Introduction: Tube Curve Tracer Ver 1.1
A curve tracer is a machine which will plot the Voltage-Current characteristic curves for a device such as a transistor or diode or something else. In this case we are talking about the plate V-I characteristics of tubes. By this, I mean a curve showing what current will flow into the plate of a tube at any moment depending on what voltage is present between the plate and the cathode. This current is also dependent on the voltages present on each of the grids in the tube. A photo of a family of plate curves is shown above for the 6AU6 pentode. There are several curves shown in this photo. Each one is for a different voltage on the 1st (control) grid. About 16 months ago I built a curve tracer for small-signal tubes to facilitate my understanding of the behaviour of some 6AU6s I was using in another project – a stereo amplifier. It was a fun build and I wrote an instructable for it.
Here's a link to it.
It also had its limitations. In this instructable I show what I did to expand its capabilities to allow power tubes to be plotted and fix a couple of small problems.
Step 1: About the Previous TCT
The existing tube curve tracer is shown in the first picture above. It could generate the plate voltage-current characteristic curve for a family of 16 grid voltage levels. The grid bias generator could generate either ½ Volt or 1 Volt steps only. Thus the maximum negative grid bias available was -15 volts as the grid was stepped from 0V to -15V. This allows only small signal tubes to be traced but not anything that needed greater bias. Also the vertical output, measuring the plate current, was at most 10 mA per 2Volts out. The vertical amp would run into the power rails at +15V so that limited the maximum plate current displayable to 75 mA. I needed to display much larger currents for power tubes.
Two main sections had to change. The stepping bias generator and the plate current amplifier. I include a PDF of the new circuit in its entirety below.
I also decided to change the pot used to balance the vertical output to the scope, R5. The pot I had was very noisy and was becoming a pain in the tush. After changing it out for the next lowest value pot I could scrounge up (50 Ohms across a 10 Ohm resistor) I show just what effect it has on output by connecting the unit up for regular operation. The 2nd photo shows a horizontal line with no plate current being drawn. I adjust the new balance pot from end to end with the storage scope painting all points on the screen. It shows that with no plate current a non-zero vertical signal can be generated from the plate voltage waveform which is a half sine wave. This would tilt the whole family of curves up or down. The vertical gain of the scope is magnified 10x to show this.
First I talk about another small problem which I had - the screen grid Supply.
Step 2: The Screen Grid Supply
The pot which dialled up the screen grid voltage was doing some funny things. I took a while to find a loud SNAPPING noise that would occur every few minutes but otherwise left the unit unharmed. It was the pot. Apparently it's not a good thing to put up to 400V across one. Sometimes the contact with the wiper was not good and when it lifted, nothing would happen but would re-connect with a loud SNAP. So the screen grid circuit was changed to allow much lower voltages on the pot.
In the old circuit (picture) the high voltage from the transformer was rectified by D4 and stored on C1. This created about 400V on C1. Thus the pot did indeed have about 300V across it. In the current circuit, the 400 volts is reduced to about 25V dc by the voltage divider made up of R10 which is two 390K resistors in series and the pot itself. Cap C2 removes high-frequency noise. The wiper of the pot drives the base of Q2 which is a high voltage NPN transistor. The lower the base of Q2 is, the higher its collector rises. The collector load, R25, needs to be a very high value resistor so that it will not draw very much power and yet not be so high-valued that it cannot drive the base of the following emitter-follower, Q5. Its value needed to be in the 50K to 75K range. I didn't have any 75K 2 Watt resistors so I used eight 1/4 Watt resistors in series. They can be seen in the photo below.Cap C1 also removes all of the ripple. In fact it time constant is so long I need to wait a few seconds for it to settle once the pot is moved.
The whole circuit was built onto the pot itself. R25 is the folded-up string of resistors just to the left of the pot.
Resistor R27 was put in should the wiper of the pot lift off the pot element. I didn't want base of Q2 to go open, thus turning Q2 off and letting its collector and the output go right up to 400V. This way if the wiper goes open (as most pots do with age) then the output will go to minimum voltage, not maximum.
