The Inside Story

By Dan Fraser
Lead Engineer ST-70

Back to ST-70

As a veteran audio electronics engineer I welcomed the news that Radial Engineering had acquired the Dynaco intellectual property and brand. I was one of the last engineers that received basic electronics education on tubes back in the 1960s. I used to repair and rebuild Dynaco products back in the 1970s. Though I worked for many years on Class-D PWM amplifiers, DSPs, digital audio distribution and other state of the art technology on the leading edge, I immediately jumped on the opportunity to spearhead the introduction of the Dynaco product line and develop the next generation ST-70, something I really wanted to do. How hard could it be to bring this legend back to life considering we already had schematics and sample amplifiers to measure? Well, it turns out this is a lot harder to do than one would expect.

For example, the Dynaco output transformers are magic when it comes to high frequency clarity, yet truly accurate build information is not to be had. When we conducted listening tests, we found that with the original tube power supply the bottom end was thin and lacked dynamics when played with modern music. It is the power supply in an amplifier that determines the quality of the bottom end. The power supply delivers the punch that modern music requires. Good low frequency response is not enough. Then there was the question of offering it as a kit. We debated this as it would save the cost or getting safety certification and they reduce the costs in production. But in today’s onerous legal environment, sending out products with high voltages that could potentially be played with by amateurs, the legal risk was simply too great. Instead, we chose to sell the ST-70 as a completed product and undertook the costs to have it properly safety certified.

Next on our plate, we felt we had to consider the countless upgrades and modifications that were introduced by the Do It Yourself (DIY) community over the years, many of which are actually quite good. There are literally tens of thousands of original Dynaco tube amplifiers still in use around the world. In fact, more quality music systems have included Dynaco tube preamps and power amplifiers than any other combination. The original Dynaco Stereo 70 is a true classic that is still regarded with awe by audio enthusiasts worldwide. If people want the old amplifier, the aftermarket upgraders do an excellent job restoring and improving the old stock of ST-70 amplifiers at a better price than a new amplifier will cost. There are however a number of people who want a better amplifier though still based on the work of David Hafler. This “new 70” reflects the illustrious history of Dynaco’s involvement with vacuum tube designs as well as the latest advances in vacuum tube technology. We like to think of the Series-3 as the update that David Hafler would have made if he were alive today. Besides, there is no point recreating the old amplifier precisely with some upgrades. Sure, it’s fun… but it simply does not deliver what today’s audiophile expects. And the more we listened, the more we tested, the greater the realisation that this next generation amplifier had to perform at a much higher level that what may have been expected in the early days of stereophonic sound. This led us to developing three different versions of the original using various design and tube configurations that stemmed from David Hafler’s immense library. The development took over eighteen months of trial, testing and tribulations.

The sum of all parts

The Dynaco ST-70 Series-3 has made extensive use of the most modem parts including: precision metal-film resistors, poly-composition capacitors and miniaturized high-capacity power supply electrolytics, devices that were unavailable at the time of the original design. These newer parts are more conservatively rated than some of the original parts. In compliance with general practices, a heavy-duty three-pin detachable grounded power cord is provided. Signal ground is the same as electrical ground and chassis and all signal connectors are gold-plated to ensure optimal signal flow.

To improve unit-to-unit consistency of the amplifier we eliminated the majority of the hand wiring and this meant using a printed circuit board for the internal components and wiring to the output tubes as well. This of course is a modern glass-epoxy board with extra heavy traces that meet today’s standards for safety. The ST-70 still employs 100% hand-inserted full-size discrete components throughout - no surface mount parts are used anywhere in the design.

Supplying the power

The original ST-70's power supply has long been the design area most vulnerable to legitimate criticism. We addressed this problem as follows:

