SET Parafeed Line Stage



It all started with a simple transformer loaded, single-ended-triode amplifier that has long been a part of the audio designer’s catalog of circuits.  It is renowned for stable performance and excellent drive capability.  Part of the attraction of the circuit shown in Figure 6 is its straightforward simplicity. In this circuit everything is in series. All current flowing through the tube also flows through the transformer and the power supply with the tube acting as a valve to modulate this current. The modulated current induces a magnetic field in the transformer, which in turn provides current to the output. The DC component of the current develops a voltage across resistor Rk that is used to bias the tube to the desired operating point. Capacitor Ck provides a low-impedance path for the signal so that it will bypass Rk and thus not modulate the voltage produced by Rk. 

Although this circuit works very well, it does have a couple of shortcomings when applied to high-end audio. Take a look at the signal current for a moment. Of course the signal must flow through the tube and the transformer in order to produce an output, but unfortunately it also flows through the power supply and Ck. In fact, Ck is only there to bypass the signal current around Rk. Ck usually needs to be a very large electrolytic capacitor so that it can pass very low frequency signals. The audible effect of electrolytic capacitors is well documented and it is desirable to eliminate them if at all possible. The signal current also passes through the power supply, and what is usually in a power supply?  More electrolytic capacitors, and perhaps other stuff that may also hurt the sound. Now consider the power supply current. This current must flow through the power supply as well as the tube and Rk, however it serves no purpose for the transformer, and in fact may be detrimental for optimum performance. So it would seem like a good idea to separate the signal current from the power supply current. In fact, this is one of my criteria for a high-fidelity amplifier. Figure 7 shows one way that this might be accomplished. 

This circuit provides two separate parallel paths for the signal current and the power supply current. The only component that is common to both paths is the triode. This is called a parallel feed, or “parafeed” circuit. Capacitor Cp provides a low-impedance path for the signal to flow while blocking any direct current. Because of the impedances in this circuit compared to those in the cathode circuit, this capacitor can be several orders-of-magnitude smaller than Ck, and thus may be a high-quality film capacitor. A transformer that is not designed to carry any direct current may be selected for use in this circuit. This is important because when a transformer doesn’t have to carry direct current it may be optimized for audio performance. I don’t pretend to know the first thing about transformer design, but from what I can gather not having to cope with direct current makes the design job easier. There are fewer trade-offs to make and it is easier to optimize the AC signal characteristics. But, what about using a “regular” transformer in a parafeed configuration? I built the circuit shown in Figure 8 with the LL1660/18mA to find out. For some reason, I’m not sure why, I decided to drop the current in the circuit to 10mA and use a 7 Volt bias point. Since there is no current flowing through the transformer, I don’t think it makes much difference. The 3mF “parafeed capacitor” in series with the transformer primary ensures that no DC current flows in the primary.  Because of the constant 10mA of current flowing through the tube and cathode bias resistor, there is no need for a bypass capacitor. 

I had great expectations for this circuit and was sorely disappointed with the sound it produced. The character of the sound was okay but the bass was anemic – puny as we say in Dixie. On examining the frequency response in the bass region, I found a large peak at 9Hz with a corresponding suck-out in the mid-bass. The bass response of the circuit is directly affected by the value of the parafeed capacitor. There are several ways to calculate the value of the parafeed capacitor. Some set the value so that the impedance of the capacitor is equal to the impedance of the transformer primary at some arbitrary low frequency, say 10Hz.  Another method sets the value so that the electrical Q of the resonant circuit is well behaved. Unfortunately, the different methods give different results, in this case ranging an order of magnitude from 2.5mF to 25mF. I had hoped to get lucky and went with the low value, but it didn’t work out. I built the circuit shown in Figure 9 to not only determine what value of parafeed capacitor to use, but also to test the result of a cathode-coupled connection and provide the capability for comparing different transformers. But let’s take one step at a time. For determining the optimal value of the parafeed capacitor I used electrical measurements as well as listening evaluation and found that there was agreement between the two. The optimal value provides a broad, low-amplitude peak in the sub-sonic region that exactly compensates for the low-frequency roll off, resulting in a composite flat response down to 3 or 4 Hz. Furthermore, there is no sonic advantage to using a larger capacitor than is needed; all that happens is a roll off in the sub-sonic region with little if any effect to the mid-bass. Interestingly, in determining optimal values for the parafeed capacitor to be used with several different transformers, I found the range to be 5mF to 13mF. Because this parafeed capacitor is in the signal path it must be of the highest quality. I recommend your favorite film and foil. The particular brand of capacitor will affect the character of the sound, but this is a matter of your taste and preference. You’ll need to be concerned with the voltage rating of the capacitor because it will see the whole B+ level. With the optimal value of parafeed capacitor in the circuit, the bass was now adequate, not as tight and deep as I would have liked but not puny like before. It seemed to me that the bass was deeper in the single ended configuration. I no longer had the luxury of direct comparison since I had to radically alter the configuration and it had been a couple of days since I had heard the single ended configuration. I wondered if this transformer having been designed for single ended operation actually needed current running through it for optimal performance?  Perhaps this would be resolved when I compared different transformers.  In spite of the bass being less than I would have desired, I knew that I was on the right track. The mid and high frequency was far superior to that of the single ended configuration. In fact, it could be that my perception of a lack in the bass presentation was only in comparison to the improvements in the mid and treble.

