Flying lead connector - Original. This will be removed.

Here it is removed


D-Connector mounted.

D-Connector from the inside. I used five shielded cables. I don't know yet how many I need for the Gm signals, I think three, so five as I have should be enough. The thin wires are Teflon cable and is for normal voltages and current. I pulled out the wiring of the flying lead connector all the way from the cable tree, so to make it look more tidy, there is alread enough cable in the L3-3

I received this L3-3 tester (#4) with a broken flying lead connector from Hungary. Now look at this, the hole is almost the same size as a D-Connector, so that sparked the idea to put a D-Connector in there, and connect all analog voltages to a computer interface. I will explain this here step by step, as the project continues.

This interface is intended only taking as much as analog signals from the tester as I can. Though some possibilities are there for controlling the tester too, I will not begin with that. This would beak my time limits, and also I have no need, since I already have a Sofia, and that can do 700Volts, 250mA. One thing I can do easily, when I have all signals, is make a calibrated Gm measurement in software. I just measure the oscillator signal once per second, and from there the PC can calculate the Gm, even when the Gm calibration pot is set randomly. After the interface works, I can also print the test reports, and the precision I get will be unseen before with any commercial tester, and will make the Amplitrex look like a toy. Specially I will take advantage of the most excellent hardware of the L3-3, the absolutely genius Gm circuit, the precision resistors, and the tube controlled current mirror that it has. It is just MADE for a computer calibrated measurement. This going to be more accurate as any other tube tester I know. Not just will this produce many digits, also the circuit itself (which equals an distortion meter....) is the finest I saw, and the calibration routine before measurement is really genius by the working principle, as is reduces precision only to the precision of the internal attenuator. Just it is a bit nasty to do every time. You need to realise the amazing engineering level of this Gm circuit, only the nasty calibration before each measurement, only an analog read out limits this a lot. So now, with a computer controled measurement, I can calibrate this by PC. No need any more to calibrate by hand, and finally no more analog panel meter. So the meter will be only for visual control.

This is not going to be like the Amplitrex, producing Gm for the 6SN7 in steps of 100. So 2500, 2600, 2700. And then when you have two Amplitrex ( as I have) they do not even give the same result. One can come with 2600 the other with 2500, for the same tube. Sorry, but I feel unhappy with such digital "so called" accuracy. Note, this L3-3 is going to be self-calibrating in software. The result will be an undisputable 5 digits. Well I hope so! So with this I have an absolute reference for Gm measurements.




The cable from the D-Connector is going to end behind this box. So I can remove the lid, and work on the connections. Like this any experiments and modifications won't mess up the tester itseld. I just need to get rid of what's on there.



Inside are two bulky diodes. They measure like a few silicon diodes in series. Well state of the art in 1975, but we have
better diodes today, mainly smaller size

Original Diodes of L3-3 - Just for fun.

It has nothing to do with the PC interface, but I tried to look inside the diodes blocks with the Tektronix 576, and we see several low current diodes are serialized in here. How many, that is hard to say. First, we see this from the forward voltage. The beginning with a silicon diode is around 0.5V, and appr 0.75V (or even more) when you reach the maximum allowed current. That would give: Five diodes. That makes also sense, as we get appr 3.8 Volts at 150mA. Also there is no smallest sign of leakage, at the maximum 1600V as I can get out of the Tek576. Then, we see relatively high series resistance, indicating again many diodes in series. Well, I replaced this by one single modern diode. We are only rectifying 500V AC. Though I can imagine abuse-proof level of these self made diodes is very high.


A small change to the cable tree, and soon the bulky diodes can be replaced by small ones. The Original wire can be used for the new diodes. This is the unit with the electron tubes on it, from the right side. All parts on this board were checked, and fine. One foil capacitor was "ok" but not as new. I would normally have left it in. Now I put a new Mundorf cap in there. (The white one)


These two little diodes replaced the big diodes behind the metal box.

Finally... The cable three from the D-Connector. It looks such a simple result now.

Note, the D-Connector and Cable can only do 50V, but I will put a 10:1 or 20:1 attenuator on each line. So it reduces the voltage to 15Volts DC, and makes the connections short circuit proof too. The factor will be corrected in software. I will store the calibration parameters for each attenuator, so I can use simple uncalibrated resistors.

