# Tubes for Synthesizers

Hi all!

It’d only be appropriate to write a tutorial on tube theory as I understand it for synthesizer enthusiasts. I’ll add more circuits as I go. Also, I say “tube” because I prefer the sound of it to “valve”, even though I’m Australian

IMPORTANT! Tubes are usually always designed to operate at a high voltage, and it’s a lot easier to run them off mains electricity. I design my circuits around a 250V DC power supply, which gives quite a bite if you touch it! I initially had them running at 150V, and they could work as low as 75V. The problem with lower voltages is you’ll get less amplification out of the circuit, as I’ll discuss.

Power:
The first thing you need to do to get a tube working is apply a 6.3V AC or DC voltage to the filament. This was known as the “A” power supply, going back to antiquity when the only electricity available was from batteries. The A battery powered the filaments, the B battery was a high voltage and powered the positive side of the circuit, and the C battery was used as a negative voltage to bias the control grid of the tubes.

In mains powered tube equipment, the A supply is usually just called the filament winding and is AC straight off a transformer. The B supply comes from a high voltage winding on the transformer, rectified and smoothed with a capacitor that can handle the voltage. Regulation may be used to improve stability. The C supply is also rectified and smoothed to generate a negative voltage.

Transformers with multiple windings for high voltage and filament voltage are available. My synth uses a Hammond 290VEX, which has windings for 342V AC (yikes!), 52V AC and 6.3V AC. I had to fork out for it though. I bought it new from Mouser.

Conducting across a vacuum:
Why do we need to waste all that electricity on making the tubes light up? Because electrons aren’t able to escape the filament when it isn’t incandescent. A cold electrode in a completely evacuated envelope is about as good an insulator as you can think of. Heating the “cathode” so it glows a dull red allows electrons to escape the metal and travel to the “anode”. Remember electrons are negatively charged, so the cathode connects to a low voltage, such as ground, and the anode connects to a high positive voltage (this is contrary to “conventional current” where we think of electricity as flowing from positive to negative, which makes things a little tricky).

A tube like this is simply a diode. It only allows electricity to flow if the anode is more positive than the cathode. If the anode becomes more negative than the cathode, it will not conduct.

When battery radios were common, “directly heated” tubes were used. These used the filament itself as a cathode, and it basically meant you had to design your circuits with all the cathodes across tubes at the same potential (though there were ways around this). “Indirectly heated” tubes are more common. They heat a small cylinder which encases the filament and is electrically disconnected from it. This allows for the cathode to be at any potential and isolated from other tubes.

Controlling the current:
The major leap to controlling the current through a diode is to introduce another electrode, called the
‘control’ grid, separating the cathode and anode. This makes a three electrode device called a triode. The important thing to remember is to slow and stop the flow of electricity from the cathode to the anode, the control grid has to be more negative than the cathode. This is the same thing as depletion mode MOSFETs and JFETs. If you want to compare this to an NPN transistor, the cathode and emitter, the grid and base, and the anode and collector are analogous. Notice there’s no such thing as a P-channel tube!

Schematic symbol

The problem is the amount of electrons that flow from the cathode to the anode is also affected by the voltage of the anode. This is why tubes need a high voltage to operate. This problem isn’t present in silicon anywhere near as much, if at all. It’s also why tubes have their distinctive ‘soft’ distortion sound - the more current the triode has to conduct, the lower the voltage on the anode (assuming a resistive load) and the fewer electrons flow, reducing the amplification and causing a less abrupt clipping of the signal.

More grids:
To solve this problem, a second grid was placed between the control grid and the anode, called the ‘screen’ grid. We now have a tetrode. The screen grid is run at a constant positive voltage. I think of it as another anode, because the electrons are drawn towards it through the control grid, but they shoot straight through the screen grid and continue off to the anode. This means the output voltage on the anode doesn’t affect the amplification of the tube, because the cathode only sees the screen grid voltage. This introduces another problem called secondary emission, where the electrons may be travelling with so much velocity they knock out more electrons from the anode when they hit it, which then travel back to the screen grid. This causes instability in certain conditions. The solution again is to add another grid, this time called the ‘suppressor’ grid, which is usually connected to ground or internally to the cathode of the same tube. This makes a pentode. The purpose is to deflect the stray electrons away from the screen grid and back to the anode.

