25 February 2024                                                    Post Number 597

Acoustat Trans-Nova Power Amplifier
Let me be frank: most solid-state power amplifier schematics bore me silly. Staggeringly similar, always the same three stages, input, voltage-amplification stage (VAS), and output stage (OPS)—this is the proven Lin solid-state amplifier topology that dominates the audio industry. No electrical engineer was ever fired for designing yet another Lin-arrangement power amplifier. But I am also sure that no one ever got a bonus, raise, or promotion due to his cookie-cutter effort. Its eventual demise will most likely be brought about due to a replacement technology, such as class-D or a true digital amplifier, not due to an alternative linear amplifier topology, much in the same way that international travel switched from ships to airplanes, not to faster and better ocean-liner ships.

In the strongest possible contrast, the Trans-Nova-amplifier topology departs from all the solid-state amplifier conventions. For example, three feedback loops are used; its VAS is a push-pull design, not a single-ended one; its output stage (inverting) delivers signal gain and requires a floating bipolar power supply and holds no current-limiting circuitry; none of its output power MOSFETs (six of them per channel) get a source resistor; instead, all its output-MOSFET sources terminate directly to ground; and no bipolar transistors are used, only MOSFETs.

Before we move on to its schematic, let's give the standard Lin topology, the 1950's solid-state amplifier design by Mr. Hung Chang Lin of RCA, a quick review. (Several times, I have been told that the Brits created the Lin topology, with either UK companies, Linn or Naim, or Mr. Douglas Self getting the credit. The Linn-Lin name confusion I can understand, but I doubt that Linn or Naim or Doug existed at all in the 1950s.)

From my post 574:

The Lin topology holds an input stage, a voltage-amplification stage (VAS), and an output stage, usually a voltage-follower stage that delivers no voltage gain, but does deliver huge current gain. The key requirement for the input stage is linearity, not gain, as it is the VAS that will deliver the bulk of needed signal gain. The differential input stage places the inverting (Q1) and non-inverting (Q2) inputs at the same voltage potential, which makes for an easy application of negative feedback. Interestingly enough, the differential input stage functions as a voltage-to-current amplifier (i.e. a transconductance amplifier), while the VAS gain stage functions as a current-to-voltage amplifier (i.e. a transimpedance amplifier), which converts the current swings into voltage swings to drive the output stage.

This is the near universal solid-state amplifier topology. Of course, we can get extra fancy, adding current mirrors and cascoded transistors and output current-limiting circuitry or by making the circuit fully complementary symmetrical, but the core Lin topology remains constant. The only features the Transnova shares the Lin topology is the differential input stage and one of its three feedback loops. By the way, "Transnova" stands for TRANSconductance NOdal Voltage Amplifier. If we accept (the American art historian) Bernard Berenson's definition of genius, "as the capacity for productive reaction against one's training," then surely James C. Strickland, the inventor of the Trans-Nova topology, was a genius.

Click on schematic to see enlargement

Strickland labels the three feedback loops A, B, and C. A stands for anisotropic; B, Black classic negative feedback, after Harold Stephen Black, whose 1927 invention of negative feedback revolutionized electronics; and C, complementary, which means positive feedback.

Here is the big-picture overview of the amplifier:

The first stage is non-inverting, but the following two stages do invert, which cancel, so the output is non-inverting. Thus, the B negative-feedback loop works the same way it would in a more conventional amplifier. Complementary feedback is simply positive feedback, which increases the first stage and second stage gain. The anisotropic feedback loop transforms the output stage into a current-to-voltage converter circuit; it functions as negative feedback.

By the way, "isotropic" means identical in all directions; invariant with respect to direction. "Anisotropic" means that direction matters. For example, a plank of wood presents different attributes depending on the direction, being stronger or more pliant in one direction, due to its wood grain; thus, it is anisotropic. If we grind up the wood and make paper from it, however, we get the resulting isotropic paper, as the wood grain no longer exists. In other words, we really only need two adjectives to describe the three feedback loops: negative and positive.

Here is the simplified version of the Trans-Nova Twin-200 amplifier:

Note relatively high output stage rail voltages ( ±72V) and the relatively low amplifier frontend rail voltages ( ±23Vdc). Each channel's output stage must get its own floating bipolar power supply, but both channels can share the same ±23Vdc rail voltages. This improves upon the circlotron, as the circlotron requires two floating power supplies per channel. By the way, floating means no direct connection to ground. I used the battery symbol to override the mind's habit of knowing only grounded power supplies. Here is what an actual floating power supply might look like.

