Emission Labs - Introduction

WHITE SPARK PROBLEM

Notes for manufacturers only.

These notes are not for end-users. Please have your amplifier manufacturer read this, in case of a spark problem.

What a white spark is:

A white spark can occur within every large size tube with Barium oxide coated filaments. (Barium oxide means the tube lights up red-orange. For a tube like 845, that lights up yellow-white, this does not apply).

The spark is caused by the plate current for a very short moment being above the maximum allowed value. This value is specified in the data sheet under MAXIMUM CURRENT. Probably you have a large surge current somewhere, causing the effect.

If the maximum current is exceeded, some tubes will produce a spark, others will not. So when you see this is linked to only one tube of a pair, this is still an amplifier problem.

With any electronic device that gets hot, such as power resistors, large transistors and also electron tubes, the MAXIMUM values are fatal limits, and when exceeded a defect will occur sooner or later. If no effect seems to occur right now when you try it, you may likely get the defect later. So don't even think of using a power device at it's maximum values.

Apart from this, it is highly recommended to do a worst case test, with all possible tolerances at it's maximum, like maximum mains voltage, and maximum ambient temperature at the same time, use maximum output signal, and connect somewhat lower impedance speakers. THEN test if your tubes operate SAFELY below maximum limits.

What happens if you exceed the maximum plate current of an electron tube?

Battery tubes are the smallest tubes made, and these will damage immediately, and it's permanent. Small tubes, like 6922 or 12AT7 will also damage immediately, and have only 20...80% of emission after the mistake. When new, they often recover slowly when you go back to normal conditions, but never will be as good as they were. These tubes are off-topic here. The next is more important:

Power tubes like the Emission Labs tubes, have always strong overcapacity of the emission. So you may not see immediate loss of emission when you make a short mistake. They recover from most errors when still new, and when the error did not result in excessive heating of the plates longer than a few minutes.

The current density of the cathode coating is physically built by many small current islands. So it's not evenly divided over the surface. When the maximum current density is exceeded, it can happen that material from the cathode is chipped off, at exactly these places (islands) where you have the highest emission. This island is now "dead" and a very small white dust particle chips off. Read in the next lines, what will happen to the chipped off material.

The electric field inside the plates is as large as 50kV/m. Because this high field gradient, this chipped off material will fly around, and get ionized. When it ionizes, that means it can get conductive. This is the spark you see. The spark has a negative dynamic resistance, meaning it it's "ohm" resistance gets lower, when the current gets higher. So this gives an avalanche effect, drawing extreme peak current. This can force the plate voltage down to 100 Volt, regardless at what voltage was. You can think of the tube acting like a 100 Volt Zener diode at that moment. This can easily blow the plate voltage fuse, and when it doesn't have that, it can blow the main fuse of the whole amplifier. This will give a very loud crack sound from the speakers. After the spark is gone, the tube will recover from it, and interestingly, it works normal again, and have no damage. Still you must avoid this situation. Here is what may cause it by mistake:

Triodes: Positive grid voltage. Bad electronic circuit. See the checklist below.
Diodes: Too low first capacitor, too fast rising plate voltage at switch on. Bad electronic circuit.

What a white spark is NOT:

A white spark is NOT a bad tube issue, and it is NOT a short circuit of the tube. Vacuum can not short circuit.

Question: Why does it only occur with a new tube, and not with old tubes?

Answer: It is easier to exceed the maximum peak current when a tube is new, when it has highest emission reserve. So you may observe this with larger tubes, and with newer tubes. But it can happen to any tube, when the electronic circuit is not good. In that case it will particularly happen when they are a high emission kind of tube.

Literature:

In electronics this effect is called "trigger".

Warning by OSRAM, from 1925, about sparking tubes
Warning by RCA for rectifiers, about sparking tubes
Table of lowest allowed transformer windings resistance (yes!)

Checklist:

What peak value do you measure for the grid-to-filament voltage at switch-on or switch-off
What peak value do you measure for plate current at switch-on or switch-off?
DC-coupled circuits need to be measured very carefully for this, they can behave unexpectedly.
Measure the surge current through the coupling capacitor, to the end-triode. What is the surge current while you switch on and off the amplifier? You will be surprised how high this value can be. This current will flow through ground via the grid-to-ground resistor, and cause a positive voltage there of 50Volt and higher. (The Grid must have a negative voltage ALWAYS)

The following are very useful items, which can SOLVE and PREVENT problems. Good luck to you if you already knew it all, and tested your tube amplifier for all of this, but perhaps one of this points may be helpful to you still. Who knows!

What can you do if an amplifier tube is sparking?

