Technical Bulletin TB-03

CURRENT SOURCE as FILAMENT SUPPLY

by Jac van de Walle

Introduction

Please let me quote this interesting text: People rather admire what is new, although they do not know whether it is right or wrong. They even prefer it to what they are accustomed to, and of which they know it is right. What is strange, they even prefer this to what is obvious. I must state here, what the mistakes are, which I wish to unteach, and what instructions I have to give to those who want to become specialists. Source: Hippocrates, in his book "Bone Fractures". Published in the year 400 B.C.

From here, we need to make a little jump. Some 20 years ago, DC heating of tube filaments was an absolute "no go" because of bad sound, which everybody could hear. Despite the low residual hum, many amplifiers were AC heated. Today, this situation has reversed. Everybody hears now the better sound of DC heated tubes.

At Emission Labs, our primary goal is, that our tubes are used in such a way, that best sound comes together with best life time expectancy, and lowest failure rate.

For this reason we need to put up front clearly in this discussion, that the major reason for problems with our tubes, in almost every case had something to do with these cases:

  1. Self build heater electronics.
  2. Unusual circuits for switching the whole amplifier on.
  3. DC coupled output tubes.

This Technical Bulletin is only about the first issue. Here it will be explained in some more detail, what happens inside the tube, how this influences the result sound wise.

It should be avoided to speak about "sound" only, and ignore the technical aspects behind it. These subjects have to understand first, before final judgement.

The items discussed will be hum, inter modulation, the bias scheme, and what this does to the lifetime and reliability of the tube.

Direct problems can be cathode sparking, and cathode chipping resulting from this can be an issue. Also broken heater wires will occur, due to accidental over voltage.

Contents

  1. What is a heater?
  2. Electrical behavior of a heater
  3. Some Basics of a "tube" heater.
  4. Type of tube heaters
  5. The intended use
  6. An experiment with light bulbs.
  7. Back to tubes
  8. An issue with hum.
  9. Biasing high power tubes.
  10. Lower Distortion with current source heating.
What is a heater?

With indirectly heated tubes, such as EL34 and many others of course, the whole problem gets less complicated, because the manufacturer can separate the heater function from the cathode function. So the cathode can be made of soft nickel, which is the easiest material to use. Whereas with direct heating, there are also such things as wire breakage or wire sagging, and actually a lot more issues. Moreover the slow warming up of the cathode, which is better as long as not too slow, is automatically achieved by the larger mass of indirectly heated cathodes. In texts, we find the word 'filament' often for DHT, and 'heater' is often used for indirectly used tubes, though I speak myself often about the 'heater' of a DHT, when it is doesn't matter anyway.

Electrical behavior of a heater

This is an important subject. We all know that resistance of the heater goes up with temperature, which is simply expressed by the temperature coefficient of the metal or the alloy. It can be expressed in percent per degree temperature. But in vacuum something else happens. The heat of the element can only escape by thermal conduction to the places were they are attached, or by heat radiation through vacuum. (Which radiation physically behaves the same as like light). Since the tube makers naturally try to avoid the loss of heat to the attachments of the wire, the only way for the heater to get rid of the heat is by radiation through vacuum. Logically, there is no heat conduction via air, but actually a gassy tube can be unveiled also by the lower heater wire temperature. So if a tube is totally dead, because of too much gas, the heater will be lower temperature. This can be seen with the eye of course, and also be seen electrically because such a heater will draw 130% or more heater current, because at lower temperature it will have lower resistance. This leads to the curious situation, such a tube draws a lot more heater current, while the heater only stays very dark red. Why do bring this up? In good vacuum, there is mainly heat radiation, and a tube heater if the color is orange, behaves like a constant current user. Even so with the right alloy, the constant current effect can be optimized, and for this purpose, so called Ballast tubes were historically made. Which were current regulators, protecting radios from too high mains voltage. These work amazingly nice and you can still buy them cheap. Even so, when you find the right type, they can be used as "current source", replacing the cathode resistor. Also at EML we use such devices in our new generation of rectifier tubes, to regulate the heater voltage inside the tube.

This is a current regulator as we use it in EML 5U4G mesh tubes


Some Basics of a "tube" heater.

When using a ceramic resistor as a heater element, it is clear that a 10% increase in voltage will result in 10% increase in current, just by Ohm's law. This is so, because they use a wire inside with low temperature coefficient.

