Technical Bulletin 08

The right way to TEST Electron tubes

For Emissionlabs, this here is an important information page. This is about the best way to test our tubes.

Meaning first: Do not use fixed grid voltage testing. This is not in line with any data sheet ever printed, or test instruction by any tube manufacturer. If you think to read this from a data sheet, you made a mistake. Only very rare battery tubes, of which the heater voltage is used as bias voltage too, should be tested with fixed grid voltage. All others need fixed anode current. (Called auto bias)

Meaning second: When using impulse testers, we must understand the errors, resulting from cold anode testing. Errors which occur only under dissipation will remain undetected. Such errors are usually catastrophic errors, causing the tube to fail completely, or blow the fuse.

So when we see somebody comparing a few tubes, test them with cold anode, saying they perform nicely to the data sheet, so they must be fine, this person missed the essence of quality control with tubes. Being up to the data sheet is step1 of the testing. Step2 is assure the tube will not fail under maximum heat. Feel free to use them at lower heat, but if a tube develops grid current at maximum heat, this is a hidden failure, and such a tube won't make it's owner happy for a long time.

Such failures will result in catastrophic errors on the long run. Such as parameter drift, G1 leakage, cathode sparking, and cathode to heater leakage. Really, when testing G1 leakage or cathode to heater leakage with a cold anode.... you will not a representative result. Specifically these two errors escalate mainly above 85% of maximum dissipation.

I do not want to criticise impulse type curve tracers for this, but I do want to say it as it is, they are good curve tracers, but useless for tests which require continuous heat.

Even so, the legendary AVO Mk4 tester, has special settings for isolation tests, to enable the user to see the difference between a cold and hot tube test. So you will see then for instance, the tube is good with cold anode, and fails with hot anode.

All of this is not complete! Please inquire if you have a tube tester you believe that fits in here, we can add it.

Here are the categories of tube testers for as far as I know, which can be used to test EML tubes. Placed here is order of usefulness.

  1. Good vintage testers for testing EML tubes, are the Neuberger 370,375. The 'Sofia' by Audiomatica is the Queen of tube testers for me, but unfortunately too few made. Funke W20, and AVO VCM163 with adapter socket for UX4 tubes.
  2. Roe-test, by Helmuth Waigel is a shooting star of the full power testers. There is nothing like it from new productiont. The SOFIA tester, which I personally like more, is not made any more since 1995, and not enough were made to serve the second hand market.
  3. Amplitrex AT1000 can be this position #3. Provided you know how to use it in PC mode. I have seen several times people not understanding the need for this, and (what a coincidence....) not having the skill for this as well. However using it in computer mode is mandatory to bypass it's many bugs and failures. When you only press the 'test' button in the stand alone mode, you will get a wrong results in 50% of the cases. So the span from weak to strong does not reflect the tube's real condition. Not due to hardware errors, but software concept errors. I will write some more about this, further along this page.
  4. Useful vintage testers, though these can not verify the EML at high voltage. Such are: Kalibr L1-3, L3-3, Hickok cardmatic, Metrix U61B. Probably several others. Such testers can at least be used for tubes which need maximum 250 or 300V plate voltage.
  5. Impulse testers are not the fully suitable for large tubes. Though Roland Dekkers latest version has the voltage and current to do those, such test can not find such quality issues which can only be found under large power. For curve tracing and parametric data yes, for quality testing no. Please let me explain some general issues about such testers here, and particular why several tube errors can only be found close to 100% maximum dissipation. Please don't understand me wrong when saying this, but we write here about how to test Emission Labs tubes, not about what is a good tube tester. Yet the item needs some minimum understanding, and simplifying it leads to mistakes. With pentodes G2 is cooled by carbon coating, and the heat back radiation of the anode is anticipated in the tube design. Besides for power tubes, a considerable part of the heater temperature is also generated by the anode, radiating back to the inside the tubes as well. Such tubes may become under-heated when the anode is left cold. Of course such tubes work, but with less emission as intended. Some tubes can be found however where this makes only a small difference, and some can be found where the difference is above 10%. For quality and matching, parameters must be checked, and these are not the same with a cold to a hot anode.

