A basic reminder about tube behave

Grounded Cathode Amp

Characteristics

• Voltage Gain: Yes
• Current Gain: Yes
• Output Phase Inversion: Yes
• Input Impedance: High
• Output Impedance: Moderate
• High Frequency Performance: Poor

Cathode Follower

Characteristics

• Voltage Gain: No
• Current Gain: Yes
• Output Phase Inversion: No
• Input Impedance: High to Very High
• Output Impedance: Low
• High Frequency Performance: Excellent

Differential Amp

Characteristics

• Voltage Gain: Yes (Half the gain of each section when used as phase splitter)
• Current Gain: Yes
• Input Impedance: High
• Output Impedance: Moderate
• High Frequency Performance: Fair to Poor (Depends on how input stage is connected.)
• As Phase Splitter: AC phase-to-phase balance: excellent. Harmonic balance: excellent.

Cascode

This is another two stage amp, this being a cascade of a grounded cathode amp driving a grounded grid amp. The "traditional" use for the cascode has long been as a VHF voltage amp. The grounded grid amp has excellent high frequency performance since it, like the cathode follower, avoids C_Miller. It does, however, have a very low input impedance, which is something you don't want to see very often. Letting this low impedance load down the grounded cathode stage kills most of its gain (ideally reduce the gain to unity or even less -- accomplished with transistors, but not tubes). The greatly reduced gain limits C_Miller to small values that don't impact the high frequency performance nearly as severely as it would with a grounded cathode stage operating at the higher voltage gains. The cascode is one of those rare cases where you can have your cake and eat it too: the high frequency performance of the grounded grid amp, with the high input impedance of the grounded cathode amp. Being a two stage amp, it also provides significantly more gain than that of a single triode (theoretically up to u², but more practically, 4 -- 5 X u).

It has long been considered that the two stage combination makes a faux tetrode: that's where the name comes from -- a combination of "cascade" and "tetrode". In an great many ways, it does tend to operate much like a tetrode, minus the screen grid "kinks", the screen current and its inevitable partition noise. The grid of the upper triode is still negative with respect to its cathode, so does not pull current. As an audio amp, it still retains much of the sonic signature of triodes, but with enhanced linearity. Since the full output voltage swing is occurring across two triodes in series, each triode sees less V_pk swing than a single triode would, thus they operate over a narrower range of voltages, and don't stray as far into nonlinear territory. Tube cascodes are rarely seen in hollow state audio practice, though they are used extensively in solid state amps.

When combined with the LTP phase splitter, the extra voltage gain often allows the elimination of an extra gain stage while preserving enough gain margin for good input sensitivity even under gNFB. This, too, is an advantage in that it gets one more active device (and its distortion) out of the signal path. Though not often seen in audio equipment, the cascode LTP has long been used in precision, wideband applications, such as oscilloscope vertical deflection amps. When connected as an LTP phase splitter, AC balance is improved by using a CCS as the tail load.

If there is a downside, it's the much greater output impedance. This may not play well with the input capacitance of a subsequent stage, resulting in premature roll off at the upper end of the audio band. This can, however, be helped by connecting to a cathode follower, with its much reduced input capacitance.

The AC coupled version is known as a "folded cascode", useful when you don't have the supply voltage reserve. Otherwise, the operation is identical to that of the direct coupled version seen here.

Characteristics

• Voltage Gain: Yes (Significantly higher than that of a single triode stage.)
• Current Gain: Yes
• Output Phase Inversion: Yes
• Input Impedance: High
• Output Impedance: High to Very High
• High Frequency Performance: Excellent

                                        ***********************

The circuit uses two triodes, usually in the same envelope. Each triode is biased the same, so half the HT is dropped across each one. The lower triode acts as a common-cathode gain stage with an active load, and the upper triode acts as common-anode gain stage with an idenatical active load. This is about as close to a complementary transistor pair as valves get!
The reason the circuit is push-pull and not single ended is that the signal reaching the bottom triode causes the signal on the grid of the top triode to be in anti-phase with it. When the top triode conducts more, driving current into the output coupling capacitor, the other conducts less. When the top triode conducts less, charge stored in the capacitor is returned and flows down into the lower triode.
Unlike the mu-follower, the output should be taken only from the cathode of the upper triode.

Since the circuit is really a small power amplifier, high current valves are preferred, but the following example uses an ECC83 (12AX7) with an HT of 300V

This circuit is often used as an output transformerless (OTL) power output stage, so high current valves are preferred, but let's see how well (or badly) and ECC83 does. For perfect balance, V2 ought to have an anode resistor equal to Rk if the cathode is bypassed, or 2Rk if the cathode is unbypassed. However, the unbalance is very small for high-mu valves.

The two cathode resistors should equal*:
Rk1 = Rk2 = (2Rl + ra) / mu
Where Rl is the following load resistance. If we were driving a 10k load, say;
Rk1 = Rk2 = (20k + 65k) / 100
= 850 ohms.
820 ohms is the nearest standard.

It is usual to add a cathode bypass capacitor to the lower cathode. Leaving it out will hardly affect the gain in this case, but it would increase the anode impedance which makes the stage more susceptible to noise and increases output impedance. Since the frequency response will be hardly afftected, there is little point calculating the bypass capacitor's value carefully, any value greater than 1uF should do.

The quiescent current is given by:
Iq = HT / (2ra + 2muRk)
Iq = 300 / (130k + 2*100*820)
1 mA

And since the SRPP can only operate in Class A, the peak current delivered into the load is 2Iq per triode, making 4mAp-p in total, or about 1.4mArms- which is not as spectacular as we might have hoped. The maximum undistorted voltage across the load must therefore be 4mA * 10k = 40Vp-p, or 20mW. The maximum input signal before clipping is simply 2Iq * Rk, which is about 1.6Vp-p (so the circuit must therefore have a gain of 25).

Alternatively, an LED or diodes could be used to bias the lower triode, which would negate the need for a bypass capacitor, but only if they provide the same bias voltage as Rk2, for optimum performance (in this case 0.82V).
Normal rules apply for the grid-leak resistor, and 1Meg is usual.

If both triodes are identical and biased the same, thel output impedance will be:
Zout = (ra + 2Rk)[ra+Rk(mu+2)] / [2ra + 2Rk (mu + 2)]
Zout = (65000+2*820) [65000 + 820 * 102]/ [2 * 65000 + 2* 820 *102)
Zout = 32.7k
This is about half the normal value for regular unbypassed ECC83 gain stage.
But of course, this figure is of limited use, since the circuit is still only capable of driving maximum current into the optimum load impedance it was designed for.

Heater considerations: Because the cathode of the upper triode will be at roughly half HT, the heater supply will probably need to be elevated to avoid exceeding the valve's maximum heater-cathode potential- always check the data sheet.