C32 class loco, flat truck and guard's van


Valves & Tubes

What good are they?


Well, what are they good for? Keeping you warm on a cold night. Letting you know that the fuse hasn't blown. Just imagine how much less air we would have to breathe if there weren't any valves around.

Being born in the forties, a kid in the fifties, and a teenager in the early sixties, I grew up with valves. Our main radio, called a wireless set in those days, was a 5 valve TRF. Built in the 1930s, it was housed in a tall console with a windup gramophone mounted in the top with a lift up lid. It had a proper earth to the water pipe and about 8ft of aerial stapled to the picture rail. Being only 6 miles from most of the high frequency transmitters but rather more from the low frequency ones, it had too much output at the high freq end and barely enough at the low freg end. The volume control was primitive, being just variable bias on the RF amplifiers. If you wound it up too far, the set would break into a wonderful oscillation, rather like the rising pitch sound you hear from a warship hooter in war movies.

It always took about a minute to come on and it was frustrating when I'd forgotton that one of my favourite serials was on and I'd miss a bit due to the delay. It was very reliable though and I think we only replaced one valve in the time period I was using it. It used 57s, 58s and the output valve was a 247. The rectifier was a UX280. It had an anode bend detector. Some time later, when I thought I knew all about radios, I replaced the anode bend detector with a germanium diode. Bad move. The result was much worse so I put the anode bend detector back. You live and learn. Well, sometimes I do.

By the time I was ten I decided I would like to muck about with wireless and so after several visits to our local tip I had a goodly collection of old radios, mostly TRFs, some with valves, some without. What was most prized were the power-transformers. Typically they were 385v-0-385v with two 2.5v windings for the signal valves and a 5v one for the rectifier. The 2.5v windings were very thick wire and so was the wire connecting them to the 5 and 6 pin valve sockets. The reason for the 2.5v heater winding was as follows.

In the beginning valves used to be directly heated and ran on batteries. The batteries had to be big and bulky to provide enough power for long periods. The ultimate aim was to get rid of the batteries altogether and run the whole set off the mains power. Unfortunately the mains power was AC and this caused the electron emission from the directly heated filaments to be modulated by the mains giving lots of hum in the output. The valve designers had the brilliant idea of changing the filaments so as to have a high thermal inertia which would even out the mains modulation of the electron emission. This meant thick wire for the filaments which meant high currents. Thus low voltage at high current was the order of the day for most of the early valves; 2.5v at 1 or 2 amps. Eventually the floating cathode was invented and this removed the need for thermally heavy heaters. But, due to the large number of sets with 2.5v heater windings, this heater voltage was kept in use when the valves stopped being directly heated. Thermally heavy heaters took a long time to get going. Eventually, in order to dramatically reduce this delay, new valves were made with 6.3v heaters at much reduced current and this is the type we mostly see today. They usually heat up in 11 seconds, much faster than the old valves' minute.

 

 
Valve Internals


 
Valve Internals
  1. Pins, top of which also support the bottom mica spacer.
  2. Glass button base which holds the pins in place.
  3. Wires which connect the elements to the pins.
  4. Top & bottom prepunched mica spacers hold the elements in position.
  5. Plate/Anode
  6. Getter support cup and sputtering ring.
  7. Getter material which after sputtering coats the inside of the glass.
  8. Glass shell with vacuum exhaust-tube at top.
  9. Heavily insulated heater wire spiral-wound to reduce hum.
  10. Sealed exhaust-tube tip.
  11. Oxide-coated cathode tube which emits electrons for the life of the valve.
  12. Control grid support rods.
  13. Close-spiral-wound control grid.
  14. Screen grid support rods.
  15. Screen grid close-spiral-wound in shadow of control grid, making this valve a Beam Pentode.
  16. Suppressor grid support rods.
  17. Coarse-spiral-wound suppressor grid.
  18. Internal vacuum, maintained by getter after sealing
Menu


 
One of my early attempts at electronics was to build an audio amp to a circuit I got out of a library book. It had a 6SJ7GT as the first amp and a 6V6GT as the output. It ran on 250v HT and used a 5Y3GT as the rectifier. Just to explain a little bit about these numbers. Some of them made sense and some of them didn't. Most of the time the valve numbers allocated just seemed to be the next ones available. However, when it came to what the valve looked like, at last some logic came into the mix. During the second world war many new valve were invented with improved characteristics. One such set came out in metal cans with the only glass in them used to hold the pins and support pillars in position at the bottom of the can. They had numbers such as 6SJ7, 6SH7, 6AC7, 6V6 and so on. Being mostly metal they were a lot more rugged than the total glass valves and so were eminently suited for wartime conditions. They must have cost a lot to make as after the war production stopped and the innards were then put into glass bodies. These had numbers such as 6SJ7GT, 6V6GT, 5Y3GT etc. Before the war valves were in quite large glass bodies with shoulders and these were still in use during and after the war. Valves using these large bodies had numbers like 6J7G, 6V6G etc. So a rule of thumb was that if the number ended in G then the valve was a tall glass shouldered body. If it ended in GT then it was a short glass straight body and if it ended in neither then it was a metal one. Of course rules of thumb dont always apply but it worked most of the time.

So back to the audio amp. By this time I had collected a few mantle radios from the tip and these contained a few of the 8-pin GT valves. As well I used to visit the local war-surplus store and spend my 50c/fortnight pocket money on the wonderful devices that were available therein. I had a second hand 6V6GT and a brand new 10yr old warsurplus 6SH7 and a similar 6AC7, both metal valves. I used the 6SH7 instead of a 6SJ7GT, it was close enough. I didn't have a 5Y3GT but I had an 80 in a GT body. The 5Y3GT needs an 8-pin socket whereas the 80 only needs a 4-pin one otherwise they are identical. I mounted all the components and valve sockets on one of the old TRF chassis so I had plenty of room. The mantle sets were quite small and very cramped and a nightmare to work with. I had to use a small plumbers iron, heated on the gas stove, for soldering so you can see why I needed plenty of room in the chassis. Some of the original valve holes weren't in the right position so I had to cut new ones. The only method I could use was to drill a series of holes with a hand drill then cut them through with a hacksaw and then file.

I was using one of the TRF transformers which had a 5v winding and three 2.5v windings. Luckily one was centre tapped. So with all three 2.5v windings in series and taking one wire from the centre tap I got 6.25v. In reality this would have even been a bit higher due to the low loading. This fed the 6.3v valves. and the 5v winding fed the 80. With all the bits wired up, I plugged in the valves and gingerly swiched on.

Now in those days the bugbear of audio amps, especially home made ones, was hum. With all the high AC voltage, and magnetic fields from the high current heater wiring, it was very easy to get hum induced into the circuit. So the sign of a good amp was little or no hum at all. If you could only hear the hum with your ear right up to the loud speaker then you had a good amp.

So I switched on and waited with baited breath and my ear in the speaker. After the required waiting period had passed and the valves heaters were up and running I could hear no hum. Well, did my head swell. My first audio amp and no hum. This was teriffic. I wound up the volume control. Still no hum. My hats were never going to fit again. I put my finger on the input, expecting to get a loud BLURP!. Nothing!. I put my finger on the grid of the first valve. This was the standard way to see if it was working as the high impedence grid cct would let your finger inject mains hum that your body always seems to pick up from the air. Still no BLURP!. My hats were fitting better all the time. I put my finger on the grid of the output valve where you can usually get a slight blurp!. Still nothing!. Now I was completly deflated. I looked at the valves, thinking that one of the heaters must have gone out. No they were all glowing merrily. Not having a meter at this time I did the usual way of seeing if there was any HT. I shorted it to ground with a piece of insulated wire. No spark at all, there was no HT. I found one of the plate connections on the 80's socket and shorted it to ground. Now we were talking, sparks all over the place. This time I looked more closely at the 80. There seemed to be more light coming from it than I would have expected. Hang on! The plates are supposed to be black not orange. The 80's plates were glowing. That doesn't look right. I hurriedly switched off. On close investigation I found a small sliver of steel going from the chassis, where I had drilled, hacked and filed, across to the heater pin of the 80 socket giving a dead short to ground. I must have missed that piece when I was cleaning up the hole with the file. As the 80 is a directly heated valve, the heater is really the filiament which is actually the cathode so a short to ground of this is a short to ground on the HT. I removed the offending sliver and switched on again. This time I was greeted with quite a bit of hum from the loudspeaker. I wasn't as good as I thought I was.

Back to the library. After further reading I discovered that you had to twist the heater wiring closely to reduce the hum radiating from it. In most amps the heater wiring is put in first and then the rest of the components after. It was a nightmare to pull the old wiring out and feed the new twisted pair through and under all the other components. If I had used anything other than the large TRF chassis I would have had to dismantle the whole amp and start again.
Menu  
 
From the above story you can see that valves are quite robust devices, the 80 having survived a direct short with no dramas. However some valves are not so forgiving. Sometime later I decided I would like to build a "portable" radio. The portable is in inverted commas as it really is a stretch of the word "portable".