    Note, cathode heat dissipation is approximately 20%, and anode heat is 80%, depending on the tube type. So this cathode of say 5 Watt, is captured inside an anode of 20 Watt. This 20 Watt gives SIGNIFICANT heat radiation back into the tube, where the grids and the cathode are. This brings up the major issue with impulse testers. Keep in mind, tube defects other than low emission, are mainly catastrophic type. Meaning, such a tube may fail completely at unexpected moments. Moreover, such failures only occur if close to 100% maximum dissipation. This may not seem a good idea to do, but some tubes like 2A3, AD1, ECC83 or C3g, are actually MADE for this. In most other cases this is indeed not a good idea, but designers seem to love it. Some catastrophic defects, appear only at full dissipation, and come suddenly, like shorts, or sparking of gassy tubes. Other catastrophic defects are not present at first, but appear suddenly above 80% of maximum anode dissipation, for a time of 5 minutes, and then rise sharply when closer to 100% emission. It should be clear, impulse testers can not verify this. Such defects can be cathode to heater leakage, or the very infamous grid emission of vintage DHT tubes. This can happen with RCA 2A3 or 45 quickly, after over heating them 15% above maximum already. So 17 Watt for 2A3. This will cause permanent severe grid damage. If used above 15 Watt, such tubes may seem to accept it. Yet after months, suddenly drift thermally. So anode current grows, and dissipation becomes 16 Watt, 17 Watt, etc. Suddenly the anode will glow red-hot, and the amplifier fuse will blow. Yet at repeated test, the tube will initially work good, but after some warm up time, the failure will repeat. Here comes the issue. If used at 10 Watt only, such a tube may work good. The problem is caused by Grid emission at full dissipation. Yet, with even cold anode, like on E-tracer, uTracer, there is no grid leakage visible, curve charts and all tube data looks good. Cathode to heater leakage causes mainly hum with AC heated tubes, but also may cause hum with DC heated tubes, because of main transformer primary to secondary capacitance, which finds a resistive path now, into the cathode resistor. I can only encourage everybody, to take a tube of which you know it has cathode to heater leakage, and measure this resistance under full tube dissipation, and with cold anode. It must be measured with the data sheet specified Voltage between heather and cathode, otherwise the measurement becomes useless. This voltage ranged from 80 to 200 Volts. Yet this is really a good learning experience when you have buy coincidence a tube with known leakage of this kind, and you will be extremely surprised to see, how this leakage appears only at 80% or more, of maximum dissipation, and is fully gone below that.
Forum babble vs data sheets.

Of course users become confused with all of this, but the greatest confusion to my opinion results when people who do not know, ask other people who also do now know. At the same time, when somebody knowledgable explains something, it is regarded an "untrue" opinion. This behavior is one of the strangest in the world of HiFi, and I have really given up trying to understand why this is. The best source of information is always what the manufacturer writes. Like when they say: "this is the test procedure... " Then, at least read it. You are free of course, to regard such information untrue mistakes. Just it becomes strange, advising people in forums to do the same thing, or not even mentioning there are some manufacturer's instructions which say the opposite. A red warning lamp should to burn, when people read pages full of facts, and then say, they have problems believing that.

A good data sheet.

For understanding how a good data sheet is written, there is no replacement for studying some good and bad examples. A bad example is JJ ECC803S though you may initially like the simplicity. But it is wrong for ECC803S, which is a life time specified tube, with specified acceptance limits and specified end-of-life limits, to ignore all of this, as JJ does. It makes me wonder what else they ignored. A much better example is Telefunken ECC803S. Or a excellent example for professional audio tubes, is TELEFUNKEN C3g. It pays off, trying to understand WHY they write the things they write it there. So not try to understand the tube curves, that is not what I mean. These curves are only for design purpose, but for quality control they are useless. Try to understand the structure of the data sheet. Because it was written for YOU, for the user. So read how to test for instance Telefunken E88CC. I am not interested in any other recommendation as Telefunken says. Telefunken does not say: USE AUTOBIAS. But they say: THIS is the cathode resistor for testing. And THIS is the grid voltage you must apply, and THIS is the plate current for a NEW tube, and THIS is the plate current for an end-of-life tube. Now look, these are clear instructions, Not only print the curves and give only average data for ECC803S, like JJ is doing. Also take good note, they do NOT say you must use a fixed grid voltage. Read this in detail, and draw the test circuit they say. You will then see, it is auto bias.

