Line Stage Prototype

Many modern components are designed for PC mounting and they make a simpler project to build compared to chassis mounting and point-to-point wiring everything. However I have found that it is a good idea to build a prototype of a circuit on a perforated board before etching the printed circuit board. You can easily mount PC components and instead of land patterns, you have the flexibility of point-to-point wiring that is easy to change if needed. I think of them as a small chassis. I have built a lot of prototype digital circuits over the years so I am comfortable using 30-gauge wire to interconnect the components, and some folks assert that the smaller the gauge wire, the better from a quality of sound viewpoint. I guess that I could have put everything on a single board but for flexibility of packaging I chose instead to make a separate board for each channel. They are identical except for which half of the tube is used on each. This provides the ability to get a new tube when the tubes are swapped between boards. Both the LM317 and the DN2540 are TO220 packages that need to be mounted on a heat sink. I am conservative when it comes to heat sinks and the large ones that I used get only slightly warm to the touch. These heat sinks and the transformer are both are about 1.7 inches long, which determined the width of the board. The length of the board will depend somewhat on the capacitor you pick for the parafeed. Some high quality film and foil capacitors are three or four inches long, although most are no wider than 1.7 inches. The length of the board will need to be about 5 inches in addition to what is needed for the capacitor. I had some old Wonder InfiniCaps that were the right size in my stash of goodies that I used for the parafeed capacitor. My board came to be 1.7 inches wide by 6.7 inches long. Photo 1 and Photo 2 show the assembly. I am pleased with the resultant package, a compact module that can be easily mounted in a variety of mechanical enclosures.

As luck would have it, the next meeting of the Piedmont Audio Society was held the Sunday after I had finished packaging the completed prototype boards.  I wanted an independent assessment from a fresh set of ears so I took the line stage along. The meeting was intended to audition new power amplifiers that some members had built but there were couple of new passive line stages there that we listened to first. Then came time to listen to my line stage. Well, how can I say it? The guys were kind in their lukewarm reception of the line stage. I was disappointed but they were right in pointing out a couple of deficiencies that I had introduced in building the latest version. I had not done sufficient listening of my own and let them slip by. First was a brittleness or hardness in piano crescendos which was definitely there, something I had not heard before I packaged the boards. Second was a buzz, or harsh hum under some conditions. It showed up only when the variable power supply was set at its maximum value of 250V and the volume control was set at its minimum value. With the volume set to a normal listening level there was no audible buzz. While reflecting on what I had changed in these last few steps, I realized that I had done all my earlier listening with a transformer ratio of 4:1 and then arbitrarily wired it up 8:1 for the new board because I did not need the gain. So I experimented with listening to the sound of the different ratios and found that to me the 2:1 ratio sounded best. The lower ratio cured the hardness in the piano crescendos. The cause of the buzz was harder to find. At first I thought that I had a ground problem, because even though I had been careful about a single-point star-ground system, one never knows when it comes to grounds. I tried all the tricks in my book and nothing worked.  In talking with Kevin about the problem, he mentioned that he had seen hums and buzzes caused by oscillations. I hadn’t thought about oscillations because I had a gate stopper resistor on the DN2540 and a similar grid stopper resistor on the 5687 as well as ferrite beads and a capacitor on its filament. I thought I had it covered but I was grasping at straws and ready to consider anything. I changed my tactics and started to get results. I found that when I hung my scope probe on the drain of the MOSFET the buzz stopped. However when I connected the scope to the output of the power supply in the separate power supply chassis, there was no effect. The two points were electrically the same; the only thing separating them was a few feet of wire. Okay, this would be easy – simply substitute a capacitor for the scope probe. Nope. It wasn’t giving up that easily. Next I tried a ferrite bead on the high-voltage wire without success. But I did notice something interesting. I had to move the high-voltage wire to install the bead and now the nature of the buzz had changed. I found that I could control the loudness of the buzz by moving the wire. I kicked myself – I had seen this before. In contrast to the present good-looking assembly, the original test beds were a rat’s nest of wires. Now everything was neat, with wires laid down and dressed. I hadn’t paid attention to what I was doing and had inputs neatly running together with both outputs and power supply wires for several inches. It was the distributed capacitance between the wires that got me. I re-wired the signals and high voltage wires with shielded wires, taking care to keep everything separated. This not only fixed the buzz, but also made a further improvement in the overall character of the sound. High-level transients in the music had driven the circuit into momentary oscillation. This was the hardness in the sound that we heard. While changing the transformer ratio softened the problem, it wasn’t the fix. I went back and listened to the different ratios and found that although they sounded different, there was not a problem with any of them. Difference now was a matter of taste. The 2:1 ratio was fuller, more voluminous, and a little softer. The 8:1 ratio had firmer and more bass and the sound was generally more tightly controlled. The 4:1 ratio was in between.