I went back to the war-surplus store and purchased what passed for portable valves. They were 8-pin GT valves and not exactly what you would call small. They were much bigger than the 7-pin miniatures then available but I couldn't afford those. I bought two 1P5GT rf amplifiers and a 1Q5GT audio output and a 10000ohm o/p transformer. I built it on the previously mentioned TRF chassis. It would have a 1P5GT Reinartz regenerative (positive feedback) detector and a 1P5GT audio amplifier followed by the 1Q5GT audio output. Now the main problem was going to be "what was going to power the radio". Of course it had to be batteries to be portable but high voltage 'B' batteries were very expensive and well out of my price range. What to do? WW2 to the rescue yet again. Most people abhore war but I, being a kid with no experience of it, actually thought that the war was teriffic. The war-surplus stores had everything I needed at prices I could afford so it's an ill wind that doesn't blow some good somewhere.

The war-surplus store had a number of ex-army dry batteries available. They were housed in khaki-green waxed-cardboard boxes and measured about 4 inches high x 5 inches wide by 8 inches long. In one of the top corners was a 3/4 inch hole with a rivet on each side. In the hole was a piece of insulation which I think was called "Presspan". This had 4 holes, each with a metal tongue, laid out in a square with two fat holes and two thin holes. The two fat holes were connected to two large 1.5v cells connected in series, thus producing 3v. The thin holes were connected to about 110 flat-cells which were also connected in series producing about 165v. It also came with a cheap matching 4 pin plug and all this for the princely sum of 50 cents. This would do me just fine. On getting it home I quickly found that the 165v section really packed a wallop. It had enough ergs to light up a 40w 240v light bulb. You couldn't do that with the usual expensive 'B' battery normally used in portable radios.

I rearranged the circuit so that the two 1P5GTs filaments were in series so they could run on 3v and the 1Q5GT was wired to use its 3v pins. Now they could run happily on the 3v from the battery. The 165v would be plenty for the HT as these valves could function perfectly on 90v.

So I assembled the receiver on the TRF chassis, plugged in the valves and then plugged in the 4 pin battery plug. Instantly the three valves went off like flash bulbs. A cold feeling came over me as I discovered I had plugged the battery connector in backwards and put 165v on the 1.5v filaments. Goodbye to 3 weeks pocket money. After that fiasco I marked the battery and its matching plug very clearly with white paint. I couldn't afford to make that mistake again. Battery valves are not rugged at all.
Menu  
 
OK. Having just showed you how stupid I can be, I'm now going to carry on showing it and wade into the "Triodes are better than Tetrode/Pentodes" war.

First a few words on the "Golden Ear Brigade." I assume that the "Triodes are Better" edict comes from this group because they reckon they can hear a difference between Triodes and Tetrode/Pentodes while we normal people cant hear any difference at all. Are these people fooling themselves or can they actually hear a difference? This will also bring me into the middle of another war between the writers who believe the Golden Ear Brigade can do no wrong and the others (mostly ex engineers) who say the GEB are talking rubbish.

Well lets look at the track record.

  1. Way back in the past, during the early valve days, Medium Fi was all the rage. Radios and audio amps went up to about 6Kc (6KHz) and nobody wanted to go any higher. Engineers had built wider bandwidth amps but the test subjects didn't like them and turned down the tone controls till the response was about 6Kc. The engineers assumed that this limit was as much bandwidth as the public wanted. However one person could not believe that people, when given the choice between wide bandwidth and 6Kc, would rather listen to 6Kc. He reckoned there must be something wrong with the valve amps and/or transmission system so he set up an experiment. He had a mechanical low pass flilter constructed and put it on rollers so he could move it about. He then set up a small live orchestra at the front of a hall and put the low pass filter in front of them. Then in front of both he placed a curtain. He filled the hall with people from all walks of life and used them as test subjects. The orchestra played various types of music and, while they were doing so, he moved the low-pass filter back and forth while an assistant asked the audience what they preferred to listen to. Because of the curtain the audience didn't know whether the filter was in place or not. The result of the tests was that the audience unanimously said they would rather listen to the non-filtered music. Using this result as proof, that there was something wrong with the valve amps and receivers, he coaxed the engineers to reinvestigate the problem. Lo and behold they found the valve equipment had much higher distortion at the high freq end of their bandwidth than they had at the middle frequencies (1kc). Once this problem was fixed the public took to HiFi in a big way. So one black mark against the engineers.

  2. When transistor amplifiers were first being built, valves were in their heyday. All the problems had been sorted and the sound they produced was all sweetness and light. All sorts of tests had been dreamed up to check that any amp would pass muster. So along came the transistor amps and immediately the engineers saw that they could do away with the biggest bugbear of the valve amp, the output transformer. It was the o/p tranny that limited the use of that all-fixing-panacea negative feedback. Most o/p transformers would not let you put more than about 15dB of feedback round an amp thus limiting how low you could get the distortion, about 0.5% was the limit. Transistors, being low impedance devices, didn't need an o/p tranny so now there was nothing to limit the amount of negative feedback which could be used. In some amps it went as high as 40dB. The distortion was down to 0.01% and getting lower all the time.

    When valves got hot they kept right on working till their glass shell melted. When transistors got hot their leakage increased and which made them get even hotter till they ran away thermally and destroyed themselves. I can remember one amp designed by Fairchild which had a medium power pair of complementary o/p transistors with no forward bias. The engineer made sure these transistors were not going to runaway thermally. To overcome the forward threshold voltage necessary to get them to conduct, this amp had at least 40dB of feedback. The idea was to force the signal to fly very rapidly from one threshold to the other so you would think the o/p signal was contiguous. Well the amp didn't runaway but everybody said it sounded awful. Eventually the cct was applied to a communications amp with small o/p transistors on tiny heatsinks putting out a few watts. That was all it was good for. This forced the engineers to realise that negative feedback was not the fix-all they thought it was and so had to reduce the amount of neg feedback in quality amps. However, even though the public was now satisfied with the sound coming from these rehashed amps, the Golden Ear Brigade still said they would rather listen to a good old valve amp. Well, the engineers heaped much dirt on the GEB, coupled with plenty of ridicule. The GEB stuck to their guns and slowly swayed the top end amp buyers to buy valve amps instead, saying that transister amps were Medium Fi at best and were ok for the masses. But if you wanted real HiFi you had to have a valve amp.

    This pigheadness by the GEB surprized the engineers and one by one they began to reinvestigate the transistor amps. It didn't take long to find that, when most of the negative feedback was removed, the resulting bandwidth was only Medium Fi. They had been using neg feedback to make up for poorly designed amps. A new style amp was designed with a wide bandwidth before any neg feedback was applied. The GEB finally said they could not hear any thing wrong with the new amps. Detailed investigation revealed that when large amounts of feedback are put round a medium fi amp, the feedback signal is delayed by the low bandwidth of the amp. This means that the output of the amp will rise rapidly until the feedback can reach around the amp and pull it down again. This is now called transient intermodulation distortion and suitable tests have been formulated to check it. The first transistor amps ever built passed all the valve amp tests with flying colours but couldn't pass the transistor amp tests (because there weren't any). This shows the futility of checking new technologies using old technology tests. Another black mark against the engineers.

  3. So its GEB 2 and engineers 0. Now lets look from the other side of the fence. The GEB reckon we need oxygen-free copper in our connecting cables. Simularly speaker cables need to be rated at 30amps at least. Loudspeakers have a couple of percent distortion. Headphones have half a percent distortion. Yet the distortion quoted sometimes for these cable products is 0.0003%. How could it be possible to hear such low distortions using such distorted listening devices? This is where I'd have to agree with the engineers. Unless the GEB have built in distortion cancelling brains, they are stuck with the listening device distortion. For them to hear such low distortions would be like trying to read a book at 100ft when your eyes can only see 10ft.

    Monster Cable Ad
    Here is an ad which proves the old adage "A fool and his money are soon parted".

    It's hard to believe that three video cables, 6 feet long with RCA connectors, could cost this much. They must have been manufactured by people who only work on 29th of February. At $50 they would be expensive. At $250 they rely on one being born every minute.

    Thirty-Amp speaker cable comes into the same category. We could assume that 30amp speaker cable might have an impedance of .01 ohms. Now I assume this low ohms requirement is to allow the amplifier to drive the 8 ohm speaker from a very low impedance source thus damping any tendency for the speaker cone to overshoot and ring. This really is only a worry at low frequencies because as the frequency rises so does the impedance of the speaker and any speaker cable will dampen the speaker sufficiently. It is only at low freq that the speaker impedance approaches its rated impedance ie 8ohms. The speakers, which most of us come in contact with, will produce doubling if their voice coil is fed with too low a frequency. This would indicate that the influence of the speaker motor (coil/magnet) is not perfect at low freq and this excessive worrying about the damping will only stop the voice coil ringing but not the cone. I think that this is 2 for the engineers so now they are even, 2 all.

