Tip #79
Q-tron PA-12 OTL Tube Amplifier
by Dick Olsher (May 2012)

PA12-Front2
Even 60 years after the concept gained some traction, the notion of an output transformerless (OTL) tube amplifier still seems mind blowing. Transformation of the high-impedance and low-current tube plate circuit into the high-current and low impedance needs of a loudspeaker appears to require the services of a good output transformer. Early OTL attempts were handicapped by low power and the need for a special high-impedance voice coil speaker. Circa 1951 Arnold Peterson and Donald Sinclair described a single-ended push-pull circuit, suited for class AB operation, that became the de facto output topology for most future OTL designs. Its major disadvantage was unbalanced drive, a problem that Julius Futterman solved elegantly in 1954. Yet, issues of reliability and damping factor continue to dog many OTL designs. Q-tron’s Hans Beijner believes that he has solved these issues with the introduction of the PA-12.

PA12B
He began building tube amplifiers when he was 15 years old, and that was some 40 years ago. He holds a MSc. degree in electronic engineering and has been working as design engineer and in various other positions at Ericsson (provider of cellular mobile systems), specializing in radio technology. Although Q-tron was founded in 1976, engaged in providing analog electronics consulting services, the idea to produce and market OTL amplifiers is a fairly recent decision. Only after the realization that his design “was performing very well compared to commercial OTLs” did he start to think about commercializing his amplifier. At this point he also “invented” a new driver stage capable of reducing distortion by a factor of 10 without increasing feedback. The PA12 uses about 28 dB of feedback which is rather low compared to Futterman’s original OTL which used more than 40 dB. I will now turn the podium over to Hans Beijner for a detailed technical overview of his OTL design. It’s clear to me, as it should to you, that this gentleman possesses a deep knowledge of the subject.

Hans Beijner’s Own Story
“I started to be interested in OTL's a long time ago when I found an article in an old HiFi magazine describing an amplifier using the Philips EL86 tube. The EL86 was designed especially for use with Philips own speakers with 800 Ohm impedance. Many TV sets made by Philips used these tubes and speakers. The article I found described a more developed version of the Philips design amplifier and data was fantastically good with an output power of 14W, a frequency response flat to 200 kHz and a distortion at 12W of 0.08%. This article was written in the early sixties and it should be remembered that there was no solid state amplifiers at that time that could show similar performance (practically all solid state amplifiers at the time were transformer coupled with low bandwidth and high distortion). I did some experiments in the seventies using various output tubes among them PL519 coupled as triode but my conclusion was that in order to get high output power and to be able to use conventional speakers I needed to use many tubes in parallel and also the heat generated by such an amplifier would be excessive.

Operated as a pentode, a PL519 can pass 1.5A at an anode voltage of only 55V. If it would be possible to build an OTL where the tube was operated as pentode the power dissipation would be much less than if operated as triode. Around 1986 I designed an all DC coupled OTL with PL519 operated as pentode. After some work with that circuit I decided that this was not the ideal solution even if it actually would work so I gave up and didn't really do any more serious thinking about OTL's until I found out about the 6C33C tube. As the 6C33C can handle twice the current of the PL519 you need fewer tubes which will improve reliability.

How do you really design an OTL amplifier with high reliability? Reliability is not what you spontaneously think about when you hear the acronym OTL. OTL amplifiers have often instead been synonymous with low reliability, using many tubes and very low efficiency.

  • Use as few tubes as possible and don’t overload them.

    Most commercial OTL amplifiers have used many parallel connected tubes in order to provide enough output power. These tubes have often been biased at maximum of their specification or even beyond that. When one tube inevitably fails, the other tubes would be even more overloaded and this could start a chain reaction often taking out several or all tubes at one time. The solution for being able to use few tubes is both to find the circuit that can give the lowest possible output impedance and to find the most potent output tube available. The OTL circuit giving the lowest output impedance is the so called inverted Futterman configuration which gives an output impedance which is less than half of the commonly used Circlotron. The most potent output tube available today at a reasonable price is the Russian 6C33C. This tube is able to withstand continuous anode currents of 600mA with peaks of more than 2.5A while giving excellent service life. The first prototype OTL I designed with this tube is still equipped with the original 6C33C tubes which are still working well despite being used for 10 years. The PA12 uses only 2 tubes of type 6C33C to provide 25W of output power, the tubes are biased for an idle dissipation of 30W which is only 50% of the allowed maximum.
     
