The K&K Audio ST70 Boards

The long-tailed pair is sometimes referred to as a “differential amplifier”. You will find that these generic terms encompass many variations, some of which have little to do with the long-tailed pair used in audio. Let’s summarize the audio-friendly features/benefits of the long-tailed pair:

Almost perfect AC balance, making tube or transistor amplification relatively unimportant. This means that if one tube or transistor pushes +10VAC, the other tube or transistor will be forced to pull -10VAC. Because their cathodes, or emitters, or sources are tied together, and placed on top of an active load, their AC will always be balanced, giving you most of the possible benefits of push-pull or balanced operation. Note: Almost perfect balance can only be achieved with active cathode loads.
Very high Common Mode Rejection Ratio, or CMRR. This is the ability of a circuit to reject noise that is common, or “the same”, on both inputs. This applies to the signal inputs and power supply inputs. So, if hum is present in equal amounts on both inputs, and on the power supply connections, they cancel out. Why? This is because a “differential amplifier” amplifies the “difference” between the inputs, not what is “common” to both inputs. Typically, noise picked up by balanced connections, and power supply noise, can be reduced by at least 60dB. Many op-amps can achieve 90dB CMRR. Very high performance IC’s can go over 100dB.
The long-tailed pair is one of the most effective audio phase splitters. Typically, a differential amplifier has both a positive and negative input. However, if you have a single ended connection, you can drive one input of the long-tailed pair, ground the other input, and the output of the circuit will be “push-pull”, creating two outputs that are out of phase with each other. One thing to remember about a long-tailed pair is that they usually operate in class-A, so both outputs are100% class-A copies of the input, but one copy is the opposite polarity. The bottom line is that the output is balanced class-A, not a mangled class-B or AB. When you hear someone talk about class AB1, class AB2, class-A or class-B, they are usually talking about the biasing of the output circuit, not the input or driver circuitry.
Especially when implemented with two triodes, the distortion is extremely low. This distortion reduction from balanced operation is mostly due to the ability of the circuit to cancel out even-order harmonic distortion. Unfortunately, balanced operation doesn’t cancel out higher order harmonic distortion. The “fix” is to use an active device with primarily even-order harmonic distortion, and as little odd-order harmonic distortion as possible. The triode fits the bill, with mostly low amounts of even-order harmonic distortion, and negligible amounts of odd-ordered harmonic distortion.
They offer moderate, to high, input impedance that makes it easy to drive by the preceding stage.
They offer moderate, to low, output impedance which gives it the ability to drive succeeding stages with low distortion.

Seeking Perfection

As superior as the HF89 circuit is to the ST70 circuit, there is still room for improvement. Both these amps use profligate feedback to maintain linearity. As good as the long-tailed pair is as a phase splitter, it still produces significant distortion in the process of splitting the phase because it is not being driven by a differential input, but by a single ended input. When the long-tailed pair, or complimentary pair, is driven with a balanced signal, distortion and noise common to both inputs will be mostly canceled out, while the audio signal is amplified. But, when used as a phase inverter, the long-tailed pair also amplifies the noise, with the signal, because it is driven by only one input. So, the long-tailed pair, when used as a phase splitter, is prone to amplifying noise and generating harmonic distortion. You only get the maximum distortion and noise canceling benefits of a long-tailed pair when the inputs are differential.

So, what are we going to use in place of the ST70 board? For me, that was an easy decision. I’ve heard different replacement boards, and there have been many, but none have the transformative powers, pun intended, of K&K Audio’s ST70 board.

K&K Audio
1004 Olive Chapel Rd.
Apex, NC 27502-6745
(919) 387-0911 voice/fax
www.kandkaudio.com/poweramplifier.html

ST70 replacement board kit prices:

Basic Kit: $285
Basic Kit without coupling capacitors: $260
Premium Kit with Cardas coupling capacitors: $375
Premium Kit with Jensen copper foil/paper in oil coupling capacitors: $425
Premium Kit without coupling capacitors: $340

I am also reviewing the unit with the shunt voltage regulator boards for the input tubes (this is an optional extra):
Shunt regulator Upgrade Kit for K&K Audio ST-70 Kit $79