Switching the parafeed capacitor between ground and the cathode provided a noticeable improvement in the sound when connected to the cathode, or cathode-coupled. Because the top of the transformer is no longer connected to the power supply, there is no noise to feed to the cathode and a bypass capacitor is not needed. 

A constant current source (CCS), as its name implies provides a constant direct current and very high impedance to the signal current.  Additionally, a CCS provides a very high power supply rejection ratio (PSRR). This means that the audio circuit is isolated from aberrations in the power supply. The CCS may be a very large resistor, an audio choke, or an electronic circuit. A large resistor will drop a large voltage, requiring that the power supply have a very high voltage for the B+. An audio choke, while not producing a large DC voltage drop, is large and heavy and still does not provide the very high level of impedance that an electronic circuit provides. A MOSFET CCS can provide an impedance in excess of one Megohm while dropping only ten volts. I didn’t bother with a resistor but compared several chokes and electronic circuits. The clear winner is a very simple electronic circuit based on the Supertex DN2540N5 N-channel depletion-mode vertical DMOSFET. This device makes it almost as easy to build a high-voltage current regulator as it is to build a low-voltage current regulator with the ubiquitous LM317. In fact, the topology is the same – a single resistor in addition to the device. While there is a simple formula that is used to calculate the value of the resistor for the LM317, unfortunately there is no such formula for the DN2540; variations in the manufacturing process make this impossible. However it is simple enough to determine the resistor value experimentally. Figure 10 shows you how.  It is not necessary to get a particular current to two decimal places. There is not a hill of beans difference in this circuit between using 10mA and 11mA, so pick a value of current that corresponds to a convenient resistance – one that can be provided with a single resistor or parallel pair. In my circuit, about 150W provided about 10mA and I found that the value determined using a 9Volt battery held closely at 250Volts. The DN2540N5 is a TO220 package and should be mounted on a heatsink. The tab is connected to the drain so the tab needs to be electrically isolated from the heatsink. One other thing: the DN2540 is prone to oscillation and needs the equivalent of a grid stopper resistor. A non-inductive, 100W resistor connected as close as possible to the gate will suffice. 

As I mentioned earlier, I tried different transformers; particularly transformers that were designed not to have direct current through them. Generally these transformers are smaller and there are various different materials used for the core. From what I can gather, core material makes a big difference. I was talking to Kevin about core materials and he told me about Lundahl’s amorphous core. They have had great success using this core in some of their smaller high-level line-input transformers. Unfortunately, Lundahl does not yet have an amorphous core line-output transformer that they specifically recommended for parafeed connection. But then in my mind a transformer is a transformer and if it can be used as a step-up input transformer, it can be used as a step-down output transformer. Of course you can’t put any current through a line-input transformer, but then that is exactly what I wanted. Kevin had a pair of LL1674 line-input transformers in stock that he agreed to let me try in the circuit. I put them in and was blown away by the sound. The music had taken on a new life. The bass was deep and solid, a characteristic shared with some other small transformers that the LL1660 had lacked, but it was now cleaner and tighter then with any other transformer. Treble was more extended and the midrange was really musical. I bought the LL1674s on the spot, knowing that I had reached the end of my experimentation and was ready to build a representative prototype of the final package. The schematic of this design is shown in Figure 11. The Lundahl data sheet says that no compensation network or termination is required for the LL1674, which is born out by the very clean scope presentation and good sound. However I did find that the sound improved subtly to my ears with a little load. As with many of Lundahl’s transformers, you can choose different ratios between the input and output. I did not need or want much gain so I chose the 8:1 ratio.

It’s no deep mystery when it comes to bass response on the transformers. It all comes down to primary inductance. The single-ended transformer is designed to allow substantial direct current, the push-pull transformer is designed to carry only enough direct current to account for circuit imbalances, and no direct current at all is allowed through a parafeed transformer. So what has this got to do with primary inductance? Well, the problem is that direct current causes a transformer to saturate, and a gap must be introduced in the core to compensate for this. A single-ended transformer has a large gap, a push-pull transformer has a small gap and a parafeed transformer has no gap at all. The gap in the core reduces the inductance – the larger the gap, the larger the reduction. So, all other things being equal, the parafeed transformer with no gap at all has the largest primary inductance which allows it to exhibit the best distortion-free bass.