Step 3: The Plate Current Amplifier
I'll take the plate current amplifier next because it was the simpler change of the two main redesigns. In order to show a larger current displayed on the scope the plate current output amplifier's gain had to change. As it was it would saturate into the rail at about 75 mA max on the 10 mA/2V output range. Preventing saturation at even higher plate currents meant lowering the amplifiers gain. I chose to reduce it by a factor of 5. Since the configuration of the amplifier, made up of U5 A and B and U6A is a standard instrumentation amplifier where two op-amps, U5A and U5B, are connected as a differential pair with common gain-setting resistors, the gain is determined by the resistance between U5 pin 2 and U5 pin 6. This resistance is made up of a pot and a fixed resistor. Two such combinations were in the original design and were chosen by flipping a toggle switch mounted on the panel. I replaced the toggle switch with a 4 position rotary switch and adding another pot/resistor combination for the new range.
The gain of this differential pair is set according to the following relation: If the pot/resistor combo is called Rx then the gain is G = (R29 + R19 + Rx)/Rx. The Rx value for the 50 mA/2V range can be found by re-arranging this equation for Rx and plugging in the known values for the resistors and the desired gain. It becomes Rx = (R29 + R19) / (G-1). The pot always needs some adjustment to bring the unit into calibration. The photo showing the chicken-head knob over the word “TEST” switches this range. The fourth position is unused but is just waiting for some circumstance to come along that will make using it necessary.
Before calibrating the unit it is only necessary to make sure that the gain adjust is wired up correctly. I put in a 12AU7 tube with jumpers and with suitable bias (0 to -15V) for that tube tested it to see if the pot will alter the vertical dimension of the display. It did. At this point I crank the pot from end to end to confirm that the range includes what should turn out to be the correct value (1/5th as tall as that of the 10mA/2V range). I will wait until there is a power triode in the socket before describing any final cal.
Step 4: The Bias Generator
The bias generator was by far the most complicated part to change. That circuit grew more than 3 X the original size. If you will bear with me I will also explain how every part of the new generator circuit works, with scope screen shots. If you can see the philosophy of operation by just staring at the schematic then more power to you and you can skip the next couple of sections.
I used the same philosophy of stairstep creation here as before: that of a 4-bit counter whose output is converted to analog with a R-2R ladder circuit..
The bias generator must accomplish several things:
1 - It must advance the state of the counter each power line cycle. It must do this when the transformer output swings negative when no waveform is being shown on the scope. Remember that the scope trace is shown only when the plate is driven positive.
2 – It must amplify the 4-bit counter output levels to that of the maximum bias level, up to 60V negative and then convert those 4 bits to an analog voltage.
3 - It must accurately generate that maximum bias voltage level so that it can be applied to the D/A conversion.
4 - It must finally drive the resulting waveform, up to 60V P-P, out to the tube grid or scope connector without distortion or loading.
The first item is a Schmitt trigger implemented by U6B. This was the one op-amp left over from the plate current amp. And therefore runs from the same 0V and +15V rails as it does. The input which comes from the 70V tap on the transformer is attenuated by about 6:1 by R73 and R74 to about 16V peak and still retains its sine wave shape. Protection diodes D19 and D20 prevent the wave from going outside the power rail limits of U6B. The op-amps +ve input is biased to half the +15V value. Thus when the input goes above +7 ½ V the output will go low to close to 0V. When it goes below +7 ½ V it will go high close to the +15V rail. Actually since the feedback resistor R77 is there, the op-amp needs to go some distance beyond the +7 ½ V(either way) before the output will switch. In this case it must travel an extra 0.75 V each time. This adds 1.5V of hysteresis. The output of the Schmitt trigger feeds the clock input of the counter.
The bias voltage supply has the responsibility of generating a calibrated voltage from less than -7V to over -60V which is to serve as the maximum bias value.
First the 70V tap from the transformer is rectified by D18 and stored on caps C7 and C13 to produce -100V with some ripple, about 1 V P-P. The two resistors R43 and R83 make sure that that this 100V is split more or less evenly between the two caps. Without this, the two caps may split the voltage unevenly and allow one to be operated over-voltage reducing its life considerably. Transistor Q12, voltage standard U9, pots R78-R81 and resistor R39 form a voltage regulator to create more ripple-free -75V at the bottom end of R48. If the bottom end of R48 (where it says “-75V Reg.”) should “rise” a bit (go toward larger -ve voltage) then the base of Q12 would be pulled “up”, too. Since its emitter is held to a very constant -2.5V by U9 then Q12 would be turned on (more than it already is) and attempt to pull its collector down towards ground. Of course, when I say “down” I mean that the collector would go positive towards ground. This pulls more current through R48, reducing the original tendency. This action reduces the ripple at the bottom end of R48 to about 0.22 V P-P. Pot R82 is adjusted to set the operating voltage at the collector of Q12 to about half its maximum safe value, about -20V. 36V Zener diode D22 is there to prevent the collector from exceeding that safe range (40 V) for any reason. Should the -75V point fall towards lesser negative voltage then the reverse would happen: the base of Q12 would “fall” a tiny bit; Q12s collector would rise towards more -ve voltage and the -75 point would be stabilized.