A tube heater is not made to be a "resistor" in the first place. It will behave like a resistor at first, but not when it begins to glow. Once the glow begins, they begin to develop some "constant current" effect. When the glow is bright orange, they are quite a nice constant current user. This means, a 5V, 1.2 Ampere heater will initially pull 4 Ampere current when it is cold. When the heater warms up, the current will be reduced initially very quick, but then this slows down, and it takes up to 2 minutes to reach it's final value of 1.2 Ampere.

Here comes a difference with a normal resistor. Suppose we have a ceramic resistor which pulls 1.2 Ampere, we can easily increase this to 1.6 Ampere, just by increasing the voltage accordingly. With a tube heater however, this will not work. Though initially at higher voltage, the current will go up the same way, but right after it will go down, into the direction of 1.2 Ampere again. This will be so, despite the higher voltage. In fact, any attempt to increase the current significantly, will lead to a broken heater wire.

The observation is, when we try to exceed the current of a tube heater, it begins to behave as a constant current user, not far above it's rated voltage, and from there of course, it will behave as a fuse.

We all know, two different voltage sources can not be out in parallel. The result will be smoke. By the same logic, you can not put two current sources in series, because in one wire can flow only one current.

Knowing, a tube heater has a constant current effect, we have to be extremely careful when we serialize it with a current source of not the same value.

Conclusion: When using a current source to heat a tube, the current source must always be voltage limited to the tube heater voltage EXACTLY. So if the voltage would go above 5V, whatever the reason, it must reduce it's current. A plain current source without protection, will lead to tube damage.

Type of tube heaters

When selecting a tube from the catalog, one can distinct three kind of heaters:

  1. Voltage specified. Such as ECC88, which has a 6.3V heater
  2. Current specified. The same tube is called now PCC88 , it has a 300mA heater.
  3. Both allowed for voltage or current supply. Such as ECC83. It can be supplied with 6,3V or 300mA. This is very rare, and with mist tubes you can not do this.

So other as what most people think, not all tubes are allowed to be current driven. Reasons have been described in the previous paragraph. If you would decide to current drive a voltage specified tube, you must ALWAYS adjust the tube voltage precisely up to the last percent. Such electronic solutions exist.

The intended use

Constant current heaters, for instance the 300mA type, can all be put in series, and have a similar heat-up behavior. Like this you can save the heater transformer, and heat the tubes directly from the mains supply. In some nice and interesting designs, even the heaters can be used as cathode resistor at the same time. Like using the bias current of tubes like KT88 to heat the pre amplifier tube with DC current, at very little electric effort. In all cases, current specified tubes of the same range (like 300mA as we have here as example) have a similar heat up behavior. That prevents one of them will act as a fuse. It is important to know this behavior is build into the 300mA series by default, and build into voltage specified tubes only by coincidence, if at all.

Nevertheless, even the special current designed heaters behave not perfect. The old timers amongst us will remember how TV's with serialized tubes often had the ECC83 on the channel selector getting a broken heater. Whereas this is very rare when the same tube is voltage driven.

The constant voltage heated tubes, differ mainly by having no particular heat up behavior. The cold current is undefined, as well as the tolerance on the hot current. So will start up can be 4x the rated current, others with 8x. Some will be at 1.3 times the rated current already after 10 seconds, others need one minute for this. Even with two of the same tube types this can have quite some differences. Only occasionally, with voltage driven tubes, the heat up behavior is specified, and can be found in added numbers or letters like GTA or GTB. 6SN7GT is just a "Glass Tube" that is what GT stands for. GTA or GTB however are different. GTA is a fast warm up tube, which special property was intended for oscillator applications. GTB is even faster.

It is evident, you can not serialize random constant voltage tubes. So a 6.3Volts ECC83 can not be serialized with an EL84 and just put this on 12.6 Volts. Some that want to learn this the hard way, can try it. The effect will be, the ECC83 will burn it's heater brightly white, before it passes away, while the EL84 does not even get warm. For better learning effects, it is recommended to take an NOS Telefunken ECC803S for this. In any case, it is not appreciated by us, to do such experiments with EML tubes, and guarantee voids in case you want to try nevertheless.