Note relatively high output stage rail voltages ( ±72V) and the relatively low amplifier frontend rail voltages ( ±23Vdc). Each channel's output stage must get its own floating bipolar power supply, but both channels can share the same ±23Vdc rail voltages.

The common-source arrangement yields screaming signal gain, but near infinite output impedance. The Source-follower arrangement delivers low output impedance, but slightly less than unity-gain. If a power amplifier runs off of ±72Vdc power-supply rails and uses the source-follower arrangement for its output stage, then we are not likely to get the full potential output voltage swings from the MOSFETs, as their relatively high gate voltages will impose about a 7-volt penalty. In other words, we would be lucky to get ±65Vpk output voltage swings. In contrast, we could probably get an easy ±70Vpk output voltage swings from the common-source arrangement with a floating power supply. (The lower the bipolar power supply voltages, the more onerous the high-gate-voltages become. There are workarounds, however, such as bootstrap capacitors, which expand the driver stage's power-supply rail voltages beyond the fixed rail voltages.)

The following is an excerpt from the Acoustat Trans-Nova Twin-200 user manual, wherein the amplifier operation is explained:

DESIGN THEORY

Your Acoustat TNT-200 power amplifier owes much of its exceptional performance to the use of an innovative new output-stage topology never used in any audio amplifier ever sold. Other amplifier output stages are typically connected in one of two configurations, either as voltage followers or as a transconductance stage (which is common in tube amplifiers).

The follower type is particularly ill-adapted to the power MOSFET, the device used in the TNT-200 amplifier, as well as in other recent entries into the high performance solid-state power amplifier market. This voltage follower topology does not allow use of either full rail-to-rail dynamic range or (power) gain-bandwidth product. The second type of output connection is also not well suited to the power MOSFET since pure transconductance-mode operation of the output stage necessitates large amounts of "long-loop" (multi-stage) negative feedback.

Acoustat's ANISOTROPIC' output configuration is a third fundamental connection alternative, which is a pure TRANSresistance-NOdal Voltage Amplifier - i.e. TRANS-NOVA stage. This output form achieves the best characteristics of the two established types, without their disadvantages. This stage has many fascinating properties such as:

• Stage input voltages cause output currents.
• Stage input currents cause output voltages.
• Full (power) gain-bandwidth product is achieved.
• Distortion, damping, phase and noise properties of a follower are preserved.
• Full rail-to-rail dynamic range of the MOSFET's are used, with NO gate-enhancement loss.
• "Gain" is described in OHMS in lieu of dB (approx. 4k Ohms here) 1 Ohm =1 volt/1 amp. The output stage, into 8 ohms, shows about 500:1 current gain and 10:1 voltage gain for an aggregate power gain of 37 dB. This high output-stage power gain results in a circuit simplicity which directly Improves sonic integrity.
• Actual driver-stage speed is increased over a source-follower connection by a factor of ten for the same driver-stage power. This surprising result comes directly from the physics of capacitor-charging energy.

This new topology, synergized with the several-hundred-megahertz (power) gain-bandwidth product of the MOSFET's yields the incredible 0.6 microsecond rise time, 165 v/us slew-rate and 400K Hz bandwidth (-3 dB) of the TNT-200 power amplifier.

The second major innovation in the TNT-200 is facilitated by the TNT-200 circuit, and is called Complement Feedback'. This circuit invention "complements" or "perfects" the deficiencies of classic negative feedback (Black-1937) so amazingly that with only 20 dB of "loop-closure" feedback, the TNT-200 yields zero (nodal) output stage distortion and zero output impedance, resulting in a damping factor of over 1000:1.

This connection also completely eliminates any need for current-limit protection against load faults. In addition, the high-speed performance of this amplifier is achieved into complex loads without any sonically questionable choke (stability) "band-aids" applied to the output path. We submit that this is truly a state-of-the-art approach to amplifier design, and further, that surviving amplifier designs of the future will follow this pure field-effect route. The speed-toughness-linearity product of the MOSFET is not even remotely attainable in bipolar devices, or in any practical application of tube technology. Best of all, the amazing technical claims are highly correlated to real-world sonic improvements, improvements that you will hear even upon first listening to your new Acoustat TRANS-NOVA Twin-200 power amplifier.