  4. This brings us to the controversial bit. If the GEB really can hear the difference, how can they?

    We all know and understand that some people can see better than others. Some have perfect vision. Others have perfect focus but have astigmatism. Others have no astigmatism but have poor focus. Some have one lazy eye and so dont see much stereo vision. If we didnt have proof for all this just imagine the disagreement on the fact that somebody could see better than most other people. They would have to prove it before anyone would believe them. I believe it's the same for hearing. Some have poor freq response. Some have built-in low level noise (tinnitis). Some have a few pickup-sensor hairs broken and hence cant hear certain frequencies clearly. Some have unbalanced ears and hence do not have proper stereo hearing. Some may have poor quality brains which cant keep up with the enormous amount of processing required to hear properly, especially through external noise (old people). All these we know about because someone as investigated them and told the rest of us about them. Well, I'd like to add another.

    In order to see properly our eyes must be focussed sharply. The sharper the focus the better we can see. Similarly to hear clearly we need our hearing to be sharp. That means no slackness in any of the mechanical parts of the ear, the drum, the anvil, the staples, the window of the cochlear. No slackness in any of the hair sensors in the cochlear. Assuming that these are all ok then we will hear well if the signals arrive at the eardrum in a clear manner.

    To show what I mean, assume a noise is coming from a point to the right at about 60 degrees to the centerline of your head. One wavefront will arive at your right ear directly from the noise source. However the other ear will get its signal from a number of different directions simultaneously. One will come around the front of your face. One will come over the top of your head and one will come round the back of your head. Of course there will be far more than three wavefronts hitting the left ear. All will be coming over different parts of your head. Depending on the shape of your head and your ears, these wavefronts may arive simultaneously or they may not. It should be obvious that, if they all arrive simultaneously, your brain will have a much clearer audio image to work from than if they all arrive at different times. The audio wavefront would be time-smeared and so your brain would have difficulty when trying to detect very low levels of discontinuites, distortions or whatever. It may be possible that the GEB actually have heads that are more symmetrical (circular) than the rest of us, which provides their ears with quite sharp wavefronts to work on. This, with all parts of the rest of their hearing system working properly as well, would put them streets ahead fo the rest of us who have various abnormalities, even minor ones. This is similar to the fact that some people, not many, have balanced faces and we say they are good looking. This is only so because the rest of us (the majority) have unbalanced faces.

  5. So if the GEB say that Triodes sound better than Tetrode/Pentodes then maybe its true. Why might this be so?

    First a quick recap of the journey of the valve.
    • The light bulb was invented. A filament in a vacuum.
    • A plate, when put in the vacuum with the filament, showed a slight current.
    • When the plate was very positive to the filament a large current would flow. The valve was now a Diode.
    • The plate of the Diode was moved further from the filament and a zigzag grid was placed between them. When negative to the filament it would control the plate current. The valve was now called a Triode.
    • If a resistor was put in series with the plate and the +HT supply, any variation in the grid voltage appeared amplified and reversed in polarity at the plate.
    • If a high freq was fed in the grid, the amplified high freq on the plate was fed back to the grid by the inherent grid to plate capacitance making the small G to P capacity appear bigger than it really was.
    • The plate of the Triode was moved even further away from the filament and another grid called a screen-grid was placed between the control-grid and plate to shield the control-grid from the plate. The valve only worked properly if the new screen-grid was connected to a high positive voltage and grounded to AC. The valve was now called a Tetrode.
    • The Tetrode valve oscillated when the plate voltage dropped below the screen-grid voltage. This was caused by secondary emission from the plate back to the screen-grid.
    • The plate of the Tetrode was moved even further from the filament and a third grid was placed between the plate and the screen-grid and connected to the cathode to repel any secondary emission back to the plate. This extra grid was called a suppressor-grid and the valve was now called a Pentode.
    • The plate of the Tetrode was moved further from the filament and two beam forming plates, connected to the cathode, were inserted to make the electron stream, flowing to the plate, dense enough to prevent any secondary emission. The spiral wires of the control-grid and screen-grid were synchronised to put the screen-grid wires in the shadow of the control-grid wires thus reducing screen current allowing more current to reach the plate. This valve was called a Beam Tetrode.
    • The filament had voltage drop across it causing the control-grid-to-filament bias to be different from one end to the other. To overcome this the filament was placed in a tube, now called the cathode, and the bias became more predictable and the heaters of several valves could now be paralleled without interaction.
     

    OK so we have the three types of valves. The Triode, the Beam Tetrode, and the Pentode. The "Triodes are better Brigade" say that Triodes sounds clearer than the other two.

    Whats interesting about this is that of the two types, that is the Triode and the others, it's the others which work more like a valve should and the Triode actually has a flaw in its operation.

    Transistors are Beta devices, in that a current is applied to the input and a current comes out, whereas a valve is a GM device in that a voltage is applied to the input and a current comes out. Pentodes are purer valves as the grid affects the plate current while the plate voltage has very little effect on the plate current. On the other hand the plate current in a Triode is affected by both the grid voltage and the plate voltage. This is why a Triode cannot produce as much output power as a Pentode or Beam Tetrode can for the same HT voltage. There is a good reason for this and it probably explains why the TABB reckon it sounds better too.

    Transistors are enhancement devices as they normally do not conduct until the base turns them on. On the other hand valves are depletion devices in that they normally conduct until the grid shuts them off.

    A Triode consists of three elements. An emissive cathode, a collection plate and a controlling grid. The cathode emits electrons and the high positive voltage on the plate attracts them to it. A negative voltage on the grid controls the amount of electrons reaching the plate. Implied in this operation is the fact that no electrons will flow unless there is a high positive voltage on the plate. If we have a Triode sitting biased in the middle of its operating range, a certain plate current will be flowing, causing a certain voltage drop across the load resistor or transformer. If we put less bias on the grid we expect the plate current to increase. But if it does it instantly lowers the plate voltage. Thus you can see that for any given cct there will be a point where the grid will be urging more current to flow but there will be insufficient plate voltage to attract it. This has two effects:-

    1. The Triode cannot swing its plate down to a low voltage with a simultaneous high plate current. This limited plate voltage swing at high currents thus limits the amount of power which can be delivered to the load. This in turn means that a Triode must have much bigger elements than a beam-Tetrode/Pentode for the same output power.
    2. This reduction in actual current over the expected current is a form of negative feedback. As neg-feedback reduces distortion, this may be why the TABB reckon Triodes are less distorted than beam-Tetrode/Pentodes where the output current only follows the grid.
     

    This poor-plate-swing-at-high-current limitation can be overcome by using the grid as an accelerating electrode to take over as the plate becomes a less efficient accelerator. This is done by using the signal to drive the grid positive. Unfortunately this has two consequences:-

    1. The grid drive voltage must be greater than with a similar Beam Tetrode/Pentode.
    2. The grid will now draw current from the signal supply which means the device feeding the grid must supply undistorted signal power; not just voltage as it does when driving Beam Tetrode/Pentodes.
     

    The plate current in Pentodes and beam-Tetrodes has very little to do with the plate voltage for a couple of reasons:-

    1. The plate is a long way from the cathode, easily ten times that for a Triode, and therefore has less affect on the electrons leaving the cathode.
    2. The main attractor for the electrons is the screen-grid whose voltage does not change as the plate voltage changes thus the plate current is basically constant and only determined by the instanteneous grid voltage.
     

    This allows the plate voltage to drop dramatically, typically down to 25v, while still having the same current flowing. Thus the beam-Tetrode/Pentodes output a lot of power for their size. As there is no internal neg feedback, as Triodes have, the transfer characteristic is much closer to the pure valve one and hence the distortion is higher, typically 5% to 10% which is rather horrendous. Circuit trickery is used to reduce this to a listenable amount.

    One such trick is the Ultra-Linear-Output. Here the screen-grid is connected to the plate to make the valve act like a Triode but, instead of a direct connection as in the full Triode implementation, the screen-grid is connected part way along the output transformer primary. This has a number of advantages :-

    1. First of all the screen current adds to the output giving slightly more power output than you would get otherwise.
    2. The Screengrid, by being disconnected from the heavily bypassed HT supply, is allowed to swing as its current varies. This begins to let the valve act like a Triode and hence provides some negative feedback as in a triode.
    3. The tapped winding provides a portion of the plate signal back to the screen grid giving us another method of negative feedback.
    4. As the screen-grid is only tapped along the winding, it does not swing too much, thus limiting the Triode type feedback, allowing the plate to still swing down to a low voltage giving plenty of power output. Far more than if full Triode connected.
    5. All these feedback loops, so produced, reduce the distortion remarkably.
     