But even two EL84 can not by "simply" serialized at 12.6 Volts, because of unknown tolerance of the heater current. This gets worse if the tubes are from different production date. So here we see already one difference between constant current heaters and constant voltage heaters: You can not always serialize constant voltage tubes, even if the heater current is the same! Whereas with two PL84 (for series heating) this can always be done, no matter what date code, or what manufacturer.

Suppose you take two 5U4G rectifiers, made for a constant 5V each. You just serialize them, and connect them to 10V. This will go wrong if the tubes come from two manufacturers, or even from the same manufacturer, but from different production sites, or other date codes. One will heat up faster than the other, and the 10V will divide unevenly. So there will be a moment where one has 6V and the other 4V. After a while they will balance at 5V or close, but the heat up was not good for either tube, since one was over heated, and the other under heated. Both is very bad for the lifetime. The one that got 6V for a few seconds will break the heater sooner. The one that was too long on 4V, will suffer cathode stripping.

However with series heater tubes, this is not going wrong, as they are just MADE for this. I am always surprised to see over and over again that most people, including professionals have not the faintest idea about this. When being pointed out this mistake, they dispute it, saying "we are doing so for 25 years and we never had a problem".

An experiment with light bulbs.

Here is an impressive experiment, that will change your mind. Serialize two incandescent lamps for the same current, but another voltage. Connect them to the rated voltage with a switch. Though technically speaking you may expect this to work, but the heat up behavior is your enemy now, You will likely burn out the heater of one of them. Then, serialize two incandescent lamps for the same VOLTAGE, but another current. This will blow the one with the lowest current.

Conclusion: The constant voltage heaters may not be connected to a current source or vice versa.

Back to tubes

We need to look at the constant current effect of a tube heater. Let's give a realistic, numerical example for the well known 300B by Western Electric. The heater voltage is 5V, at 1.2 Ampere. They never specified a tolerance for either of those values. What is sure, they write this is a voltage specified tube. This means, best lifetime you occur at 5V, not more and not less. There is a strong tendency by users, to take 10% tolerance for anything without tolerance specification. However this is not so. Each 0.1V you deviate, in either direction, will reduce the lifetime significantly. (The military handbook says: exponential to the power 13). So if you are 2% off, the tube will not reduce lifetime by 2% but >20%. Most likely +/- 5% is the (deadly) limit. So be nice to the tube, and give it exactly 5V, it is the best you can do. Make very good note, that a few percent is very much. Now, what current will it draw at 5V? Well, most of the time 1.2 Ampere, but not always.

Add to this the natural tolerance. But...how much is that? If you ask the so called 'experts', a 300B is supposed to draw 1.2 Ampere, period! So they want zero tolerance on the current draw. Then, If you ask the experts about the voltage, they tell you 10%. This 'expert' expectation is a very crude error. It is based on no data sheet, no literature, and not in line with anything will observe with real, physical tubes. Even so, if there is tolerance on the 300B heater specs, it is in current draw, and not on the voltage.

In real life, most WE 300B at 5V will draw from 1.2 Ampere, but some will draw 1.1 or even 1.3 Ampere as well.

A real measurement.

Now let's see what happens, if you give only 1.2 Ampere to a tube that would use 1.3 Ampere at 5V. (so that is 8% above average, for this particular tube). It is evident, the voltage will be lower than 5V. How far below will it be? Well the only thing that counts is a measurement. I did so, on Western Electric tube. The result was: 4.45Volt. This is 12% lower. If the heater would be an Ohm's resistor it would be 8% lower, but due to it's constant current user effect, the effect is larger. I used a 1958 NOS Western Electric tube for this.

Conclusion: A current source may only be used when the current is individually adjusted to the tube in such a way, that the voltage across the tube is always 5V. Any other devices should not ne used.

An issue with hum.

When connecting the DC module with one end to ground, and the other end to the filament, this greatly increases the sensitivity for the tube for AC ripple. So you might be disappointed with the result. In this case you are amplifying the AC ripple with the tube gain now, instead of rejecting it with the classical (differential working) connection scheme. If you are not familiar with common mode rejection, we can not explain it here quickly, but just look at these diagrams and use the right scheme.

Biasing high power tubes.

If feeding a DHT heater, with with DC voltage, this causes irregularity of the grid voltage inside the tube. This is so, because the grid voltage is defined from cathode to grid.

