This is for all those tube amp enthusiasts and hackers out there. I wanted to build a tube stereo amp that I could be proud of. However in the course of wiring it up I found that some 6AU6s just refused to bias where they should.
I have a 1966 copy of the RCA Receiving Tube Manual and having designed electronics of all sorts for about 30 years, I understand that the published data on a device need be taken with a wee grain of salt sometimes. But the tube data published in these books is definitely NO guarantee of behaviour in a real circuit for any one specimen.
I like the little plate curve family charts, as in the picture above, in the book and THAT is what I wanted to see for the tubes I had. Using a tube tester, even a well-calibrated, high quality one will only give you one data point on one plate curve amongst that family. And you don't even know which curve it is. It's not very illuminating. Buying a curve tracer on the market can be expensive and rare (You may find an old TEK 570 on EBAY once a year for $3000 or up) and finding one locally is out.
So I decided to build one.
P.S. I have completed some enhancements to this TCT here:
Step 1: The Circuit Design
I needed a circuit that would be relatively simple but would provide a high plate and screen grid voltages as well as a stepping control grid voltage with steps of ½ V, 1V each, etc. For the plate drive I used a half sine wave straight off a high voltage transformer winding since I realized that the plate current would follow the same characteristic path going up the wave as coming down. The wave form need not be precise, calibrated or any particular shape as long as it rose and fell in a non-abrupt fashion. It did not even have to be consistently the same shape each time it rose or fell. The shape of the resulting curve is determined solely by the characteristics of the tube under test. This eliminated any need for a precision high-voltage ramp generator but I still needed to acquire the transformer for this...
I wanted to have several tube sockets for the various existing base types but eventually settled on four: 7 and 9 pin miniature plus octal sockets. I also included a 4 pin socket to allow testing old rectifier tubes.
The stepped bias generator is a cheesy 4-bit R-2R ladder type digital-to-analog converter driven by a counter advanced by the 60 Hz wave from another winding on the transformer.
The filament voltage came from a transformer ripped out of an old ReadRite tube checker from the 1940's which provided many filament voltages from 1.1 V to 110 V AND a switch to select them.
Finding a switching method to accommodate all of the various and sundry tube base pin-outs proved to be futile at best so I avoided the whole issue and used patch cords with each numbered pin and each drive signal brought out to 5-way banana connectors. This gave me ultimate connection flexibility and prevented me from going mental trying to figure out a good switching method.
Finally, the biggest concern was measuring the plate current. I didn't measure the cathode current since it is the sum of ALL element currents including the screen grid. The place where the plate current is measured (at the plate) was elevated to about 400V at the top of the wave. So after dividing the plate voltage down to 0-6V with a resistor divider so OP-AMP ICs could work with it, a large gain, very-well-balanced differential amplifier was needed. The LMC6082 dual precision OP-AMP did this very well and to boot its signal range includes ground so it could be wired up as single-supply.
Both plate current and plate voltage readings were then output on BNC connectors to an oscilloscope operating in A-B mode so the final chart of these two quantities could be plotted against each other.
Some people have written asking for a clear copy of the schematic since the one that shows up was pretty fuzzy. I have removed it and replaced it with a PDF version. The green line encloses all of the circuit on the small hand-wired circuit board. A couple of parts of the circuit are expanded upon in step 7.
There were a couple of surprises in the build and I will talk about those later.
Step 2: Making the Front Panel
I decided I would build it on a 19” x 7” x 1/8” thk aluminum rack panel I happened to have laying around. It would be later supported by a wooden box made from scrap shelving.
The first photo above shows some of the major parts placed on the panel to determine a good arrangement. The large open space represents where a hand-wired PCB would be put on standoffs. Several arrangements were tried. After covering the whole panel in painters tape and marking drill points, (all I had were a couple of Greenlee chassis punches and a small drill press to make holes with) I drilled all holes. Note: always start with a small (1/16”) pilot hole, even in aluminum and work up to the larger size in steps. I used three sizes of drill bit to do the 1/2” holes for the banana connectors. The use of a center punch is a good idea too.
