Vacuum Tubes Part
Note: Part 1 is here, part 2 is here, part 3 is here, part 4 is here and part 5 is here.
began this series with a basic tutorial on electricity, then moved on through
passive components to tubes. To illustrate a quick and easy way to build a tube
design, I suggested beginning with the values in the back of the RCA Receiving
Tube Manual, then began exploring variations on a theme, building first a
single-ended preamp, then mirror-imaging it into a balanced circuit, describing
some of the choices made at each fork in the road, which is something that
frequently gets left out. There's a strong tendency for DIY articles to present
a finished design as a monolithic whole that sprang, unalterable, from its
creator's mind. Well, it just ain't so. There are lots of ways that you can go
with a circuit and it's a rare day when you find yourself with one, and only
one, solution to a given design problem... excuse me... "challenge," as we're
supposed to say these days.
The circuit, as we left it, consisted of a differential
front-end, analogous to the single-ended front end in Part 2. It's now time to
complete the circuit by adding cathode followers as an output stage so we'll
have a low output impedance and be able to interface with an amplifier,
minimizing the effect of the impedance of the interconnects. The obvious thing
to do is to simply add two cathode followers like the one in Part 3. And,
indeed, that works and works well.
Most articles would stop here, spend a paragraph or two
thanking the audience for coming and asking that you tip your waitress well (did
you ever notice that performers are generous as long as it's your
money?) and other such metaphorical hand-dusting, preparatory to leaving the
stage. Yeah, we could do that... we could... but that would leave me with an
unscratched itch that hopefully bothers a few of you, as well.
Supposing you wanted an output stage that could both source
and sink current? Remember that resistor-biased cathode followers can only do
the current thing one way. When the signal goes the other way, you're left
facing a resistor — a passive component. Again, I want to stress that the vast
majority of circuits out there have exactly this sort of cathode follower and
they sound fantastic. There's no reason to suddenly think that they somehow
sound less wonderful than they did yesterday. But still...what if...?
Have no fear. It can be done. In fact, the very fact that we're dealing with a cathode follower output gives us more options than we had with the front end. Back in Part 3, I glossed over the idea that we could use a current source to bias the cathode follower, then went on with the standard resistor bias option. Let's circle back and pick up that option, just for fun. The front end got a solid state current source because we were looking at a mere 3 to 4 Volts under the cathodes of the differential. Now that we're looking at cathode followers running under a 300V rail, tubes suddenly become a lot more practical. The cathode of the follower will be at about 150V or so, more than enough room to justify sticking in a tube. The neat thing about tubes and JFETs is that they can be biased with a simple resistor. In fact, the tube current source circuit looks much the same as the JFET current source we covered in the previous article.
Note that the resistor is back to 1300 Ohm, rather than the
620 Ohm value we used in the differential. That's because we're only biasing a
single tube, just like in the beginning when we were looking at a single-ended
front end. And, like then, we'd expect the tube to bias somewhere in the range
of 3 to 4 mA, depending on the individual tube. Incidentally, unless you've gone
to the trouble to match your tubes, I'd suggest using one tube for the front end
differential and another for the output stages, rather than using one tube for
the left half of the circuit (i.e. half the differential and its associated
cathode follower) and the other tube for the right half of the circuit. Why?
Because the tubes are matched internally; they're not necessarily going to match
as well from tube to tube. Matched performance is crucial for the proper
operation of a differential and certainly won't hurt for the output. So, what
happens now that we've got tube current sources?
The bottom tube, the current source, sets the operating
current. Since, by definition, the two tubes are running at the same current
(they're running in series and all components in a series circuit carry the same
current), the cathode follower now shares the same bias.
Let's say, just for symmetry, that you wanted to go with solid
state current sources, to match the one used to bias the differential. It can be
done, but it will take a bit of fine tuning to make it work out. The current
source's output device (the MPSA18 in the front end) will need to be exchanged
for something a little beefier.
If there's one thing solid state excels at, it's current.