What else can they say. You really need to read the TELEFUNKEN C3g data sheet, and draw for yourself the auto bias circuit which is needed to test C3g, exactly the way they say it. The break through for yourself will come, when you realize this is an auto bias circuit indeed, and understand how this circuit works. After that you will realize how badly wrong fixed grid voltage testing is. Not only because (I hope you noticed...) no manufacturer on the world ever wrote you have to test tubes with fixed grid voltage. But because fixed grid voltage will not reflect the Gm of the tube at the INTENDED bias point.

So far, for this much too long introduction.

At Emissionlabs, all we can do, is say how to test our tubes.

  1. Short instructions for testing of EML triodes
  2. What causes misunderstandings
  3. Fixed current Testing vs. Variable grid voltage.
Short instructions for testing

EML Tubes must always be tested at the VOLTAGE and at the CURRENT as we say. This can be done with a classical auto bias circuit, but it may also be done with a tester, or some other set up with variable grid voltage, such that the SPECIFIED CURRENT is set exactly. At this CURRENT, all other parameters have to be tested. Most of all, Transconductance and grid leakage, and for who is interested, Gain and Plate impedance can be tested during the same test.

Furthermore, testing should only be done after thermal balance. So, directly after adjusting the anode current to the EML specified value, it can be seen it begins to change slowly. This may take some 5...15 minutes, and only when the tube is fully warm, including the socket part, the test results become significant. This means the anode current has to be re adjusted until it becomes stabile.

Here is the same instruction, listed in steps.

End of Life

If even after longer waiting for final stability, the INITIAL parameters appear to have changed. The user who sees this, wants to have a life time prediction. What counts for this, is the CHANGE of those parameters, compared to initial values, which are written on the box.

We see it over and over again, that "specialists" which build tube testers, know nothing better to do, than comparing random tubes with data values as published, while testing the tubes with fixed grid voltage. This is so dead wrong. THIS IS WRITTEN NOWHERE. IN NO LITERATURE, AND MOST OF ALL, IN NO MANUFACTURER's DATA SHEET. This is ignoring the fact, that a data sheets represent only average data. As you will see, im those average settings, not only the grid voltage is fixed, but also the anode current, and the anode voltage. So from this alone, you may as well use fixed anode current (called auto bias) and see what grid voltage that needs.

This for instance is from the RCA 2A3 data sheet. (Link here):

But this applies for all 2A3. Anode Voltage 250V, Grid Voltage -45V, Plate current 60mA. Please check for this, you will quickly find it.

Where does it say: Grid voltage must be "fixed" and plate current is variable? It does not so say this anywhere! You may as well say plate current must be fixed, and grid voltage is variable! Please look up the original data sheet by RCA! At EML we can not speak for RCA, but we should not "read" things from their data sheet, which RCA has not written.

The use condition of a tube depends mainly on the change versus the initial factory values.

Some suggest that any tube which is not "average" has a certain "problem". Well, this may indeed be so. But you see, it may as well not be so, because you can not tell. This is why this method is so wrong. It tells you not much. A 2A3 which appears to have only 50mA (vs average 60mA) can be totally unused and it was only born that way, due to anode distance tolerance.

If so, such a tube will have normal lifetime as any other. In the same way, an unused tube may appear to have 85mA, suggesting it has 140% lifetime, or whatever this 140% is supposed to refer to. Also this will disappoint you, if you think such a tube well last shorter. You see, if it was that easy, to prolongate tube life by building them with too small anode distance, in order to have them test at 140%, all manufacturers would have done so ever since. Things don't work that way.

That is why we invest so much time here, explaining what is the the right way to test tubes.

How to estimate the remaining life time of tubes.