    Of course the TABB still reckon that pure triodes work better than fiddled beam-Tetrodes/Pentodes. There is possibly one more reason why this might be so. If you look at the plate characteristics of most Triodes, the lines on the graphs are curved but also quite smooth. Whereas the lines on most beam-Tetrode/Pentode graphs are also curved but have some vestigal kinks as well where the valve approaches cutoff. It seems the Dynatron oscillation region of the beam-Tetrode/Pentode operation has not been completely eradicated by the beam-plates or the suppressor grid. Maybe this is what the TABB can hear?

    If you took a normal Pentode, laid it horizontal, shrunk yourself, then went for a walk along the cathode and looked up, you would see the control-grid wires. Then you would see the space between the grid wires. Looking through that space you would see the screen-grid wires. If you were lucky you may see through the screen-grid wire spaces and see the plate as you walked along the cathode but mostly you wont because it is obscured by all the various grid wires. Similarly the electrons, emitted from the cathode, cant see the plate either and must negotioate this obstacle course on their way to the plate, bashing into most of the grid wires on the way. This creates noise and puts a limitation on the current flow. The electrons land on the plate in a not-so-dense cloud, so that electrons knocked off the plate (secondary emission) have no trouble escaping and flowing to the back side of the screen-grid, especially when the plate voltage is much lower than the screen voltage. This causes the infamous Dynatron oscillation region. The initial reaction of the valve makers was to put another grid, between the plate and the back of the screengrid, to reflect any secondary emitted electrons back to the plate. This appeared to work but it put yet another layer of wires in the path of the electrons and made the path to the plate even more obscure. The noise got worse and the current was limited a bit more. Naturally the company, who invented this, patented it so stopping other companies from using it without paying a royalty. As is usually the case, people dont like having a financial gun held to their head so the other companies beavered away to figure out some other method of getting rid of the Dynatron region.

    One company's smart prople observed that birds have a lot of trouble flying into the face of a stiff wind. If they could just get the electron stream to be a bit more dense, then the secondary emmitted electrons would have a hard time flying from the plate to the back of the screengrid in the face of a stiff electron wind. But how to get the electron stream more dense.

    The valve makers reckoned that the stream was diffused because it kept bashing into the screen-grid wires, making it deviate from a straight course to the plate. They decided to make the screen-grid wires have the same pitch as the control-grid wires and placed the screen-grid so that its wires were directly behind the control-grid wires. So now, standing on the cathode and looking towards the plate, you would only be able to see the grid wires and not the screen-grid wires. This meant the electrons, which managed to pass through the control-grid spaces, now had a straight run to the plate. They noticed that the Dynatron effect was much reduced but still large enough to cause some oscillation. Now what.

    They reasoned that the remnant Dynatron effect was coming from the electrons which travelled through the control-grid and screen-grid wires near the vertical mounting pillars on either side of the valve. At this point the electrostatic fields were not uniform thus letting the electrons stray all over the place producing a less dense stream at these two places. They decided to place two shield plates around the two groups of mounting rods to stop any electrons passing through these regions to the plate. They were called Beam Forming Plates but that is a misnomer. They really are Shield Plates. They dont form the electrons into beams but just stop any electrons that are not in the beams. This means that the plate current now is only made up of the electrons which are already in the beams.

    Now that the electrons were arriving at the plate in dense sheets from the spaces between the grid wires, the secondary emitted electrons were unable to get away and were swept back to the plate from whence they came. However you may ask "What about the parts of the plate which are behind the grid wires?. They wont have the dense electron stream to prevent secondary emission". Good question. The answer is simple. The electrons travelling in the dense sheets actually dont want to be there. Electrons repel each other so the sheets slowly spread out and become less dense as they travel from the screen-grid to the plate. If the valve makers have done their job properly, the edges of the sheet-shaped electron streams will have spread just enough to join the edge of the next electron stream, as they reach the plate. Thus forming one large mass of electrons that will stop the secondary emission from all the plate area which is actually receiving electrons from the cathode.

    If you look at the plate characteristics of all the beam-Tetrode valves, you will see a slight kink in each graph where the plate voltage gets substantially lower than the screengrid voltage. It seems that this method of stopping secondary emmision is far from perfect. Mind you some of the suppressor-grid Pentodes also have the kinks so they are not much better. As I said before maybe the TABB can hear this discontinuity in the characteristics.

    Some valves are called Beam Power Pentodes, I've no idea why. If you look inside them you cant see any suppressor-grid so they are really Beam Tetrodes. Maybe it's just to get around some patent or other.
    Menu  
     
    Ok, this Ultralinear Output is all very well if you have the required tapped output transformer. But what if you haven't. Is there any other trick you can do? Well, actually, there is, but being a trick it might not work as well as expected.

    There were three advantages with Ultralinear Outputs, two concern increased negative feedback and the other a slightly increased power output. If we ignore the slightly increased power output and are willing to put up with slightly less output, we can access the other two advantages without having a tapped o/p tranny.

    Ultralinear Output
     
    The trick is fairly simple. All you need is a ganged twin 10k 3 watt wirewound pot. When you have found the values you need, you can replace the pots with fixed resistors.

    Ultralinear Output without the tapped transformer.
     
    You currently have an output transformer with a centre-tap. Wire both pots, as shown in the centre diagram, across each half of the o/p tranny. Make sure the pot rotation is such that both wipers are at the plate ends at the same time. Disconnect the screen-grid connections from their HT point and connect them to the wiper of each pot. Make sure the HT for the transformer is not higher than the screen-grid HT, otherwise you might over run the screen-grids and melt them. I hope you can see that we have created an artificial tap on each half of the output transformer primary winding. The pot wipers can move from the HT at one end all the way to each plate at the other end. Thus the valve will be operating as a Pentode at one end of the pot and a Triode at the other end. You can adjust the pot while you are listening or testing the amp to see if there is any improvment in distortion.

    You will get less power out for a number of reasons:-

    1. The screen-grids will be running on less HT than they were before.
    2. The two feedback systems will reduce the gain of the amp.
    3. The driver will have to push the output valves harder.
    4. The o/p valves wont be correctly matched to the load.

    In more detail:-
    1. This will be the biggest problem. The screen-grids are normally fed with full HT. After our trick is inserted they will have a series resistor which will have a voltage drop across it. This will reduce the voltage applied to the screen-grids which will in turn limit the amount of plate current available when the signal on the control grids reaches zero bias. There is something we can do about this but it depends on the complexity of the HT filters in the amp.

      Ultralinear Trick with separate screen-grid powersupply.
       
      Looking at the above diagram, the power supply cct across the top is a representative cct of an average amp. Most power-supplies will have one or two chokes and/or resistors in series with the raw HT from the rectifiers, accompanied by an appropiate number of resovoir caps in the usual places. The closer we get to the rectifiers the more voltage we can pick off.

      This occurs for two reasons. Firstly, there is a DC voltage drop across each choke and/or resistor in the string due to their DC resistance. Secondly, the AC ripple, sitting on the DC voltage, gets higher the closer we get to the rectifiers. We can takeoff some of this DC voltage plus ripple, which will have a peak value higher than the o/p HT, feed it through a diode to a Pi filter, consisting of couple of caps with a resistor strung between them, to produce an auxiliary HT just for the screen-grids. Once the correct position for the pots is found, they can be replaced with fixed resistors, and the resistor in the Pi filter can be selected to produce the correct value of HT which the screen-grids had before we put the pots in. This is possible as the screen-grids usually only draw a small current. Typically 6mA for small valves and 15mA for large ones. This means that the auxiliary supply only has to provide 30mA at the most.

      A 47uF and 100uF 400v cap should handle most amps. Dont forget ot put 47 ohm 1 watt resistor in series with the diode to limit the ripple current load on the original filter. The diode only has to handle the ripple voltage so a 100v 1a diode should be plenty.

      Make sure a bleeder resistor of about 220k 1watt is fitted from each HT rail to ground and wait for the caps to discharge before you go poking around. Caps can hold a charge for a long time if no bleeder resistors are fitted. Do not discharge the caps by shorting them out with a screwdriver. While it may give you some spectacular flash-bangs, it will most likely do damage as well. It will destabilise the dielectric gas film and cause the cap to draw large reforming currents every time you turn the amp on. The sparks will also partially melt the wire and burn off the protective tin coating letting corrosion get under the tin layer.

      Do not be tempted to run the screen-grids at a higher voltage to get more plate current out of the valves. You could easily over run the screen-grid power rating and melt them.

    2. This really shouldn't be a problem as most, if not all, amps usually have excess gain.

    3. This could be a problem for those amps which only have enough drive for the normal cct. Once the twin neg feedback systems reduce the gain of the o/p stage, they could run out of drive voltage. You might be able to increase the HT to the driver valves but be careful you dont push them beyond their maximum ratings.

    4. There's not a lot we can do about this problem as we can assume the amp was designed with the o/p valves correctly matched to the load. However, there is one thing in our favour. The valve data graphs, which normally show the distortion and power output versus load impedance, are usually broadly curved and hence the mismatch shouldn't cause too many headaches, fingers crossed.