Pentode schematic symbol. The suppressor is internally connected.

I’ll continue this story in further posts.

14 Likes

When I was studying electronics as a child, people would explain these new fangled “field effect” transistors by comparison with the familiar thermionic valves which could still be found in televisions and radios. They were quite different from the junction transistors in transistor radios. Now we’ve reached the point where valves have to be explained by comparison with FETs. Plus ça change…

5 Likes

Part 2: Circuits

Now we know how tubes work it’s time to build up an arsenal of go-to circuits. All resistors I use are 1W 5% tolerance and the power supplies are +250V and -80V.

DC Amps
Probably the circuit I use the most is a DC amplifier. It works just like an op-amp, except the non-inverting (positive) input is always at ground. If you’re going to use this as an integrator or want to change the DC offset, you’ll need to change R4 to a 220 ohm resistor and put a 500 ohm resistor in series with it.

This is basically exactly the same as the Heathkit EC-1 analog computer’s amplifier module, though I changed some of the values to make it faster at the expense of power consumption. It’ll give about +/- 60V output swing.

It’s really important to mention a quirk about this circuit. If the negative rail is regulated by a transistor or zener diode circuit, the output will swing to that rail when power is applied, then as the tube warms up it’ll settle to its operating condition. If this is a problem (and it probably will be) use a DC blocking capacitor on the output.

Sawtooth Oscillator
This circuit can be used to make a V/Hz oscillator if you don’t mind cheating. It’s a boring old integrator but with a DIAC in parallel. The DIAC closes when the voltage across it is greater than 32V, which rapidly discharges the capacitor. When the capacitor is discharged the DIAC opens and the cycle start again. The DIACs properties seem to be very independent of everything else, so tuning remains stable across changing supply voltages and temperature.

The only problem is the bass response, which needs to be tuned with the DC offset pot I mentioned before. It’s pretty easy though, you just need to get the higher octaves to be in tune and the bass octaves can be adjusted by ear. The offset control has a negligible effect on high frequencies.

Because the DC amp is an inverting amplifier, the positive control voltage will make the output ramp between 0V and -32V. I’ve added a high value capacitor and resistor to the output to remove the DC offset.

The angle of the ramp is controlled by the V/Hz input and the values of the two pots, which determines the frequency. V/Hz oscillators can be tuned by multiplying or scaling the control voltage, rather than adding an offset like in V/Oct oscillators.

Pulse Waveform
This is straight forward enough. Using a DC amp again, with zener diodes in the feedback path, will amplify and clip the sawtooth waveform into a pulse waveform. This circuit doesn’t have a DC offset because the output is clipped in both directions by the zeners. I’d still use a blocking capacitor because of the problem I mentioned above about the negative rail.

I have included the previous sawtooth oscillator for clarity. You can still have a sawtooth output on this circuit, just include the blocking capacitor before the output (but not between the two amps).

Triangle Waveform
More DC amps. This one inverts the sawtooth waveform with an added offset, then picks whatever voltage is higher, be it the original sawtooth or the inverted sawtooth.

Exponental converter
It’s almost unfortunate that tubes are too linear. I found it to be impossible to make a V/Oct exponential amplifier out of them. So this circuit is a traditional 1V/Oct exponential amplifier, but the current output is connected to a tube DC amp. This gives a V/Hz control voltage which can be connected to several oscillator modules. The best part is the modules will track each other very closely, assuming the bass offset is set correctly, and the output can be as high as 60V which gives you all the octaves you’ll need. I haven’t labelled the components on the exponential converter because there are better circuits out there.

Stay tuned!

9 Likes

Ha, next to the recent posts on FPGA technology inspired synth components (which couldn’t be based on more differing technology than your work), this is one of the most interesting reads on this forum for quite some time! Please continue writing and explain some more about how you used tubes to make a synth.