  • Lower distortion without extreme feedback by the use of distortion cancellation.

    As OTL amplifiers are not equipped with output transformers there is almost nothing to stop you from using very large amount of negative feedback, much more than what is normally used in tube amplifiers with output transformers. Original Futterman OTL amplifiers use more than 40dB of feedback and are still unconditionally stable. As these OTL amplifiers are push-pull connected even order harmonics are largely cancelled and the result is that 3rd order distortion dominates. I started to play with the idea that it would be possible to actively cancel the 3rd order distortion and as the even order distortion already is relatively low it would give the possibility to build an amplifier with lower feedback but with even lower distortion than the original Futterman OTL. The result is the PA12 with more than 15dB lower feedback than the Futterman but distortion is more than 10 times lower reaching levels as low as 0.01% at 1W output power.
     
  • The Inverted Futterman Circuit

    The Futterman circuit was of course invented by Julius Futterman in 1954. This circuit provides balanced drive to both tubes by returning the cathode of the phase splitter to the output, this makes the circuit balanced and both output tubes actually operates as ordinary cathode coupled amplifiers, (even though Futterman thought that they worked as cathode followers). The output impedance of this circuit is Rp/2, where Rp is the plate resistance, so for two 6C33C the output impedance will be ~80/2 = ~40 Ohm.

    The inverted Futterman circuit was first described in 1955 in a paper titled "Analyses of Drivers for Single Ended Push Pull Stage” by Hiroshi Amemiya, who at the time was a doctoral student at Tokyo University. In the inverted version of the Futterman, both tubes act as cathode followers and the output impedance is Rp/[2*(1+µ)], where µ is the tube gain, so for two 6C33C this will be ~80/(2*(1+2.7)) = 10.8 Ohm. This is a major advantage over the standard Futterman circuit in that the output impedance is ~4 times lower which means that fewer tubes need to be used. The balance in a well designed inverted Futterman stage can be made as good as in a normal push-pull amplifier, i.e. all even harmonics cancel and it is quite easy to get as much as 40dB cancellation of 2nd order distortion.
  • PA12-schematic
    The way the inverted Futterman configuration improves balance can be described as follows: In the series-connected output stage the lower output tube acts as an ordinary cathode grounded stage with gain while the upper tube works as a cathode follower where gain is below 1. By returning the cathode resistor of a split load phase inverter to the speaker output and by connecting the cathode of the phase splitter to the upper output tube and the anode of the phase splitter to the lower output tube, gain in the two output tubes can be equalized. The signal coming from the anode of the phase splitter is reduced by negative feedback from the speaker output but this feedback is not affecting the cathode signal from the phase splitter as the cathode potential is controlled mainly by the grid potential. The result is that the signal coming from the anode of the phase splitter is reduced just the right amount so that the excessive gain of the lower output tube is compensated for.

    Circuit description of the PA12

    The basic design is 10 years old. Distortion cancellation and active noise cancellation has been added more recently. The input signal is amplified by a 12AX7 tube where both triode sections are connected as a SRPP stage. The signal from the SRPP is connected to a cathode grounded stage using half of a 12BH7 tube. The anode of this stage is DC connected to the grid of the 2nd half of the 12BH7 which is configured as a split load phase inverter. The outputs of the phase inverter are connected to the grids of the two series-connected 6C33C output tubes in the inverted Futterman configuration. In this configuration the anode of the split load phase inverter is connected to the lower output tube and the cathode of the split load phase inverter is connected to the upper output tube, the other side of this cathode resistor is connected to the speaker output. A 5751 tube is also included in the driver circuit and performs distortion cancellation lowering distortion by more than 10 times. A 5814 tube is used for active noise cancellation and reduction of hum and noise to very low values.

