QUestion about diode mod for rectifier

[schoolie]

The Ceriatone layouts have diodes in series with the rectifier plates, but they have a 0.01uF/3000V capacitor across the rectifier plates. I copied this, but now I'm worried about the cap shorting and taking out the PT. Is this a bad idea? Should I just have two of these caps in parallel with the diodes? I'm assuming this cap is there to reduce switching noise. Is a cap even needed with a fast diode?

Thank you!

[R.G.]

The Ceriatone layouts have diodes in series with the rectifier plates, but they have a 0.01uF/3000V capacitor across the rectifier plates. I copied this, but now I'm worried about the cap shorting and taking out the PT. Is this a bad idea?

A shorted cap here would do nothing that's not already being done by the rectifier plates on amps with no "backup diodes". No harm if they short. Another question is whether the diodes they parallel are rated at over 3000V. If they're rated for less, they'll break over first. Likewise, is the rectifier tube rated for 3000V? If so it'll fail first.

[EFK]

I add those diodes on pretty much every rectifier socket although I'm not pro enough to completely understand why. I don;t use any caps, either across the plates or across the diodes individually, and I can't hear any difference so my perspective is, one less part to fail.

Here's an interesting question for someone really smart: using the example of a 5V4, for example, the datasheet specifies a max capacitance in a C/R/C input of 10uf, unless resistance is added and you do all the math to calculate how far you can increase the capacitance. My question is, does the addition of these diodes do ANYTHING at all insofar as allowing more input capacitance without increasing resistance?

[schoolie]

Thanks for your replies! I'm a little dense, so bear with me Just to make sure I understand, the capacitor is connected to GZ34 anode pins (4 and 6), and not in parallel with the diodes. If the cap shorts, that won't be an issue?

Thanks for your help

[R.G.]

I add those diodes on pretty much every rectifier socket although I'm not pro enough to completely understand why. I don;t use any caps, either across the plates or across the diodes individually, and I can't hear any difference so my perspective is, one less part to fail.

The "why" is that the silicon diodes back up the rectifier tube. Tubes burn out and short. When your rectifier tube shorts, it has a good chance of killing the power transformer, which is the single most expensive part in the amp. If the diodes are there, then if the rectifier tube shorts, the silicon diodes take over, preventing AC from reaching the filter caps, killing them, and becoming a short across the high voltage winding. This probably saves the power transformer.

In normal operation, the silicon diodes add no noticeable coloration or effect. If the tube rectifier shorts, the B+ rises 30-50Vdc and the amp keeps on working. They turn a rectifier tube short from a tragedy into something to be taken care of when you can. The output tubes may get hotter, so you can't just leave it, but it's not an immediate disaster.

I have not messed with paralleled caps.

Here's an interesting question for someone really smart: using the example of a 5V4, for example, the datasheet specifies a max capacitance in a C/R/C input of 10uf, unless resistance is added and you do all the math to calculate how far you can increase the capacitance. My question is, does the addition of these diodes do ANYTHING at all insofar as allowing more input capacitance without increasing resistance?

They do NOTHING for the amount of capacitance the tube rectifier can stand. This is limited by the electron emission capability of the rectifier tube cathode. They're just a disaster backup.

Thanks for your replies! I'm a little dense, so bear with me Just to make sure I understand, the capacitor is connected to GZ34 anode pins (4 and 6), and not in parallel with the diodes. If the cap shorts, that won't be an issue?

Sorry - I understood this was parallel with the diodes. Can you post a schematic of the amp so I can see the actual circuit? If the caps are not parallel to the diodes, what I said may well not make sense at all. Got a schemo?

[schoolie]

This is the layout diagram of Ceriatone's 5f6a

[Gingertube]

The diodes are in series with connection to the anodes. What this mod does is help protect the tube rectifier in that the diodes block the peak inverse voltage which means that the tube rectifier doesn't have to handle it.
The mod became popular when a whole lot of inferior modern production tube rectifiers, 5AR4 in particular, hit the market. They arced internally creating a short. This mod stopped them from arcing.
The cap is not essential. It has been put there to help suppress the switching noise of the added semiconductor diodes. The switch noise will not get through the tube rectifier BUT it will couple back through the transformer to other windings which can cause the switching noise to appear on heater windings or worse still on a winding used for a bias supply. The cap is specified as a 3,000 Volt device purely to reduce any likelyhood of it failing.

With a 350-0-350 winding you could THEORETICALLY get away with a 1,000V cap. (The voltage they see is 350 x root2 x 2, twice the peak voltage)
A 2,000V cap would THERORETICALLY cope with windings up to 700-0-700.
The 10nF 3000V caps are cheap. The damage a failed, lower voltage cap could do is expensive - so stick with the 3,000V cap or leave it out.

Use Ultrafast Soft Recovery diodes (the UF family and similar) to limit any switching noise and leave out the cap would be my recommendation 0h and make sure that the semiconductor diodes you are adding have the required voltage rating, otherwise you've just shifted the problem and not fixed it.