I remember reading the Acoustat white paper for their amplifier decades ago, but Google failed to deliver a link to it, sadly. After reading it, I instantly made a translation of the design into a tube-based equivalent, which took some effort, believe me. First, however, let's do a translation into bipolar transistors, which will require adding a few additional parts.

Click on schematic to see enlargement

The 100-ohm emitter resistors improve the input transistors linearity. Transistors Q3 and Q4 were added to replace the zener in the original VAS circuit. Transistor Q3 and the diode form a current-mirror circuit that relays input transistor Q1's current conduction down to the 1k resistor at Q6's base. Transistor Q4 simply cascodes Q3 and probably is not essential. (With higher bipolar power supply rail voltages, it usually is needed.) By the way, the VAS stage still functions in a push-pull fashion still inverts its input signal at its output. Transistors Q7 and Q8 were added to drive the output transistors and to unload the VAS stage. The capacitor that shunts feedback resistor B is needed to ensure stability. (I didn't add the two needed Miller capacitors, aka Cdom capacitors, also needed to ensure stability, but I feared making the schematic too complicated.) The output transistors need to use emitter resistors to prevent current hogging. Unlike the lateral MOSFETs used in the Trans-Nova amplifier, which exhibit a lovely negative temperature coefficient, the bipolar output transistors exhibit a mad positive temperature coefficient. Indeed, the output stage could use some current-limiting circuitry, as bipolar transistors are not nearly as forgiving as MOSFETs.

Another possible translation into bipolar transistors is the following:

Click on schematic to see enlargement

(Damn shame that the RMAF ceased as this amplifier circuit would have made a fine entry in the private Annual RMAF Unanticipated Amplifier Challenge .)

Translating the Trans-Nova into tubes runs into a snag: tubes only come in an N-version, with no P-version available. In other words, electrons only flow from cathode to plates, never the reverse. My workaround was to forgo the push-pull feature of the Trans-Nova's output stage and stick with single-ended operation.

Click on schematic to see enlargement

The input stage signal still enters a differential amplifier and a single triode functions as the voltage-to-current convertor stage, aided by the positive feedback from its plate to the input stage. Note that the output pentode's cathode is grounded and that the output transformer's primary DC resistance functions as a cathode resistor to cathode-bias the output tube; moreover, no negative power-supply rail is used. Also note that the pentode attaches to a floating 400V power supply, but its screen attaches the grounded fixed 350V power-supply rail. In other words, the pentode operates as pentode, not a triode-connected pentode. Three negative feedback loops are in place, each serving the same task as in the original MOSFET-based amplifier. What I like about this design, other than that 99% of tube-loving audiophiles and musicians will make no sense of it, is that feedback resistors A and B work to make the bottom of the output transformer's primary almost as ripple-free as its direct connection to ground.

The clever thing to do is to choose feedback resistor C's value to zero the voltage at the connection with feedback resistor B. If the VAS triode's plate voltage is 250V, then feedback resistor C's value should be ten times that of feedback resistor B's value, as resistor B sees one tenth the voltage drop as resistor C. As I look over this circuit, I see the VAS stage in the middle as being the weakest link. One possible improvement might be to replace the grounded-cathode amplifier with a cascode amplifier.

The cascode topology offers higher gain and higher output impedance—both desired features for the Vas stage in this amplifier. Another, but far more complicated alternative is to cascade the input differential amplifier outputs into a second differential amplifier.

The MOSFET on the left acts as a capacitance-multiplier, the one on the right acts as compliant-constant-current source. Some may balk at the intrusion of solid-state devices, but the inclusion of the constant-current source already snuck some solid-state into the amplifier. (Of course, we could add a negative power-supply rail and create tube-based constant-current sources, after we replace the two MOSFETs with two triodes. This would work, but it would mean that we must use at least two heater power supplies, possibly three, due the huge voltage differentials between cathodes.)

(I wish that I had written this post while Mr. Strickland was still with us, as I am sure he would have enjoyed seeing it. By the way, he designed the tube-based, high-voltage direct-coupled amplifiers for Acoustat.)

Class-G Trans-Nova Output Stage
I wondered if a class-G topology was possible with the Trans-Nova output stage with is floating bipolar power supply. It can be. At first, however, I was sure that it must involve a complicated arrangement that monitored the output voltage swing and switched on the needed higher power-supply rail voltages. Then it hit me: voltage relativity. Is a 9V batter a positive 9V voltage source or a negative 9V voltage source? It can be either, depending where we assign the ground connection. Likewise, where we assign the ground connection in the output stage is up to us.