  • Design choices

    The power supply is conventional, voltages are not regulated, just filtered and all heaters run on AC, hum and noise can still be kept to very low values as care is taken to given that the design and layout is done correctly. Toroid type transformers are used in the power supply because of size and lower radiated noise. Components are chosen for its performance qualities and not by name, for instance many capacitors marketed as being especially made for audio purpose are available with different brand names for more reasonable prices. The chassis design is made after the principles of not having any visible screws and there are no components except tubes visible; many tube amplifiers have transformers visible but this is an OTL so I thought that the image would be spoiled if any transformers could be seen.

    Concerning damping factor, or rather output impedance, I aimed for about 0.4 Ohm which results in a small effect on the frequency response of most 8 Ohm speakers. I could easily gone further but the more feedback you apply the more complicated it gets to make the design stable without adding a lot of compensation networks, as of now the amplifier is unconditionally stable with any load and there is no overshoot on square wave response.
    PA12C
    The power supply consists of two toroidal transformers, the big one supplies all heaters and the output stage anode supply, while the small one supplies anode voltage for the driver tubes plus all bias voltages for the output tubes.”

A Few More Technical Details
A monitor interface box, supplied with the amplifier, connects to the multi-pin connector on the back panel. Using a digital multimeter, bias current and DC offset can be monitored. Pots, accessible through the back panel, allow adjustment of the bias current and DC offset voltage for each channel. The Manual describes the process in some detail, though it does take patience and a bit skill to nail the recommended settings as there is interaction between the pots. It would have been helpful to have the pots labeled, something that Hans assures me will be done for the next production run. Bias readings between 100 – 150 mV are equivalent to a bias current in the range of 200 – 300 mA. My preference is to zone in on bias reading in the range of 125 to 130 mV. Lower bias currents appear to reduce soundstage transparency. With a bit of practice it is possible to zone in on this bias current range while maintaining DC offset below about 10 mV. The good news is that there is little drift in bias current and DC offset over a period of even several weeks.

Without a blocking cap in the output, and lacking a perfect balance between the push and pull power tubes, there will always be some DC current flowing through the loudspeaker. The DC offset during normal operation is not regulated by any special circuit but is said to stay below ~50mV even when tubes are ageing. There is, however, a DC offset protective circuit that works off a window comparator and trips the anode supply when it senses a DC offset in excess of 0.8V.

The amplifier runs hot to the touch, especially the chassis top. Hans tells me that the amplifier’s power dissipation is 340W, a combination of heater and anode dissipation, with about 170W for the heaters and the rest for the anode supply. The heaters in each 6C33C are series connected and are powered by AC 12.6V from the main power transformer. Even so, each tube draws 3.3A on 12.6V. I used the amp on a dedicated 20A circuit in my listening room. When the power switch is depressed a soft start circuit is activated that switches on the mains supply after a delay of about 3 seconds. Then after a further 90 seconds delay the HV anode supply is switched on and the amplifier is ready for operation. Hans tells me that the next production will include a few updates, improved cooling by better circulation around the output tubes, and an even better stabilized operating point to further reduce bias current drift.

Sonic Impressions
The PA-12 was going to have to sink or swim while driving the Basszilla Platinum Edition mk2 DIY speakers. The Basszilla, being a 96 dB sensitive loudspeaker, is quite comfortable with low-power tube amplification. And field feedback had it the coupling of the Basszilla with a Futterman style OTL sounds superb, so I expected this to be a successful marriage. As you will shortly see, I was not disappointed.

What struck home rapidly was the PA-12’s startling textural purity. I’m a huge fan of the 6C33C power triode, but I’ve only heard it previously in the context of conventional push-pull amplifiers, as in Balanced Audio Technology designs. While such designs do a good job in eliminating even order harmonic distortion products (except when the phase splitter is unbalanced), they typically end up with a residue of odd order (mainly 3rd) distortion products that imbues textures with a forward, or in the extreme, an aggressive presentation. There was none of that in evidence with the PA-12. Violin overtones flowed with an incredible sweetness of tone. It felt as though this amplifier was providing a closer approach to the intrinsic sound of the 6C33C – high density textures and robust tonal colors. It never sounded anemic. The PA-12 was able to deliver a big tone sound when the program material demanded it.