  • We designed an entirely new and larger dual primary power transformer with improved line regulation (due to added iron in its core). This improves bass response and enables the ST-70 to be set for either 115 volt and 230 volt operation for global distribution.
  • The original – now long discontinued 5AR4 tube - has been replaced by solid-state rectifiers. This reduces the 5AR4's inherent losses while improving regulation and eliminating a failure-prone tube in the process. The benefits include more low end impact, improved efficiency and significantly greater reliability.
  • The original power supply filter capacitors have been made far larger; as large as we could fit in the space available while observing required safety clearances. This advancement has been desirable for a long time but has only recently become practical with the evolution of next generation capacitors that have been developed for computer power supplies and other switch-mode circuits. This benefits the ST-70 with significantly better low-end dynamics and improved damping factor.
  • The troublesome selenium rectifier used in the bias circuitry has also been replaced by a solid-state rectifier with additional supply capacitance. This of course improves reliability.
  • The bias reference voltage is now derived from an IC voltage regulator for consistency. This improves the stability of bias settings for more consistent tube performance.
  • The filter inductor has been reduced in value to complement the increase in filter capacitance with a resulting reduction of DC resistance. This makes available more voltage for use by the outputs and somewhat higher peak power output for a more impactful bottom end.
  • The final filtering stage of the operating B+ voltage has been split into four streams, each with its own RC filter network to greatly enhance channel separation while lowering noise.
  • The PC board has been carefully routed and optimized with advanced circuit development software to ensure that ground return currents from the power supply remain separate from the filtered power that operates the output stage and the separate ground returns for the driver stages. This reduces channel crosstalk, lowers overall noise including rectifier noise, and lessens overall low-level harshness or grit for more “air” when listening at low levels.

All of the above changes serve to improve "stiffness" of the power supply by lowering the effective power supply impedance and improving regulation while increasing available voltage for maximum safe power output. This ultimately makes for a tighter bottom end. It is Dynaco's position that full electronic regulation of the plate power supply does not result in cost effective sonic improvement.

Transforming the output

The output transformer inside the Dynaco ST-70 is often considered to be one of the primary contributors to its legendary sound. Before we set out to build a ‘new version’, we invested heavily by purchasing various ST-70s ranging from original to updated. This laid the foundation and set the goal in developing the right transformer for the job.

We began by taking these apart and then scouring the planet to find possible solutions. After testing various ‘off the shelf’ output transformers we found that we could not find any that had the high-end response of the original David Hafler designed A470. The square wave response of the original transformer at 20kHz is telling. In looking at the old part we felt that this was part of what made the original amplifier so well loved by audio fans. The picture at the right is shows the square wave response at 1 watt output as are all the rest of the tests herein. They are representative of the other tests we did at various power levels to clipping.

The bottom end is like this at 40 Hz square wave. The angle of the tilt indicates the low-end response. The less the tilt; the better the bottom end. Tilt is caused by both the frequency response and the ability of the power supply to deliver current at low frequencies.

Following is the frequency response of a professionally upgraded ST-70 with the original ST-70 transformers in it. The amplifier used to obtain these results already had upgraded coupling capacitors and a somewhat upgraded power supply.

Our first attempt at a custom transformer looked like what you see to the left. This is as good if not better than every off-the-shelf transformer we tried. Please note the vertical scales are different. The original is set to 2V per division while this one is at 5V per division. 20 kHz at the left. 40 Hz below.

Though the top end was disappointing, the bottom end was stunning in comparison. The greatly improved power supply is a large part of it but still the transformer contributed a lot as well.

In our second try after exhaustive work by our transformer vendor over many months including testing of the original A470 we got what you see here at 20kHz. While the rise time is not quite as fast, that was the price to pay to get the excellent bottom end. The issue is that there is no overshoot and no ringing or distortion of the wave shape that cause issues. Our listening tests confirmed that the rise time was not an issue. The lack of ringing is what makes the high end magical. Here is what we got with the new transformer.

Besides preserving the high frequency and square wave response of the original transformer, we were able to extend the low frequency response to do justice to the dynamics and range of today’s music, recorded with the modern recording technology that is now prevalent. As far as we can tell, we feel these may well be the finest output transformers made for any tube amplifier, anywhere near this price point.

Two speaker output impedances are now offered for 4-ohm and 8-ohm operation via a recessed selector switch and feedback is now taken from whichever impedance tap is selected by the switch.

Driving the circuit

It is important here to reiterate the philosophy behind the development of the ST-70. When conceptualizing the product, we asked ourselves a simple question: What would David Hafler do if he were in our shoes today? The answer is simple: he would use the very best components, try to keep the costs under control and deliver exceptional value.

At this point, we ran into a brick wall. The original 7199 driver tube with pentode and triode sections that was serving as both voltage amplifier and phase splitter has long been out-of-production and unavailable as are all triode-pentode types except for expensive or hard to come by new old stock (NOS) tubes. David Hafler’s approach would be to find a solution that would enable the ST-70 to be produced without part shortages while delivering the very best sound possible. This drove us to searching for and testing various options until we could uncover the best overall solution.