Other than where the output is taken and where the ground connection is made, this topology is identical to that of a conventional source-follower output stage. Thus, the class-G operation works in the same fashion, in spite of the flipped ground and output connections.

Why bother with class-G operation? Several reasons come to mind, all of them good. The first is that the main output MOSFETs Q1 and Q2 will dissipate much less heat at idle, as the voltage drop across them has been halved. Second, and certainly more important to me, is that we can mix technologies, using rare and expensive lateral MOSFETs for Q1 and Q2, and common, cheap, and vastly more powerful vertical MOSFETs for Q3 and Q4. The best of both worlds.

At idle, the main output MOSFETs Q1 and Q2 draw their current from the center two 36V floating power supplies. As Q2 draws more current than Q1, the output voltage swings positive, as Q2 pulls up the string of floating power supplies positively. (Do not forget that this output stage inverts the input signal at its output.) Once the output voltage swing approaches +28Vpk, the commutating rectifier D4 is no longer forward biased, and rectifier D2 takes over, relaying the current flow from MOSFET Q2 to Q4, which allows the output voltage to further swing positively. When the output voltage swings negatively, it is due to Q1 drawing more current than Q2, with Q3 engaging, i.e. drawing heavy current, when the output voltage swings negatively beyond around -28V.

My contribution to class-G operation is the addition of the 300-ohm resistor that bridges Q3 and Q4 sources. This resistor prevents Q3 and Q4 from ever turning off, which in turn prevents many headaches, such as those that arise from switching glitches and oscillation bursts. Does this mean that this is a class-A output stage? Good God no, as the main output MOSFETs Q1 and Q2 will turn off and Q3's and Q4's current conduction only gets relayed to the loudspeaker when the output exceeds ±28Vpk.

I would actually use a different voltage split in the floating bipolar power supply, as music reproduction exhibits relatively low output voltage swings on average, with occasional big swings.

This floating bipolar power supply offers a nice 1-to-3 ratio between lower and higher voltages, as 72V is three time greater than 24V. The actual capacitor values and voltage ratings should differ, depending on the capacitor's location within the circuit.

When charged up, all the capacitors pack around the same amount of joules of energy. The assumption is that music playback is the amplifier's function, not shaker-table activation. If we want to add some overkill, we can use 10kµF/63V capacitors in place of the 4.7kµF capacitors.

Acoustat's ANISOTROPIC' output configuration is a third fundamental connection alternative, which is a pure TRANSresistance-NOdal Voltage Amplifier - i.e. TRANS-NOVA stage. This output form achieves the best characteristics of the two established types, without their disadvantages. This stage has many fascinating properties such as:

• Stage input voltages cause output currents.
• Stage input currents cause output voltages.
• Full (power) gain-bandwidth product is achieved.
• Distortion, damping, phase and noise properties of a follower are preserved.
• Full rail-to-rail dynamic range of the MOSFET's are used, with NO gate-enhancement loss.
• "Gain" is described in OHMS in lieu of dB (approx. 4k Ohms here) 1 Ohm =1 volt/1 amp. The output stage, into 8 ohms, shows about 500:1 current gain and 10:1 voltage gain for an aggregate power gain of 37 dB. This high output-stage power gain results in a circuit simplicity which directly Improves sonic integrity.
• Actual driver-stage speed is increased over a source-follower connection by a factor of ten for the same driver-stage power. This surprising result comes directly from the physics of capacitor-charging energy.

Nice attributes, but why can't we derive the same features with a fixed, not floating, power supply? The alternative arrangement is simple enough:

Click on schematic to see enlargement

The cascade of stage polarities remains the same: non-inverting, inverting, inverting. All three negative-feedback loops are still in play and still serve the same functions. Yet, a fixed (i.e. grounded, not floating) bipolar power supply is used, and it can feed two (or more) channels. As shown, the circuit won't work well without lateral MOSFETs, as the output MOSFET idle current will drift higher as the vertical MOSFET heat. The workaround is to replace the 5V zener with a transistor-based voltage-multiplier circuit that tracks the heatsink temperature change. Yes, the same circuit used in 99.999% of solid-state power amplifiers.

This got me thinking about the about the fantastically interesting auto-bias IC for MOSFET-based output stages, the LT1166.