Since a stable -75 V is now created, then any point along one of the pot strings, with R78 thru R81, is also stable and can be selected to provide a stable maximum voltage for the stairstep generator. Four voltages are created at the wipers of the pots: -7.5V, -15V, -30V, -60V and then selected with the switch. Cap C10 removes any residual noise. But this voltage needs to be buffered to produce a low impedance source. U8 does this. Since 741 op-amps cannot withstand more than 44V max across its power pins, then it needs to be helped. Transistors Q13 and Q14 are driven by the output of the op-amp to shift the + and - voltage on the 741 power pins so that those voltages are near to the buffered pot wiper voltage, be it -7.5V or -60V. The 741 is actually run on about 36 V between + and – power pins. It can be shown mathematically that regardless of what voltage is input to the 741, this 36V difference never changes but just shifts up or down as needed within the total 100V available. The transistors Q13 and Q14 also can stand only 40V max across them. This places another limit on the maximum signal amplitude that can be handled. This fact comes into play in the output buffer, U7A.
Step 5: The Bias Generator - More
The 60 Hz positive edges from the Schmitt trigger advance the count of the 4-bit counter. I generate a 16-member family of curves with this 4 bits. But each of them switches from ground to only +15V. They need to switch from ground down to the max bias level (-7.5V or -60V or whatever). The transistors Q8 thru Q11 do that. When the emitter of each is pulled high by an output bit it is turned 'on' because current is drawn from the bit through a 15K resistor and then through the emitter-base junction of the transistor to ground. When the transistor is turned 'on' it pulls current through it and the 120K collector resistor on its collector to the regulated, buffered max bias voltage at the output of U8. When each of the four bit outputs of the counter go low to ground then each of Q8 thru Q11 are turned 'off' and those collector voltages fall to the negative bias voltage.
When each of Q8 thru Q11 are 'on' the collectors may be pulled significantly above ground potential. The diodes D23 thru D26 prevent this. They limit that travel to about 0.6V above ground.
The next section is the D/A itself. This is a very standard circuit that has been around since digital logic was invented before I was born. Either end of the 510K resistor string could have been chosen as the output but only the end used as shown produces a regular stairstep wave form. So that end is fed on to the output buffer and the other end has another 510K resistor tied to ground through R68. This resistor fills in for all bits lesser than lsb which are absent. The D/A is made up of 2 values of resistor which should be exactly 2 to 1 value ratio. As you can see mine are not but are close enough for me. IMHO making them any more exactly in 2:1 ratio has no return on investment and is so pointless.
The output impedance of any R/2R D/A string, regardless of how many bits are involved is always equal to R.
Since the four amplified bits were clamped at 0.6V above ground then the output waveform travels to +0.6V, too. I shift the whole waveform down by this amount with a tiny amount of negative current injected through resistor string R69 thru R72 from the -75V point. The value of this string is adjusted by trial and error to just eliminate the +0.6V. I get it down to about +0.05V
The output buffer is similar to the max bias voltage buffer described in the last step with a small difference. Since it must output 0V to -60V, it needs some elbow room above and below that and so runs from +15V to -100 total power. And since this exceeds the amount that the opamp and transistors in the previous circuit could stand before smoking, I doubled up on the totem-pole transistors to spread the voltage amongst them and limit the voltage applied to each. Believe me, I did not realized I had to do this until mother nature made it clear when I turned it on. The opamp chosen here is a FET-input version of the 741 so that it will not load down the waveform from the D/A by drawing any current into its inputs through the D/A resistors. Even 2 uA input current would shift everything by 1 Volt.
The 820 Ohm resistor, R84, saves the op-amp if the grid drive connector gets accidentally shorted to ground. Actually it serves another very subtle purpose, too. With the transistors in the path of the power to the chip, the op-amp becomes less stable than when used in the conventional way. This means that any reactive (especially inductive) load on the output of the op-amp can send it into oscillations. This is the kind of thing that can drive you crazy if you have no idea what's causing them. Having the output wired through several inches of wire to its destination can be such an inductive load. So to isolate this inductance from the op-amp then a small value resistor of several hundred Ohms is inserted into the output wire near the op-amp. Putting the resistor at the other end of the wire would be useless. Thus, the op-amp is protected from both accidental shorting and from load-related oscillations.
