to "back-up diodes" placed going into the rectifier socket, is that correct? It's easy enough to place a pair of diodes on each HTS line into the rectifier tube socket, but do any of you place parallel 10 uF parallel capacitors as noted here?

http://www.valvewizard.co.uk/bridge.html

Thanks.

[trobbins]

You have wandered in to what can be quite a technical topic, especially regarding the 10nF (not 10uF) parallel cap shown in Merlin's schematic.

My view is you don't need such caps, but should use two series 1N4007 for winding voltages above 290-0-290, and three in series above circa 440-0-440V. This is a link to a discussion paper where that topic is discussed in Section 2 - it gets technical and detailed - https://www.dalmura.com.au/static/Power ... 20amps.pdf. There are a few that ardently state that balancing parts are always needed - my view is different for this type of valve rectifier application.

[ViperDoc]

Thanks for the correction and link, much appreciated!

Nothing like a little light reading before bedtime.

It was interesting to read about the 5V CT claimed to be present on older, high quality audio transformers and the balanced rectifier output generated and sent to the choke input node. Is this notably better to the point that any of you spec 5V CTs in your iron? Clearly only applicable to valve rectification. Would this balance apply at all to improved sound quality, or perhaps more to valve rectifier stability and life expectancy?

[trobbins]

Wherever you look, topics can have detailed underpinnings - but that doesn't mean they can present a tangible and noticeable difference. People wax lyrical if there is more or less oxygen in their copper wiring, or if the label and colour of a capacitor is good enough for them to use in an amp.

"life expectancy" is a nebulous term that doesn't mean much nowadays for valves, as nobody takes statistically significant fault data of does a rigorous assessment on the operating conditions, and people offer opinions on poor life because one or two valves failed on them.

People don't choose a power transformer because it's secondary HT windings have equal dc resistance, nor do they buy a valve rectifier that has matched plate voltage drops to within some tolerance, nor do they periodically check the plate voltage drop balance of their rectifier valve, or the leakage current level at PIV.

Just my 2c.

[Stevem]

Here's a way to ( 1) get tube rectifier sag . ( 2) ease the current load on the tube for a reduced chance of failure and greater life and (3) should the rectifier short out you just yank it out, replace the blown fuse and continue the gig playing on the diodes.

This wiring would be for octal tubes like 5U4, 5AR4, 5Y3, 5V4 and a few others.

[TUBEDUDE]

You have wandered in to what can be quite a technical topic, especially regarding the 10nF (not 10uF) parallel cap shown in Merlin's schematic.

My view is you don't need such caps, but should use two series 1N4007 for winding voltages above 290-0-290, and three in series above circa 440-0-440V. This is a link to a discussion paper where that topic is discussed in Section 2 - it gets technical and detailed - https://www.dalmura.com.au/static/Power ... 20amps.pdf. There are a few that ardently state that balancing parts are always needed - my view is different for this type of valve rectifier application.

If the diode is doubled, for protection from the high voltage, shouldn't the caps be used to insure equal splitting of that voltage?

[pdf64]

Here's a way to …

That codswallop keeps gets wheeled out whenever this topic arises.
4 silicon diodes and yet none are in series with the valve anodes; what a nonsense
Silicon diodes in series with the valve’s anodes will tend to reduce the likelihood of arcing within the valve, and if it does occur, reduce its ability to sustain itself.

[ViperDoc]

I'm looking at this for a GZ34 in a 5F8A HP Tweed Twin, if that makes any difference.

The diodes on Stevem's drawing looks to me like a parallel solid state rectifier supply that is not switched. Do the 250R 25W resistors drop the SS rectifier voltage to the same level of the tube rectifier voltage? Not sure how those voltages would blend at pin 8. The series diode arrangement into the tube rectifier supplies pre-rectified DC to the tube rectifier plate, and supply protection to the transformer winding, right? Seems like these are for different purposes. Very interesting, though.

[martin manning]

SS and vacuum rectifiers do not have the same voltage drop in forward mode, but the voltages could be equalized in reverse bias. Seems like some odd behavior would result if you tried to do that with paralleled caps. I would just put two Si diodes in series with each vacuum diode on the socket to protect against a rectifier short. If you like, get some with higher PIV than 1N4007's, say 1.2kV or greater, at 1A.