I don't know for sure if this arrangement will work, thus the big question mark. Will the 0.33-ohm drain resistors ruin the otherwise low output impedance? The drain output impedance approaches infinity, so the addition of this trifling resistance makes no real difference; the amplifier's low output impedance is due to negative feedback loops A and B. A bigger potential problem is the voltage drop across these resistors, which will rob us of watts. The workaround is to shunt the drain resistors with low-voltage-drop rectifiers, such as a Schottky rectifier or an MUR420G.

One last idea. Since the Trans-Nova amplifier's frontend can run off much lower power-supply rail voltages, why not use a high-quality OpAmp?

I don't know for sure if this arrangement will work, thus the big question mark. Will the 0.33-ohm drain resistors ruin the otherwise low output impedance? The drain output impedance approaches infinity, so the addition of this trifling resistance makes no real difference; the amplifier's low output impedance is due to negative feedback loops A and B. A bigger potential problem is the voltage drop across these resistors, which will rob us of watts. The workaround is to shunt the drain resistors with low-voltage-drop rectifiers, such as a Schottky rectifier or an MUR420G.

One last idea. Since the Trans-Nova amplifier's frontend can run off much lower power-supply rail voltages, why not use a high-quality OpAmp?

The previous design that retained the Trans-Nova negative feedback loop "A" produced a THD of 0.0394%, while the last design that forwent all the Trans-Nova features delivered 0.0409%, meaning that it seemingly lost the shootout. Why seemingly? I would much prefer to listen to this last design. Why? Look at the odd harmonics.

The previous design that retained the Trans-Nova negative feedback loop "A" produced a THD of 0.0394%, while the last design that forwent all the Trans-Nova features delivered 0.0409%, meaning that it seemingly lost the shootout. Why seemingly? I would much prefer to listen to this last design. Why? Look at the odd harmonics.

All harmonics are below -60dB, but the distribution favors greater suppression of even-order harmonics over odd harmonics. Typical solid-state power amplifier harmonic structure, in other words. Next, here are the last design's results, the one without any Trans-Nova flavoring.

Mercy, this is what I want to listen to, as the ear-lovely cascade of harmonics resembles that of a single-ended power amplifier. Note that the 7th harmonic is almost 120 dB down, compared to the 64 dB down of the Some Trans-Nova Attributes amplifier. This is not twice as good, but one thousand times better, as dBs are logarithmic units; each 20 dB difference is a tenfold difference.

Why the difference? The extra negative feedback loop (A) in the previous design probably made for greater balance, hence greater suppression of even-order harmonics. (Or, possibly, just having two fewer transistors helped.) Does this say anything useful about the original Trans-Nova topology? Not really, as it is possible that the inclusion of its positive feedback loop undoes some of the forced balanced. I really need to run SPICE simulations on the original Trans-Nova amplifier. By the way, in my SPICE circuits, I used ±30Vdc bipolar power supply for the two output MOSFETs, which were IRF240 and IRF9240 types, not lateral types.

In addition, we need to look at later development of the Trans-Nova topology, for example, in the Trans Nova Twin 120:

Click on schematic to see enlargement

Note that the input stage differential amplifier now sports JFETs, not MOSFETs. Also note the inclusion fuses in the output stage. Later, Strickland joined Hafler and designed the Hafler 9300 power amplifier:

Click on schematic to see enlargement

Note the complementary balanced/symmetrical input stage and VAS stage; also note the use of bipolar transistors and the cascoding. The output stage still holds lateral MOSFETs and floating bipolar power supplies.

Music Recommendation: Sienna Dahlen & Bill Coon, Balladextrous
Since I am always on the outlook for jazz albums for those who dislike jazz yet might enjoy, discovering Balladextrous was a true find. Bill plays guitar and Sienna sings—that’s it. But it is often the spare musical arrangements that allow great stereo systems to shine. In addition, it takes greater talent when there are only two of you. Her voice is smooth and expressive, the last adjective being the essential one for a great jazz singer. Bill Coon's playing of the guitar is light and subtle. In other words, he is not just some session musician. Good music, good recording, great songs—what's not to like? Amazon Music offers the album in 24-bit, 96kHz, as does Tidal. Speaking of Tidal, I spotted that they are now offering Holly Cole Trio's album, Blame It on My Youth, in 24-bit, 88.2kHz MAX/MQA resolution; also worth hearing.

//JRB

Did you enjoy my post? Do you want to see me make it to post 1,000? If so, think about supporting me at Patreon.

User Guides for GlassWare Software
Just click on any of the above images to download a PDF of the user guides.