SS rectifier diodes?

[guitman321]

Hi all.

In the Bandmaster AB763 and other BF amps what type diodes were used, and why are there 6 of them. Would two 1N4007s do the same job?

[pdf64]

It may be that in the 60s, high voltage silicon diodes weren't available / cheap, so series combos can be used to bump up the total voltage capability.
In regard of 1N4007, the voltage that the diodes are subjected to may exceed 1kV.
See The Valve Wizard

[guitman321]

Thanks pdf64. The Valve Wizard does indeed refer to the use of the 1N4007.

[g1]

Would two 1N4007s do the same job?

The Valve Wizard does indeed refer to the use of the 1N4007.

He shows 4 of them, not 2.

[Enzo]

Considering that 100 1N4007 costs $3-5, why not spend two extra nickels and remove any doubt?

[R.G.]

Stacking modern 1N4007s is probably OK, but there's one thing to be aware of - how equally the series diodes share both static (DC) and dynamic (AC) voltages.

At DC (or low frequencies, like power line, which is tantalizingly close to DC for these purposes) the DC voltage across each diode in a series string is not determined by the number of diodes, but by the reverse leakage of the diodes. If one of two diodes happens to leak ten times as much as the other, then it will only be holding off 1/10 of the applied voltage, while the less-leaky one will have to hold off 9/10 of the voltage. Whatever the ratios of the leakages are, that's the inverse of the sharing.

This can get ugly if you happen to be close to the sum of the reverse voltages per diode and one is leaky. The leaky one lets the less-leaky one get broken over, and if the broken one shorts, then the whole voltage is applied across the remaining leaky one, which then dies too.

Since diode datasheets never say what the min and max leakages are, usually just a maximum, you don't have the data to figure out what might happen. What everyone does is to presume that if you need 1100-1200V, then two 1N4007s is good enough. Mostly you get away with it. Sometimes not. High voltage leakage is dangerous to measure.

A trick that works is to swamp the leakage of the diodes with a paralleled resistor. If the diodes leak 5uA (which I pulled from the Vishay datasheet) and you parallel the diodes with resistors that let 50uA flow when at the full reverse voltage, then the ratio of the leakages on the several diodes doesn't matter - it's completely swamped by the resistor "leakage" and you can force voltage sharing.

A similar thing happens with diode reverse capacitance and AC signals. The (nonlinear... ) diode capacitance causes the diodes with the higher capacitance to eat less of the dynamic voltage than the ones with low capacitance, so the low capacitance "good" ones die first, followed by the high/bad capacitance.

Swamping with fixed capacitors fixes this.

These are the reasons that high voltage multiplier strings often have each diode paralleled with a resistor and a cap.

The closer the reverse voltage is to twice the reverse voltage of the rectifiers you use ( for two diodes ) the more you need swamping. If you only need 1200V and you use two 1N4007s, you're probably OK, although leakage can still get you.

Just a thought. I have all these electronics scars...

[guitman321]

Hi again
Thanks for all the info. I went with the 4 diode approach. So far so good.
Chuck

[gnappi]

Considering that 100 1N4007 costs $3-5, why not spend two extra nickels and remove any doubt?

Good advice, I always wonder why people ask for advice then go ahead and ignore it. I bought 100 1N4007's for $3.25 delivered :-)

[Leo_Gnardo]

R. G. that's the "white paper" on the subject, very informative & thanks for that.

What do you think of UF4007 instead? FREDs? Still need swamper R/C's?

FWIW for at least 5 years I've been using UF4007 in Fenders & similar, 3 in series on each leg, when replacements are needed. No failures yet, amps sound just fine, scarcely more expensive than 1N4007 iow still very cheap.

[trobbins]

For peace of mind I bought a 100 UF4007 rather than 1N4007. There certainly is a reverse recovery difference if you want to test for it with a simple signal generator test, and perhaps if you have a slamming rectifier application that pushes an ss diode in to reverse recovery territory then it may be noticeable, depending on how well you manage the rectifier layout and filtering. I parallel them for any application needing a bit more continuous current. It can also be worth it noise wise if you add an extra one in series with the number that are needed for PIV, as the parasitic capacitance is less.

If there were bypass caps across ss diodes (presumably for noise), then the UF4007 allows you to remove those caps.

[Leo_Gnardo]

I parallel them for any application needing a bit more continuous current. It can also be worth it noise wise if you add an extra one in series with the number that are needed for PIV, as the parasitic capacitance is less.

Thanks for your post trobbins. There's a UF540n series, n from 0 to 8, 3A rectifiers for higher current application. Haven't used any yet but I expect same advantages as the UF400n series.

IIRC, first ran across UF fans (and FRED too) on Audio Asylum, you're over there too, right?

[R.G.]