[R.G.]

Good rollup of several topics.
1> CT power supplies need really high voltage diodes. The first filter cap charges to the peak of the AC wave from the secondary. The other half of the secondary is an equal amount below ground, so the voltage across the off-side diode is twice the peak of the AC voltage. This is about twice the DC output voltage. A 400V dc supply means that the diodes need to withstand a minimum of 800V, twice the no-load DC on the first filter cap, >> with some reserve << if they are to live a long life. 400V divided by 1.414, the peak-to-rms factor for a sine wave gets to 283V. Yay! - Merlin got the math right! For a 450v supply, that creeps up to the diodes needing to withstand 900V every other half cycle. 1000V diodes look really marginal. And when you add in the possibility of an inductive transient from the wiring, you can pop 1000V diodes on a 400V supply for lack of voltage rating margin.

2> Putting diodes in series is tricky. If you need to withstand 1000V, you put two 1000V diodes in series and get 500V across each one, right? Probably not. Diodes leak a bit.The voltage across two reverse-biased diodes is probably not divided equally. Instead, it divides inversely with the reverse leakage. If one diode leaks twice as much as the other one, the voltage divides with the low-leakage one holding off twice as much voltage as the high leakage one. In today's uniformly-great-silicon world this is a better bet than it used to be, but three in series would be a better bet. That's because a really, really good almost-zero leakage diode puts itself at risk because a slovenly, high leakage diode in series can put nearly all the divided DC on an ultra-low-leakage one.
To avoid this situation, put a very high value resistor in parallel with each diode. Even a several-megohm resistor is much more leakage than the diodes, but less leakage than would hurt the supply. And the resistors force equal sharing on the diodes. In this case, you can get away with two 1000v diodes - for low frequency dc, no spikes.

Diodes have capacitance too. This is a murkier than leakage because diode makers say less about it, especially on ordinary low frequency diodes. They typically only specify a maximum capacitance, if they specify it at all. Again, the high capacitance diode in the string will hold off less of a spike than its really, really good partner, and the partner can then be punctured by a spike. The cure for this is to put a swamping capacitor in parallel with the diodes too. This forces the AC voltage on each diode to be equal.

3> Hum and buzz capacitors; it's common to see parallel caps on low frequency diodes with the explanation that they keep rectifiers quiet. Ordinary low speed silicon diodes have a "slam-off" quirk. When the voltage reverses, the diodes take a short time to turn off as the reverse voltage sweeps carriers out of their junction, then they turn off HARD. This sudden current stop lets any inductance in the wiring create a flyback spike, and this rings with the parasitic diode capacitance and the parasitic wiring capacitance. This ringing comes out as a squark of RF conducted and/or transmitted to the circuits around the wires. It is possible to snub the transient by using a series R-C snubber on each diode. But it's not very practical, as a snubber that's not tuned to the wiring and diode resonance may do not much. But many "designers" put one of these on every diode in hope. Actually, mostly the parallel anti-transient diode and snubber on each diode at least run the RF resonance down in frequency. A better solution is to use the fast and soft turn off diodes now available. "FRED" types, Fast Recovery Epitaxial Diodes and such, are a good solution. They don't slam off.

4> "just-in-case" diodes on tube rectifiers; IMHO, every tube rectifier needs a set of suitably rated silicon diodes in series with each anode. Tube rectifiers fail and short sometimes. If you have a backup string of silicon, this will not be fatal to the filter caps, your tubes, or the power transformer when it happens. The silicon will simply take over the job that the damaged tube can no longer do, and the DC voltage on the first filter cap will rise a bit. If you're playing, you may not even notice.

[trobbins]

If the diode is doubled, for protection from the high voltage, shouldn't the caps be used to insure equal splitting of that voltage?