    If you think that the DC currents, flowing through the pots/resistors, are affecting the amp, you could put a 10uF cap in series with each end of the pot/resistor going to the plate connection, as shown in the diagram. The caps wont need the full HT rating as they only have to take care of the voltage drop across the primary output winding which shouldn't be more than 50v so a non-electrolytic could be used. If you have added the separate screen-grid power-supply, you will have to recalculate the value of the Pi filter resistor.

    Remember the pot is connected to the +HT so try not and get a shock if you can help it.
Menu  



Now to something completely new, er, old. Well it was new to me and I suspect it will be new to a lot of others as well. This idea was originally designed to only work with real Pentodes.

Back in 1943 a fellow named James Crawley was working for RCA in their valve circuit laboratory. It being wartime he was trying to do something that would make modern managers salivate. He was trying to do more with less. That is, he was trying to make a better amplifier while reducing the number of used components in the process. He came up with an interesting cct which he imediately patented under RCA sponsorship. This cct has multiple methods of local negative feedback and uses less components to boot. It seems to have been lost in the mists of time as I have never seen it in any amplifier and it only came to light when I was reading a 1947 aussie radio handbook. I tracked down the patent and it is written in rather quaint language, such as "one skilled in the art" (of electronics). It is heavily sprinkled with the word "degeneration" which nowadays means negative feedback.

This cct is brilliant in its simplicity but very complex in its operation. It may even solve the current problem of getting an ultralinear amplifier without the special transformer. This probably sounds too good to be true but see it for your self. It is meant to be used to get one Pentode to drive a second Pentode but, looking at the cct, the second Pentode could be a beam-Tetrode instead.

James Crawley: Triodyne
In keeping with the era I have called this circuit a " Triodyne ". The values are Mr Crawleys.


Looking at the cct you can see a number of interesting points:-
  1. The cct uses AC and DC coupling.
  2. There are a number of feedback mechanisms.
  3. It saves two resistors and two capacitiors over a normal two Pentode cct.
  4. The second Pentode acts like a partial Triode.
  5. The first Pentode must have a seperate suppressor connection.
In more detail:-
  1. The plate of the first valve V1 is DC connected to the screen of the second valve V2. This saves the screen resistor and the bypass cap for the second valve.
  2. The plate of V1 is AC coupled to the grid of V2. This is stock standard normal.
  3. The grid of V2 is DC connected back to the suppressor of V1. This method of feedback uses no extra components. The normal method would and so saves another resistor and cap.
  4. The screen of V2 is AC coupled back to its grid.
  5. The screen of V2 is unbypassed.
So what does all this mean?
  • Points A + B mean that V2 is acting like a Triode with the same signal being fed to both the screen and the grid. If you connected the screen to the grid directly, V2 would not work as the plate of a pentode is too far away from the cathode to do much electron attracting when the screen has no high voltage on it. This method allows the screen to have a reasonably high voltage while still being fed with the signal.

  • Point E means that V2 begins to act like a Triode. That is, as the screen current increases, the screen voltage will drop causing the plate current to be less than it would have been if the screen voltage was fixed. This is internal feedback, same as you'd find in a triode.

  • Point D + E means that, as the plate current varies in V2, its screen current will also vary. This will cause a voltage drop across the screen resistor which will be fed back to the grid, producing neg feedback in V2 via its grid.

  • Point A + E will also affect the plate current of V1 as the screen resistor has two currents flowing through it. The plate current of V1 and the screen current of V2. If V1 current increases causing the screen voltage to drop, V2 screen current will drop thus opposing the current change from V1. This is more negative feedback.

  • Point C allows the signal from the plate of V1 to be AC coupled back to its suppressor, here being used as a supplementary control grid while still doing its primary roll of suppressing secondary emission. This is more neg feedback.

  • Point C also means that the plate-to-suppressor capacitance of V1, which is higher than the suppressor-to-screen capacitance, will be reduced because the same signal is present on both the plate and its suppressor thus cancelling out the capacitance between them. This improves the high freq response of the amp.

  • Point A + B + D allows a very unique way of valve operation. Previously I said that valves were depletion devices in that they normally conduct and the grid can only back them off. Thus, as the signal momentarily decreases the bias, the valve passes more current but only to the limit set by the attractor element, ie plate or screen. But this cct, by placing the signal on both the grid and the screen, allows the valve to start acting like an enhancement device. As the signal momentarily swings in a positive direction, it decreases the grid bias, which will let the plate current increase, and it simultaneously increases the screen voltage which will make the amount of available plate current increase even further. Thus the valve will never run out of plate current as far as the signal is concerned. This extra increase will continue all the way till the grid reaches zero bias.
 
This is all rather brilliant when you consider it was worked out under wartime conditions with all the pressure that entailed.

 
For those of you who thought my description of the Triodyne was hard to follow here are Mr Crawley's own words and his original diagram. Good Luck.

James Crawley: Triodyne
James Crawley: Triodyne Description

There a couple of possible reasons why this cct fell by the wayside. First, the miniature double-valves only have 9 pins. Secondly, the second valve must operate with its screen voltage lower than the plate voltage. The first meant that the Pentodes in Triode/Pentodes do not have enough pins to allow the Pentode to have a seperate suppressor connection, unless two other elements are commoned as in the ECL80. This meant that single valves have to be used which increases the valve and socket count. The second means that Beam Power tubes: Tetrodes and Pentodes, which normally have the screen running at the +HT, will not work at full power if you have to drop the screen voltage.

Notwithstanding the above, maybe this circuit should be given a run for its money again by people who are not afraid to try something different. And that's you, isn't it?
Menu  



If it really is you then here is a starter cct based on the Triodyne principle.

Triodyne amplifier basic diagram.

Beside the Triodyne circuitry, there are one or three unusual things to note in this cct:-
  1. Feedback is taken from the output valve plates and fed back to the screens of the input valves in an attenuated form.
  2. Adjustable overall feedback from the voice coil is applied to the grid of the second input valve.
  3. Last and probably the one to raise most eyebrows is the network around the cathodes of the input valves.

In more Detail:-
  1. From the original triodyne circuit you can see that the first pentode uses only the suppressor in an unusual manner. This leaves the screen-grid and the control-grid available for other duties.

    As I mentioned above, the Triode has a flaw in its operation which, funnily enough, makes it sound better. It is the fact that, when the grid requests more current, the plate voltage falls which means that, as the plate is the only available attractor, this drop in plate voltage reduces the amount of current available. This is a form of negative feedback and reduces some of the inherent distortion which comes with a valve.

    In a Tetrode/Pentode the plate is at least 5 times to 10 times the distance from the cathode that it is in a Triode. This means it would be useless as the main electron attractor. The screen-grid takes over this role and works well, as its voltage is normally static, regardless of the plate current. Any small changes in the screen-current are supplied from its bypass cap thus its voltage is fairly stable. Unfortunately this means that the negative feedback inherent in the Triode is missing in the Tetrode/Pentode and so its internal distortion is not reduced in any way.

    There are a couple of ways to make a Tetrode/Pentode act like a Triode:-

    1. Just join the screen-grid to the plate. This is about as close to a Triode you can get. However the advantage of the inherent negative feedback also comes with the Triode's disadvantage of limited output voltage swing when high currents are involved.

    2. Another way, which allows you to fine tune just how far you want to go towards a Triode, is to partially unbypass the screen-grid resistor.

      If you take the normal screen-grid resistor and replace it with a potentiomener of the same value nothing has changed. Now move the bypass cap from the screen-grid pin connection and connect it to the pot wiper instead. With the wiper at the screen-grid end, the valve acts like a Tetrode/Pentode. With the wiper at the HT end the valve now has an unbypassed screen-grid and acts like a Triode. Any increase in plate current, caused by a positive going grid, will also have an accompanying increase in screen current and, as there is no bypass cap to supply it, this increase will have to come through the resistor causing the screen voltage to drop. This does the same thing, as in a Triode, reducing the current available and so produces a form of negative feedback. However this time you can adjust how much you want by moving the wiper towards the screen end.

      As the screen is now unbypassed it is available to be used as an input and so I have fed back some of the output valve plate voltage suitably attenuated. Of course the amp maybe so good that it wont need this extra feedback. :-)

  2. The second input valve has a control-grid doing nothing so I've fed back the voice coil voltage through a pot to make it fully adjustable. This could be used to reduce any problems with the output transformer. Or, if you are one of those people who dont like global feedback, you can reduce the feedback to zero with a twist of your wrist.

  3. You all know about the long tail pair where the cathodes are commoned and taken to a large negative voltage via a high value resistor. This basically provides the pair of cathodes with a constant current. Thus, if one valve wants more current, it has to pinch it from the other one in the pair. This guarantees a truely balanced signal output from the two plates.

    What you may not know is that Alan Blumlein first used long tails on single valve stages before he applied it to the cathode coupled pair.