Voltage, on the other hand, can be a bit of a problem, and the MPSA18 used last
time, while a worthy part, is only rated for 45V. That's okay when you're only
going to ask it to cover a few volts, but we'll need a part that can take 150V
without flinching. What about the MPSA42 I used for the "brain" of the current
source last time? That part is rated for 300V — should be a good candidate for
the job, right? Unfortunately, it's not such a good choice. Here's why. First
off, it's never a good idea to run a part — any part — right at its limit.
But, what's the problem, asks the fellow in the second row, it's rated for 300V
and you're only asking it to take 150V should work just fine. In fact,
everything probably would be okay, as long as the circuit was idling, which
would leave the MPSA42's collector at around 150V. However, a circuit that is at
idle isn't going to do us any good and once the output starts bouncing, the
voltage will increase as the output signal swings positive, potentially exposing
the transistor to something much closer to its rated voltage. You can make the
argument that any reasonable amplifier circuit will only need a few volts at the
input to achieve full output, but what if you run across one of the so-called "hot
followers" where you'll need the preamp to generate the entire voltage gain?
Then there's the fact that tube circuits can do funny things at warm-up or the
possibility that you could have some sort of unexpected signal at the input,
like a shorted cable, and things could get nasty. Plus it's just good practice
to leave some elbow room.
The other fly in the ointment is that the current we've been
biasing at, around 3 to 4 mA, times 150V will leave us with power dissipation of
a little over 0.5W. In theory, the transistor can take the heat as long as you
use a heat sink. But it would be better still to choose a device with a bigger
case; the TO-92 case is just not that good at getting rid of heat. Consider also
that the transistor will be used in close proximity to tubes, which generate
quite a bit of heat on their own, so the ambient temperature in the circuit will
be higher than normal, which in turn will make the transistor even harder to
Fortunately, the TO-220 case is ideal for what we need and if
we open the door to MOSFETs, we'll find a number of perfectly good candidates.
We can use the same general topology as the previous current source, replacing
the MPSA18 with an IRFBE30. There's nothing particularly special or unique about
this part, other than the fact that it's rated for a healthy 800V — it just
happens to be a part I've used in the past and am familiar with. Feel free to
substitute any N-ch MOSFET you like. I'd suggest a part rated for 500V or above,
just to be on the safe side. As of this writing the IRFBE30 is available from
either Mouser or Digikey for $1.59 apiece, so they won't break the bank. While
you're at it, order a small heat sink, just to help the part stay a little
cooler, although the TO-220 case can dissipate 0.5W comfortably.
Whether you go with tube or solid state current sources, the circuit will perform well. It will accept either balanced or unbalanced inputs and — with a current source biasing the differential — will always give balanced output, even with a single-ended input. Now I'll toss out some more ideas to see if I can get your creative juices flowing.
It is well known that there are phase problems with some recordings. The effects are subtle, but audible. The differential version of the circuit gives you the option of choosing your phase according to the needs of the recording. If your system is single-ended, you can install a switch at the output that allows you to select either side of the circuit. One is inverted. The other is non-inverted. Choose the one that gives the bass drum more "oomph." The same trick works if your system happens to be balanced. Just swap both outputs, one for the other.
Here's an idea that should appeal to those who want to explore negative feedback. The voltage gain of a circuit is determined, in part, by the size of the load resistor — the larger the resistor, the more gain you can get. However, there are practical limits as to how large a resistor you can use. There is a way to get a nearly infinite load...use a current source. Well, okay, if you want to be pedantic about it, it is technically a current sink, rather than a source, given that electrons flow into it rather than out, but still... it is an elegant way to approach the theoretical amount of gain available from a given tube. That will, in turn, give you more gain to burn for more negative feedback.
I'm writing this before Christmas and all the usual
traditional tales are in the air. As with Dickens's Ghost, the time for this
series is drawing to a close. Before the end, however, I'd like to ask you to be
careful when working with tube circuits. The voltages are high enough to hurt
you. With that caution out of the way, I urge you to have fun, learn lots, and
enjoy the music.
Oh, and remember to tip your waitress and waiters well this