Dozuki said:
You probably mean higher - the answer is not so simple, though. CCS, properly implemented has very high impedance:
1) It presents a very high input impedance for the tube stage, so it's operation is primairly governed by parts in it, and not by the load
2) It makes the bias current independent of the rail voltages. The Vgs multiplier is not as good a voltage source as a Vbe muiltiplier, due to lower MOS gm, so a higher current through it increases the bias voltage.
3) It requires a good DC servo as the two oposing CCS have a very high 'gain' - a very small difference in the currents will stick the input to the rail with the 'stronger' CCS. This can be avoided if the input is given an actual load, but this can be much higher than the resistors the CCS replace, because the current for the Vgs multiplier does not need to pass through it.
4) For certain topologis on the tube side (eg CCS loaded plate), the use of CCS biasing is mandatory because any load degrades the tube linearity.
In general, the more the better (within reason) - higehr current means less distortion at HF as the available current to charge the gate charges a nonlinear cap. The higher the available current, the lesser the nonlinearity is in comparison. For topologies with global NFB you have extra critera, low current means bigger lag, and can, if too low, lead to oscillation. Typically 5-10mA is used for the above MOSFETs. The current is proportional to input capacitance. This can be somewhat relaxed if the global NFB is very low or zero.
Well, that mostly depends on what kind of power you want out of the SS stage as that defines the required driving voltage. The smaller this is in relation to the +B of the tube stage, the more linera the tube stage will be in relation to it. For isnatnce, you would need >80VPP for 100WPC into 8 ohms, obviously this may be a problem with a +B of only 120V, though some topologies will be better at such an attempt than others (SS CCS loading or bootstrap CCS for instance).