Using a silicon diode rectifier instead of an AZ31 valve

[Old Sparky]

I'm thinking of using a 1N4007 to replace an AZ31 valve whose heater is o/c, the idea being to wire the diode to inused pins on the valve base, leaving the valve plugged in for appearance's sake. Having already replaced the 100 ohm surge limiter resistor, I would just wire the 4007 in series, then check the HT voltage against the quoted figure on the service sheet for the Roberts P5A inbeing repaired and, if necessary, increase the surge limiter resistance. So far so good, but, in an old Mazda Data Book I have found a diagram showing 2 x BY105 diodes replacing a UU5, in which two 105s in series are used, each having a 1Meg. 1Watt resistor in parallel, with a 1n, 2kV capacitor in parallel. Since a 4007 is 1kv VRRM @ 1A only one diode will be needed, but do I need to fit the parallel R & C? I've never before fitted a parallel resistor in such conversions. BTW, The P5A uses the AZ31 with both anodes commoned and the primary of it's mains transformer feeds the anodes via said surge limiter resistor. The only secondaries are a 4v & 6v heater windings.

[Skywave]

When several power diodes are connected in series for rectifying a high a.c. source - such as you refer to - it is necessary to ensure that the reverse voltage across each diode is the same - or very nearly. If the reverse-biased resistance of those diodes could be guaranteed to be equal - then that equality would be assured. But that reverse-biased resistance will not be equal - so high ohmic-value resistors are wired in parallel with each diode to equalise that reverse voltage. The usual value of those resistors is in the range 220 kΩ → 470 kΩ.

The purpose of then connecting low-value capacitors in parallel with each diode/resistor combination is to enable short spikes of voltage (usually from the incoming mains supply) to by-pass the diodes. Such voltage spikes could 'punch through' the P-N junction of a diode bringing about its destruction. The usual value for those capacitors is in the range 1,000 pF → 0.01 uF; 1,000v working is usually adequate.

When the voltage source to such a silicon rectifier arrangement is derived from the H.V. winding of a transformer, as an alternative to those capacitors, sometimes an R/C series network is wired across the secondary winding: the purpose of that is the same: to present a low-impedance path to high-voltage very short duration spikes.
When only one (or pair, in a bi-phase rectification scheme) of diodes is used, that R/C network is then still required for same reason.

Al. / Skywave.

[Alan Douglas]

These days, the resistors are unnecessary. The PN junctions go into reverse avalanche if their prv is exceeded, which is no harm as long as the power dissipation is sufficiently low. Therefore two in series will automatically share the voltage. Current being the same in both, it can never be more than the reverse leakage of one diode.

[Skywave]

Alan - the requirement for those voltage-equalising diodes may not be necessary with modern rectifier diodes, but from personal experience, they are necessary with the 1N400x series. But even with diodes where those resistors may not be required, a few low-power resistors are cheap - especially compared to a burnt-out H.V. winding of the power transformer, should a bi-phase rectifier scheme fail with all diodes short-circuited ! Yes - I've been there!

Al.

[Alan Douglas]

1N4000x aren't made the way they used to be. Some years ago I measured prv of a few with a Tek curve tracer, and found they were all around 1600v no matter if they were labeled 4005 or 4007 and maybe even some 4001 (I don't recall). In the old days they would have been graded after manufacture but now they're all alike.

Resistors typically aren't rated for the voltages involved, and add one more failure point.

[Skywave]

I am reasonably sure that a fair number of 1N400x diodes that I do have in stock are quite old, but I still maintain that inserting voltage-equalizing resistors across a chain of rectifiers is well-worth doing. As for those resistors failing on account of inadequate voltage rating, that is something I have never experienced when using them for this purpose, but I do understand and appreciate the importance of the voltage-rating of resistors, especially those with spirally-cut tracks.

So, in conclusion, I repeat what I said earlier: my remarks are based on experience, spread over 45 years. It seems that you and I will have to agree to disagree on this topic.

Al.

[Alan Douglas]

Well the world would be a dull place if everyone agreed on everything.

[Old Sparky]

So, the consensus of opinion is that voltage equalising resistors should be fitted if two 1N4007s are used in series. According to the PIV (VRRM- what is the difference?) rating of the 4007 one should be sufficient fed from 240vac via a 100 ohm surge limiter, but two may be better. I've fitted balancing resisttors across two series capacitors in the past, but it never occured to me that they'd be needed with rectifier diodes until I found my old Mazda valve Data Book. I'll certainly fit the parallel capacitor, though. (PS I've been involved with radio & electronics for almost 48 years, 24 of them in the Car Audio repair trade from 1971-1995 - where does the time go!!)

[Nick]

You can use PSU Designer II (available free from http://www.duncanamps.com/psud2/) to model the replacement of the AZ31 (60mA, 500Vrms max) by silicon - although I don't think that the AZ31 is directly supported by PSUD2, a similar equivalent tube may be used from the supported list.

PSUD2 is a great resource for this sort of thing - its use is pretty much standard in the audio design world.

EDIT: FWIW, if you buy pukka (rather than suspect cheap IN400x from eBay or a grab-box at a ham fest or whatever), I've never known one fail when operated within spec. - an 1N4007 here is well within its SOA.

EDIT: The P5A HT supply is derived directly from the mains (no isolation). It may be a nice thing to do to put in a 1:1 small xfrmr for safety.

[Alan Douglas]

So, the consensus of opinion is that voltage equalising resistors should be fitted if two 1N4007s are used in series.

No censensus; they're not required, nor are "balancing" resistors for two series electrolytics. They are inherently self-balancing.

Regarding series electrolytics, I ran an experiment four years ago and posted the results in the Antique Radio Forum:
http://antiqueradios.com/forums/viewtopic.php?f=19&t=153805&p=1264569#p1264569
I just repeated it a few minutes ago (same capacitors, but at 550V) to get a new photo since the old one was lost.