Cheers,
Ian

[schoolie]

Thank you for your reply, Ian! I am using NTE558 diodes which are rated at 1500Vr and have 250ns recovery time. I guess I don't need the capacitor, so I'll snip it.

[R.G.]

I suspect that if the tube rectifier is intact, since it's in series with the silicon diodes, it controls the turn off, and doesn't give the silicon a chance to show the "slam off" that causes buzz and RF ringing with just silicon diodes. That might well only happen when the tube rectifier shorts.

That could be a good thing, since one of the issues with using silicon backups is that you can't tell when the tube rectifier shorts. The difference is subtle, mostly an increase in B+ and more power, less sag. Getting a buzz in the sound would provide another indicator without having to concoct a circuit indictor.

[trobbins]

If both sides of the tube rectifier short then that wouldn't be a concern per se. If one side shorted anode to cathode, then mains fundamental hum ripple would increase, due to assymetry in the rectified pulse current levels (and hence ripple voltage waveform). I think that will incur a higher rms ripple current, and hence not good for the cap.

R.G. - I think the diode turn-off transition would be dictated by the silicon diode(s). As the 'diode' voltage changes polarity, there will still be current flow through the series diodes (due to the load level) - and the ss diodes will halt their current flow according to their reverse recovery characteristic - whereas at that point in time the tube diode will probably just be slowly changing its effective off resistance.

[R.G.]

If both sides of the tube rectifier short then that wouldn't be a concern per se. If one side shorted anode to cathode, then mains fundamental hum ripple would increase, due to assymetry in the rectified pulse current levels (and hence ripple voltage waveform). I think that will incur a higher rms ripple current, and hence not good for the cap.

Yes. There are some failure modes which are bad. One aim of adding backups is to keep any failure from happening. Another and valid one is to make them less severe. I suspect that one-side-shorted tube rectifier is very, very bad for the filter cap, probably sudden death. Bigger ripple current caused by asymmetry of the rectifier is less good than the no-failures, balanced case, but less bad for the cap than one half of the tube rectifier shorted, which is what would otherwise happen.

R.G. - I think the diode turn-off transition would be dictated by the silicon diode(s). As the 'diode' voltage changes polarity, there will still be current flow through the series diodes (due to the load level) - and the ss diodes will halt their current flow according to their reverse recovery characteristic - whereas at that point in time the tube diode will probably just be slowly changing its effective off resistance.

I don't think so. The tube rectifier is a bigger impedance at all current levels than the silicon diodes, and I think it's also slower in both turn on and off. The slowest element in a series chain dominates the turn on/off characteristic, I believe.

[trobbins]

The slowest element in a series chain dominates the turn on/off characteristic, I believe.

Nah - the ss diode will suddenly move from very low resistance, to taking the available reverse withstand voltage (ie. highest resistance element in the current path), as the full diode current is passing through it's pn junction. The recovery time and level of reverse recovery current reached will depend on the forward current prior to turning off - so is somewhat linked to the amps load current. As such, there can be benefit from fitting eg. UF4007 diodes rather than IN4007s.

[R.G.]

Nah - the ss diode will suddenly move from very low resistance, to taking the available reverse withstand voltage (ie. highest resistance element in the current path), as the full diode current is passing through it's pn junction.

That is true when the ss diode is what is resisting reverse current. In this instance however, the current is actually forced to soft-off by the tube rectifier. This keeps the ss diode from seeing a sudden reversal. Instead, it's current is tapered off (relatively speaking) by the tube rectifier's coming out of conduction.

But I speculate, based on the circuit. I'll have to see if I can lash up something to measure the RF squarks from diode-off with
(1) tube rectifier
(2) tube plus 1N4007
(3) tube plus UF4007
(4) UF4007 alone.

This may take me some time to lash up and test.

[trobbins]

Hmmm, this is interesting. At the start of the diode turn-off region, the AC waveform is coming down towards the voltage of the capacitor. At one point in time T1, the capacitor current is zero, and the diodes are conducting the load current, and the on-voltage of the tube will be likely quite a few volts. At the point in time T2, when the AC voltage is the same as the capacitor voltage, no current is flowing through the diodes (assuming leakage inductance is not significant). The current will have transferred over from the diodes to the capacitor between T1 to T2. T2-T1 is very short for just ss diodes as the AC voltage change is only about 0.7V (assuming no capacitor ESR/ESL), and so the dI/dt is high. But with a valve diode in situ, T2-T1 will be somewhat longer, as the AC voltage must reduce probably by 40V for 100mA load for a 5Y3GT. This assumes no dynamic change in the rectifier tube V-I characteristic. So I agree R.G., the dI/dt would be softened to the point probably where reverse recovery effect is negligible even for a 1N4007. The impact of winding leakage would also be much reduced. Thanks for that! But it would be great to see a tube diode V-I dynamic characteristic.

[R.G.]

This is one of those imponderables that seem right when thought about, and also seem to be verified in simulation, but where the actual effect is subtle. I suspected this based on my thoughts about the circuit, and simulation seemed to support it, but the exact operation of diodes, cap, tube rectifier and any strays is going to depend heavily on the actual parts in place. I'm still going to try to lash up one of these to run tests on. May take a while.