Step 6: Construction
Being as lazy as ever,I constructed the new circuit on the same 3” by 5” proto board I had already used. I cleaned off the previous stairstep circuit and started putting parts for the new circuit on. You can see that I ran out of room after completing most of it.
The final output buffer had to be put on its own piece of proto board. And, as usual in my slovenly style, I just epoxied it onto the end of the existing board. On testing it, I accidentally blew it up with the scope probe and so it had to be ripped off and re-built. You gotta watch that scope probes don't short together points that the machine just doesn't appreciate being shorted together.
Step 7: Calibration
The first thing to calibrate is the voltage on the collector of Q12. By monitoring that point and adjusting R82, -20V approx. is obtained. Precision is not needed here. As long as the collector has enough elbow room to work in without saturating the transistor or making the zener diode come into play then its good. It was lucky that I used a 10-turn trim pot as the its value is quite critical. When a scope is put on the collector of Q12 then a large, inverted version of the ripple at the -100V point is seen.
Provided the circuit works to generate the stairstep, the output at the grid drive connector was put on the scope. Now I adjusted the resistors in the light pulldown string, R69 thru R72 so that the topmost step was as close to 0V as I could manage. No pot here – just substitute value after value until it's as right as can be made.
Next, are the four max bias pots, R78 thru R81. They are easily set by scoping the output stairstep waveform at the grid drive connector and cranking each one until the bottom (most -ve ) step is correct. I show all four bias output waveforms in one shot on the scope by manipulating the beam brightness and the selector sw. while the storage scope paints the screen.
Once the bias generator is operational then the plate current gain can be calibrated. With a 6080 triode in the test socket, I used the special jumper with the switch on it. The switch inserts a 10 Ohm resistor into the cathode of the tube. I put a scope probe on the cathode end, not the “drive” (ground) end of the cathode jumper. Then pressing the “Push To Test” button to get current through the tube, I could see the current drawn by the tube as a series of “hills”. The largest of these shows 2.25 V at its most positive peak making the cathode current 2.25/10 = 225 mA. With this figure recorded, I re-connect the setup for regular operation. Now, pressing the button shows me the expected family of plate curves but with wrong amplitude. Finding the spot on these curves which corresponds to the earlier measurement allows me to adjust R11 to the same value. Doing this I must not change the switch on the cathode jumper wire or that would change the conditions of the test.
Another note: The transformer is not quite beefy enough to supply the filament current to this tube. With the fil. switch at 6.3V I measured only 5.23V. But switching it to 7.5V gets 6.4V at the filaments. Thus that is what I used for testing. Normally starving the filament would make all the plate curves shrink a bit but this tube is specifically design to resist that. At least that's the claim I read about it somewhere.
I also took the opportunity to touch up the calibration of the other plate current ranges, R15 for 2V/1mA and R16 for 2V/10mA out.
Note that although the photo shows only 8 “hills” actually all 16 are being generated but the scope sweep was triggered only every other time.
Step 8: Trying It Out
The tube used for the calibration was a 6080 twin power triode. This one is designed as a regulator in power supplies and that's just what it was used for when I ripped it out of the old Tek 545 oscilloscope. Thus, it could take a lot more current than other triodes.
Note in the shot of the plate curves captured that the curves all don't reach the same maximum voltage (horizontal extent). This is not a design feature. It just shows that the power transformer cannot supply that much voltage at higher currents. Its internal impedance comes into play. The lower the current drawn, the lesser this effect is but it can be seen to some extent in the scope shots of the first instructable. It also reminds me that it was a good idea to include that “Push To Test” button so as not to continually stress the plate supply or the tube itself.
I will point out as I did in the first instructable that the exact waveform of the plate drive is irrelevant as long as it has no abrupt jumps. Since the characteristic shape of the relation between the tubes plate-cathode voltage and its plate current is determined solely by the device under test (the tube) then, should the exact waveform applied to the plate have wiggles and swoops these non-linearities would only change the speed or time that the scopes' beam travelled up and down the curve but not cause it to deviate from the aforementioned characteristic path. Thus, for anybody who thinks they need a precision high voltage ramp generator or such for the plate/collector/anode/whatever then you can just as well save yourself the effort and aggravation of going that far. It makes no difference.
I found while looking at the grid drive stairstep waveform on a scope that putting a tube in the socket or pulling it out made absolutely no difference to it as should be the case.
I still find it a fun device to build and play with and still no software! Who knows how long my luck may hold out?