A remarkable combination of transient speed and treble finesse was on display. Struck cymbals shimmered and decayed with life-like conviction. There was no hesitation at the point of attack. This is an area where cap-coupled Futterman designs stumble. No matter how you slice and dice it, having a large DC blocking cap between the output tubes and load is detrimental to transient speed and bass definition. No such problems exist when the output stage is directly coupled to the load. Treble detail flowed naturally, organically, and with complete control. Bass definition was exceptional, no doubt aided by a low source impedance. Bass lines were generally well controlled with superb pitch definition. If there’s any complaint I would issue against the bass range, it would be a slight reticence in terms of bass slam. At the risk of stating the obvious, the PA-12 clearly lacks the insane current drive capability of a Pass Labs or Ampzilla solid-state power amp.

The soundstage received very special treatment as well. Layers of veiling were peeled away making it exceedingly easy to step into a recording’s acoustic space. The depth perspective was finely layered and image outlines were palpably rendered within a cohesive left-right perspective. Musical lines boogied along with plenty of kinetic energy. Microdynamic nuances were allowed to bubble to the surface convincingly, making the listening experience so much more emotionally involving.

Most impressive was its exceedingly low noise floor and the resultant ability to resolve low-level detail. Recording chain attributes such as mike quality, tube vs. solid-state electronics, and miking distance were readily revealed. On heavily processed multi-tack recordings, vocal overdubs, reverb effects, and pan potting tricks were easy to spot. It would be fair to say that the PA-12 resolves as much detail as I’ve have ever experienced with any amplifier irrespective of cost.

If you have ever wondered how a tube amp might sound without gratuitous second and third order harmonic distortion, well the PA-12 provides the answer; in a nutshell, stunning transparency, clarity, and detail resolution. It is, in the finest meaning of the term, a reference-grade amplifier in that it does not hide program material imperfections behind a euphonic veneer. I know that it’s a class AB design, but my ears tell me a different story: it sounds much like a fine pure class A amp! Speaking of class A amplification, it seemed like a good idea to pit the PA-12 against some of the single-ended amplifiers I happen to have on hand. First challenger was Pete Millett’s Glowing Hybrid which uses an op-amp output stage. No output transformer here, but the PA-12 came across as more liquid sounding, with more palpable image outlines, and it captured the sensuous quality of live music with greater conviction. It outdistanced a 6B4G based SET amp in terms of soundstage transparency and bandwidth. Only when it came to T-Rex, a 300B based SET amp, was the PA-12 edged out, and only in the midrange where T-Rex is the champ when it comes to reproduction of dynamic nuances.

Conclusion
I’ve auditioned plenty of OTL amps over my audio lifetime, and that includes the David Berning ZOTL types. I can confidently state that the Q-tron PA-12 beats them all. The combination of an inverted Futterman topology together with distortion and noise cancellation circuitry has resulted in a tube amplifier that redefines what is possible in a push-pull design. If your speakers and listening environment are happy with 25 wpc, then the PA-12 is the amplifier to own. It delivers the full promise of OTL technology. A toast, cheers, skål, to Hans Beijner on a fantastic achievement in both technical and musical terms. Highly recommended for audiophiles and music lovers alike.

Specifications
Output power: 25W into 8 Ohm at less than 1% distortion
             20W into 4 Ohm at less than 1% distortion
Distortion, THD at 1W: < 0.01%
Distortion, THD at 10W: < 0.1%
Frequency response: 5Hz - 150kHz, -1dB
Power bandwidth: 10Hz - 100kHz, - 1dB
Rise time: < 0.5us
Hum and noise: < -100dBA
Output impedance: < 0.4ohm
Dimensions: 430 x 250 x 290 mm, (W x H x D)
Weight: 20 kg
Price: €8,000 in Sweden; US retail to be determined

Company Information
Email: sales@qtron.se
Website:
qtron.se