We began by recreating the design using a pair of the EF86 Pentodes and a 12AU7 dual triode as the phase splitter. This worked reasonably well. However, we discovered that the EF86 is only available in limited supplies and did not feel secure with this reality. This would not only inhibit production requirements, but also make it difficult to impossible for Dynaco owners to find replacement tubes down the way. Yet even tough quality and price would be called into question, we built one anyway and called it the 2T.

As we delved further into the original ST-70 design, we came to the conclusion that this driver circuit was likely developed by David Hafler to meet a target price point based on the available parts that were obtainable over 60 years ago and would not reflect what he would have done today. Over his illustrious design history David Hafler had developed a number of more advanced designs – many of which were never commercially produced. In fact we discovered a design that used three dual triode tubes with a 12AX7 as a voltage amplifier and a pair of 12AU7 tubes as long-tailed pairs to act as a phase splitter. This configuration cost a little more to produce than the original design and had very good sound quality. However, the long tail pair phase splitter has unequal gain on the two phases which require a trimmer to balance the phase for lowest distortion. To make sure it has been set correctly, it requires carefully adjusting the trimmer in the field and this is extremely difficult. We set the course and built a prototype and called this version the Triple Triode (3T) design.

We then proceeded to investigate the advanced Williamson circuit that Hafler had drawn up as illustrated here.

This Williamson design used a dual triode as a voltage amplifier and cathodyne phase splitter. Another dual triode is used as an additional voltage amplifier and output tube driver. The cathodyne phase splitter used here and in the original ST-70 had the advantage of being as accurate as the resistors used in the circuit. With today’s proliferation of high tolerance 1% resistors, this made investigating this option completely viable. We called this version the Quad Triode. (4T)

It is important to note that our goal was to deliver the best sound possible, regardless of the tube configuration. The fact is, whether we employ 2, 3 or 4 driver tubes in the design makes very little difference to the overall cost when building an ST-70. Well over half the bill of materials costs (BOM) are in the transformers, case and labor. So we took a six month detour and built prototypes of all three designs (2T, 3T and 4T) for comparative listening. To ensure fair testing, all three designs used the same output transformers and power supplies. We set them up for blind testing and conducted multiple tests. To ensure there would be no listening bias or sonic memory to contend with, we built a switcher that instantly switched between amplifiers and carefully made sure each had the exact same output level. In measuring, we found after warm-up and bias adjustment the triple triode (3T) unit with the long tailed pair phase splitter had the lowest distortion by a very tiny margin but it had to be laboriously set up with an Audio Precision in order to bring the distortion lower than the quad triode. The benefit of the quad triode (4T) and pentode-triode (2T) designs is that they do not require balance/trimmer adjustments. The quad triode (4T) design measured a slightly higher distortion through the critical 500mW to 5W range as compared to the triple triode (3T) while the pentode-triode (2T) design measured to have about twice the distortion of the quad triode (4T) design.

During the listening phase of testing process, we found the sound quality of the pentode-triode (2T) design to have the lowest sound quality. Still good - but not as good as the other two. When comparing between the triple triode (3T) and the quad triode (4T) the mid and high end seemed indistinguishable. Both were excellent. However the quad triode (4T) with the Cathodyne phase splitter was felt to have a punchier low end. Likely because the dedicated drivers were able to feed the EL34 grids from identical lower impedances giving it more punch.

After considerable deliberation, we decided that while the quad triode (4T) design cost a little more to produce, the improved sound quality was worth the extra.

All of these variations followed David Hafler’s original design philosophy and are still valid today. The low open-loop gain principles are maintained, thereby minimizing Transient Intermodulation Distortion (TIM). TIM distortion is regarded in many circles as the type of distortion to which the ear is most sensitive. This is also why we chose to use lower gain 12AU7 tubes for all four driver circuit tubes versus the more common 12AX7 as the excessive gain and subsequent dynamic distortion levels appeared higher.

Staging the output

The output tubes remain the fine EL34/6CA7 combined with the Ultra-Linear principle Hafler and Keroes developed under US Patent 2,710,312. However the troublesome cathode resistor that was originally used has been replaced by a separate high quality metal film part for each cathode.

A further improvement in distortion, particularly in the higher frequencies, was achieved by moving the feedback point from the 16 ohm tap of the output transformer to the actual loaded tap by the double pole impedance selector switch. The switch uses one side of the switch to direct the 4 or 8 ohm tap to the output terminals. The other side directs the feedback point to the tap selected and changes the feedback divider ratio so the overall gain remains the same whichever tap is selected. This also improved the transient response and subsequent dynamic musicality of the amplifier.