In the picture a spool of wire stands in for the filament voltage switch since it was not yet separated from its transformer.
Holes were drilled for two transformers at this point.
The toughest hole to make was the 9-pin socket hole since I didn't have a punch of that diameter but had to use the one for the 7-pin socket hole then file it out to the larger size. That was a job.
The only rectangular hole was for the power switch. It was filed out from a round hole as well.
Step 3: Assembling the Panel
The first thing to do before any parts were on it was to label as many of the items on the panel as I could before mounting any parts. This was done with some old transfer LetraSet lettering left over from school days. As far as I know this can only be purchased in England nowadays. I then covered it in three coats of transparent spray Varathane coating. I don't know how durable this will be over time but so far so good... The steps on the filament switch were later done by hand since I had no lettering of an appropriate size.
The light beige coloured fuse holder is in the upper right near the power entry hole where the cord goes. Below that are the neon pilot lamp and the ON-OFF switch. You may or may not notice that the switch looks to be in the up position but in fact says OFF. This switch is an English DPST power switch. All power switches there are UP=OFF/DOWN=ON not like here in North America where it is the other way around. The logic used when setting electrical code for ON/OFF switches here is that when one accidentally falls against a switch it is more likely to apply downward force than upward force and so was deemed safer if whatever is controlled by that switch be turned OFF not ON. I have no idea why England is vice-versa but I liked the switch anyway. When thrown it gives a very solid “Thunk”.
The G2 V switch is to select the voltage supplied to the screen grid. This would later become a pot. The G1 Step switch selects the size of the grid step (currently) either ½ V steps from 0 to -7.5V or 1V steps from 0 to -15V. The two BNC connectors labelled H and V are vertical and horizontal signals to the scope. The G BNC connector is the grid drive waveform so that it can be seen if desired. The drive voltages are the red 5-way Banana connectors and the black ones are, of course, wired to the socket pins. All of the correspondingly numbered socket pins are in parallel.
The PUSH TO TEST button closes the connection to the plate of the tube under test so that it will be drawing current only when asked to do so. No point turning your back just to find out only by smell that something ain't right! (Wouldn't be the first time for me.)
Step 4: Assembling the Circuit Board
The board is a chunk of perforated fibreglass about 2" x 5". I made a guess as to board size and just started sticking parts on it. My method is to build a bit – test it – build a bit more – test it, etc. This prevents one bad part/circuit from destroying many more with it all in a flash.
The screw terminal strips are held in place with 2-part epoxy glue since there is no copper circuit on the bottom to solder it on with as is the usual case.
The circuit was hand-wired using PTP technology. That's “point-to-point” technology. Crude but any acronym makes it sound high tech, right?
Just to the left of the small heat sink can be seen two identical 1megohm resistors. These are what I first used for the plate current voltage dropping resistors R3 and R4. As will be seen in step 7 these had to be replaced.
The circuit is not pretty on the bottom but then I wasn't going for neatness in this step.
Step 5: O Yeah... the Patch Wires
I chopped some unusable meter test leads into approx 7” lengths and soldered banana plugs onto both ends. Those leads are made with some great flexible wire you would have to go a long way to buy. The plugs: one red and one black as you can see. The red one is for the drive end and the black one is for the socket pin connector end not that it matters but it seemed better that they match the the colours of the connectors I had. I'm so fashion conscious.
Knowing that I would have to be able to confirm the plate current measurement calibration with a completely different method I made a patch for the cathode with a difference. I show it with a small box with a switch. Inside the box is a 10 Ohm resistor which can be switched into the circuit or out of it. The cathode “drive” is actually just connection to ground (0V). When the resistor is switched “in” a scope can be put on the cathode end of the patch and the actual cathode current of a triode can be measured to confirm what its plate is drawing, This assumes that the grid is always at a negative voltage. Normally the resistor is switched “out”.
