In the world of hi-fi, many believe the notion that all that can be done in terms of circuit topology has been done; and thus, the widely held belief that only improvements in part quality can lead to improvements in sound quality. So while the Transcendental OTL is highly regarded, it certainly is not perfect; change is possible. Of course, if a recommended set of changes are great enough, the original topology disappears. So let's try to retain as much as possible of the original amplifier, while bearing in mind what we have covered so far.
    From the last issue, we saw that the patented drive equalization technique, the bootstrapping of one of the cathode follower's B+, used in the Transcendental OTL was a bust. But we have seen how to balance the drive signals and configure the long-tail phase splitter for power supply noise canceling. Applying these techniques will require an additional adjustment, however. The examples of the long-tail phase splitter based output stages that we have covered have all assumed that the first grid does not see any power supply noise. If the grid does see any, this noise will be amplified along with the signal. We know that no mater how carefully we filter the power supply, some power supply will be superimposed on the input signal to the output stage. So what we must do is configure the circuit so that the noise is largely rejected.

     Here again is an example of how by purposely interjecting noise into a circuit we can lower the noise at the output. Although seldom mentioned, the long-tail phase splitter, much like the differential amplifier, contains a good CMRR mechanism. If both grids see the same signal, the common signal is barely amplified compared to the same signal applied to only one grid. So what is needed (in order to lower the noise at the output) is to ensure that the second grid see the same noise that the first stage sees.

    How do we balance the noise at both grids in this amplifier? The 12AX7 based first stage functions as a voltage divider to the power supply noise at the top of its plate resistor. In this instance, the division equals about 30%. In other words,  10 mV of noise reduces to 3 mV at the 12AX7's plate. The same division can be achieved by using a 70k and a 30k resistors in series. The problem with using resistors is that they divide AC and DC voltages equally, whereas the triode and plate resistor do not, the DC division being closer to 70%.
    Remember we need to ensure that phase splitter's both grids see the same DC voltage or the long-tail phase splitter will not work well if at all. The solution is found in using two capacitors in series and a 1M biasing resistor. The resistor forces the capacitor junction to the same voltage as the first grid sees.

< PREVIOUS

NEXT >