Below are simulation results in pentode mode with or without NFB, then in triode mode with and without NFB.
|
EL34 tubes in Pentode mode, with NFB
|
|
EL34 tubes in Pentode mode, without NFB
|
|
EL34 tubes in triode mode, with NFB
|
|
EL34 tubes in triode mode, without NFB
|
|
the last 2 stages, purely symmetrical, receive ideally split signal. Each driver tube is biased equivalently with the stock configuration but separately, and sees a separate long tail cathode load resistor.
|
This suggests that if the phase is split ideally, then the last 2 stages deliver very clean amplified signal.
The next data is for signal going into the stock EICO-HF87 differential splitter, except not through the input 12ax7, but directly to R10. The value of R5 was optimized using SPICE for minimum of 2nd harmonic
|
the last 2 stages, stock long tail splitter, triode mode on the output tubes, no negative feedback. R5 is optimized at 27.5k for lowest 2nd harmonic. The value is dependent on mu of 6sn7 tubes, and your mileage may vary.
|
This suggests that if the first stage is very clean, then the stocj splitter does a pretty good job. How things can be improved further? One direction is constant current sink on the splitter instead of the 18K resistor.
To approach that, first CCS on the pure balanced variant. Below is the data from a silulation of the balanced 2-stage topology with a constant current source feeding both driver tubes. One apparent change is 2x increase in amplification but harmonic numbers are very similar:
|
the last 2 stages, purely symmetrical, receive ideally split signal. Each driver tube is biased equivalently with the stock configuration and sees a CCS.
|
So, what if we inject a CCS into the original schematic? Once again, not bothering yet with the first stage, feeding signal to R10. Note this requires R5 of 30k. That's what SPICE says, I'd still think it needed be somewhat lower than that. Anyways, the data is good:
|
the last 2 stages, stock, except the long tail resistor (R7) is replaced with a 7mA current source. Both shoulders need practically equal load.
|
The improvement is visible but not significant. Does it mean that advantage of CCS over a long tail is just a grand illusion? Not exactly. Here is, just for fun, a "short tail" variant: R7 is substituted by a 100V voltage source and a 3.6k resistor, which gives the same baising to the tubes. Below is simulation data. Note that the optimized R5 for this case is much lower: 20.5K !!
|
the last 2 stages, stock, except the long tail resistor (R7) is replaced with a 100V voltage source and a 3.6K resistor, forming a short tail config. R5 is 20.5K, as required for minimal 2nd harmonic.
|
Hence it definitely pays to have a "long" long tail, however it is also true that R7= 18K and R5 = 27.5K provide data that is fairly close to CCS-tail.