Earlier I had listened to various 5687s and several variants but had not auditioned either the ECC99 or the 6H30, both of which I had heard good things about. I had good success with my earlier method of swapping tubes without a protracted delay between listening sessions so I wanted to use that procedure again. Unfortunately these tubes have a pin configuration different from the 5687. The ECC99 has the same pin arrangement as the 12AU7 while the 6H30 is like the 6DJ8. However, upon closer study there are some similarities in pins 1 through 5 that I could take advantage of. Pins 1 through 3 are a triode with the same pin assignments for each tube. Pins 4 and 5 are filament pins for each tube. I was only using one-half of a tube for each channel so the fact that pins 6 through 9 were used differently for the second triode did not matter. Since I was using a current regulator for the filament, the nominal voltage did not matter and all I had to do was rig a switch to control the amount of current provided. It worked like a charm. There is a marked difference in the sound of the three tubes. Compared to the 5687, the ECC99 objectively has a little better detail and foundation. The big difference is that it sounds much more powerful, expressive and lively. However, it is almost bigger than life and I wondered if the presentation would wear thin over time. This is my subjective assessment and I must say that a lot of people like this tube. The only objective down side that I can point to is that the pair of tubes that I have is slightly microphonic. In contrast, the 6H30 is more tightly controlled, provides a presentation with less fanfare and is less dramatic than the ECC99, but perhaps is more realistic. I really like this tube – it does everything well. I tried several variations of bias point and current and settled on 5Volts and 20 mA. With the 5687 and ECC99 I could wire each channel to use a different half of the tube and provide the ability to get a fresh tube by swapping the tube between channels. Unfortunately, you can’t do this with the 6H30 because there is a single, shared filament. Rather than waste half a tube, I tried both triodes wired in parallel, which provided an even better presentation. First I simply wired the two plates together as well as the two grids and the two cathodes. The circuit immediately went into oscillation that was easy to recognize from my earlier experience. There was a grid stopper, but the extra fraction of an inch between the two grids was enough to cause a problem. Although adding the second grid stopper fixed the problem I did not stop there. In addition to the grid stoppers, I added plate stoppers. I also added a drain stopper in addition to the gate stopper on the CCS. I also added capacitors across the two power supplies and the cathode bias resistor. I expect that these precautions would have been valuable even if I had not paralleled the two sections. I ended up with the circuit shown in Figure 12 with 40 mA shared between the two triodes at 4.85 Volts bias. As part of the final design process I went back and re-assessed the parafeed capacitor. It turns out for this circuit the optimal value is 5mF.  As before, use a high-quality film and foil capacitor that suits your taste. My subjective preference here is the Kimber Kap.

Happy with the design, I proceeded to layout and etch the printed circuit board. I reduced the size of the heat sinks to be more in line with what is needed and added some holes for mounting the board as well as tie points for cable shields if needed. I wrestled with what to do about the parafeed capacitor. It took up a large portion of the board, and I wanted to make provisions for a variety of different capacitors. My solution was to provide maximum flexibility by moving the capacitor off-board, allowing for connecting it to tie points on the edge of the board.  The final board is shown in Photo 3.