Wear out of a tube takes place in four phases as follows:
  1. The required grid voltage, needed to bias the tube at the factory plate current, becomes less negative. This is normal, and indicates only use. Wear out and use is not the same. Use hours may be significant, and yet wear out can be low. Use hours can recognized by this change in required grid voltage in the very first place. Though a small loss of transconductance goes along with this, mainly the grid voltage changes, and this can begin to take place after some hundreds of hours already.
  2. The above phase continues for quite some time, during which the tube would test strong still. At the end of this phase, a loss of transconductance begins to occur but no loss of plate impedance yet. Now the tube enters the third phase soon
  3. In this third phase, plate impedance begins to rise. This is a beginning wear out, though the tube will still work, and this phase can take quite long as well. As long as the gain is not changing, the tube is still in this phase 3, and for this reason (gain=ok) it will work good in most applications. The next phase is characterized by loss of gain.
  4. Gain of the tube is the multiplication of plate impedance and transconductance. For a very long time, the plate impedance will marginally go up, while transconductance will marginally go down in the same rate. This keeps the gain constant. The lifetime comes near the final phase however, when also gain begins to reduce. Not only the natural function of the tube (Gain!) begins to get lost, simply problems will accelerate now very fast, sound problems will soon appear.

Some important informations:
What causes misunderstandings
Fixed current Testing vs. Variable grid voltage.

Our target is to offer tubes that are well matched, in a way which can be verified by others. At EML we begin with factory testing, which is for emission, for quality and for data limits. Then match the tubes on the AT1000. We are not using the AT1000 because we think this is the best tester. It is not such a good tester at all. Only it is the most widely used tester. And though AT1000 has many shortcomings, it can actually be used for matching, if you know how to bypass the errors is makes by default.

AC Heating vs DC heating.

Tubes can be tested either way. Just test result is another for DIRECTLY heated tubes. That is because half the DC voltage effectively influences the grid to heater voltage. So a directly heated tube will give another result, if DC heated or AC heated. (If DC heated, the tube will draw less current at the same grid voltage)

EML Test method:
Test result for Gm (2). Small expert talk

This is an ever lasting subject. There are a few methods, which differ by precision of the result. We begin with the best method.

  1. Transconductance is a dynamic parameter, so to say an AC signal parameter. For best accuracy, it is s measured by applying a distortion free sine wave signal to the tube, with an oscillator, and then monitor the output signal of the tube, while using a band pass filter. Testers like the Russian L1 or L3 testers are doing this, and some other high class vintage testers. The reason why this is the best method, is any distortion elements from the output signal gets filtered out, as these come in the form of higher harmonics. Also any hum and noise is filtered out. So hum, noise and distortion make the output signal artificially too large, any make the output signal (and so Gm) look larger than it is.
  2. An oscillator method which uses no filtering, is better than non at all, but it is not ideal.
  3. A lower class method is just use a DC change on the Grid 1, but this doesn't filter out any of the above mentioned signals, and some testers suffer a drop of the plate voltage in such a case too. So in case of any difference with the first method, it is clear where it comes from.
  4. By software, it is possible to derive Gm by calculating it via the Barkhausen formula, but that may have resemblance with method 3. However when with clever algorithms it may be possible to make it look more like method 1.
How Gm depends on the working point ( a lot)

1602 Electron trube by RCA

Keep in mind, Gm depends heavily on the plate current, as you can impressively see from the curve on the left. This is from the 1602 which is nothing but a better 10Y by RCA.

As you can see here, Gm varies from 520 to 2100 depending on the plate current. So a huge difference of 1:4 So it should be clear when somebody says he 'has measured Gm of 1600', this by itself has nothing to say. Also results from tube testers, saying 'this is the Gm' but not saying the working point are meaningless too. If this Gm of 1600 was measured at a plate current of 60mA, the tube is bad, and when measured at 10mA, a Gm of 1600 is very high. . From this it becomes evident that saying 'this tube has Gm of 1600' is a useless information, as long as we do not know the plate current. The Gm value can only be compared with the data sheet at the SAME operating point where the data sheet specifies it. So as long as you aware of these potential error sources, it is ok. It would be totally wrong however to measure Gm at any random current, like done with fixed grid voltage testing, and then compare this with the data sheet value for a bogey tube.

Why fixed grid voltage biasing is useless for tube testing.

Here is a numeric example. The tube in this example is the RCA 10Y (1602 special). The data sheet specifies Gm of 1330 at 10mA, which should happen with an average tube at a grid voltage of -23.5V and 10mA plate current. Of course no two tubes will have average values. In order to compare Gm of an unknown tube with the data sheet value of 1330, you need to set this tube for 10mA, by adapting the grid voltage to whatever value is needed.