When the switch is flipped back and forth during a test the difference in plate current can be seen with the whole family of curves shifting up and down a bit. The effect is so small (maybe 2-4%) that it makes no real difference to whatever the motive in measuring the tube is but does illustrate that even a 10 Ohm resistor in the cathode can make a visible change.
Step 6: Marrying the Circuit Board With the Rest of It
The board uses screw terminals to connect wires so that I could remove the board for further construction/changes after testing parts of it. I put it on hinged standoffs on one end and straight ones on the other end so that I could lift it for access to the other side for quick measurements or changes without needing to disconnect a million wires.
For the most part, heat was not a concern but I put the low voltage positive regulator on a small heat sink for the sake of safety. Those 3-terminal regulators such as the 7805 that I used can dissipate about 1 Watt with no heatsink but it is always good to keep things cool when there is any chance to do so cheaply. Its ground terminal is biased up to +10V with a 2N3906 transistor and a couple of resistors. This gives the +15V that the differential amplifier runs on. This is a good way of getting any voltage you like from one of those common regulators. Variability or programmability can be had the same way by using a pot or D/A converter in place of one of the resistors. Since a variety of AC voltages are available from the Xfrmr it was easy to choose a voltage for this regulator. 25V was it. And since it draws so little current half wave rectification did fine to supply the regulator.
As you can tell from the picture, I began lacing up the wiring instead of bundling them all with plastic ties. I have always admired the look of a well-laced harness and wanted to try it here but there was no lacing cord to be found anywhere. Maybe some of you know where it can be had. I used some embroidery thread suggested by my wife pulled over a lump of wax. I used the standard lacing knots for my harness. For those willing to learn this arcane art, Googling “harness lacing” brings up a couple of how-to sites.
The old ReadRite tube checker had an interesting method of calibration. By putting the ends of a ceramic pot across part of the primary winding and connecting the wiper to the line voltage source, the voltage that the tester operated at could be adjusted above or below nominal to take care of local variations in wall voltage that may happen from time to time. (Remember this stuff was designed and used during WWII era times.) Well, this pot just had to be included here since the transformer was designed so that neither end of that part winding was at nominal line voltage and so could not be used as-is. That pot, which gets fairly hot, can be seen as the white object held by the perforated plumbers metal strapping near the transformer.
By the time I got to discovering what all of the anonymous leads on the old ReadRite filament transformer were, I discovered, of course, that it had a high-voltage winding! So my plate voltage source was solved and I eliminated one transformer.
Step 7: A Little More About the Circuit
The Bias Generator: In order to keep things relatively simple and low current, 4000-series CMOS logic was used. This stuff which was ubiquitous in the 1980's will work on any voltage from 3V to 18V. This means that the power can be anywhere in that range, it can change if needed and in fact will work even if there is large amounts of ripple or other noise on it. It's great for battery-powered applications. It can still be had today at any of the usual outlets (Mouser, Digi-Key,, etc.) even if they are not making all of the types they used to. It also draws next to squat power. So I used a 4040 12-bit counter I had lying around as the 4 bit counter for the stepping of the bias voltage. The step size is changed by changing the power rail voltage for it. Since the tube bias voltage must be negative the counter is operated between ground as its positive rail and a negative rail for the other end. The “VDD” pin is thus grounded. A TIP 107 with a bias network similar to the 7805 supplies the minus supply volts to the chips “VSS” pin. A panel-mounted switch with pots for each range calibrates the maximum bias generated. The counter drives a cheap R-2R resistor ladder to make a simple Dig-Analog converter and then out to the banana connector it goes.
The Plate Current Amplifier: Since the plate current is sensed with a 100 Ohm resistor, R1 in series with the plate, its voltage is elevated to about 400V. It was made smaller with two resistor dividers, one for each end of the 100 Ohm resistor. It is shown as R3, R4, R5. R6 on the schematic and the small-value pot and placed near the Push To Test button on the schematic. The pot balances these two dividers so that the output of the amplifier reads zero when zero current flows in the plate of the tube. I first used some old large value resistors for the R3, R4 but when I tried it out the curves I got looked more like word balloons than single lines. I include a pic of what I saw. You can also see that the display is a little crushed into the baseline. I changed these resistors to more modern 5% resistors and re-calibrated. Same thing but a little less. Each curve on the display takes 1/120 second to trace with the scope spot first going up the curve then coming back down the same way. But between those two excursions the resistor would heat then cool enough to change their value! Resistors will change value depending on temperature, not much but will do so. I didn't think it could happen that fast but changing them again to 1% metal-film types largely solved the problem.