7 Likes

6 Likes

3 Likes

Just went to bookmark the thread. Discovered I’d done it with the first entry. Super stuff!

Part 3: VCAs

Before I begin I’ll mention in part 2, I say to use a 500 ohm resistor in series with a 220 ohm resistor to adjust the DC offset of the amplifier. It should read, “500 ohm variable resistor.”

Now for some signal processing!

There are a few ways to make a voltage controlled amplifier with tubes. There are methods to do it with just one pentode, but I’ve not had much luck with them because they don’t reject the control voltage very well if the rise/fall time is too quick (CV feed-through, it makes a pop sound when there’s a sudden change in control voltage).

The method I’m going to discuss is very similar to the way it’s done in transistor VCA’s - a differential pair with a current sink to control its gain, followed by a differential input amplifier. That’s normally an op-amp in transistor circuits, which are excellent at converting differential signals to single ended signals. This method isn’t ideal either, CV feed-through can still occur, but multiple input stages can be mixed together. More on that in a bit.

The differential pair can be as simple as this. Vbb is the bias voltage. -CV is the control voltage, which needs to be pulled a few volts more negative than Vbb to amplify the input to the differential output, and a few volts more positive than Vbb to cut the signal off. The circuit is more useful with a positive control voltage, and this can be done with a pentode which I’ll explain in a bit.

You might have noticed there are no resistors on the anodes of the triodes. I did that because I draw them on the output stage. This is because it’s possible to chain a number of input stages together onto the one output stage to make a voltage controlled mixer. The two triodes need to be matched for good performance however (that is, don’t use a dual triode tube with one of the triodes completely worn out). I might make a drum machine using this method some time.

Another thing to point out is I’ve found both inputs to the differential pair need the same impedance (a mixture of resistance, capacitance and inductance) for good performance, so rather than connecting the other half of the pair directly to Vbb, I’ve used a 1M resistor and a 100nF capacitor to match the input side.

To convert from a differential signal to a single ended signal, we need the next circuit. It looks identical, but the lack of a resistor on the anode of the second triode means it does something else entirely. I’ve included the resistors to generate Vbb here, along with the anode resistors for the input stage.

Here is the complete circuit. The additions here are the triode buffer on the output, so it can drive other modules, and the pentode to control the current through the input differential pair. Remember in part one, the pentode doesn’t care a lot about the voltage on its anode as much as it cares about the voltage on the screen grid. This means they can be used as relatively good current sinks, which is what is being done here. The 6BL8 tube is a triode and pentode in one, so the whole circuit is three tubes.

I’ll be discussing my tube filter next!

7 Likes

Loving this. Keep going. Many thanks.

1 Like

This is what this thread reminds me of:

1 Like

<3 Abney Park. Even if this particular song is way low on my list!

FYI you can edit your previous post to correct it.
(And probably should.)

1 Like

Could you maybe elaborate on that a bit? What does it do?

Hrm I’m not able to find an edit function…?

There should be a pencil beside the link symbol on your posts.

I’m seeing it on my replies, but not on the original post. Perhaps there’s a problem?

Might have to do with “trust level”, new users can’t edit their posts after 24H.

1 Like

Probably should’ve expanded on that! The idea is the triode without the anode resistor is an cathode follower (much like an emitter follower), so the cathode voltage follows the grid voltage. This is the non-inverting (positive) half of the differential pair. Because the other (inverting) triode’s cathode is connected to the non-inverting cathode, that point follows the non-inverting voltage. If either of the inputs deviate from the other’s voltage, the difference will be amplified and will appear on the anode.

This is all of course somewhat imperfect due to the triode’s low gain. Also, the output is a bit dependant on the common voltage between the two inputs. This can be improved if there’s a current sink, such as a pentode, instead of a cathode resistor. This is called common-mode rejection.

3 Likes

@ 256byteram - can we hear it, please?!

2 Likes

he put a track here

1 Like