How do I want to measure????


How do I want to measure, that is a good question. It depends on what equipment is available for a reasonable price. Let me first bring to your attention the Agilent 34401a Multi meter. This is now a previous model, but it was an industry standard for many years. And actually what was "perfect" 15 years ago, shouldn't be so bad now. It has 6 1/2 digits. The Vacuum Fluorescent (VF) display was a step back into time, this is vacuum tube technology as you may remember from very early pocket calculators. . So it has a real cathode glowing, which you can see in the dark. However it looks very nice, power consumption is very low, and long term reliability is a lot better than anything LCD or anything like that. This meter can measure micro Ampere and mega Ohms at higher sensitivity as I need, and even frequency and temperature. On Ebay you get them around 600 Euro now. I will try to do even leakage measurement with it.

Here is the back. You see once more the same connectors. That is for a permanent set up, so you don't have all the cables hanging out on the front. It is fully programmable via parallel or serial Interface. But how to measure 10 signals? I needs a relay board, or 10 meters. Do I want to buy 10 meters, or build a relay board? No.... not necessary. This already exists by Agilent, It is called 34970A Data Acquisition unit. (See next box)
The AGilent 34970A Data Acquisition unit, is using the electronics of the 34401. It has no connections at the front, but is has two card slots from behind. The most used card is probably the 20 Channel multiplexer. This is simply a 20 channel relay board, which makes it a 20 channel Multi meter, with all 34401a capabilities. You can program each channel for one of the Multi meter. settings, and store that inside the 34970A. Suppose channel 1 is for 300Volt DC, and channel 2 is for 10mA AC, you can program this, same as with a 34401A, and store it. Now comes the nice part. There is a knob on the bottom right, and you can rotate it from Channel 1...20. So when you turn it in "1" the meter displays a DC Voltage, and when you rotate it to "2" the meter displays 10mA AC current. On the 20 channel card are 20 wire pairs, of which only two pairs are used in this example. It can even measure temperature with a thermocouple. All you need to do, connect it to wire pair 3, and set channel 3 for temperature.

Here is the 34970A Data Acquisition unit from the back, no cards are inserted. A nice collection of cards exists. Even a 20 channel relay card. I am not going to do so, but with such a card, you can even do some switching to the L3-3 tester. Even switch it on and off. However, the idea is not to make the L3-3 fully programmable. I only want to measure the tube data digitally and all at once, and be able to print a quality report, which looks the way I want to myself. Yet a small option I do have in the back of my mind, that is an external grid voltage. That is possible. Some of the cards for the 34970A have an analog output too. So with that I can plot a tube curve. With the AT1000 digital tube tester, the anode impedance is measured by: Ra = (Ua2-Ua1) / (IA2-Ia1). It does require to change Ua a few Volts, and I don't think (now) I want to do this. With the L3-3 I hope to find an alternative way, as I can measure very accurately any voltage or current inside, DC or AC. In case you want to buy one via Ebay, buy it with the cards together. Often the cards come for free with the unit so to say. Second, with the 20 Channel input card you can immediately test it, as you should when you buy it on Ebay. You don't need a PC for that, you can test it in the stand alone mode.

This is an example how to use the multiplexer cards, but I intend to do that more nicely. I just will use a D-connector cable that cost 10€ on Ebay. I have one, inside cabling looks very nice. I can't do it myself of 10€. So it's going to look a LOT nicer as on this picture example here.

The 34970A Data Acquisition Unit, you can indeed hand-program it. Actually this is easy to do, and you can store the results. Sure for the beginning, this is the way to start. Later, I want to program the 34970A via computer, and read also the results with the computer. So then, on my computer screen, I can arrange all voltages and current, with virtual meters, the way I want it. I can even use a virtual scope, or do a distortion measurement. Actually the L3-3 is like MADE for this! It features this low distortion oscillator (by means of a band pass filter), and the output is connected to Grid1. Then, the anode current is put through another band pass filter inside the L3.3 which is already there. All I need to to is compare the signal before the second band pass filter with after the filter. The difference is the distortion. This is a typical example of an analog distortion meter, they all work like that. Specially with the software calibration I have in mind, this distortion meter will become very precise.


What do I want to measure????

