Lewm, the bypass cap is only necessary for the single ended version. The differential LTP version of course does not need a bypass for the cathode CCS as the signal passes from cathode to cathode directly. I have a complete schematic for the beta version of the K&K Audio Sonus Veritas Venice which used this circuit for its input stage (BTW the Venice is fully differential). I only have a "paper" version of this schematic and don't have a scanner unfortunately. Maybe if you talk to Kevin about this he will be willing to email you the schematic.
The schematic tends to look pretty busy due to all the current sources and shunt regulators. One plus sonically is the only caps affecting the signal are the two RIAA caps that are shunt connected across the two balanced signal paths. No caps in series with the signal path. No bypass caps either of course and the PS is well isolated from the signal (CCS "tail" impedance between cathodes and ground, separate CCS and shunt reg. between B+ and PS for each gain stage). Also, the only series resistance in the signal path are the 10 Ohm grid stoppers.
One way of thinking about this input stage arrangement is to consider it a parallel feed design without a parafeed cap (we aren't trying to block DC in the shunt resistors). The shunt connected RIAA resistors are now the tubes'e plate load with bias supplied by the plate CCSs. One issue that Kevin had when he went to this arrangement is that the tubes' capacitance now enters more into the RIAA calculation. This required some rejiggering of the RIAA capacitor to obtain an accurate response.
The schematic tends to look pretty busy due to all the current sources and shunt regulators. One plus sonically is the only caps affecting the signal are the two RIAA caps that are shunt connected across the two balanced signal paths. No caps in series with the signal path. No bypass caps either of course and the PS is well isolated from the signal (CCS "tail" impedance between cathodes and ground, separate CCS and shunt reg. between B+ and PS for each gain stage). Also, the only series resistance in the signal path are the 10 Ohm grid stoppers.
One way of thinking about this input stage arrangement is to consider it a parallel feed design without a parafeed cap (we aren't trying to block DC in the shunt resistors). The shunt connected RIAA resistors are now the tubes'e plate load with bias supplied by the plate CCSs. One issue that Kevin had when he went to this arrangement is that the tubes' capacitance now enters more into the RIAA calculation. This required some rejiggering of the RIAA capacitor to obtain an accurate response.