Differential Parafeed Line Stage


Single-ended circuits, by their very nature are unbalanced with respect to their “ground” reference.  As shown in Figure 13, the input signal is taken as the difference between ground and the grid, while the output is the difference between the plate and the positive terminal of the power supply. The positive terminal of the power supply is theoretically at signal (AC) ground, however because of the limitations in physical devices, a real AC impedance exists. To make maters worse, this AC impedance is not fixed but rather varies with frequency.  Another problem with this circuit is that the input is susceptible to signal noise. One way to alleviate these problems is to use a differential circuit. There are several ways to implement a differential amplifier; the classic push-pull configuration is shown in Figure 14. The differential circuit is really two single-ended circuits, side by side. The important difference is the input and output references. Now the input is referenced between the two grids while the output is referenced between the two plates. This input scheme will provide better rejection of common-mode noise. Another benefit is that distortions in the circuit, being the same on both sides, will cancel on the output. Of course for this to work the signals on the two grids must be different – in fact opposite. An electronic circuit or a transformer may provide this inversion of the signal on one of the grids with respect to the other. Because an electronic circuit will suffer some of the ills that you are trying to remedy, I believe that a transformer is the best solution for this function. I really liked the Lundahl LL1674 that I had used as a parafeed output transformer and since this transformer was originally intended to serve as a line input transformer, it was natural for me to select it for the phase splitting task. 

As much as the differential circuit is an improvement over the single ended circuit, it still suffers the earlier problem of sharing much of the circuit for the power supply current and signal current. The obvious solution is to use a parafeed topology for the push-pull circuit as shown in
Figure 15
. This shows the input circuit with a transformer to provide the phase inversion. The parafeed circuit works as before with the pair of CCS isolating the power supply from the signal and the parafeed capacitor isolating the output transformer from the power supply current. Think of it as two single-ended parafeed amplifiers sharing a common output transformer. In fact I built the first prototype using a pair of the SET line stage boards. I felt that this would give a more valid comparison when auditioning the new line stage. The result was immediately obvious, a more lively presentation and better impact. I knew that I was on the right track. 

 Lynn Olson described another parafeed variation in his article “Ultrapath, Parallel feed and Western Electric” in Vacuum Tube Valley issue 16. I think of the circuit shown in Figure 16 as two single-ended-triode amplifiers, one driven in opposite phase from the other, bridged with a transformer between their plates. I re-wired my test circuit with this circuit not expecting much. After all, to my thinking it was a minor twist on an old horse. I was surprised at the audible difference between this circuit and the earlier one I had been using. I anticipated a subtle difference but this was anything but subtle. The sound was much cleaner and clearer without sacrificing any of the dynamics I had come to appreciate. The most obvious difference between the two configurations is that the circuit shown in Figure 15 has two signal current loops, while the circuit shown in Figure 16 has a single signal current loop. This is significant because with the classic push-pull circuit, the transformer is mixing the two signal current loops magnetically in separate input windings. Any imbalance between the two windings or between the two signal currents will produce an aberration of the signal at the output of the transformer. The single signal loop sidesteps this problem and may be responsible for at least some of the characteristic that SET fans applaud. Another aspect for this circuit is the advantage of a shared cathode resistor which ensures that the two triodes are biased at the same point, further aiding balance in the circuit. As with the push-pull circuit, the triode is the only thing common to the power supply current loop and signal current loop in the circuit shown in Figure 16. 

During the development of the differential circuit I experimented with several ways of implementing a volume control, two of which are shown in Figure 17. In a single-ended configuration the signal is connected to the top of the attenuator, the bottom of the attenuator is connected to ground and the output is taken off of the wiper. Configuration A is two of such attenuators, one for each phase. There are two advantages to this configuration. First, the impedance of the attenuator, reflected through the input transformer to the input, is constant for all settings. Second, since it takes a stereo potentiometer to implement this configuration, you can use a separate one for each channel and thereby achieve a balance function. The disadvantage of this configuration is that it is more expensive, requiring either two stereo potentiometers or a single quad potentiometer to implement the volume control function. The attenuator shown in configuration B takes a different approach. Decreasing the resistance in the potentiometer attenuates the differential signal on the two grids. This decreasing resistance is reflected back through the input transformer so that, unfortunately, the input impedance varies with the setting of the potentiometer. However, the two resistors located between the transformer and potentiometer limit the minimum impedance to an acceptable value. The advantage to this configuration is that it requires only a single stereo potentiometer to implement the volume control functions for both channels. Aside from the technical differences, there are audible differences between the two configurations. I found that I prefer the sonic characteristics of configuration B over those of configuration A as well as those of several others I tried. Perhaps, and I am speculating here, there is something going on here similar to that in the output with one signal loop versus two signal loops. There are two signal loops in configuration A, one for each half of the differential circuit. Each loop goes from ground, through one half or the transformer, through a potentiometer and back to ground. Differences in the two halves of the transformer as well as between the two potentiometers will affect the differential balance. There is a single signal loop in configuration B, through the transformer, through the two resistors and through the potentiometer. Since everything is in series, differences between the two halves of the transformer or between the two resistors do not produce an unbalance in the differential signal. 