Filtering out the bad stuff

The original Dynaco ST-70 made use of roll-offs within the active circuitry to tailor the frequency response of the amplifier, reducing open-loop gain at the frequency extremes. This open-loop gain reduction resulted in the characteristic distortion increase at the frequency extremes in the original design. Extensive testing and analysis revealed that roll-offs within the feedback loop may not be the most appropriate form of response limiting in all cases. Dynaco installed a passive input frequency limiting filter network to limit bandwidth from 20 Hz to 30,000 Hz. This filter allows the removal of the tailoring components within the active circuitry and permits the active gain circuitry to be flatter across the frequency spectrum while minimally affecting the important phase response. The audible benefit is a great improvement in distortion components at the frequency extremes while still maintaining adequate bandwidth. This network requires the drive impedance from the preamplifier to be less than 5K ohms for proper top end response. Even tube types frequently achieve less than 5K using cathode follower output drivers. This also minimizes the effect of cable capacitance between the preamplifier and the power amp. This is acceptable for most modem day preamplifiers though many older pre-amps such as the Dynaco PAS-3X may have some matching issues. It is possible to change the input filter capacitors to permit higher impedance drives if absolutely necessary.

Due to the extended low frequency response, the improved power supply, and the inherently lower damping factor the tube amplifier provides, operation with acoustic suspension speakers may result in an exaggerated low frequency response. Hence a high pass, 3 position filter switch is provided with a bypass position and two different cut-off frequencies that gently roll off excessive bass at with a gentle 6dB per octave slope. This not only accentuates low mid punch, but helps reduce neighbour listening fatigue when the amp is played loud!

Each channel features a dual triode device operating as a voltage amplifier and “Cathodyne” phase splitter. Feedback is applied to the Triode Voltage Amplifier followed by a second dual triode acting as a driver stage to provide additional open loop voltage gain. Each channel drives a pair of EL34/6CA7 power pentodes operated in the classic Hafler Ultra-Linear mode. The power pentodes drive a custom-made output transformer with both the extended high frequency response of the original transformers and extended low frequency response. The impedance taps on the transformer are selected with a rear panel switch.

The printed circuit board material of the original ST-70 was exposed economical phenolic which deteriorated with age creating unpredictable electrical changes resulting in sonic differences. The ST-70 Series 3 uses military-grade, double sided epoxy fiberglass PC material and shields the recessed PCB with a solid metal chassis above it. The board is populated by high grade modern parts. Sensitivity to internal wiring placement is eliminated since wiring to the output tubes is accomplished via the printed circuit board. In fact, nearly two-thirds of the trouble-prone original point-to-point wiring has been eliminated!

The bias adjustment is now done by adjusting a set screw and using the dual front panel indicator LEDs. After a warm up period and with no signal playing, one merely sets the pair of LEDs to equal brightness. To accommodate unequal aging between initially matched pairs of output tubes, separate adjustments are provided for all four tubes.

Physical Fitness

As a final call to action, the styling of the amp came into question. The original ST-70 was a pretty utilitarian box that was OK in the 1950s but it did have a certain cachet with the cooling cage on top. So we went to town to create a better looking amplifier that would retain the original ‘base and cage’ design, yet bring a more modern and sexy look to play. We made the ST-70 wider to accommodate the clearances needed to address the high voltages for modern regulations and to match the width of the upcoming Dynaco ST-1 tube preamp. Both are about 17” wide to follow the generally accepted standard in use by many manufacturers today.

The front panel is especially modernized. It features adjustable recessed bias controls with pairing LEDs to optimize each tube and separate channel volume controls enable the user to balance the output for odd-shaped rooms. The rear panel is made with ultra high quality gold plated RCA connectors and heavy duty gold-plated ‘No Touch’ speaker binding posts - compatible with modern safety standards. A modern IEC power cord connection has been added is now capable of both 115/230V operation with a simple rear-panel switch.

The end all

Recreating and improving upon a true classic was a lot more work than any of us expected. This year and a half journey proved to be challenging and exhilarating. I am pleased to have played a significant role in the design and am comfortable is saying that this is the finest ST-70 ever produced whether it is measured technically or qualified sonically. We think David Hafler would be proud.

Dan Fraser
Senior Design Engineer

Back to ST-70