Atma-Sphere S-30 MK 3.3 OTL Power Amplifier

Quality and Quantity

Equipment report
Tubed power amplifiers
Atma-Sphere Music Systems S-30 MK 3.3 OTL
Atma-Sphere S-30 MK 3.3 OTL Power Amplifier

It would be no exaggeration to label most tube amplifiers today as derivatives of designs that originated in the 1950s. And output-transformerless (OTL) tube amps are no exception. With the benefit of hindsight, it is obvious that the early 1950s were an incubator for various OTL types. From the many competing designs, two emerged that have withstood the test of time. The Futterman circuit is by now celebrated as a classic, and in its model H-3 incarnation experienced a short commercial run during the 60s and 70s. In parallel with Futterman’s work, a major OTL alternative emerged that eventually became known as the “Circlotron” in 1955, thanks to the marketing folks at Electro-Voice; a totally cool name that for me evokes a mental image of a cyclotron particle accelerator. It seems that this circuit was invented independently by several designers. In the U.S., Cecil T. Hall filed a patent in 1951 (granted in 1955) for an OTL, described as a “parallel opposed power amplifier.” In Finland, Tapio Köykkä filed a patent in 1952 for a similar topology (but transformer coupled), which he dubbed “push-pull parallel amplifier.” And in Japan in 1952 Satoshi Shimada published an OTL circuit referred to as “cross shunt push-pull.” In 1954 Alpha M. Wiggins at Electro-Voice filed a patent for a transformer-coupled Circlotron design, which strangely did not reference the Hall and Köykkä patents, but paved the way for Electro-Voice to market several Circlotron amplifiers in the mid-to-late 1950s without apparently much commercial success. Some 30 years later Atma-Sphere’s Ralph Karsten filed a U.S. patent for a circlotronic OTL amplifier and went on almost single-handedly to refine and popularize this design.

One minor complication introduced by a Circlotron is the requirement for two independent high voltage supplies to power each section of the output stage. However, its major advantage over the Futterman circuit lies in its inherently balanced topology, with both the push and pull sections of the output stage operating as cathode followers. Futterman had to resort to positive feedback in order to balance his single-ended push-pull output stage, and mistakenly believed that both output sections were performing as cathode followers. In reality, both sections worked as common cathode circuits, raising the intrinsic output impedance relative to that of a Circlotron. Futterman decided to apply negative feedback aggressively in order to reduce both output impedance and distortion products.

Atma-Sphere’s Circlotron-based S-30 OTL amplifier reviewed here is by now a mature product. The latest iteration, the Mk3.3 Series of amplifiers, essentially incorporates the driver circuit of the next amp up in the lineup from the Mk3.2 version. That is to say, the S-30 now has the driver of the M-60, while the M-60 now has the driver circuit of the MA-1, and so on. Five 6AS7G twin power triodes are used per channel and are biased for Class A operation. The 6AS7G certainly qualifies as an OTL output tube due to its low plate resistance; you just need a lot of them in parallel due to its rather modest 13 watt plate dissipation per section. Atma-Sphere now uses mostly the Russian 6H13 rather than the Chinese 6N13. Both are 6AS7G equivalents, and according to Ralph Karsten, the Russian tube is a bit more robust but otherwise should sound identical. If all goes well and the amp isn’t pushed hard all the time, Ralph expects the lifetime of the power tubes to exceed 10,000 hours. Each half of the output stage comprises five triode sections connected in parallel for increased current delivery. Even without global negative feedback the theoretical output impedance should be about 14 ohms. Karsten uses a miniscule 2dB of feedback to further reduce the output impedance, the official impedance spec being 7 ohms, while I measured 10.6 ohms at 1kHz.

The reason to dwell on this topic is that an amp’s output or source impedance interacts with the loudspeaker’s impedance magnitude to alter the speaker’s frequency response. These EQ effects are magnified whenever the load impedance is on the order of the amp’s source impedance. Instead of delivering a constant voltage, high source impedance amps begin to approximate constant power delivery. That’s not necessarily a bad deal for a vented speaker with large impedance peaks in the bass range. Constant power means a bass response boost for such a speaker, and in some cases, welcome bass extension. That, for example, was the basis of the Fisher Z-Matic circuit in the 1950s, which allowed the user to adjust the amplifier’s source impedance continuously in order to control the speaker’s bass response. My Basszilla DIY speaker expects to see several ohms of source impedance. But at 7 to 10 ohms the effect was to over emphasize the deep bass. The solution was to insert acoustic foam plugs into the bass-reflex vents. That resulted in a Goldilocks bass balance—just right for my listening room!

The Basszilla was right for the S-30 for two other reasons. With a sensitivity of 96dB, its appetite is satisfied with merely a handful of watts, so the S-30’s 30 to 45Wpc represented more than an adequate power reserve. In addition, its impedance magnitude is reasonably benign. It’s important to realize that any OTL is a specialty amp that requires care in partnering with a prospective speaker. Sadly, we live in a world where the norm is to deconstruct an audio system into individual components; by and large, system-matching has become a lost art. Limiting current delivery into a potential speaker load is a major consideration and there’s no question that any OTL would be more comfortable with a 16 ohm rather than an 8 ohm nominal load. The reality is that most speakers today inhabit the 4 to 8 ohm impedance universe—16 ohm speakers are almost impossible to find. The next best thing is an 8 ohm speaker, and by all means, avoid 4 ohm loads. Ultimately, the sonic destiny of any OTL is closely dependent on the associated loudspeaker. What you think it sounds like will depend on your choice of speaker.

Right out of the box my review sample exhibited a fairly loud buzz in the left channel. It was loud enough to be heard at the listening seat during quiet musical passages. Karsten indicated that this issue has occasionally popped up in the field after an amp left the factory and attributed the cause to a power supply resonance, an interaction between the inductance of the power transformer winding feeding the rectifiers and the capacitance at the rectifier junctions. Apparently, standard silicon rectifiers when properly snubbed are usually silent even with high-efficiency loudspeakers. But when an S-30 buzz issue arises, the fix is to replace them with a high-current HEXFRED bridge rectifier that once neutralized is super-quiet. The buzz was too annoying, and so my sample was off to the factory for replacement of the rectifiers. This did the trick. The right channel was dead quiet and the residual buzz in the left channel was now almost inaudible to the point that I had to move my ear next to the speaker to hear it.

The tube complement is totally octal. There are three 6SN7s per channel, apparently of Chinese origin, which are configured as voltage gain and phase-splitter stages. The S-30 accepts both balanced and unbalanced inputs. Since none of the preamps in my current portfolio are balanced, this review was limited to using the RCA inputs. When using the RCA inputs, make sure that shorting jumpers are installed on the XLR connectors between pins 1 and 3. The turn on sequence is also worth mentioning. It’s essential, at least if you wish to extend output-tube lifetime, to first turn on the power switch on the left side of the chassis. This allows all of the filaments to warm up, and Atma-Sphere recommends waiting one to two minutes before activating the standby switch on the right, which applies high voltage to the power triodes. Since the S-30 is directly coupled to the speaker without a blocking capacitor, it is possible for significant DC offset current to flow through the load. DC offsets for both channels can be checked using the front-panel meter and nulled out using dedicated pots. This meter was good enough to tweak the DC offset voltage to less than 50mV. If you desire more accuracy connect a digital volt meter (300mV DC scale or better) to each set of speaker terminals. This way you can adjust the DC offset to near zero. But in all honesty, this is probably overkill since DC offsets of even 50mV are perfectly fine.