 During the development of the SET Line Stage, I experimented with adding a small amount of load across the secondary of the output transformer to quiet a small amount of very high frequency ringing in the transformer. This use of a load is controversial with some swearing by it while others condemn it. I must say that in cases like this I let my ears be the final arbiter. To me, a little load sounded better so I used it and carried the load into the differential design with a resistor across the secondaries of both the input transformer and the output transformer. Early on in the development of the differential Line Stage it also sounded better with the resistors present, however as the design evolved through many refinements, it progressed to the point where it sounded better without those resistors. So, the lesson learned here is that some designs may benefit from adding a secondary load, however, I believe that it is really compensating or masking problems elsewhere in the circuit.    

I had been using the 6H30 dual triode, which I had selected after listening to many candidates during the SET Line Stage development. Kevin Carter of K&K Audio, the US Lundahl distributor, told me that he had heard good things about the 6N1P dual triode. Looking at the data sheets I thought, “no way can the 6N1P outperform the 6H30, it doesn’t have the muscle.” However I learned long ago that we don’t listen to the data sheets so I decided to try it anyway. It was a simple enough experiment – the pin configuration was the same and a switched resistor was all that I needed to change the idle current. Well, the 6N1P may not have the muscle of a 6H30 but it has the finesse. The music was clearer and more detailed. There was no indication that it was deficient in the power category - the dynamics were excellent. The difference was subtle; I still think the 6H30 is an excellent performer and wouldn’t hesitate to use it where I needed its drive capability, but for this application the 6N1P is the clear winner. 

Let’s bring it all together now by taking a look at the schematic diagram in Figure 18. Most of what you see there will now be familiar, although there are a couple of minor points that we have not yet considered. If you remember from the SET article, I was plagued by problems of oscillations in the circuit and went to great lengths to eliminate them. In addition to gate stoppers and grid stoppers, I had added plate stoppers and drain stoppers. I am not sure why, but the differential circuit is much better behaved in this respect and only gate stoppers (R7, R9) and grid stoppers (R4, R5) are required to curb any tendency to oscillate. Lastly, the circuit at the bottom of the schematic diagram provides a regulated filament current for the tube.  Photo 4 shows this circuit packaged on a printed circuit board. 

Okay, I had chosen the circuit configuration and tube and it sounded really good, but I still had a few things that I wanted to experiment with before I put the lid on the project.

The differential Line Stage was now even more revealing and sonically clear than the excellent SET version I had been using. I could easily hear a difference between various passive components and it was time to do re-evaluate the components that I was using. The capacitors were easy to choose: C1, C3 and C4 are there for noise decoupling and have little if any affect on the sonics of the Line Stage. The parafeed capacitor, C2, definitely does have an effect and must be a high quality capacitor. My preference here is the Kimber Kap, however your ears are different from mine so you may prefer the characteristics of another brand. Choosing the best sounding resistors was an entirely different matter. It took many hours of critical listening to audition various combinations of different types of resistors in the nine locations other than the potentiometer and those in the filament regulator circuit. My first thought was “but where to start?” Well, tantalum resistors have a reputation for being the best so that is where I chose to begin this leg of the journey. I populated all locations with tantalum resistors and let the Line Stage play for a few hours to let everything settle in. It sounded great! The word “aristocratic” came to mind to describe the sound of these resistors. As good as it sounded, I was not about to let well enough alone – I replaced all of the resistors several times, first with carbon film, then exotic metal, then wire wound, and several others. None sounded as good as the tantalums. Was I done? No, not quite. Working from a base of tantalum resistors I replaced each resistor one at a time with each of the different types. So, for example, I would have eight tantalum resistors and one carbon film resistor. This stage of the experiment proved very interesting and worthwhile. It turns out that different locations are best served by different kinds of resistors. When I got through I had a mix of Caddock, Kiwame and Mills resistors with no tantalum resistor in sight.