Here is a drawing that hopefully is self explaining.

A (general) linearity problem.

This graph is from A. Barkhausen 1953. He gives here an example for a DHT tube with five heater wires. Each has of course the same curve, but they are biased differently. From the horizontal axis you can see, he did so for a 4 Volt tube. The heater voltage 4V is distributed along those five heater wires.

Influence of the heater voltage on the anode voltage, for each heater wire.

The heater voltage of 4 Volt in this diagram is subtracted from the anode voltage. This has to be done separately for each of the five heater wires, which are each in the drawing. The fat line is the average value. Even when the tube would have ideal properties, meaning each line would be fully straight, still the averaging effect would bend the bottom part. You can do this as a practical exercise, you will see it is so.

This irregularity has some problematic effects.

In the above drawing picture, you can see -60V on the left side, and -55V on the right side.

  1. This means, the peak grid voltage of he whole tube is limited to -55V. By itself this is not a problem, only the data sheet curves would indicate -57.5 Volt for this, and the functioning of the tube may not be fully as expected.
  2. A more nasty problem is the uneven heat dissipation in the tube. This really creates a considerable difference between the heat development on the right and the left side. If a tube is not specified for DC heating, this is the reason.
  3. The curves in the Barkhausen diagram look all identical, but just when looking at real tube curves you will see they are not. So for each heater wire a tube is in parallel working to the others, and the total result is added up. This will introduce some distortion.

With this technical bulletin, I have tried to generate some more sensibility that DC heating of a tube is generally not an advantage, but it is the most convenient way to eliminate AC hum.

Solutions can be AC heating with:

  1. Radio frequency (for instance 200kHz)
  2. Ultrasonic frequency (for instance 30kHz)
  3. Subsonic frequency, from 5...15Hz.
Lower Distortion with current source heating.

After all the talking about the disadvantages, people that use current source heating report lower harmonic distortion. How can this be? The answer is unexpected: This is a form of negative feedback!

Assuming, the current source is high quality, is has a high internal impedance. So, with a 300B tube, the current source would just inject it's 1.2 Ampere into the heater, regardless of anything else. Even regardless if some other small current was already flowing. In that case it just adds up. The AC signal current by definition will leave the tube via the heater. Suppose we have 60mA AC current, what will this do to a 4 Ohms heater wire? This will give a voltage of 240mV if just flowing from the left to the right. In reality it the first (of 8 wires) will see only 1/8th, the second wire 2/8th etc. So we have to calculate with en effective 120mV, instead of 240mV.

This 120mV gets effectively Subtracted from the grid voltage, and what we have here is simply a form of negative feedback. It is less than 1% feedback, but it can explain the tiny differences that were reportedly measured by some.

This kind of feedback also takes place when creating a virtual cathode, by means of two centering resistors. These are generally higher value than the resistance of the heater wire, only to prevent those from being heated up. However the negative feedback effect with those gets much larger. If two resistors of 50 Ohms are used, each would see half of the 60mA AC current, resulting in 1.5 Volts negative feedback. This is a few percent already. For this reason, if you don't want this feedback, they should be shorted with capacitors. When doing the same thing with a current source, a much larger capacitor (across the 4 Ohms heater) must be used.

Loss of efficiency with current source heating.

This may be easiest explained by the simple fact that 120mV AC signal is applied to the electrical resistance of the heater wire. So we loose here 3 milliwatt. Not very much, only a theoretical amount. With voltage source heating, the AC signal is shorted by the voltage source, and so it can not develop energy in the heater resistance.

End of this Technical Bulletin

+ + + +

Here is a most interesting link about this subject too. There is a small in error in this model, saying you can simulate a five-heater-wire tube by but in parallel five of the same tubes with 1/5 of the current. That is because the tube not working like this in the cut-off range. It does work like that when all filament are generating pulling current. But it does not work, when three of the five are cut off. The explanation is (once again...) in this Barkhausen diagram here. So you can see there, the five single curves are much more linear as the composite (fat line) curve. Yet take 1/5th if this fat line, would take also 1/5 of that unlinearity. Whereas the composition should be made from the five "better" lines. The writer oversees this situation, but for the rest his approach is magnificent, and so is the rest of his website. I noticed that website is very old, so you should consider to make a copy of interesting articles.