That is the classical approach, and is typically applied for long series strings, and for applications that have significant dynamic waveforms with significant dV/dt glitches/disturbances (as a high dV/dt forces more current through a capacitor, and any difference in diode capacitance forces a difference in voltage across the diode). But that is not the operating condition in a mains frequency valve rectified power supply. Many people just point doggedly to the classical approach without appreciating the application. At the high inverse voltage stress region, the diode(s) supporting reverse voltage are seeing the lowest level of mains dV/dt (ie. the rounded top portion of the sinewave). The most significant internal disturbance occurs at a somewhat lower inverse voltage as the other diode turns off, but again for a valve rectifier with a HT winding that exhibits significant resistance and shunt capacitance, then even that disturbance is typically benign. An external mains side disturbance could glitch through, although its primary side level will depend on your environment, and the transformer attenuates any such glitch based on its internal construction - that is a risk, but is best considered by other practical issues like do you specific lightning protection built in to your main board.

Using 1N4007 diodes from the same batch provides the first level of managing any differences between diode junction capacitance, which is well under 5pF as PIV is approached, and can be viewed as an RC network (with the diode leakage effective resistance) that doesn't clamp the voltage division ratio.

I'm looking at this for a GZ34 in a 5F8A HP Tweed Twin, if that makes any difference.

The diodes on Stevem's drawing looks to me like a parallel solid state rectifier supply that is not switched. Do the 250R 25W resistors drop the SS rectifier voltage to the same level of the tube rectifier voltage? Not sure how those voltages would blend at pin 8. The series diode arrangement into the tube rectifier supplies pre-rectified DC to the tube rectifier plate, and supply protection to the transformer winding, right? Seems like these are for different purposes. Very interesting, though.

The 5F8A has nominal 300-0-300V PT secondary, so I'd say two series 1N4007 would be reasonable. There is in-built design margin when recommending 280-290V changeover to two series 1N4007 - by far the largest margin is that 1N4007 have a typical measured PIV up near 1.4-1.5kV for anyone ever testing them so far (not just my testing).

Yes Stevem's drawing has its benefit in reducing the peak current that a valve rectifier plate is exposed to (eg. if the DC load is too much). But it comes with operating caveats, not the least is that PT voltage is not even considered as a risk given that nigh on all the valve base pins are put to use.

One way to view a valve diode in series with two ss diodes is that the leakage current of all three diodes will dictate the voltage division across the three diodes. It may well be that the valve diode has lower leakage current than the ss diodes and hence supports all the available PIV stress - but note that the ss diodes still conduct that leakage current, and so will drop some voltage across themselves. That partitioning of voltage drop across diodes would vary with time (as the valve becomes more gassy or metal leakage across mica increases), or as ss diode temperature increases (there is a strong relationship between leakage current and die temperature). Also note that the leakage current can never increase more than any one of those diodes allow, even if one of the ss diodes was at its PIV level where its leakage current would increase rapidly if it was allowed - that same situation occurs with the valve diode when it becomes leakier (ie. it doesn't arc because its leakage current can't rise to an arc level).

[ViperDoc]

Many thanks for all that detail, wow. I'm distilling that down to putting two 1N4007s in series on each HTS lead into the rectifier plate pins is A-OK.

[ViperDoc]

I forgot to ask earlier, when using back-up diodes into the tube rectifier plate pins, are in-line HTS fuses redundant or unnecessary? Thanks as always.

[pdf64]

It seems a good idea to have a delayed acting fusing to protect the winding, eg from a reservoir cap short, and fast acting after the reservoir cap, so that valve problems are dealt with quickly.
Given that with the the doubled up protection diodes, a short across the HT winding seems somewhat unlikely, so a single fuse between the winding CT and its return to 0V common may be reasonable.
If you get rid of standby, that fuse can be a fast acting type, and the fast acting fuse after the reservoir cap would be superfluous.

[Colossal]

Given that with the the doubled up protection diodes, a short across the HT winding seems somewhat unlikely, so a single fuse between the winding CT and its return to 0V common may be reasonable.
If you get rid of standby, that fuse can be a fast acting type, and the fast acting fuse after the reservoir cap would be superfluous.

Pete,

What troubles me with fusing a CT is that a PT can still remain energized after a fault. Is that not the case here? I would think fusing the secondaries would be preferable, or barring that, have a single B+ fuse downstream of the rectifier.

Thanks