**Bypass Current Source vs Resistor**
With my very limited understanding of electronics, this is how I interpret Knut's suggestion of using a simple resistor in lieu of the bypass CCS.

It was stated that using a resistor will reduce the frontend gain, which is of course true.

The question is, by how much.

In the original Hirage circuit, with 0.6A output stage bias, target voltage across the 1k drain resistor is ~1.2V.

That means its current should be 1.2mA.

This is of course the same current through the JFETs and the cascode BJTs, as well as the 100R trimpot.

Datasheet of 2SK170 indicates that the transconductance at 1.2mA is ~16mS, or an equivalent resistance of 60R.

Assume also the trimpot is at its mid position, i.e. 50R for each half.

The input stage gain is 1K/(60+59) = ~9.

In our example of GR grade JFETs, with 50R at its source, the drain current is ~2.3mA.

Since we only want to allow 1.2mA to flow through the 1k resistor, we need to bypass 1.1mA.

Because TP2 is at the base of the cascode transistor, voltage across TP1_TP2 is ~6.6V.

We can either connect a current source of 1.1mA across TP1_TP2, or a resistor of ~6k.

The above is of course all DC.

Suppose we now apply a small AC signal to cause the JFET to swing 0.1mA.

In the original circuit, and in case of an ideal current source, all of that current wing can only go through the 1k drain resistor.

So the input stage gain is, as stated above, ~9.

But the cascode transistor also has a small emitter resistance, approximately 20R at 1.2mA.

In case of a 0.1mA ac current, TP2 will change from its DC value by 2mV.

Which means that the current through the proposed bypass resistor will also change by 0.33µA.

This has to be subtracted from the current through the 1k resistor, which now only swings 0.0997mA instead of 0.1mA.

That, if my sums are correct, corresponds to a loss in open loop gain of 0.33%.

The exact value depends also on the Idss of the JFET used.

So it is still desirable not to use, e.g. BL grade JFETs, unless you are using a very good CCS and not the resistor.

But I am sure Knut will correct me if I am talking nonsense.

Patrick