Reducing Output Stage Sensitivity to Noise
In the last post, the effort was about reducing power supply noise supplied to the amplifier from the get-go. This time, the effort is to make the output stage less sensitive to the noise to begin with -- this because the output stage receives the least amount of filtered B+ voltage in the design. This effort can have additional benefits as well.
From a theory standpoint, the output stage will be least sensitive to power supply noise if the output tubes are perfectly balanced with respect to current flow. The B+ voltages are applied to the output stage in a "common mode" fashion, so that both output tubes of the push-pull circuit receive the same power supply noise that rides on top of the B+ voltage. Since both sides of the push-pull stage receive the exact same noise signal then, the noise component will cancel out in the OPT if the output tubes are perfectly balanced with respect to DC current flow.
In the original design, Eico counted heavily on this basic principle of common mode cancellation through balanced operation. They did not go so far as to specify that the output tubes be matched pairs, as this was first and foremost a kit unit with compromises made to maintain economy. But there is no doubt however that the use of matched output tubes will certainly help towards achieving that end.
To minimize output stage sensitivity to power supply noise then, the tubes must operate with a balanced current flow through them. But balancing the quiescent current flow in the output stage has two other benefits as well, one of which is quite significant.
For starters, when the DC current flow is balance through both halves of the output transformer, not only is is very immune to common mode power supply noise, but it also makes the OPT less susceptible to magnetically induced noise as well.
In the HF-81, the power transformer is quite large in comparison to the OPTs, so there is plenty of magnetic noise impressed upon them. The closest OPT is turned 90 degrees to help minimize any coupling effects. But with so large a power transformer, some coupling can still take place -- and it does. Just give a listen close up to the speaker of Channel 2 when the unit is first turned on. Before the tubes warm up, you will hear the hum of magnetic coupling taking place between the power transformer, and this closest OPT. In the subject factory built HF-81, this OPT was mounted well off of perpendicular to the power transformer, which allowed the OPT to then pick up more noise than an optimum mounting would. For this OPT then, proper orientation is important. Channel 1's OPT is mounted with the same magnetic orientation as the power transformer, so the only help it receives is from being mounted at some distance from the power transformer.
In both cases however, having the DC current flow balanced through each half of the OPT primary winding biases the winding to a higher energy level than that of the available magnetic noise from the power transformer, making the winding rather immune to the magnetic noise impressed upon it. How well balanced the current flow is determines how immune the winding is to the external magnetic noise impressed upon it.
From an audible standpoint however, providing a balanced quiescent current flow through the OPT always enhances the low frequency performance of the OPT. With balance current flow, LF distortion drops to a minimum, and maximum LF power transfer is obtained. When the currents are not balanced, the magnetizing effects in each half of the primary winding don't effectively cancel each other out. Depending on how great the mismatch is, this can cause the core of the transformer to saturate earlier than it would with an otherwise balanced condition. Therefore, balancing the quiescent current flow through the OPTs not only makes them have greater immunity to common mode injected and magnetically coupled noise, but also allows them to achieve optimum LF performance as well. The OPTs of the HF-81 have a bigger core than many EL84 based amplifiers, but are hardly the biggest by a long shot. Balancing the quiescent current flow in them then can have a significant impact on their LF capabilities.
But there are other benefits to balanced operation also. Invariably, in push-pull output stages, one tube begins to become a current hog, drawing more of the current, which in common cathode resistor cathode biased designs, inherently causes the other tube to draw less current. This causes the hog tube to wear out faster, and even potentially over dissipate itself which can lead to its demise even faster. As a result, one popular modification is to raise the value of the cathode bias resistor to lower the overall dissipation levels in the output stage -- not only to account for today's higher line voltages -- but further to the point that even if one tube becomes a hog, it won't go chernobyl in the process.
Raising the cathode bias resistor slightly to account for higher line voltages is fine, but raising it further to account for the "what ifs" only serves to reduce power output and increase distortion. Therefore, maintaining a balanced current flow through the output stage also serves to allow maximum performance from the output stage, while achieving maximum tube life from it as well.
So while balancing output stage quiescent currents produces many benefits, Eico didn't provide any means to accomplish this other than through the use of matched output tubes -- which are rarely "identically" matched when new, or if they are, rarely stay so over time. Therefore, some means of accomplishing balanced output stage operation would be a very worthwhile improvement for the HF-81, with positive impacts on hum, LF performance, and tube life.
A suitable DC balancing circuit was then devised, along the same lines as those used by Williamson in his original amplifier. I modified his basic design with component values as appropriate for EL84 type tubes, and improved it by including matched 10 ohm 1/4 watt resistors in each cathode leg, which allows a simple voltmeter to be used to adjust for a 0.00 volt difference between two test points (the two cathode terminals), indicating perfect quiescent balance. This is similar to the same approach Heath adopted in the W-5M, but only requires one cathode bypass cap. The 1/4 watt resistors not only allow for an easy DC balance adjustment to be made, but also provide protection for the OPTs should a tube decide to become socially unacceptable.
A terminal block was mounted atop the rear of the chassis to effect the test point connections, allowing for easy balance adjustments to be checked or made, without removal of the bottom cover. Channel 2's balance control occupies the former position of the hum balance control (which is why it was moved), while a new mirror image hole was drilled to accommodate the balance control for channel 1.
Pics include:
1. The output stages modified to include DC balance capabilities in each channel. The stock value cathode bias resistors were maintained, while additional resistive elements of the balance circuits effectively raise this value to about 185 ohms. This causes the tubes to operate at a conservative 80% of the Design Center dissipation rating for the tubes.
2. Top side, the two DC balance controls are readily apparent, as is the terminal block at the rear of the chassis to allow for easy meter connections to make the adjustments. The shielded cable to the panel lamp is also visible in this pic.
Next up, quieting down the line/tone amp stages.
Dave