First we start with this standard data, from the Muiderkirng book. This datasheet shows the way an ideal tube would bias. The complete book you find on www.4tubes.com.

Here is a detail from the same datasheet. Look at those nice, parallel lines, with their well defined cut off points. So the lines do not point all to some vanishing point, as with real life tubes. The distance between each line is the same, regardless where you look. This means the tube has no distortion. Now with amplifying devices, distortion and gain go hand in hand. So for a tube with very high (100x) gain, some distortion is NORMAL. However, not so when looking at those GE datasheets here. Wow! they have no distortion. How did they do that? Well for drawing such curves they used a flexible ruler. That is a plastic bar, with lead inside, and you can nicely bend it in any desired shape to draw curves. Anyway, these curves show how an ideal tube must look like, but how ideal are they in real life? Lets see what I measured with the AT1000 curve tracer, and a nice collection of NOS and NEW made tubes.

GENERAL IMPORTANT REMARKS ABOUT ECC83:

Ever since I test NOS ECC83 and 12AX7, I find that the average of new production is not at 1.2mA as by the text book. The average is always around 1.1mA. Also for new tubes, and 1.0mA is normal too. And don't ask me why that is, but it is so with NOS and NEW tubes. The dictionary definition of "normal" is: "occurring most often". So this somewhat lower current must accepted as normal, and in a good auto bias circuit, such tubes bias completely normal, at the correct values.

The test data was taken with the Amplitrex AT1000. This digital unit has 10 bit resolution at full scale, which is 0.1% but specially at the end of the scale resolution gets kind of "digital" so it shows "steps", which can be seen at the lower end of some curves. The low scale is 10mA, so digital steps are 0.01mA. No problem at 10mA, but it gets visible below 0.5mA of course. Also the transconductance is always ending on "00". So steps are like 1500, 1600, 1700, etc, and a tube with 1600 -1700 on both systems is in reality for instance 1641 -1692. No problem, as long as we know it. I verified the AT1000 results against the higher resolution Russian L3-3 and the AT1000 is very accurate.

When you compare the graph with the test data, of course you will find the test data from the table in the curves, since it's from the same tube. The curves are made from System1 , not System2 . Look just above the number 195 at the blue line. The measured value is right above that number 195. So not at the left or the right. It's a bit unusual, but it's how the AT1000 works, the software for the graphs is a Microsoft module, called MSCHART, the Amplitrex builder told me, so that explains it.

ABOUT ECC83 TEST DATA IN GENERAL:

Look at the test data. The tube is set for -2Volt grid, and then at 250V plate the tube is tested So the plate current is exactly 1.2mA on both systems. This is no "Must", and in reality most tubes have a deviation of +/- 0,3mA and are fine tubes still. Also Tesla can have some deviation. The transconductance is exactly right 1.6m/V. Next thing is something O observed more often: Gain is a bit lower at around 80...85x. I don't know what that is, but I seldom find an ECC83 with a gain if 100x and when I do, it's a bit funny testing tube anyway. Plate impedance is a bit lower than average, this is normal too, and plate impedance goes up somewhat while a tube ages, and transconductance goes down. These two things together keep the gain constant while the tube ages, since gain is Rp x transconductance. So the data table is very nice

TUBE CURVES

Each tube curve by itself is a genetic fingerprint, meaning it holds the data for the other curves too. So we draw only tree curves, it is enough, and otherwise the plotting of all those curves would take too long.