That is not easy to answer now! I do not want to over-engineer the hardware. From designing computer controlled equipment in the 1980's, I remember one principle very well, and that is: Everything that can done in software or in hardware, you should do in software. To give a good example for this: The D Connector (and the printer cable with it) are made for 30 Volt maximum. That is not problem to measure 500 Volt with it still. Just use a 1:20 voltage divider and send the result to the 34970A Data Acquisition, and from there to the PC. Only, of course 1:20 divider resistors can not be found. Making a 1:20 divider yourself will give some error still, so you may need a calibration potmeter to get exactly 1:20. The software solution can use two uncalibrated standard resistors. Let's say 100k and 2M2. Of course the 100k is loaded with the impedance of the 34970A Data Acquisition unit also. So in the end some divider ratio is produced that you can not predict accurate. All I need to do, make one calibration measurement. So the computer will say in calibration mode: Apply a voltage of approximately 500 Volt AC. That is what I do with some transformer. I measure this, and suppose I see 511,122 Volt. I just enter this number, and the computer (actually being send appr 25,xx Volt by the 34970A) will know this is 511,122 Volt, and can calculate and save the correction factor. So I am calibrating this is software, and not in hardware. Even so, this obsoletes the nasty calibration of the Gm test. I can just leave the hardware error existing, and calibrate the software quickly before making a new test.

  1. Make as little changes to the tester as possible. This is the highest rule. So only tap signals passively.
  2. The idea is, to set the tester to "transconductance testing" (in German: "S" for Steilheit). This applies both AC and DC to Grid 1. With this setting, using the panel meter, you can only measure transconductance of course. Yet all other functions are active. So Anode Current, Anode Voltage, etc, it is all present inside the tester at some points. It can just not be displayed on the panel meter. HOWEVER, with the computer interface this can all be tested at the same time. So there is no more need to rotate the parameter switch. All parameters, including "S" can be measured simultaneously
  3. Transconductance calibration is not needed any more, as we will simply also measure also the oscillator output voltage directly, at Test point TP8. We can just design a formula in software, where the oscillator output voltage goes in as well. In other words: When the oscillator output voltage goes down for instance 10%, normally also the "S" result goes down 10%, but when we divide the result by the oscillator voltage (witch linear correction factor) the result will be auto calibrated. The linear correction factor is a calibration constant, which needs to be found only once. So when calibrating the PC interface.
  4. The band filter function is to reject anything that is not 1400 Hz. That will be 50 Hz hum, white noise, and harmonics, so 2800 Hz, 4200Hz, 5600Hz. Also 50Hz, 100Hz, 150Hz. So the 1400 Hz is chosen far enough from area of 50Hz harmonics. Basically this describes also the difference between a true, real dynamic Gm measurement, and such lower level measurements just determining Gm (here called "S") as a DC difference in Anode current, caused by a DC difference in Grid voltage. The last includes also curve distortion, which can be very much with some tubes. How significant this distortion is, you can simply see when you go up 1V in grid voltage, or go down 1V in grid voltage. If the result is not the same, this is caused by curve non-linearity, which again is the source for harmonics. We can now measure what the band filter is removing. The input is TP18 and the output is TP27. The lower the ratio of TP27/TP18, the more distortion was in the signal at TP18. Of course distortion is only valid for that one signal level the L3-3 is using. (120mV Grid signal). You need a calibration constant to find out what the ideal ratio of TP27/TP18 is. I haven't thought very deep about this yet, but I suppose you can use the calibration potmeter for that in some way. As the normal calibration routine (without computer) works around this problem too. If not, the best (ideal) ratio of TP27/TP18 must be found only once, using a high quality oscillator.
  5. show "live" on the PC screen, following values. Live means, when I change one parameter, like the grid voltage or heater voltage, I can see all other parameters change simultaneously:

    Plate Voltage + Plate Current
    G1 Voltage +, G2 Voltage
    G2 Current
    Heater Voltage + Heater Current
    In the End: Print this as a test report.

  6. Optional: WIth a small change in Anode voltage (must be done hand) , also Plate impedance can be calculated, and from this gain can be calculated. The change of plate voltage doesn't have to be precise, as the software can do the rest. It's enough just to change it a few Volts.
  7. To be continued