    Back in the early days, an original primitive valve amplifier had the filaments /cathode grounded and the grid was fed with a fixed negative bias voltage. The circuits worked quite happily at first but as the valve aged its bias requirements changed and the amount of fixed bias became incorrect causing the plate voltage to drift quite dramatically with valve ageing.

    The method used to over come this was automatic biasing in which the valve current, passing through the cathode resistor, produces its own bias. Thus, as the tube characteristics change with age, the bias adjusts automatically to allow for it.

    If you analyse the operation of automatic bias, it really just reduces the gain of the stage at DC while the bypass cap maintains the AC gain at maximum. If the gain of the stage is reduced then its sensitivity to bias change is reduced but it is not removed completely.

    Alan Blumlein set out to try and remove the stage's sensitivity to internal characteristic bias change once and for all. He reasoned that if the bias voltage requirement change could be made a small percentage of the voltage applied to the automatic bias resistor then the stage would effectively be insensitive to any bias change at all. He increased the size of the auto bias resistor and fed it from a large (in relation to the valves actual bias voltage) negative voltage making, what we would call today, a constant current generator. The valve basically sits on a constant current generator which is set to the current required from the valve. Regardless of what the valve does as it ages, it must adjust itself to pass the current fed into its cathode. Once the stage is wired up this way, its gain effectively drops to zero, hence it does not pass any bias change on to the output cct. Unfortunately the stage will not pass any change in grid voltage either so the stage is basically usless as it is. However, when you put the normal bypass capacitor on the cathode, the AC gain comes back to normal while the DC gain stays at zero. When the grid tells the valve to draw more current, the extra current must come from the bypass cap, as the constant current generator in the cathode cct will not provide any increase. therefore any signal current will have to come from the cap whose size will set the low freq response.

    So getting back to the Triodyne amp, the two input valves have effectively got constant generators in each cathode cct. This sets the cathode current and thus the valves do not need a balance control as they must always pass the required current until they are too weak to do so. The caps, connecting the two cathodes, allow the signal current to flow to and fro between the valves thus causing them to act as a longtailed pair, but only to AC. If you are unhappy with electrolytics in the signal path, you can either try to find suitable plastic caps or put the cct back to a normal longtailed pair but you will have to fit a balance control.

Fig 3 shows the practical circuit.

Triodyne amplifier circuit diagram.

You will notice that the output valves are actually video output valves. The Triodyne cct requires that the output valves must run with their screen voltages much lower than their plates. While Beam Tetrodes will run this way they normally run with the plate and screen running off the same HT whereas all the video valves, I looked up, run their screens lower. This would seem to say that the video output pentodes would suit the Triodyne cct better. Some possible valves to try are 12BV7, 12BY7, 6CL6, 6CH6, 6CK6, 6CW5, 6AG7.

The majority of the video o/p pentodes are not Beam tubes and hence do not have their grid wires synchronised and therefore have higher screen currents. They have one thing going for them though in that their screen currents dont vary as much with plate current change as the the Beam tubes do. Of course there's nothing to stop you trying Beam tubes as the output valves. Some suitable tubes to try would be 12A6, 6V6GT, 6Y6GT, 6BW6, 6CM6, 6AQ5, 6BQ5, 6AM5.

You have probably spotted the main thing that most likely sent the Triodyne cct by the wayside. The plate load resistors have to pass the screen currents of the following valve and therefore are much lower in value than they would be thus each stage gain is much lower. Add to that all the neg feedback built into the Triodyne cct and the final amp will hardly have much overall gain at all, considering the number of valves in it.

It is also highly likely that all the negative feedback will remove the distortions, which create the "valve sound", completely and so make a valve amp sound like a transistor amp. Perish the thought.

Menu


I realize that some people just dont like capacitors in amplifiers at all so here is a cct for you.

DC Coupled valve amplifier circuit diagram.
The first and most obvious thing in this cct is that there are no caps anywhere in the signal path. That should keep the purists happy. However, it also means that the powersupply has to be rather fancy. The powersupply problems have been minimised somewhat by using a string of 5 watt zeners to provide the multiple voltages necessary. All the zeners dont need to be 5 watt types but I've made them that way so as to be able to easily change the voltages on the valves without overloading any of the zeners. In order to limit the powersupply voltage to a usable value, the other necessity is to find valves which will work satisfactorily with 50v across them.

The initial HT is 370v. 20v is dropped across the two 91 ohm current limiting resistors, letting the zeners run at 350v. The o/p valves run at 250v. The driver valves run with 50v between their plates and cathodes as do the voltage amps. The -33v supply need only handle a couple of milliamps.

The current limiting resistor has been split in half to provide the best place to put the HT bypass caps. The impedance at the centre is actually higher than at either end so the caps will work best at this point.

All the valve pairs are long tailed which should help the balancing. The voltage amps have a control in their cathodes to set the o/p valve current balance.

The current in the driver valves is set high so as to be able to drive the input capacity of the o/p valves.

Local feedback is applied in two ways:-
  1. Each valve section has part of its cathode resistor unbypassed.
  2. Each valve plate provides a neg feedback signal to the cathode of the previous valve.
Variable global feedback is provided from the voice coil to the spare control-grid of the second voltage amp. This can be used to partially alleviate any problems with the o/p transformer or, if you dont like global feedback, it can be removed with the twist of your wrist.

Any high freq oscillation which may occur can be stopped by the appropiate grid stoppers.

A 12AX7 should work in place of the 12AY7. A 6DJ8/E88CC/6922 should work in place of the 6CW7/ECC84.

This would be a good time to point out that the circuits in Figs 3 & 4 are theoretical only but, with a little bit of component value fine tuning, I see no reason why they should not work as described when built.
Menu


 

While reading the 1947 Aussie Radio Handbook I mentioned above, I came across a rather interesting circuit for an amplifier of about 10 watts output.

High Quality valve amplifier circuit diagram with a difference.
There are a few unusual aspects to this cct.
  1. The output valves are wired as cathode followers.
  2. The audio output transformer is actually a mains power transformer.
  3. The output valve cathodes are shorted together.
  4. The cct has base and treble controls of the old style feedback variety.
  5. The component values are of the old style.
In more detail:-
  1. Apparently, according to the article, by using cathode followers, the output impedance of the valve part of the amp is very low (around 500 ohms). This then allows a poor quality transformer to be used, as the valves can provide enough drive to overwhelm whatever poor characteristics the transformer has.

  2. The author said that we will be surprised at just how good this amplifier sounds.

  3. The original published cct had the output valve cathodes drawn as though they were heaters and the unused ends were shorted together. I think this was a drawing error. I have shown this connection as a dotted line in case the amp is working in an unusual fashion. I leave it up to you to decide whether to short them or not.

  4. The base and treble controls work by modifying the feedback around the first two valves. They will probably have only a limited effect.

  5. The component values are those which were available in 1947 and you will have to change them slightly to fit the modern value system.

The cct otherwise is quite straight forward with two voltage amplifiers having adjustable neg feedback around them followed by a concertina push-pull driver feeding two 6L6s running as cathode followers on a fixed negative 20v bias (set with the o/p valves not plugged in). It is interesting to note that, when Tetrode/Pentodes are wired up as cathode followers, they actually run as Triodes and therefore should have reduced internal distortion. Adjustable global feedback is also provided from the voice coil to the first voltage amp.

Even though the published cct had only one 6.3v heater winding, it might be a good idea to run the 6L6 heaters on their own floating 6v winding as the cathodes will be swinging close to if not above their heater/cathode breakdown limit.

One thing not mentioned in the original article was the resistance of the 385v windings and its effect on the bias of the o/p valves. Each winding will act like a self biasing cathode resistor and produce some voltage drop which will bias the valves off slightly. One would hope that the fixed bias of -20v takes this into consideration.

There could be some confusion about how much power this amp can deliver to the load. In the original article the author stated the o/p was 12 watts, however he didn't show how much HT there was. Neither did he state the load impedance or which taps on the o/p power-tranny he used.

The original power-supply has a 350-0-350 tranny with a 5Y3 rectifier. The filter is a 20uF followed by a 1k ohm choke and another 20uF. With the amp drawing around 110mA, the voltage on the first cap would be about 380v. This 110mA load-current will cause a drop of 110v across the 1k choke. This leaves us with 270v feeding the amp. The o/p valves are cathode followers so they have no gain. The driver is of the concertina variety so if we allow 50v across the valve then we have 110v across each of the load resistors. A pk to pk swing of 109v gives you 38.5v rms. Thus the o/p 385v winding gets 1/10 of its normal voltage. If we use all the low voltage secondaries plus the 40v tapping on the 240v winding in series, we have a 50v winding which will produce 1/10 the o/p, that is, 5v rms. If we feed this 5v rms to a 2 ohm load, we get 12 watts. The wire in the 240v winding is not very thick so using it to feed a low impedance loud speaker is not particularly efficient. I think it would be better to feed in more volts and use a lower voltage tap or maybe even using a different o/p power-transformer, say a 285v-0-285v one.