It would be wrong, to set the grid voltage to simply -23,5V because there is no 'must be' for this value. In fact, plate current is so variable, that minimum and maximum values are not even NOT specified that way. So it becomes totally silly to measure plate current still, and compare it with the average value of a bogey tube. Any 100% strong, factory new tube, at -23.5V will draw anything from 7 to 14mA. So it would be wrong, to let the tube draw whatever current from 7...14mA, measure Gm at that plate current (whatever it is), and compare this with Gm plate current at 10mA.

Seeing this from anther perspective, you can test a tube at any place on the Gm curve you like, but not compare such a measurement with another one, which is NOT on this curve. Logically, the only way to synchronize this, is by testing any 10Y at 10mA, and compare this with the curve at the left at 10mA. Or, alternatively do the whole process at 20mA or any other current you like.

Having understood the above, it should be clear, a Gm measurement should be made always at the data sheet specified DC current, and never at a specified grid voltage.

Using the Roetest

This is the Roll Royce amongst the new production tube testers. It comes close to the Sofia, but it is a kit. You can't buy it as a ready made product. If you can afford it, take this one,

Using the AT1000

LINK to Important Note for non-USA users of AT1000

This is written here is some detail, because we use the AT1000 at the factory for the data on the tube boxes. We do not do so, because it is such a good tester. The reason is, it is commonly used, and indeed it is one of the very few testers that can test under full anode power. AT1000 from before a certain date, have errors in the internal data tables for all directly heated tubes. Please communicate with Amplitrex directly, how to repair this. (We are not a service address for them) However when you set the tester to AUTO BIAS, which is required to repeat the factory test, this problem plays no role anyway. However you will soon see, sometimes (not always) tubes that test 'strong' on Auto Bias, may test 'weak' on Fixed Bias, or vice versa, so you can already see one of the methods is not right. Use only: AUTO BIAS.

AT1000 Stand alone mode or computer controlled mode.

Make good note, the stand alone mode has generally lower precision, because the tubes warm up only the heater, and not the anode, before a measurement.

Stand Alone mode

Computer controlled mode

For higher precision, the computer controlled mode should must be used, which allows heating up of the entire tube until thermal stability occurs, which is the only really good way to test a tube. For new tubes, the difference between stand alone more and computer controlled mode will be small. Used tubes however, benefit more from thermal stability. So, a quick test in the stand alone mode, will make used tubes look less good as they are.

Refer to the manual, so you know how to do the following:

Using full power analog testers, DC heated

This should repeat EML test data precisely. We have tested this with the Russian L3-3 and it works nice.

Using full power analog testers, AC heated

This should repeat EML test data precisely, as long a you correct the grid voltage for half the heater voltage. We have tested this with the Metrix U61 and it works nicely. Unfortunately it limits at 250V DC. (and a little bit outside if you try with an external voltmeter) Yet it can do not many large tubes.

Using E-tracer or uTracer

Please click here to read under Impulse testers.

Using the Sofia curve tracer

Mentioned here to honour this tester, but you can not buy it anymore. This tester has a possibility to heat the tube under full power. After this, a 10 curves chart is made in just a few seconds. The result is, all points of the curve charts are made with a fully hot tube. No other tester I know of, can do this. The AT1000 takes a few minutes for a full tube chart, and while doing so, the tube cools down, and the charts become jeopardized by that. This is why the Sofia is our one and only favorite for tube charts. It can deliver unusual high RMS power to a tube. Unfortunately such a tester is not made any more. The lowest class are impulse testers, which are not capable of heating up the anode at all. The AT1000 at least can heat up the anode before you begin with a chart.

Using the Russian L3-3

This is a perfect one. Measurements have reference quality. Unfortunately it is limited to +300V, but when using an external voltmeter, and not too full DC current, it can be used above 400Volt, provided you have the right test cards for this. We have test cards for EML 5Z3, 5U4G, 80, 81, 274A, 274B, AZ4, 2A3, 2A3-Mesh, PX4-mesh, AD1, 300B, 300B-Mesh, 20A, 20B and 20B-V4.