In this case, I felt compelled by a pile of tube-related parts I had harvested after ripping an old 500-series Tek scope apart, a wish to play with test tube circuits (test-tube circuits?) on the bench with a handy bench power supply built just for this purpose, and not much else to do. In order to construct some tube circuits on a “breadboard” style set up and be able to manipulate or dial-up the supply voltages at will, I would need to have a bench supply that would put out low as well as high voltages, not to mention grid bias voltages, filament power and would display what those voltages were. Being able to also display the current drawn from any of those would be a bonus. It might also look pretty cool.
As I said I wanted to put together circuits that were “breadboard” style, sort of like those awful-but-ubiquitous (don't get me started) white push-in proto-boards that you see everywhere anybody is messing around with electronics. But for tubes. At least as far as capability goes; I don't really mean to use those white things for circuits with 350 V floating around them. But the breadboard part is for another instructable and not covered here.
23 Apr 2014: I finally got around to a breadboard here:
By the way, the term “breadboard” comes from the 1920's and 1930's when young people were just getting into an exciting new thing called “radio”. They were building their own crystal radio receivers from scratch and when they needed a platform to mount coils and cat-whiskers and such to, the handiest way to do it many times was to “borrow” a real honest-to-goodness bread-cutting board from Mom's kitchen, sometimes when she wasn't looking. I'm just guessing, but She probably wasn't too impressed. Yeah, hackers existed then, too, and the term just stuck.
I also assumed that I would need to build it for minimal cash outlay, hence the old tube-based scope.
First, my source of parts...
Step 1: My Source of Parts
These monsters are entirely fixable by any knowledgeable hobbyist with a spare carcass for parts and will still put in a decade or two of good service as long as you don't mind having a 500-Watt heater next to your bench and giving it at least half an hours worth of warm-up time before use.
They can be had for not much more than a song when you can find one. The surplus houses will list them for $100 - $150 but best to try the swap-meets or ham-fest events first. Most of them are swept up by collectors (one Tek freak has almost 100 of them stacked up) but they come available from time to time. Depending on model, service manuals can usually be had online but otherwise are sold for about $25 - $35..
Above is a couple of shots of one of the ones I have and is not the one which I have cannibalized for this project, a Tek 545, but is a Tek 535 which is very similar.
Step 2: Design Goals
I eventually reduced the wish-list requirements to:
2 regulated high-voltage DC outputs adjustable from less than +100V to greater than +400V, metered,
1 regulated bias voltage output adjustable from 0V to -70V approx., metered,
1 regulated +6.3 VDC output capable of delivering about 3A,
1 regulated low-voltage DC output adjustable from about +1 VDC to greater than +15 VDC and capable of delivering up to 1A, metered,
Since the 545 power transformer had the windings for it, I also included:
1 123 VAC unregulated, unmetered output, and
2 6.3 VAC unregulated, unmetered, floating outputs,
I thought 2 high-voltage, individually adjustable outputs would be good to have so that the B+ and screen operating voltages for pentodes could be separately played with.
It seemed that a low-voltage DC output would be handy to power any solid-state stuff that I may have to put in.
The 6.3 VDC output was there to power filaments with DC and to eliminate that as a source of power-line hum.
Floating the two 6.3 VAC filament outputs, i.e., not being biased or tied to any other signal (such as ground) could be essential when dealing with any circuit, such as a cascode amplifier where one or more tube section is biased up by more than 100V. In that case, using a filament source which is tied to ground, would violate the heater-cathode voltage spec. of many tubes and cause an arc or other damage.
The metering need not show all of the above values simultaneously but should, at least, be switch selectable.
At this point, I still had to admit that this was a pretty (over?)ambitious project. But since I had no masters to answer to this time and it began to look like I had the parts to do it, what the hell... Besides, even if it did wind up being more than I would ever need, it would still be a cool build experience. That being said, on with the design ...
Step 3: The Circuit
A full schematic is found below. It can be roughly divided into the sections: transformer and filament circuits, high-voltage regulators, bias regulator, low-voltage regulators and metering circuits. I won't describe the whole shebang here all in one go, resistor by wingnut, since I know you want to get on with the good parts. I'll describe the operation of the sections in the order that I wired them up.
I characterized the transformer by first matching up terminals in winding pairs with an Ohmmeter and then measuring the DC resistance and open circuit voltage available from every secondary. This gives enough information for SPICE simulation without getting too hairy with details such as leakage inductance, etc. Each winding can be modeled as an AC voltage source of 1.414 x (open-circuit RMS Volts) in series with a resistor of value (DC res.)
The HV and the bias regulator sections were based on the design used in the Tek 545 power supply which claimed some fairly impressive ripple figures (10 mV P-P for the 350V and 3 mV P-P for -150V) for their outputs.
The bias regulator (-100V) looks pretty complicated. OK, I admit that the bias regulator was probably way too complicated for this project but I just wanted to build what they had done and see if I could get near the performance they did (I got close).
The LV sections are standard (for me) three terminal regulator stuff and the metering section design was dictated by the two meters I had on hand which would not draw too much current from the output as to load it down or be so physically small I was squinting to make them out.
Are you with me so far? Good... lets get on with some metal bashing!
Step 4: The Chassis
As I have done before (see my Tube Curve Tracer project, here:
I used a flat 1/8” thk aluminum 19” rack panel I had in the old “glory bin” as a “chassis”. This one already had a few holes in it but I figured that they would be covered or obliterated by new holes to come. The plate would lie flat, level with the ground and be supported on four sides by a wooden box with no bottom. The box, as before too, would be fabricated from leftover veneered particle-board shelving.
On to the plate ...
The first picture, above, shows what I started with. So after cleaning it with Javex and covering it in painters tape (which won't leave glue when you rip it off even after months), I tried a few arrangements of the parts and marked the positions of all the holes except the small mounting holes for the tube sockets. I figured that the best way to place those was to allow the sockets themselves to tell me once their large holes were cut. When drilling holes always use a centre punch and a pilot hole. And its best to use two or even three sizes of drill to work up to a large hole.
All of the big holes for the sockets, the two lights and the big caps were cut with two Greenlee punches. The next shot shows one in a vise. Rotating the metal around the vise with the punch bolt-head clamped in the vise like this was the only way to get enough torque on the punch to cut through the 1/8” thick aluminum. Believe me, I have done a couple of holes with a wrench only and its a killer. If you try this, may I remind you that the punch will leave one side clean and one side scratched up. The clean side is the side shown up in the photo so that photo shows the wrong way to do it since the side of the metal that was to be up in the finished article was the down side in the photo. You can see the scratches around some of the tube sockets.
The four counter-sunk holes which show brightly in the next photo are for a small perf-board on standoffs that I knew I would need for some of the discrete circuitry.
The large cutout for the power transformer was done by drilling two largish holes in opposite corners and then threading a fat scroll saw blade through it. The (power) scroll saw was able to cut the material but patience was needed to allow it to do so slowly. After some filing ... Voila! Then the small mounting holes for it were laid out according to measurements taken from the transformer and drilled. Some 7-pole terminal strips were screwed on where I thought they would be needed. They were held on by the tube socket screws. Their mounting tabs had to be custom-shaped on the bench grinder to make them fit.
Now all controls, binding posts, the power switch, etc are marked, in this case with press-on transfer Letraset lettering. Another painstaking task. Too bad I couldn't just feed the plate through a printer ... hmmm. Next, it's covered with 2 or 3 coats of spray-on Varathane and allowed to dry overnight between coats.
And then all parts are mechanically mounted.
About the large filter caps: There are two pair of large filter capacitors, C1, C7 and C2, C8 which are stacked circuit-wise so that their combined voltage tolerance exceeds 700V. Thus, the top one of each pair must be fully isolated from the plate. Fortunately, I got two 125 uf/350V electrolytic cans from the Tek 545 which were covered in black paper. They can be seen in the 7th photo above. But the bottom of them was still exposed metal so I fashioned an insulator from some thin cardboard (Christmas card) and electrical tape and squeezed it between the cap can and the metalwork. That cap can is raised to about 275V DC above ground and the ad-hoc insulation does just fine (so far).
Step 5: The Transformer Primary and Filament Connections
This transformer has no less than 7 (count 'em seven) 6.3VAC filament windings. I needed some of them for the low voltage stuff and, of course, to light the tubes but it left me with 2 high-current windings to dedicate to output including one rated for about 9 A (!) and another for 5 A.
The tube filaments I wired with the usual twisted pair as the audio buffs do to reduce hum pickup. This may or may not be important here but since it was free, I did it. Twisting the pair is also easily done by taking a length of about 12 feet of wire and holding the two ends together in a vice (or some such and grasping the midpoint in the chuck of a small portable drill. The job is done in about 15 seconds giving about 5 feet of twisted pair.
The 2nd shot here also shows the bridge rectifiers and filter cap cans for the HV sections wired up. The bridge rectifiers are actually four individual 1N5408 diodes. I might have used 1N4007's if I had any but these were the only ones I had that would stand the voltage. I used 1N4004's elsewhere. Again, I'm trying not to buy stuff but later on I find that I will have to break down for a few small bits (mostly 2 Watt resistors).
These big series-wired caps were biased with a 47K/2W resistor across each to make sure that the voltage across them was split more-or-less evenly. I also took a hint from Tek and fed each pair with a 10 Ohm/2W resistor (which I bought from a local hobby house). After I had them in and started to apply voltage I found them going off like popcorn. They were carbon film types! They don't conduct much current or drop much voltage (less than 2V) but if a microscopic crack develops inside them that crack suddenly has several 100's of volts across it and will arc over until the resistor is toast. This may take a minute to do but ends with fireworks! Replacing them with wire-wound types cured that one (Yeah, yeah, I should have known...). I show three of them (out of 4) which failed.
Step 6: The Low Voltage Stuff
The LV stuff was put together next since it's fairly independent, not relying on anything more than what comes from the transformer to work..
Most of it was built on the perf board that mounts on the standoffs. Since I wanted to turn the LV stuff off when I wanted, then the variable LV, the fixed 6.3V , and the two floating 6.3 VAC outputs all went through a relay, K1, that was controlled by a single switch, SW7. The coil of the relay was paired with a catch diode, D25 which “catches” the coil current when the switch is turned off as all DC relay coils should be. This diode provides a path for the coil current when the switch opens so that very large voltages will not develop across the coil as its magnetic field collapses. If this were not done then any electronic switching, such as a transistor, etc. could be destroyed. In my case, the arcing might have toasted the switch contacts after a couple of years. The relay mounts in a socket that is epoxyed to a cutout in the board.
The several LV supplies need to be connected to the rest of the unit so a couple of rows of screw terminals were applied also with epoxy glue.
The +18V, U4, supply was built and as with the fixed 6.3V supply, the ground pin of the three-terminal regulator is biased up, in this case by 13V to make it regulate at +18V out. The 7805 is on no heat sink since it supplies only a couple of milliAmps to the -100V regulator.
The fixed 6.3V regulator, U2, is an old LM209K 5 V regulator in a TO-3 case mounted on the aluminum plate. Interestingly, this part has no terminal for the GND pin other than the case which must be insulated from the plate and heatsink. So I put a solder tag on one of its screws. Having discovered this late, I find I have to drill a new hole and put in a grommet for the wire that goes to this tag. Putting a hole in now meant that metal chaff could be flying everywhere so I made a simple barrier with masking tape before drilling as I show with another hole I needed later. This regulator can put out about 1.5A and with 4-5 volts of overhead needs a heatsink capable of dissipating 7W or more.
The variable LV regulator, U3, was a similar three-terminal guy but with a low reference voltage built-in, about 1.25V. This means that with a pot the output can be anywhere between 1.25V to as high as the incoming which in this case is about 24VDC. I limit it to +20V. The pilot lamp was also wired here.
This is not the original place in the circuit for the pilot lamp. It WAS going to be a 6V bulb powered from a filament winding but when I saw that the lamp socket was firmly grounded to the frame and I had NO grounded 6.3V winding available I took a look at what bulbs I had and cobbled up a cheesy way to hold a 30V bulb in the lamp holder. Otherwise I would have had to ground one of my “floating” filament sources and that was just unacceptable.
Step 7: The -100V Bias Regulator
This circuit encompasses V13, V10 and V11 and surrounding parts. OK, here is where I went nuts with the circuitry – I admit it. But the circuit in the Tek manual looked so cool I had to try it. This circuit was drawn from the Tek 545 manual verbatim in philosophy and architecture but with a lot of the values changed since the 545 regulated at -150 and used as precision reference, a 5651 reference tube (which I had but decided not to spend on this project). I only needed to generate -100V and used a couple of zeners to make the 85V reference I needed. It SPICED up really well, showing only a couple of millivolts of ripple, However SPICE is NO guarantee of success – it just shows that the circuit MAY work and usually only if a bunch of other things (that Mother Nature will pointedly point out) are taken into account.
In my case, the circuit oscillated like the clappers at about 60KHz. Now, a fastidious engineer might mathematically analyze the circuit to establish where its dynamic loop gain and closed-loop response fails its stability criteria, then calmly devise a surgical fix and nail it on the first try. I didn't have that much time or patience. I looked for the usual suspects and started shoving small capacitors in places hoping to discover what would make it shut up. I found that putting a 47 pf cap across V10 from plate to grid did so, mostly. So I soldered in 100 pf to be sure (what the aforementioned engineer might call “putting a pole at the origin of the Nyquist plot”). The output has 10-15 mV P-P of ripple. Not quite as good as the 3 mV claimed by Tek back in 1956 but good enough for rock 'n' roll in my book.
The -100V is calibrated once only with R84. I guess I should have mentioned before this that its output is not adjusted from 0 to -100V or at all. R84 is the small, black, trim pot in the second photo hovering over the left socket of the group of three. Its a pretty mess but its nice and compact. The regulated -100V point is that point in the picture where black and orange wires come to it from the bottom and a bunch of resistors fan out from it to various spots on the sockets.
The white capacitor can seen in the upper-right corner of the first shot is the filter cap C20 which had to be isolated from ground. Wrapping it in thick paper (more Christmas cards) and putting it in a clip did it.
Variability of the bias output voltage is provided by the following circuit made up of Q6, U5 and associated pots and stuff. See the next step.
Step 8: The Bias Output Circuit
This circuit was built on a couple of terminals added after I started wiring. I tried to show it in another shot.
The price to pay for this arrangement is that the 741 (or TIP30) cannot go within a few volts of either ground or 100V. I don't mind the 100V part so much as the crucial 0 to -10V range that is necessary for all of the small-signal tubes. That's why pot, R77, is there and switch, SW1B, to choose between the R77 wiper or the 741 output. This pot is the smaller one beside the bigger one.
Some measured values:
Output ripple (up to 10KHz BW): <2.5 mV P-P on high range down to -70VDC
<1 mV P-P on low range down to -10 VDC
Wideband noise (up to 1 Mhz) adds < .5 mV P-P
Output impedance of high range: about 9.6 Ohms
Output impedance of low range: <1K Ohm
Metering this voltage means that the voltmeter would need two ranges as well, so that switch, SW1, has two sections, one to select the output voltage range, SW1B, and one to change the meter scale, SW1A. This meant another small hole to be drilled for this switch and the use of a tape wall. (See pix for LV stuff.)
Step 9: The High Voltage Sections
The power supplies of those old scopes were regulated and still have a lot to teach about efficient, effective design. So much of the philosophy for the high voltage and bias sections were drawn from there. The high voltage regulators consist of a bridge rectifier and following filter caps, the pass elements being the three 12B4s in each regulator and a feedback error amplifier, the 6AU6A. The error amplifiers' grid is connected to a sample of the output voltage through the 100K pot and the 12K resistor, R14. If the output should droop due to load (for instance) then so would the 6AU6A grid. This would cause the 6AU6As plate to rise and turn the 12B4s on more, correcting the output. And vice versa if the output rose.
Now, any feedback system like this relies on the gain of the error amplifier being very large in order to make sure that small output changes are responded to. The greater is this gain then the smaller the output changes are. This is why the 6AU6s have such a high-valued plate resistor – 2.2 MegOhms. This gives them a gain in the 100's. It also explains why a pentode is used and not a triode since triodes cannot have gain larger than mu (about 100 for the 12AX7, the best known high gain example).
The function of C5 or C6 is to make sure that there is a high-frequency path to the 6AU6A's grid from the output so that quick changes are responded to quickly. If this cap were not there then the travel time from output around the loop back to the output might be enough to send the circuit into high-power oscillations and that would not be good.
As the pot is rotated with the wiper moving upwards towards R14 then the 6AU6 grid is pulled more positive towards the output. Thus, the 6AU6As plate is pulled down to keep the grid in the vicinity of its cathode, ground. So turning the pot changes the output voltage and all the above arguments still apply.
Since both HV sections run off the same secondaries of the transformer, I can turn them on or off by interrupting it with SW3A. The same switch has another pole to light a lamp (page 3 of schematic) which warns that the HV is on.
One important factor is the heater to cathode spec of the 12B4s. The voltage between these two must be kept less than 200V total, DC and ripple. The obvious solution: use a floating filament winding on the transformer and tie it to the cathode. Then the output can travel wherever it may and it just takes the heater winding along with it. This is why there were so many 6.3VAC windings on these amazing transformers – those scopes had many tubes biased at a variety of voltages up to -1350 VDC!
The two HV sections are the two groups of four tube sockets in a row, left and right of the transformer.
The shot, here, shows the progress of the wiring, so far. Getting kinda messy, huh? I cable it up at the end.
The resistor values I used around the 6AU6s grids allowed the regulator to take the output right to the max if the pot was turned that far, about 575V. However, the output filter caps, C3 and C4 were rated only for 450VDC max (and really for long life you shouldn't let it get too near that). So I raised R27 and R28 to 27K as seen in the schematic. How does this work? The regulator is really just a tube-based OPAMP operating in inverting mode. The positive input of this “OPAMP” is the 6AU6A cathode which is grounded. The negative input (if I just talk about HV supply -1 ) is taken through R27 and R37 to its input signal, -100V. The output is the regulated Vout. Thus any noise on the -100V shows up on output magnified by the ratio Vout/100V. That's the reason for the small noise filter R37 and C24. Raising R27 then reduces the amplification range of the “OPAMP”. The output can, now, only go from 70VDC to 430VDC on each supply.
Some measured values:
Output ripple (up to 10KHz BW): <160 mV P-P up to 430VDC regardless of load
Ripple is lowest, 30 mV P-P at 95 VDC out
Output impedance is: 3.85 Ohms at 100VDC out
3.1 Ohms at 250VDC out
10.3 Ohms at 450VDC out
The first shot, above, shows how messy it's getting. The second shot is of the HV-2 section and was taken after tying up the wire bundles. C2, with the paper and electrical tape insulator is shown upper right and C8 is the one hiding under the LV perf board section.
Step 10: The Meter Faces
I decided to meter all four variable outputs for both voltage and current draw. The two meters were to go on the front of the box and to switch select the outputs metered. Since I didn't allow room on the top plate for the switches they had to go next to the meters which, I guess, was just as well. I selected a spare piece of PCB to serve as the switch plate. I cut it to 4” by 6”, drilled two rotary switch mounting holes and spray painted it with white enamel. Then some more of the Letraset was applied as you can see in the pics. I drew lines on with a Sharpie but it bled somewhat into the paint – ugh!
The two switches were mounted and short, connectored wire bunches soldered on. These connectors will mate with others on the top plate. Both switches are 2 pole, 6 or 7 throw types. Two poles are needed for the current selector since it has to switch to both sides of a series current-sensing resistor on each source. Two poles are not strictly necessary for the voltage selector but came in handy when measuring the negative bias voltage as it allowed the meter movement connections to be reversed in that case. In any case, a tougher problem came to light. How to make appropriate meter faces.
The original meter faces were completely wrong for me. One was marked in Knots and wind direction and the other was salvaged from an old Simpson 260 V-O-M. Anybody remember the Simpson 260? Another oldie and goody from the 60s. I sketched what I was looking for then disassembled and extracted the two meter faces. I scanned them into my PC and used PAINT to get rid of everything I didn't want including numbers but kept the biggest linear scale. I used this scale to make smaller copies with the same centre. On the current-measuring scales I needed a linear scale with 3 major divisions and 5 (or so) tics between each. For this I used a trial copy of Meter Basic 3.01 by Tonne Software ( http://www.tonnesoftware.com ), printed it, rescanned it and cut/paste it into my Paint image and finished it with new numbers, “BSD” marking, etc.
Now, I had to print the result onto sticky-backed paper I could peel and stick onto the metal meter scales...
All I had in the way of sticky labels was some sheets of binder labels 1 ½” x 3”. My wife solved this one. She said to print the scales onto a sheet of the stickers, cut them out to the outline of the meter faces, then cover them on the printed face with green painters tape. Next pierce them with tacks through the mounting holes. This I did. Now with the sticky labels held together with the painters tape, I peeled off the wax paper backing for the whole meter face and positioned each new face onto its plate, trying as best as I could to align them with the mounting holes using the already pierced holes in the new paper scales. Once done, I peeled off the green tape and Voila! - new scales on old plates. Then I re-mounted the plates onto the meter movements.
Meanwhile, the 260 comes with the meter housing and front switch plate and bezel all moulded into one piece of Bakelite. After completely emptying out the 260 body of all parts the switch part had to be hacksawed off and then finished on the bench grinder. (This part put LOTS of bakelite dust into the air not to mention requiring the grinder wheel to be redressed.) The HEPA air cleaner earned its keep that day.
The glass front of the Simpson 260 was gone so I cut some Plexiglas to fit, polished it good and screwed it in. Putting it back onto the meter caused the needle to wave around and be generally weird. This puzzled me until I realized that the plastic had become charged with static electricity and this charge drove the needle all over the place, so out with the air cleaner again, this time with IONs turned on. Passing this air over the meter face fixed it perfectly.
Step 11: The Meter Circuits
Basically the meter circuits are all the same for voltage and also for current. On each output, a low value current sense resistor is in series with the load current. The meter movement for current, together with a series resistor to limit current through it is wired in parallel with this sense resistor. For example the LV output (2nd sheet of the schem.) uses a 0.2 Ohm current sense resistor, R46. Two connections, one on either side of this resistor then travel to the meter through the switch. One of these connections has a 1.5K resistor, R48 and 5K/1T pot, R49 for calibration of the reading.
Each voltage meter connection is taken from the point downstream of the current sense resistor, right on the output. This means that the meter falls to zero if the selected output is turned off by means of switch or relay or whatever. For each voltage to be metered, a resistor divider, such as R47 and R50 for the LV source reduces the voltage to the vicinity of 1 Volt. Then, again, a series resistor and pot, R51 and R52 limit the current through the meter movement and allow calibration.
Only for the sake of convenience did I put the two HV voltage cal. pots and the two HV current cal. pots at the selector switch. The circuit arrangement remains the same. The Bias cal pots (all three of them – one for current and TWO for voltage) can be found at the Bias range select switch on the plate. The LV cal pots can be found on the perf board.
You know this stuff can be calculated. It really can. But when you wire it up and try it out, it never fails but that the meters just won't calibrate correctly to an independent DMM . Always some resistor value has to be changed to make the pots swing the meters into correct reading. Just sayin'. And this was the story after measuring the FSD current and voltage of both meters and being very careful with calculations. Oh, well.
Step 12: The Box
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 11 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 Sharpie
The last shot, above, shows the box being tried on for size with the plate..
Step 13: Mounting the Meters
Mounting the meters into the front wood piece of the box used no finesse whatsoever. The meters housings that projected out the back of the plastic were measured and the measurements transferred to the wood panel with what I hoped would be enough clearance to let them sit well without showing any of the holes. Then after sketching the outline of the cutouts and drilling a 1/2” access hole into each, I jig-sawed it into submission. Similarly with the switch panel.
The switch panel already had mounting holes so they were easily transferred to the panel and drilled. The mounting studs of the meters were a bit more complicated to locate on the panel but by using a sufficiently large drill bit and holding my mouth just right it was done. I even got them right way up.
Step 14: Cable Lacing
This took about 3 hours to do.
Step 15: At Last...
I have a Lab Tube Power Supply. It may be some time before I have the breadboard part sorted out since there are other things which may get in the way, but once I do I promise to report on it.