The big bugbear with concertina drivers is the difficulty in getting enough drive to really push the o/p valves. The cct values are as published in the handbook but I think that the cct would work better if the first 6J5 was replace with another 6SN7. The first 6SN7 could then provide both voltage amps and the second 6SN7 coud be used as a long tailed pair to provide much more drive voltage to the cathode followers. Of course, if you can get the o/p tranny to reflect a low enough impedance, the cct may work quite satisfactorily as it is. You need 9 volts RMS across an 8 ohm voice coil to produce 10 watts. That's 25 volts peak to peak.

One thing to remember. In 1947 the voice coils in use were 2.3 ohms, 8 ohms and 12.5 ohms, or so a Ferguson ad in the 1947 aussie Radio Handbook says. By the time the late 1950s rolled round these impedances had disappeared and 8 ohm loudspeakers were very thin on the ground. The normal impedances in use then were 2 ohms, 3.5 ohms and 15 ohms. The 2 ohm speakers were for replacements in old wireless sets. The 15 ohm ones were mainly for HiFi speakers. The vast majority of the speakers in use were 3.5 ohms. You need 6 volts RMS across a 3.5 ohm voice coil to produce 10 watts. That's 17 volts peak to peak.

By the 1970s the 2 ohm speakers had disappeared, 3.5 ohm ones were beginning to fade and 8 ohm ones started making a comeback, most likely pushed along by the appearance of transistor amplifiers. Now, due to the rise of 12v car stereos, 4 ohm speakers are making a big comeback. You need 6.5 volts RMS across a 4 ohm voice coil to produce 10 watts. That's 18 volts peak to peak.

According to the author, the full 240v primary winding of the "o/p tranny" could be used to drive a 500 ohm line.

Another thing to remember. Power-transformers came in a variety of voltages. 385v-0-385v, 350v-0-350v, 285v-0-285v, 250v-0-250v, 225v-0-225v, 150v-0-150v. Low voltage secondaries could be 1.25v-0-1.25v, 2.5v, 2.5v-0-2.5v, 3.15v-0-3.15v, 5v, 6.3v, 6.3v-0-6.3v. Writing these another way 2.5v, 2.5v centretapped, 5v, 5v centretapped, 6.3v, 6.3v centretapped.

With such a wide variety of "output transformers" this circuit could match just about any thing you wanted it to.
Menu


For those of you who are not too happy about using a power transformer as an audio output transformer, here is a follow up article about this unusual subject. It's from the same 1947 aussie Radio Handbook.

Power Transformers for Impedance Matching.
A power transformer used in 4 or 5 valve radio sets can be used as an output or matching transformer. Due to the presence of four windings there are many useful matching ratios. A few typical examples will illustrate this.

Typical old style Power Transformer.
Assume a transformer with a 240v primary winding and three secondaries, 385v-0-385v, 6.3v, 5v. Such a transformer is shown in Fig.6. The open circuit voltages of the 5 and 6.3 volts windings would be closer to 5.25 and 6.5 volts.

Let us now consider this transformer with C & E as the primary terminals and F & J as the secondary, G being connected to H. The voltage ratio is then 2 x 385/(5.25 + 6.5), which equals 65.5. The square of this figure, which is 4300, gives the matching ratio. This ratio allows a satisfactory match between the voice coil of a small speaker (3.7 ohms) and the 15000 ohms plate load of a battery set output tube, because 15000/3.7 = 4050, which is near enough to 4300.

Another case is an amplifier with two 6L6 valves in a class AB1 output stage. The plate load is 6600 ohms. Six speakers, each with 2.3 ohm voice coil impedance, are to be matched. The transformer in the above example can be used as follows: the valve plates go to C & E and the +HT to D, while G is connected to H, and the line to the speakers terminates at F & J. The speakers are arranged in 3 pairs in parallel, each pair consisting of two speakers in series. Each pair has then an impedance of 4.6 ohm and all three pairs in parallel, that is 4.6/3, equals 1.53 ohms. 6600ohms/1.53ohms = 4300, which is exactly the right matching ratio, as shown above.

Let us now try another connection. Let A & B be the primary terminals and H & J the secondary, the other windings remaining unused. The voltage ratio (or as we deal with open circuit voltages, the turns ratio) is 240/5.25, which equals 46. The matching ratio is then 46 x 46 = 2100. This enables us to match a 5000 ohm load (say a 6V6) to a 2.3 ohm speaker because 5000/2.3 = 2170 or near enough to 2100.

Another example is the matching of a 5000 or 6000 ohm load to a 500 or 600 ohm line. The matching ratio required is 10 and the voltage ratio (square-root of 10) is 3.16. The ratio of CE/AB, (2 x 385)/240, which equals 3.2, is near enough for this requirement.

Many other combinations are possible and often handy to meet a special requirement. It is realised, however, that a matching transformer must fulfil other requirements too. It is beyond the scope of this note to go into details, but the following can be stated from practical experience. No serious distortion due to high flux density will be heard, if at 50 cps the RMS voltage on each winding is approximately half the value this winding would show in its normal application as a power transformer, and if on and above 100 cps, the voltage on each winding does not exceed the value this winding would show in its normal use as a power transformer. The efficiency and regulation will be slightly worse than with a special transformer designed for the purpose, but often better than cheap output transformers. The same can be said with regard to leakage inductance which will reduce the high frequency response.

The biggest disadvantage is the reduction of primary inductance by the DC flowing through the winding in the case of single ended output stages (eg. 1 & 3). But even this is usually tolerable because normally power transformers have a longer iron path than output transformers, thereby making the lack of an air gap less pronounced. The low flux density inductance of the 240v winding is usually in the order of 3 to 5 Hy. and that of the high tension winding 30 to 50 Hy.
Menu  

Well, now you know. Unfortunately, nowadays, valve power trannys are rather thin on the ground and very expensive when you do find one. However, there is possibly another avenue to explore. Various electronic stores sell very nice toroidal power transformers and some of them have twin 120v primary windings. They come in different power ratings and secondary voltages.

Toroid power transformers used for impedance matching.
 
The above table shows the variety of toroidal transformers available and the possible impedance matching they may provide. Placing the unused secondary in parallel with the used one may increase the efficiency slightly.

Some overseas parts suppliers can supply very nice normal style EI transformers with twin 115v primaries and a selection of different secondary voltages. These are different to the toroids above and so provide even more possible impedance matching solutions.

EI power transformers used for impedance matching.
 
Brands include such names as Hammond and Tamura. The EI magnetic path is a bit longer than the toroids and so can withstand slightly more out of balance currents in the two primaries.

Well, now that you have a selection of matching transformers to choose from, let's look at a few circuits which use them.

Valve amplifier using modern power transformer as output.
Here is a circuit to get you started, theoretically speaking.

A few things to note in this cct:-

  1. The driver is a long tailed pair so as to generate more drive voltage swing. The long tail occurs because the grids are running with elevated bias. This is provided by partial DC coupling from the previous valve plate. A balance pot in the driver plate cct allows the AC drive to the o/p valves to be equalised.

  2. The currents in the output valves need to be matched pretty closely. The reason is that toroid trannys have a very short magnetising path and cannot stand as much out of balance DC current as a conventional power tranny can. The valve currents can be monitored via the voltage drops across the two 1ohm resistors. You may have to fit a balance pot to the fixed bias supply and feed the grids separately, as in Fig 7a, to accurately match the currents if the valve's internal current matching is too far out.

    Valve amplifier using modern power transformer as output.
     
  3. The resistance of the 120v windings will have to be taken into account when calculating the amount of fixed bias to apply. The 120v windings on one brand of 10 watt toroid have a resistance of 250ohms each. If the o/p valve current is 40mA then each winding will have 10v across it and the valve will already be biassed at -10v before you apply any external fixed bias. In this case you would only have to apply -6v externally. One would assume that the higher powered the toroid is, the lower the winding resistance would be. This resistance will increase slightly as the transformer heats up so that will have to be taken into consideration as well. You will have to make sure that the voltage drop, at the operating current, is less than the bias required on the o/p valves. In fact it would have to be quite a bit less if the valves are supposed to be working on a fixed bias. In the worst possible case you may have to apply a fixed positive bias to get the o/p valves to draw the required current.

  4. If you love 6SN7s and are so inclined, the 6BX7 could be replaced by three 6SN7GTA in parallel.

With these toroidal trannys having twin primaries and twin secondaries the circuitry possible is only limited by your imagination. You can even make a circuit which will hopefully keep the "raincoat" brigade happy just in case they are starting to feel left out.

Valve amplifier using modern power transformer as output in a Macintosh arrangement.
Here the twin secondaries provide local feedback to each of the o/p valve cathodes. The ac drive balance is adjustable. Similarly a balance pot is provided to equalise the o/p currents which can be monitored via the voltage drops across the 1ohm resistors.

