Too much filter capacitance?

[casey73]

I know that there are maximum filter capacitor rules when following a rectifier tube. My current project will use a 5AR4 and 60UF is the recommended maximum capacitor immediately following it. Are there any "rules of thumb" regarding the maximum capacitor values down line from the plate supply?

[maxkracht]

Aside from diminishing returns with unnecessarily large values, there shouldn't be any harm. I believe the limiting value for the rectifier is to prevent arcing. Fully discharged caps draw a lot of current at startup and the rectifier can't handle it. It's not a big deal on the later filter stages with a resistor or choke to limit current. Pretty sure you can get away with bigger caps on the first node with an inrush current limiter, but don't quote me on that.

[R.G.]

Max is right. The limitation on the first filter cap is because high rectification pulses will damage the rectifier tube.

The isolating resistances in the rest of the B+ string prevent them from causing these pulses, so they can be much larger. The purpose of these isolation caps is to prevent feedback from stage to stage through the B+ line.

Damage to the rectifier tube happens when the rectification pulses try to pull more electrons than the rectifier cathode can provide without being damaged.

[Stevem]

So does stacking two filters to get to what a tube recto can handle uf wise delay power supply sage any better then one filter of the max value?

[maxkracht]

I sometimes add an extra rc filter before the plate on SE amps, say 100-200 ohms between caps. Maybe even add the same resistance on the ground side. I'm guessing this is enough to push the max value. I have heard the resistance of the HT winding is an important factor in how much capacitance a rectifier can take, so you probably don't need a ton of resistance to stay safe. It's pretty common to see the max capacity rule broken by 20+ uf without visible long term consequences, but that could just be luck and a sturdy rectifier.

The real question is, why do you want a tube rectifier if you don't want sag? If your amp still hums too much with the rectifier maximum, you have other problems.

[Reeltarded]

I agree with Max.

Build it with a socket to make it easy later if you drop the cap, but bridge silicon over it.

[R.G.]

Vacuum tube rectifiers are limited by how many electrons their cathode can boil off without degrading the barium and strontium oxides that coat the cathode. Oxide coated cathodes can produce a huge amount of electrons at low (orange glow) temperatures as long as the cathode surface coating isn't destroyed by <a number of things>. Trying to suck more electrons than the cathode can produce is one of the things that leads to cathode poisoning and exhaustion, either sooner or later. Sooner happens in gassy tubes, later happens eventually, which is why tubes have sockets so they can be replaced. High peak currents exhaust the "electron cloud" that somewhat protects the cathode surface from impacts by gas ions and eventually degrades the cathode.

Rectifying sine(ish) AC waveforms causes a sudden peak current as the incoming transformer voltage rises above the filter cap's voltage. The filter cap voltage runs down during the portion of the AC half wave where the transformer voltage is lower than the cap voltage. This sudden-peak-and-ramp down is what makes the ripple voltage.

When the rectifier turns on, the only things that limit the current into the first filter capacitor are the AC mains impedance (!! low !!), the transformer winding resistances, the rectifier resistance/V-I curve slope, the connecting wire resistances, and the ESR of the first filter cap. We deliberately try to make all of these limiters as low as possible, so as to make ripple voltage smaller. That means we are as a side effect making the rectifier pulse currents as big as possible. Vacuum rectifier tube specifications on maximum filter capacitance is a boiled-down number for people who don't go calculate all these effects. The tube makers decided that as long as the cap was smaller than X, the tube life would be good enough and the ripple voltage would be good enough.

Using stacked filter caps changes this a bit by having two capacitor ESRs in series. With modern low ESR caps, this isn't much of a limitation, as capacitor designers have been trying to design ESR out of electros for the last hundred years or so. Stacking caps won't help. Simply using modern low ESR caps is bad enough.

Putting a small (1-100 ohm) resistor in series with the rectifier helps the rectifier, but makes ripple voltage worse.

Using solid state diodes and a faking resistor works great. But high capacitances still have the other effect of a stiffened, non-sag B+.

A two stage filter cap, with a first capacitor of some modest amount, then a resistor and a second large filter cap works great for the rectifier and the cap, but drops the B+ a bit. Using a BFI (big freakin inductor) between the first filter cap and the second-first filter cap works great, doesn't drop much DC voltage, but costs you an inductor.

There is another option. I was thinking about all of this when I designed the Tube Amp Current Clamp.
http://www.geofex.com/Article_Folders/m ... 0Clamp.pdf
This setup lets you simply set the max current you'll allow through your rectifier tube and have done with it. This includes both normal peak currents and turn-on inrush.

[maxkracht]

Putting a small (1-100 ohm) resistor in series with the rectifier helps the rectifier, but makes ripple voltage worse.

So, NTC thermistor after the rectifier for minimal cost/consequences if you must exceed the maximum value?

[R.G.]

Each time the rectifier cathode sends out too much current, the cathode life is reduced a bit. Do it a lot, lifetime comes down really rapidly. Do it a little all the time and lifetime is slowly sucked away. Even normal, not-too-big filter caps will eventually wear out the cathode. Bigger caps wear it out faster. An NTC would help with the power on surges, but won't help too much with the constant wear of a too-big cap.

Of course, it's better than just lumping in a too-big cap and not doing anything to soften the blows.

[TUBEDUDE]

I'm a proponent of the BFI method. Reducing the shock of strong charging pulses while reducing the D.C. voltage minimally. And adding resistors to the secondary winding if the windings impedance is too low.

Ànd pre-rectifying the voltage by installing sandy state diodes on the rectifiers plates, as protection from rectifier failure, might be a good idea.

[maxkracht]

Each time the rectifier cathode sends out too much current, the cathode life is reduced a bit. Do it a lot, lifetime comes down really rapidly. Do it a little all the time and lifetime is slowly sucked away. Even normal, not-too-big filter caps will eventually wear out the cathode. Bigger caps wear it out faster. An NTC would help with the power on surges, but won't help too much with the constant wear of a too-big cap.

Thanks, I was still operating under the assumption that it was mainly the extra current from startup that caused problems. A bit of stress every cycle makes sense.



Do ss diodes before a tube rectifier generally improve the tube’s lifespan or decrease damage from high capacitance in any way? I generally add them by default, often with a pair of fuses before the diodes, but just as fault protection like tubedude says.

[R.G.]



Do ss diodes before a tube rectifier generally improve the tube’s lifespan or decrease damage from high capacitance in any way? I generally add them by default, often with a pair of fuses before the diodes, but just as fault protection like tubedude says.

Sadly, they're just fault protection. I think you're smart to add them by default, because they are GREAT fault protection and cheap.

I wish I knew if there was a damage threshold, a current peak below which tube rectifiers led a prolonged life. This makes me speculate that there are some scholarly papers from the 50s on safe cathode current density for oxide cathodes. There almost has to be, somewhere.

I hypothesize that using SS diodes into a pre-first filter cap, and from that into the tube rectifier (with both sections paralleled to cut current density) and the original first filter cap would cut the peak currents a lot. The tube rectifier then only has to contend with the average current and ripple voltage on the added front end rectifier/filter cap. This is a lot lower than actually rectifying. The relative values of the pre-first filter cap and original filter cap will need some juggling to get the lowest peak currents on the tube rectifier. I probably ought to do some simulations on how much this helps.

[maxkracht]

I probably ought to do some simulations on how much this helps.

I would be very curious to see the results. Always looking for simple ways to make things more reliable.