For peace of mind I bought a 100 UF4007 rather than 1N4007. ... I parallel them for any application needing a bit more continuous current. It can also be worth it noise wise if you add an extra one in series with the number that are needed for PIV, as the parasitic capacitance is less.

If there were bypass caps across ss diodes (presumably for noise), then the UF4007 allows you to remove those caps.

The same issues apply to the UF diodes in series as apply to the normal diodes in series. They can be unbalanced in both DC and dynamic impedances when being turned off and when holding off voltage. Parallel resistances force the diodes to equalize their reverse voltage at near-DC conditions (like AC power line without transients) and parallel caps force the diodes' internal capacitances not to matter for transients so they share spike voltages equally.

It's not really a noise issue, it's a diode reverse voltage issue. UF series diodes recover from forward conduction so quickly that they don't get the sudden slam off that ordinary diodes can, but they are not immune to reverse voltages being unequally distributed. As I said, this matters most when you are trying to block a voltage that is just over N times the diode reverse voltage rating with N diodes. If you don't use equalizing resistors and caps, you can get N diodes failing on N-minus-a-little-bit voltages applied to the string. The resistors and caps fix this.

Beyond that, you can add additional diodes in the string to take care of this. However, you never know how many additional diodes to add because a few really good (i.e. low leakage and low capacitance) diodes can then let one just-meets-specs diode fail by shoveling the voltage load to that diode, which then fails. If the resulting diode string with one diode failed cannot support the voltage, the string starts dying one diode at a time.

Usually modern diodes are pretty good about this, but if you're doing conservative, I-know-this-will-work-because-I-calculated-all-that design, you have to use the resistor and capacitor swamping components to series diodes where one diode is not rated for the whole voltage. Otherwise, you're counting on luck for a parameter that is NOT specified in the datasheets.

Using luck is what you do when you can't do the calculations or can't make unspecified things not matter.

Good engineering practice dictates that anything that cannot be controlled must be made irrelevant.

[pdf64]

For peace of mind I bought a 100 UF4007 rather than 1N4007... I parallel them for any application needing a bit more continuous current..

I suspect this is 'non-ideal' practice.
As the individual forward voltage / resistance, and particular point of switching, may result in the current (instantaneous / average) not being evenly distributed.
I think that good practice when paralleling such devices may be to add low value individual series balancing resistors.

[trobbins]

Most designers I guess would manage PIV sharing, and current sharing, with in-house derating factors applied after worst-case applied levels like mains voltage variation. Over time, with feedback, those derating factors would provide a level of confidence that new builds would have adequate MTBF. Trying to ascertain whether a PIV derating factor is appropriate is pretty difficult unless failure rate is gross, due to difficulty in getting quality feedback and external influences like mains swell or transients. A derating also include the hidden derating from the manufacturer in providing devices that don't fail until beyond the rated levels used for design. In Oz VAC nom is 230V, but was 240V. I just do single restorations on old amps, so my worstcase rule-of thumb is VAC up to 264V on a mains that could be down at 230V during testing of a PT secondary voltage. As such, a nominal 250-CT-250VAC secondary winding could have up to 813V PIV on a diode, so I'm still happy to use a single xx4007, but would likely use two in series above that.

Implicit in that is using two diodes from the same purchased batch. Where PIV was a concern, it would be wise not to mix purchased batches, or buy dribble quantities which would often mean using mixed parts for applications.

For common amp supplies, I can't see that dynamic effects come in to play for diode PIV, unless the focus is on surviving transient disturbances. VAC-CT-VAC secondaries would be more prone, as one half-winding is not clamped. I personally would avoid sharing circuitry for amp applications, due to noise increase, and if robustness was a concern then would add another diode in series. The comment on noise alleviation by using series diodes is just one arrow in a large quiver, but worth noting if trying to track through noise contributions in an amp, and implicitly makes the PIV topic more robust.

As I only do valve work, there is little opportunity for using two UF4007 in parallel. Using PSUD2 to see if peak is getting towards non-repetitive level is always worthwhile. For me, a higher powered B+ doubler turn-on surge may get above 50% rating. I can see that UF54xx is pretty much pro-rata with 3x UF4007, so well worth it if the application warranted something beyond what 2x UF40xx could handle. I can see that a toroid PT with full-bridge and large capacitance for B+ would need the peak rating of a UF54xx.

Can't say I frequent Audio Asylum much.

[trobbins]

Nah, I can't really align with that view. For my applications, the only current related concern is mains turn-on - the first charge up pulse. If PSUD2 indicates a turn-on peak of perhaps more than 50% of non-repetitive rating for 8-10ms half sine (which is the target of that spec) then adding two diodes in parallel seems very practical to me. If the surge was above the peak rating for say a UF4007, then using a UF5408 is the practical path to follow I'd suggest.