Toroidal trannys have very good cores, are close wound and have good coupling between primary and secondary. They should work quite well as push-pull output transformers. The new style EI power trannys use quite good steel in their laminations and so are better than you would think. Why dont you give either of the two types a try?
Menu  

When you are building a class A amplifier and are not sure which valves are biassed properly, you can check this rather easily if you can lay your hands on a few milliammeters. You simply wire a milliameter in the plate cct of each valve, as per the following diagram. A signal is then fed to the input and the gain control increased. Take note of which meter varies first. This is the stage which will limit the fidelity.

Simple way to check class A amplifier bias.
 
If the meter kicks up, either the grid bias is excessive or the plate voltage is too low. However, if the meter gives a downward kick, the bias is insufficient or the plate voltage is too high. Adjust the bias and/or voltage until the gain can be further increased without meter movement up to the maximum o/p of the amp. As stated at the beginning this method only applies to class A amplifiers in which the plate current normally remains steady.
Menu  

Well, from a valve lover's point of view, this is where the page drops down a class level and therefore, shock-horror, here are a series of solid-state amps designed by a valve lover using Solid-State's closest thing to a valve, the Mosfet. You'll recognise the ccts as a rehashed version of the 1947 6L6 cct above. The thing to remember is that the triode Mosfets will never sound as good as Triode vacuum tubes because the Mosfets have characteristics more like a Pentode than a Triode. They have no internal negative feedback like a Triode Valve has.

Power-Fet amplifier using modern power transformer as output.
Here is a very theoretical circuit to get you started.


The one thing that makes these Power-Fets different to valves is that they are enhancement devices whereas valves are depletion devices. This makes the cct somewhat difficult to design as they are halfway between voltage depletion operated valves and current enhancement operated transistors. I've cheated slightly in this cct by turning all the Power-Fets on permanently and by using the ageold valve circuitry to bias them off again. This way the valve circuit will still work, in that it thinks all the devices are depletion operated and the autobias will operate as it normally does.

A few points to notice:-
  1. Power-Fets usually need about +4v to be applied to the Gate with respect to the Source to turn them on. This is done by connecting the ground return cct to -10v. This effectively put +10v on the gates. The autobias resistors in each Source cct will reduce this to about +4v. The 6SN7s used in the original cct needed about -6 to -8v so the autobias resistors should theoretically be the right value.
  2. The long tailed pair gets its long tail by DC coupling its Gates to the previous stage. This lifts the Gates to about +30v and thus the Sources to +26v. This in conjunction with the -10v produces about 36v across the tail resistor setting the tail current to about 7mA.
  3. The output Power-Fets are fed with a fixed negative bias of approx -6v. This needs to be adjustable so that the current in the o/p fets can be set to 60mA. This can be measured by the voltage drop across the 1 ohm resistor in each Drain. You should be able to find two Fets matched closely enough out of the six.
  4. One final thing. Depending on the current rating of the powerfets you use, their input capacitance will be a lot higher than a valve. This means that the gate circuit resistors may have to be reduced and the coupling capacitors increased to improve the frequency response. We are lucky that the output stage is actually twin Cathode Followers as they tend to cover up this higher capacitance.

    This capacitance problem is worse than I first thought. Not only do power FETs have high input capacitance but their Miller feedback capacitance is high as well. Unfortunately the Miller feedback capacitance will be multiplied by the gain of the stage and so become very large. On top of this the same FET made by different manufacturers have different capacitances. See table below.

    Mosfet Manufacturer Input Cap. Output Cap. Miller Feedback Cap.
    IRF710 International Rectifier 170pf 34pf 6.3pf
    IRF710 Fairchild 135pf 35pf 8pf
    IRF730 International Rectifier 700pf 170pf 64pf
    IRF730 SGS-Thomson 700pf 140pf 13pf
    IRF820 International Rectifier 360pf 92pf 37pf
    IRF820 SGS-Thomson 315pf 52pf 7.7pf
     
    From the table you can see that the IRF710 would be the FET to use, or if you cant get it, the SGS-Thomson IRF820 would be next best, though they both still have capacitances much higher than a valve. Luckily, by using Source Followers in the output stage, the effect of these high capacitances is pretty much neutralised. However the other stages will need to be looked at more closely with respect to the increased capacitance.

As the above cct is a very theoretically one you will have to fiddle with some of the resistor values and you may even have to put smaller Fets in the various positions as the Power-Fets may not work too well in the milliamp range.

Below is the an updated version of the previous solid-state amp which uses Mosfets.

Power-Fet amplifier using modern power transformer as output.
 
This theoretical circuit is a bit more practical and allows you to balance the currents and drives to the output FETs.
A more important change is the fact that here the first two stages are now using the age old cascode circuit which overcomes the capacitance problem somewhat. You could replace the first six IRF820s with six smaller low capacitance MOSFETs such as the 450v Zetex ZVN0545a or Supertex VN4012L. These only have 1 watt or less power dissipation but should be ok for the low current stages. As the input capacitance of these smaller mosfets is right on the edge of being just small enough, you might even get away with using only four of them as in fig 11 below.

Power-Fet amplifier using modern power transformer as output.
 
Here are the characteristics of a few fets, some of which may be useful.
 
Mosfet   Manufacturer   Input Cap. Output Cap. Miller Feedback Cap. Maximum Voltage Drain  Current  Power Dissipation Impedance Gate     Voltage 
BUK456 Philips 1000pf 80pf 30pf 800v 4 amp 125 watt 3 ohm 2v to 4v
BUZ41 Harris 1500pf 110pf 40pf 500v 4.5 amp 75 watt 1.5 ohm 2v to 4v
BUZ60 Harris 1500pf 120pf 35pf 400v 5.5 amp 75 watt 1 ohm 2v to 4v
BUZ74 Siemens 450pf 50pf 20pf 500v 2.4 amp 40 watt 3 ohm 2v to 4v
BUZ77 Siemens 460pf 55pf 20pf 600v 2.7 amp 75 watt 4 ohm 2v to 4v
BUZ78 Siemens 320pf 40pf 20pf 800v 1.5 amp 40 watt 8 ohm 2v to 4v
BUZ80 SGS-Thomson 650pf 82pf 28pf 800v 3.4 amp 100 watt 4 ohm 2v to 4v
BUZ90a Siemens 780pf 110pf 40pf 600v 4 amp 75 watt 2 ohm 2v to 4v
IRF710 Fairchild 135pf 35pf 8pf 400v 2 amp 36 watt 3.6 ohm 2v to 4v
IRF730 SGS-Thomson 700pf 140pf 13pf 400v 5.5 amp 100 watt 1 ohm 2v to 4v
IRF820 SGS-Thomson 315pf 52pf 7.7pf 500v 4 amp 80 watt 3 ohm 2v to 4v
IRF830 SGS-Thomson 610pf 120pf 10pf 500v 4.5 amp 100 watt 1.5 ohm 2v to 4v
ZVN0545a Zetex 70pf 10pf 4pf 450v 90 mA 700 mW 50 ohm 1v to 3v
VN4012L Supertex 110pf 30pf 10pf 400v 150 mA 1 watt 12 ohm 0.6v - 1.8v
Valve Manufacturer Input Cap. Output Cap. Miller Feedback Cap. Anode Voltage Anode Current Power Dissipation Grid Cutoff Heater Power
957 RCA 0.3pf 0.7pf 1.2pf 135v 2 mA 300 mW -10v 62.5 mW
12AX7 General Electric 1.6pf 0.5pf 1.7pf 330v 3 mA 1.2 W -4.1v 0.945 W
6BX7 General Electric 4.5pf 1.2pf 4pf 500v 60 mA 10 W -60v 4.725 W
6AS7 General Electric 6.5pf 2.2pf 7.5pf 250v 125 mA 13 W -185v 7.875 W
Valve info for Triode single sections only.
 
There's a few interesting things which can be gleaned from this list.
1. First is that all the powerfets need approx the same gate-source voltage to conduct 1 mA, regardless of their current rating or manufacturer. However it must be noted that they all need a gate-souce voltage of 10v to allow them to pass their full rated current.
2. Second, the wide two-to-one gate threshold voltage range would make it very difficult to get a closely matched pair.
3. Third, most of the high current powerfets have horrendously high capacitances and only the BUZ78, IRF710 and the IRF820 come even close to being useful in audio amps. (There are others but getting them in a non-surface-mount case is almost impossible.)
4. Fourth, Valves, no matter how big or small, being naturally high impedance devices, have very low capacitances. Solid state devices are inherently low impedance devices and so have high capacitances. This makes it difficult to substitute one for the other.

Unfortunately all powerfets are designed for switching which means their transfer curves are not particularly linear. However all curves look linear if you use a small enough portion, so keep the current swing a very small percentage of the total allowable current and they should be linear enough for audio. They'll never be as good as valves, though.  

Happy Experimenting.
 


  ©  Gary Yates   Locofonic Recordings Australia  
This page first written 20-3-2006 last updated 3-9-2007.