As I’ll get to later, the K&K boards, along with Dynaco iron, might make for the best amplifier bargain of all time. The K&K board still only sports two 9-pin miniature tubes, but instead of having a compromised triode-pentode tube, like a 7199 or 6GH8, the K&K board uses an input phase splitting transformer from Lundahl. The transformer, which also offers modest gain, is followed by a 6n23p dual-triode, configured as a long-tailed pair with active cathode loads. The two 9-pin tubes are used as drivers for the output tubes; though the 9-pin tubes also amplify the signal, I prefer to think of them as driver tubes. So, the signal goes into the input transformer, where the single-ended input is converted to a balanced output, which then drives the 6n23p, which add both voltage and power to drive the output tubes. The compromise-in-a-bottle 7199 pentode/triode tube is replaced with a transformer with almost no audible distortion or noise, and a 6n23p, the Russian version of the 6922. By the way, the input transformer also eliminates ground loops and RFI. While cheap transformers have their own peculiar types of distortion, even mediocre examples produce distortion that is very difficult to hear, and only bring attention to themselves at extremes of frequency and loudness. The units from Lundahl are very good quality, have low levels of harmonic distortion and absolutely no spurious noise.

“Ah, but what about all those heat sinks? Is it a hybrid?” No, although we have active transistors on the amplifier, they are in a support role only. What we have here is an ultimate expression of the “long tailed pair”. Both the 6n23p driver tubes, and the output tubes, are configured as long-tailed pairs. The K&K design maximizes the performance of the long-tailed pair by presenting very high AC impedance at the two cathodes. Consider the HF89 long-tailed pair: instead of an 18 KΩ cathode resistor of the HF89, the K&K circuit presents an AC impedance of several megohms, preventing AC “leakage”, and forcing almost perfect AC balance. It does this with an active “constant current sink”, which is the same as a “constant current source”, but used on the cathode, instead of the plate. The constant current sink regulates the amount of DC current that the tubes are allowed to use, while presenting an almost ideal AC impedance approaching infinity, meaning several million ohms is close enough to infinity in an audio product. If you added this constant current sink to the HF89, the plate load resistors would then be the same at 33KΩ, instead of 28.75KΩ and 33KΩ, because the AC balance would be nearly perfect.

Because the phase-splitting is done by a high quality transformer, and because all the tubes are run in very linear Class-A, with most of the distortion canceled out because all the tubes are acting as long-tailed pairs, and because the input transformer blocks RFI which might cause instability, and because the amplifier has moderate “open loop” gain, the amplifier does not need feedback and is stable without feedback. Without the use of feedback, you don’t need a high gain input pentode or 12ax7, since the 12ax7 has a gain of 100. The use of feedback requires the use of high gain tubes, which requires even greater reliance on feedback, just like acoustic feedback, electrical feedback tends to promote itself. Higher gain usually means greater instability, or susceptibility to instability, and greater likelihood that it will oscillate. By ditching feedback and choosing a low distortion circuit, you don’t need very much gain. The gain of the input transformer can be either 0 or 6 decibels as selected by the user, and the amplification factor of the 6dj8, at least on paper, is 33. That’s significantly less than the gain of the 7199, but because feedback has been eliminated, you do not need as much gain. Loop feedback, as employed in most classic audio gear, is the audio equivalent of a Ponzi scheme.

Hopefully I have convinced you of the efficacy of the design. In part 2, we will look at tearing down the stock ST70, fitting larger transformers, powder coating, and board assembly. In part 3, we will listen to the results.

Manufacturer’s Comment:

Constantine,

It’s rare that I have a wide smile on my face throughout a review reading, but Phillip’s emphatic style and the accompanying imagery did it. His explanation of the topologies illustrated was very good and pretty extensive, enough for those more technically interested to benefit from.

Inevitably in the course of a lengthy review period things can change and with this kit they have (or more correctly, will shortly). The kit will continue to be available and has the same circuit topology that Phillip describes with virtually the same parts set, but the new version will have enhanced abilities in that the output stage circuitry can be operated at higher voltage and current resulting in higher power output. The Basic version, which was rarely purchased, has been discontinued and the prices for all of the Premium variants will rise by $20, reflecting very modest parts cost increases. In addition a K&K Audio power transformer that fits the ST-70 footprint will be available shortly to facilitate the use of the most excellent TungSol KT-120 tubes. In addition, it will be possible to fit a Lundahl output transformer in place of the Dyna A470.

Best regards,

Kevin