Alright, now it's time to talk about the input stage.
There were several problems with the input stage of the SOHA:
1. A difficult CCS situation
2. The need for a trimpot to adjust the plate voltage every time tubes are changed
3. Inability to run tubes other than 12AU7s
The first problem to tackle is #2. How can we automate the equivalent of the trimpot adjustment? Well, the first necessary item is a way to sense the voltage at the plate of the triode. We must be able to measure this to compare it against a reference so that we can adjust some control point on the triode (such as the cathode resistance).
But, remember that an important design parameter is the impedance that the input stage sees. We want to keep it very high. Now the only way to sense the plate voltage is to put a resistor there connected to some active circuit (like an opamp servo). And this resistor will always reduce the load on the tube. Unless we make it something like 10M in which case it will introduce noise. Furthermore, using an active mechanism like an opamp servo introduces complexity that we really shouldn't have in the input stage (but it's ok in the buffer where we MUST control the DC offset).
So, what do we do? Well, we need another kind of sensing and reference scheme. For the SOHA II, the sense/reference scheme is a common cathode amplifier (a sibling of a differential amp) that uses two triodes connected at their cathodes. Like this:
Now much explanation is required so hang in there.
First, the tail of the triode pair is loaded with an adjustable CCS. Don't panic yet because the trimpot has a different function than the SOHA. This CCS is typically set to run at 2mA.
Second, the plate loads are a current mirror with the control (diode) side of the mirror on the first plate. This current mirror divides the current between the two triodes exaclty in half (within component tolerances) setting the current in each triode to 1mA.
The behavior of these two CCSs never changes unless the stage goes out of its operating region (say by being way over driven). The fact that they never change is really important. Why?
Well, look at the circuit on the first plate. It consists of a 220R resistor, the BE junction of the BJT, and an 18k resistor in series. Remember that these three components are always seeing 1mA of current which means that they will always drop approximately 19V. If the B+ is at 60V (the design center) then the plate voltage on the first triode will be 41V.
So far so good.
Now, what happens at the second triode? Well, the tail CCS adjusts the cathode voltage so that the first triode will pass 1mA with 41v on its plate. This is the same cathode voltage that is on the second triode. Now since the second triode has the same cathode voltage as the first triode and since it is conducting exactly the same current, 1mA, then its plate voltage MUST be the same as the first triode, namely 41V.
Of course, how close the second triode's plate voltage is to 41V depends on how well the triodes are matched. In general, triodes in the same package of a functioning tube are farily well matched. Furthermore, the circuit will tolerate some amount of mismatch. The voltage on the second plate won't be exactly 41V, but gents these are tubes and a few volts in either direction really doesn't matter. Still, the closer the triodes, the closer the plate voltages.
If you look carefully you will also see that the first triode operates as a cathode follower because its plate voltage is held constant. The cathode follower drives the second stage which is wired as a grounded grid amplifier. So, this is not a differential amplifier. One othe feature of this configuration is that it does not invert the phase of the signal.
Notice that there is no tube type on the schematic. This is because the CCS and mirror do their jobs no matter what type of triodes are there. The only thing that changes is the cathode bias that will set that triode's operating point to be 41V and 1mA.
Back to the trimpot. The trimipot only needs to be set once to establish 2mA in the cathode circuit. After that you can roll tubes until you drop and the circuit will do the right thing without your having to make any adjustments.
And if there are tubes that are way out of balance, you'll just hear it in when the second triode hits the rails too soon. Bad tube.
Also, any triode pair will work providing that they can operate at the low voltage, low current conditions. A good rule of thumb is that the triodes are safe if the cathode voltage is at least 300mV above ground.
The current mirror's dynamic half presents over 3M resistance to the plate. This resistance is in parallel with the buffer's Zi (about 950k), making the actual plate load about 720k. This is high enough for any triode pair that I can think of using in this amp.
So here we have it. The replacement of the SOHA front end that removes its problems and lets you tube roll, not just different tube of the same type, but different types of twin triodes too.
The last step is to put the buffer together with the input stage to make a single channel amp. There is one more feature when this happens that will have to wait until that post.