The amplifier is a conventional differential amplifier as used for instrumentation but with a gain-changing toggle switch to give it two ranges of output and two pots for range calibration. This gives 2V/1mA and 2V/10mA output scales.
The screen grid drive circuit is simply a filtered pot hung off the rectified plate voltage source with a high voltage transistor as emitter follower to drive voltage into the banana connector. The filter is fairly slow and takes a couple of seconds to settle when the pots knob is moved.
Step 8: Operation
I turned it on.
After the smoke cleared... the circuit worked surprisingly well. I found that the balance of the differential amplifier needed about 20 minutes warm-up time to settle down fairly well. After that time the 25 Ohm balance pot needed to be adjusted to give a very horizontal line on the scope when no plate current flows. After a while of adjusting this on the board every time I used the unit it was removed to the panel and appears as the medium-sized brown knob near the red banana connectors. I don't know why I didn't do that sooner.
Shown are a couple of screen shots of curves obtained.
Since each curve on the display is generated in 1/60 of a second and there are 16 to a scan before it repeats, then scans come at about 4 scans per second. This flashing works but is not really fun when trying to make a measurement. One solution is to capture each plot with a long time exposure on the camera. Or... use a storage scope. What you see is an oldy but a goody – a HP 1741A analog storage scope with variable persistence. The display will bloom after a while but for about 30 seconds presents a very watchable chart. It will store a screen, undisplayed, for hours. It does OK.
Shots of curves for a 6AU6A pentode as well as a 6DJ8 triode are presented.
The 6DJ8 has scale factors of 50V / division horizontally and 10 mA / division vertically while the 6AU6A has scale factore of 50V / division horizontally and 2.5 mA / division vertically. These scale factors are a combination of the output range of the curve tracer and the vertical sensitivity dialled up on the scope. Zero in all cases is the lower left corner of the screen.
These were taken simply by holding the camera near the scope screen. After putting up with this for a while I decided to take drastic action and cobbled up a REALLY cheesy method of holding the camera attached to the scope....more plumbers strapping. The camera mounts into it with a short 1/4” bolt through the bottom into its mounting hole. Aiming the camera amounted to twisting the strapping just right. Obviously, I cant show the camera in this mount since it was needed to take the shot!
Step 9: The Box and Final Article
The box, like all other parts of this project was put together out of scrap material on hand. It is a simple four-sided box with no bottom but screw-on rubber feet. The pieces were jig-saw cut out of a spare particle-board bookshelf which had 3 sides covered with the same veneer as the top and bottom sides.
The cuts were made keeping in mind that the edges with veneer should show on the front of the box. Unveneered edge was inescapably shown on the back and bottom. The pieces are held together with particle board screws left over from some Ikea kitchen cabinets from 10 years ago. The screw heads are covered with white plastic push-on screw head covers from the same source and then coloured black with a permanent marker.
The box took about 2 and ½ hours to make.
Step 10: Finally
The unit has answered my questions about the biasing of 6AU6As and allowed me to adjust my amplifier design to take old tubes into account. Simply put they conduct more poorly as they age.
Obviously the unit could be enhanced with more bells and whistles. It would be good to have a digital panel voltage meter which indicates the screen grid voltage dialled up with that knob amongst others. Also more and higher control grid bias ranges or step sizes. And while we are at it how about capturing the plot to internal memory so that it can be uploaded to a PC. Perhaps the curve tracer could be Windows based and come with a mouse. Then tests could be done from any place with internet connection. Or maybe not.
P.S. I have completed a couple of enhancements to this TCT here: