Fun with Vacuum Tubes


Recently my friend Steve cleaned out his basement. Besides some solid state gear, he gave me two pieces of surplus tube equipment he had collected over the years, from the MIT flea market and from Ebay.  Steve collected them based on their interesting front panel controls. One was a large Rutherford Pulse Generator from the mid 60's, the other a US Navy Radar test set, probably from the late 50's, At first I questioned the sanity of more basement clutter. But then I realized that by stripping them down I might get a good baseline set of parts to do some playing with tubes. I have an old Philbrick tube op-amp that I have never fired up. It needs a +/- 300V power supply and heater voltage. Sure enough, these two pieces of equipment could provide the parts. I have not used tube circuitry since high school, and possessed none of the transformers, tubes, high voltage caps, power resistors, etc. that would be needed. Plus there is a resurgence of tube circuitry for audio and music. This could be fun.

I began with the tube pulse generator. It was a 11" high,  rack mounted, weighed about 40 pounds, and had a few interesting parts.
It took a couple of hours to completely strip the chassis and end up with a shoe-box size collection of parts. The sheet metal will get recycled. My plan is to revive the power supply  board since it can be used to power just about any tube project. Here is the power supply transformer on a mounting bracket and the board. I cut the board to keep the capacitors and rectifiers but reduce it's size. I removed the tube linear regulator section. If I had a schematic I might have kept it, but to reverse engineer it seemed far more work than it's worth.

Power

The Radar test set below was made by Lewyt Corporation. Alexander Lewyt was apparently famous for among other things, a vacuum cleaner patent. The test set consists of two chassis connected by 3 cables. The larger bottom chassis is a tunable 900MHz Klystron tube microwave generator. It has a 1600V DC supply with an oil-filed capacitor and high voltage rectifier tube, and lots of rectangular waveguides and complicated mechanicals. It is truly a thing of beauty. Scary too. The large bottom chassis is available if you want it. It has mad scientist written all over it.

The smaller top chassis is yet another pulse generator which I guess controlled the Klystron. It has 8 tubes, a small 640VCT transformer and the requisite caps, 3 neat, hand wired turret boards for all the discrete components. The front panel has 5 pots plus a couple of rotary switches. The top chassis is just about the right size for a tube guitar amp, and it's transformer is about right for a 10-15W amp. There are schematics for dozens of classic Fender and Marshall amps on the web. Hmmmm...
Radar LoveRadar top
Check out the tunable waveguides inside.
Radarbottom


Caveats:
Needless to say working with AC power line and voltages in the 100's of volts can be dangerous and potentially LETHAL.  You may blow something up or start a fire.  I *STRONGLY* recommend that you:
ALSO be careful of your scope probes. Make sure you know their voltage rating before probing tube circuitry. When I was a young engineer, I fried two new $200 HP scope probes trying to measure the plate voltages on a tube amplifier. My manager was unimpressed. I suggest buying one or two old giant Tek scope probes that are rated for high voltage (below).

Power Supplies

A tube power supply generally consists of an AC transformer that provides a center tapped high voltage (B+) winding and heater voltage windings. Most rectifier tubes such as the 5U4 use a directly heated cathode, meaning that the heater acts as the cathode. This requires the heater voltage to float at the B+ voltage, which requires a separate heater voltage, typically 5VAC, from the rest of the tubes which want their heaters to be at low voltage. So if your transformer has the 5V winding and you have room for a big octal tube, go for it. I lean towards using semiconductor diodes where possible. A pair of 1N4007 diodes is much smaller and cheaper than a 5U4. The 5U4 has a higher internal resistance, and thus has a larger voltage drop. Some transformer / filter capacitor combination's require this extra drop, and will cause excessive high voltage if semiconductor diodes are used.  The transformer windings also provide some resistance, but watch the high voltage, especially with no load.

Speaking of the capacitors, high voltage electrolytic caps warrant discussion. Surplus and vintage gear may have dead or dying filter capacitors. These caps are large, often chassis mounted. There are often multiple capacitors mounted in a single case. Usually they are labeled, but you can also tell by seeing how many leads. They generally fail in two ways. Sometimes they open circuit, in which case they are pretty useless and need to be replaced or paralleled. One test is to apply a lower voltage (20- 40V) from a separate supply and see if the voltage decays slowly. If it drops away in less than a second, the cap is likely open. But there may be multiple caps on a B+. Often the 2 or 3 caps are separated by power resistors or large inductors. You may need to isolate each cap to test it out.

There is another common failure mode for electrolytic caps. After years of being unused, the caps lose the ability to be charged at full voltage.  When this happens they draw lots of current and the rectifier, transformer and caps all get hot and can fail. The caps can overheat and blow off some electrolyte. You may see the rectifier tube arc out as the caps exceed its current rating. When I powered up each supply, I found that the caps would not charge and got quite hot after a few minutes. I was powering up only the 260VCT winding, having disconnected all the rest. I use a small power meter to monitor power on the AC line and sure enough, it showed >150W of power! Where was this power going? Into 2 of the 4 caps, which got very hot. These were 250V 80uF caps, several wired in parallel. I connected them to a lab supply, and sure enough, they drew 100's of milliamps even at 80V. But then the current slowly dropped to zero over time (about 5 to 20 minutes), and I could then increase it to 90V before they drew excess current. I continued this drill for several hours until they were able to accept the full 200+ volts. Then they worked fine after that.  The caps needed to be conditioned or "re-formed" before they could be fully charged. To condition them, charge them slowly, monitoring their voltage and the charging current. Watch the temperature too. What I found was that they required up to an hour of charging under these controlled conditions, then they were fine.

It is often recommended to use a variac to slowly ramp up the AC voltage on a vintage piece of tube gear the first time you power it up. This is another way to condition the filter caps. I researched this condition and it is called Electrolytic Capacitor "re-forming". Using a current limited supply, either a series resistor or a variac, is the common way to resolve this. Here are a couple of good sources that cover the subject:
http://www.nmr.mgh.harvard.edu/~reese/electrolytics/
http://www.vcomp.co.uk/tech_tips/reform_caps/reform_caps.htm

The power board shown above is working OK. Ultimately I unsoldered all of the caps to test them in isolation. There were four 2x80uF 250V, and three 2x20uF 450V caps. The 250V caps provide +/- 180VDC. the 450V caps provide +/- 320VDC. They were wired to provide both + and - supplies but by simply changing the ground could provide +180V and +360V.

It turned out that one of the 250V and two of the three +450V caps were dead open. I removed the parallel defective 450V, and there is currently no working -320V supply. But the good news is that 450V axial caps are readily available from Digikey and other suppliers. Axials are more convenient than radials, but not as available. And more expensive.  Jameco has axials for cheap. To replace open filter caps on old tube gear, one trick I have used is to just wire a modern cap across the leads of the large chassis mounted cap. Modern caps are much smaller and may fit below the chassis. It is better to disconnect the old cap though, but this requires more care with 400 volt connections.

Guitar Amp: Swords to Plowshares

A guitar amp is an appealing first tube project. There are hundreds of schematics availableon the web for classic Fender and other amps in every range from 5w to over 100W. Check out Schematic Heaven for Fender and other schematics. It has virtually every version of every Fender Amp. There is a resurgence of low power tube amps because guitar players are realizing that to drive a high power amp to get nice crunchy distortion requires deafening sound levels. A low power amp can more easily be used for its distortion characteristics.  The low power amps are single ended, the 10-20W amps use push-pull pentodes such as 6V6GT. The higher power amps use dual or quad 6L6GTs

Another appeal is that only one channel is required, unlike a hi-fi amp, making the project lower cost and easier to build. There are plenty of parts such as tubes and output transformers available on Ebay. For this project I plan to use the top chassis of the Radar test set. I'll use the transformer, filter cap, chassis, front panel, knobs, controls, and the turret boards. The transformer is 630VCT at 90mA which is about right for 450V DC caps and a 10-15W output power.  I'll probably use a solid state rectifier. I'll need an output transformer, 6V6GT output tubes and various connectors and discretes.

So which amp design to pick? It is limited in power by the power transformer at about 90mA, and it's voltage is fixed at about 350VDC. That means either a single ended (Triode) or a push-pull 6V6GT design. A larger 6L6GT design would exceed the transformer power.

The Fender Deluxe 5E3 looks good. It meets all these needs, and here is the schematic and original layout, in .PDF, courtesy of Schematic Heaven (also shown below).  This classic amp was built in the 50's and has a solid following. It uses a 12AY7 dual front end and can accept 2 inputs with either high or low level on each. The tone circuit is primarily used on the primary input, but it apparently can be used as a low pass filter on both inputs. Fenders are famous for interactive controls and I suspect this amp is like that.  It uses a 12AX7 as a phase splitter to drive the two 6V6GT pentodes. There is no local or global feedback, just lots of open-loop tube gain. The output stage bias is set by a single 250 ohm 5W resistor wired to both cathodes. This eliminates the need for a negative bias supply. The original power supply used a 5Y3GT rectifier and three 16uF 450V filter caps. Its heater chain has one side of the 6.3VAC grounded. All nice and simple.

Fender 5E3 schematic

From the Radar test set, I like the military looking battleship gray front panel and knobs. I may keep the front panel with its crazy labeling, maybe add some hints as to what the knobs really do, but probably not. If I decide to change the front panel, a new one could be built and would simply mount with 4 screws. I'll definitely keep the turret boards. The original design had 3 turret boards and was quite crowded. I'll try to use just 1, mounting some components to either the tube sockets or to the front panel controls. The original design used a military circular AC input connector and 2 BNC connectors for outputs. It had six 7 pin tube sockets and two 9 pins. The original tubes were mostly RF types plus a couple of Thyratrons, not so good  for an audio amp. It also had a lovely 10 Henry power supply inductor, that black lump below the transformer. The might be an improvement over Fender's design and will help keep the AC ripple (hum) down. My design and component needs for the amplifier:
Since the existing socket holes are mostly in the way of the new parts, I'll lay out a new rear panel after I get the transformer, and cut out the old panel

The first step was to strip the old chassis. This was a fair amount of work, with most of the hours spent stripping the tube sockets and the turret boards. It took the better part of a snowy January Saturday.  I began by cutting lots of wires to remove the turret boards. What you don't see under the turret board is a sea of wires, mostly one big harness, all laced up with nylon twine. Fortunately the bottom panel was removable so access to the tube sockets was good. Then many hours with a large soldering iron and solder sucker. The stranded wires could mostly be heated and would wiggle free. But there was a lot of solid wire which required bending while the solder was melted. You can see the original work on the turret boards. I hope to come even close to its craftsmanship.

The original rectifier is a 6X4W dual diode with indirectly heated cathode. This is a convenient part since its 6.3V heater can be grounded. I left it in temporarily to test out the filter caps. Sure enough, it arced and sparked when first powered up. The caps were leaky and drawing too much current. So I hooked them all in parallel,  wired in that big, (green, vertical mounted,  5K ohm resistor in series with the caps, and monitored the high voltage. It took several hours to condition the caps this way as I watched their voltage slowly increase from 150V to the final 350V. I need a variac. I monitored the AC power and it went from 37 Watts with the caps getting pretty hot down to less than 10W. Welcome back to life,  little capacitors.

Here is the chassis before stripping. I had cut the wires to the front panel controls.
Chassis Before
And after. Nice and clean.
After

The transformer and output tubes from Ebay are on order as are the parts that Digikey carries: film caps, electrolytics, 1/2W and power resistors, connectors and pots. Digikey's  1Meg pots are $11, so I ordered the 3 pots and the 1/4" jacks from from Jameco. The pots were $1.49 each. Note that the volume pots are after the preamp stage, and are wired in a 'shunt' mode, not the traditional voltage divider mode.  I couldn't figure out whether to use linear or log pots for the volume so I bought both and ended up using the log pots. The tone control is a linear pot.

I'm not fond of the heater wiring on the original Fender and on many tube products. They grounded one side of the tube chain and used the chassis as the return path. The heaters draw a few amps, and running this current through the chassis can induce millivolts of 60 Hz hum into the audio. Also the wiring is unbalanced, so every hi impedance input can pick up hum. Twisted pair heater wiring is better, with a single point ground. The classiest way is to use a hum balance circuit: either two 100 ohm resistors to GND or a 1K or so pot wired with one end to each side of the heater chain, and the wiper to GND. Then you adjust the pot until the output hum is minimized. I'll go for the twisted pair and if the hum still bothers me, add a balance circuit.

Here is a .pdf schematic of my own version of this classic amp .

Update Jan 11 '10
All the parts except the output transformer arrived this week, so this weekend I made the chassis changes and did most of the wiring. After removing the transformer and cap, the original rear panel was cut out with a jig saw, and a new rear panel was fabricated from 0.06" aluminum. The only parts that were in their original locations are the transformer and capacitor. Since I didn't have the output transformer yet, I left a good amount of space for it. I bought two hole saws for the four tube holes: 3/4" and 1 1/8".  The hole saws in my drill press worked out perfectly.
amp rear

I realized that the AC transformer location was wrong for a guitar amp. Typically the inputs are on the left and the outputs on the right, with the signal from from left to right: preamp -> tone stack -> phase split -> output tubes -> output transformer -> power stuff.  I didn't want the AC transformer with its magnetic field, 120VAC input and 630VAC anywhere near the input stage. So I reversed the rear chassis to put the AC transformer on the right. The sides are almost symmetrical but not quite. A few new holes are needed in the sides to line up with the PEMs in the bottom. The front panel is symmetrical so no problem there.

For the turret board, I made a drawing with dimensions for the turrets in Visio, then added the components and connections by pencil. Much quicker to draw the wiring and various size components by pencil. It turns out the Fender 5E3 parts fit fine on one turret board. My board had a small wire hole for each turret, so you could route all the wires below the board and up up through the hole. The original radar set was built this way, but it is a whole lot of work to keep flipping the board over to thread every wire in from the back side.  I decided to use these holes for only the local wiring such as power and ground. The other wires connect primarily to the tubes and front panel and would be wired above the board.   I used the following color coding to make it easier to wire and to check.
Without the output transformer I am reluctant to wire the output tubes. You are not supposed to run a pentode screen grid without the plate voltage because the screen will then try to be the plate and dissipate too much power. Their heaters are wired and working. I wired the front panel controls and the preamp tubes, up to the output tube grids. I still don't have a scope probe that can do above 150 or so volts, so I can't really probe the plates. I found a 2KV rated x100 scope probe made in China on Ebay for under $30. I may get one. A Tek P6009 or P6015 will do the job but for real money.  Fortunately every plate (except the outputs) is AC coupled to the next guy's grid so I just probe the grid for AC. A DMM will check the plate DC level just fine. An easy way to see if a tube is working is to check the cathode voltage. It should be in the .5-2V range for a 12AX7 or 12AY7. I applied a 200mV p-p 1KHz sine wave to the input, and voila. there was gain. The phase splitter, (the second 12AX7) was outputting both non-inverted inverted, so it is working OK. The only problem is that the 12AX7 is a bit intermittent. I haven't swapped tubes to see if it's the tube, socket or my wiring yet. But this is progress. Can't wait to get the output transformer.

I should mention that I first checked out the DC power supply with the 1N4007 diodes driving the 10 Henry inductor. It worked fine with B+ about 300V. I can't really measure the final DC or the ripple voltage until I get the output tubes loading up the supply though.

I had two other 12AX7s, so I tried them. I get a little more output and bias from them. I suspect they are all old and weak and plan to borrow some from a friend.

Update Jan 14 '10: What is real?
The transformer came in. No documentation but it's pretty easy to figure out the connections. A transformer for a push-pull amp has a center tapped primary. This one also has two "ultra-linear" primary taps for fancier 6L6GT amps. These are taped off. There are two secondary taps, I assume 8 and 4 ohms, but I'm not 100% sure. Size-wise, it is a bit bigger that I expected, so it doesn't fit perfectly in the chassis. I'll have to mount it at 45 degrees. Without mounting it, I wired up the output stages and the transformer, connected up a speaker, and voila, sound. Connected a guitar and strummed a few chords. With an 8 ohm power resistor load I got about 7 watts before clipping. I'm hoping for 12-15W.

As far as the clipping out of the phase splitter, I had two other 12AX7s, so I tried them. I get a little more output and bias from them. I suspect all 3 tubes are weak and plan to borrow some more from a friend. When looking at the speaker output, it seems that the phase-splitter stage clips before the outputs do. This implies that the phase splitter does not have enough output. It could also be that the output tubes are weak, requiring too much drive, but I see single ended clipping, not symmetrical. This will take some looking into. Right about now I'm wishing I had a tube tester. I'll jury rig up something.

To test a 12AX7 you need a socket, some bias and output loads. You can do a DC test and/or an AC test. And you need to measure both sections of dual tubes. Since a 12AY7 has the same pinout, the 5E3 preamp stage would make a decent tube tester for either type. Measure the two plates DC and the single cathode to get the bias, maybe change the cathode resistor to test at different currents. Then by measuring the output AC voltage with a known AC input, you get the gain of the tube (Vo / Vi) . Divide by the plate resistor (in Kohms and you get mA/ Vi, which is the gm. Cool!

Here is a spreadsheet of a handful of tubes I tested this way. The amp is basically working, but I have come to realize that there are too many unknowns in this project: the surplus power transformer voltage and impedance, the effect of the solid state rectifier and compensating resistor, a handful of used 10-50 year old tubes of unknown condition, no knowledge of the "correct" DC voltages, a non-original output transformer. Not to mention my minimal experience with tube amps, so I really don't know how it should sound. Way, way too many unknowns, and I wanted at least one solid reference point. One stage whose bias, gain and distortion would be 'right' So i broke down and bought a $20 12AX7 at Guitar Center on the way home last night. It turned out the best 12AX7 that I had was about equal to it. But I felt better knowing this little fact.

Update Jan 16 '10: Actual listening
Mounted the transformer and am doing final tests. I found a Fender 5E3 schematic with DC voltages. Turns out that my beloved 10 Henry filter inductor is causing a big loss of voltage. Makes sense, an inductor input filter takes in the full wave DC and outputs its average level.  In all my years doing e-lectronics I had never used an inductor input AC filter. With a capacitor input filter, the output voltage is basically the peak voltage which is higher by about 1.4. So instead of getting close to 400V, it's closer to 300V. But because the rectifiers are solid state,  I cannot simply drive the first capacitor. The 330VAC x 1.4 is 462VDC with no load, too much.  The original design used a rectifier tube which has considerable resistance. So I looked at the curves of the 5Y4 and guessed that 1-200 ohms 3W would do it. 68 ohms worked out pretty well.

The Rosetta stone for getting this all set up was a 5E3 schematic with all the DC voltages.  My thanks to you, unknown 5E3 person.

In increasing the voltage, I let the smoke out of one section of the 3 x 12uF filter cap. I unwired it, found the bad section and cut its pin off. I checked the other 2 sections up to 400V and they seem OK. If the remaining sections give trouble, out it goes. I found some axial filter caps, 22uF 450V at Jameco for under $2. I ordered a handful and will replace the radial caps when they come in. Just in case, there is a spot on the turret board for the extra cap.

Another good test of the power supply is to measure the voltage with your DMM set to AC. My DMM, a fluke 75, uses AC coupling it shows the ripple voltage. Ideally I'd use a scope, but ain't got the right high voltage probe. This isn't an accurate measurement since a non-RMS DMM assumes a sine wave, and ripple looks more sawtoothy. But it's better than nothing. I measured nice low ripple on the second and third caps, but the first one measured 40VAC which is close to 100Vp-p. ! I paralleled a new 22uF 450V cap, and it dropped to about 6VAC. That's strike three for the big chassis mount electrolytic. As soon as my Jameco order arrives, it is out of here.

Here is the "completed" unit without the inductor.
All working!

What's left? I mentioned axial filter caps, the input and output jacks are not the shorting types, the second input and its volume control are not wired up since it needs a shorting jack, there is no power switch or power light yet and the fuse is hanging off wires.  And I'm fresh out of 1/4" phone plugs for the speaker connection.

I also plan to add a headphone jack. This is not so simple since the amp needs a full 8 ohm load always. So I plan to add a switch and an 8 ohm power resistor to load the amp when headphones are used. This way I'll get a reasonable idea what the amplifier distortion sounds like even with the phones on.

How does it sound? Pretty good. I'm not much of a guitar player, but have been enjoying playing along to Yes, Clapton, Pink Floyd, etc. I'll ask a few real guitar players to check it out.

And of course there's the small matter of a cabinet and speaker. Having friends who are guitar electronics nerds is real handy. Steve also gave me two sweet 12" Jensen speakers from a Fender Twin Reverb that I helped work on. Let the sawdust fly.

Update Feb 4 '10: Case
The case went well. The sides are 3/4" birch ply and the horizontal pieces and the speaker baffle are 1/2" mahogany ply. I had some Okume and some mahogany ply lying around. The sides and the bottom and middle piece are rabbeted to accept the baffle. With all those rabbet and dado joints, glue alone should do the job. But the top will bear all  the weight when a handle is used to pick it up, so I added a few nails to the top joints. Here is the case being dry-fit to make sure all is good.
 Dry fit Rear dry fit
And with the speaker hole cut in the baffle, glued and clamped.
Many ClampsAmp front
The finished cabinet. No paint or covering or grille cloth yet.

I have this old speaker with holes in the suround, and two old Fender speakers. I kinda like the range and bass on the old holey one, but I'll try the Fender ones in the new cabinet. The holey one has a larger dome / voice coil, which may explain the improved bass.

One problem: one time when the amp was on for a while, there began a slowly increasing hum. I checked the back and one of the two 6V6 output tubes was getting real hot. I suspect some kind of thermal runaway. I checked the plate current on the two output tubes, and sure enough, that one was significantly higher than the other. That's what happens when you buy unmatched tubes on Ebay. I also suspect the cooling for the amp is not good enough. I'll add some holes in the middle panel to get cool air in to the area where the tubes live. I'd rather not add holes or a grille to the top.

Output Stage Bias Measurement
If I was smart, I would have added a one ohm resistor in series with eack 6V6 cathode. That way you can directly measure the cathode currents without having to deal with the high voltages on the plates. Another way to get plate current is to first measure the resistance of the two halves of the output transformer (with the power off, of course) , then measure the DC voltage acoross each half (carefully since it's 400V!!!). Apply ohms law (I = V / R) and you have the plate currents. In my case the transformer measured 130 and 138 ohms (should I be concerned about this mismatch?). The voltages were 3.85 and 6.22 respectively for currents of 29.6mA and 45.1mA. Not real good matching. The cathode resistor,  250 ohms,  measured 20V or 80mA. The extra 80 - 29.6 - 45.1 = 5.3mA must be the screen curents.  Another trick to directly measure the screen currents works if the screens are the only thing connected to the middle B+ supply.  Measure the three B+ supplies and then calculate the currents in the 5K and 22K resistors. Subtract these 2 currents and by Kirchoff's current law, the reaminder is the screen curents. I got 10.4mA - 4.0mA = 6.4mA which matches pretty well with the missing 5.3mA. Cool.

When I feel like spending another $50 I'll buy a nice new matched pair of 6V6s. Fathers day is coming.

Back to Dave's Home Page

This page was last updated 2/4/10