First: Don't work around measurements. The reason is very simple: If you would have known, you would not have made the error! You can find such errors, by measuring everything. This means also the things you thought you didn't need to measure, because you expected them to be good anyway. Here is one typical verbal conclusion, no good engineer will accept: ".... I tried tubes of several brands, and I never had a problem. So the amplifier must be OK....". Another typical verbal conclusion is: "....The problem is linked to one tube of a pair, and so it must be the tube, and the amplifier is OK....". Well, perhaps this is true, and perhaps is not. The real answer can ONLY come from measuring it. Is the amplifier OK, then it will pass the following tests:

Disconnect the coupling capacitor to the end-triode. If the spark is gone now, you have at least localized onedesign error.
A standby circuit sometimes solves the problem, but... when made wrong can also be the cause of unexpected, hard to understand problems. Don't even think of using a standby circuit, when you have no good way to measure surge currents that may result from it. You try to do something good, but may be doing something very bad to the tubes.
Overdriving DHT power triodes is definitely not allowed. Will the amplifier allow for excessive overdrive with the volume at maximum? If yes, this is an elementary design mistake, and you need to correct this. Overdrive must occur in the driver stages first. So the end tube is not in an overdrive condition yet, while the driver stage starts to clip or distort. This allows the user to reach the overdrive level, and effectively not overdrive the end tube.
Reduce the grid-to-ground resistor to the data sheet value. 2 Mega Ohm is fine for miniature tubes, but not a good value for a 300B tube. A low enough grid-to-ground resistor will protect against the coupling capacitor surge current. At switch on, the voltage across the coupling capacitor is zero, and then this large capacitor gets charged within a fraction of second. Charging a capacitor will make a DC current flow through the grid-to-ground resistor. When this resistor is too high value, POSITIVE grid voltage results from it. The tube will now try to produce the absolute maximum plate current it can do, and at high enough plate voltage that will (of course) give a spark in the tube. The better quality the tube is, and the newer it is, the higher the chances on a spark. This is why we specify limits to the grid-to-ground resistor. What counts is the R x C product, so also read the next paragraph.
Change the coupling capacitor to the smallest value you need for good frequency response. Oversized coupling capacitors are the number one problem makers for DIY, but also professionals make this mistake often. It can damage DHT tubes. Calculate the correct value, and verify it by a measurement. The coupling capacitor to the end tube is where you need to limit the low frequency response of the amplifier. Do not oversize this capacitor, and leave it up to the transformer in a natural way to limit the low frequency response. Whatever you do, always measure the peak-current in the end tube at switch on. Perhaps you will be surprised how large it is. (And again... we have the maximum value for that in the data sheet).
A grid series resistor can partially help to reduce the plate peak current at switch on. This resistor will give an RF frequency roll off, which is why designers put it in. It prevents ringing, and also it reduces interference from radio stations. This is why you see this resistor in professional designs. What you can do for HiFi is choose the resistor low enough to get a frequency roll off outside the audio range, but choose it high enough to limit the grid surge current (into the grid, via the coupling capacitor). The R x C product here is coming from: Series resistor value, and tube grid capacity multiplied by tube open loop gain. (For Miller effect). So the value must be chosen with brains. Not just take "some" value that you see in other circuits. Often these people just copied it also from another design, and so on. Measure what the resistor does. Where does it make the roll of start? When at 10Mhz, it's useless and wrong chosen. When at 30kHz/ 3dB, it's too close to the audio range of course.
Please Calculate your designs, and verify it by measurements, and measure the switch-on surge current, and also the switch-off surge current. Measure systematically everything. Trial-and-error designs are one thing, but if you don't at least measure the results of what you produced, you don't know what the circuits really does! You may be heading for defective tubes without knowing it, and when it does happen, not realize where it comes from.

Checklist for the amplifier:

What peak value do you measure for the grid-to-filament voltage at switch-on or switch-off ?
What peak value do you measure for plate current at switch-on or switch-off ?
DC-coupled circuits need to be checked very carefully for this, they can behave unexpectedly
Measure the surge current through the coupling capacitor, to the end-triode. What is the surge current while you switch on and off the amplifier? You will be surprised how high this value can be. This current will flow through ground via the grid-to-ground resistor, and cause a positive voltage there of 50Volt and higher. (Grids are supposed to have a negative volt age...)
Since you must count with people overdriving your amplifier even if you warn for it. Make sure the overdriving and limiting takes place in the driver tube, and not in the end tube. This gives good protection to the end tube.

What can you do if a rectifier tube is sparking?

Measure the peak current, and check see if you exceed it.
Check if the HV windings of the transformer do not have too low DC resistance. This is a widely unknown specification, but it exist for ALL rectifiers! If the resistance is too low, add a series resistor.
Make sure the plate current is NOT higher than normal at the start-up moment of the tube, while the filament voltage is lower than normal.
For the peak current during start up, be aware that with a C-L-C filter, capacitors C1 and C2 are in parallel, because the choke is not working at this moment. The choke is probably saturated, and then becomes a copper wire only. Add a series resistor in the choke circuit, to prevent problems like this.