The Sound
Depending on the choices you’ve made, the K&K ST70 can be one of the finest sounding un-sounding amplifiers in the world. It is neither “tubey” nor “colored.” It has very little sound of its own. Because it uses a transformer for a phase-splitter, there is none of the noise and distortion of the stock circuit. Because it has a better power supply, there is virtually no audible noise or hum through the speakers. Because it uses zero loop-feedback, there are none of the problems associated with that. Because it has very symmetrical push-pull balance, the sound of the various components is minimized. Because it runs class A, there is no power supply droop like in class AB amplifiers, unless they are equipped with really amazing power supplies, usually have an element of confusion or muddle as the changing power supply voltages change the mu and amplification of all the tubes, all the time; or the sound of hands randomly clapping other people’s hands. Because the K&K ST70 is a DIY product, it doesn’t cost huge money, even if somebody does the work for you. From inception, it is a remarkably simple circuit, with lots of support around the circuit; no complicated circuit, just complicated support structure, like the metal bracing inside the Statue of Liberty.
Subjectively, I feel like I am listening to an amplifier that has no shortcomings, other than absolute power and deepest bass. You cannot overcome the power limitations, or the limitations of the size of the transformer core. On the other hand, most listeners don’t have full range speakers flat to 15Hz, or listen to organ recordings at realistic volume. So, you won’t hear any kind of power or bass limitations on your Diana Krall records, I suspect.
It is hard to make comparisons by memory, but I think I actually prefer the stock Dynaco transformers. Yes, the Lundahl has better detail, imaging and sparkle. But, the remanence of the affordable Dynaco output transformers injects some smoking-jacket luxuriousness to the sound that I find reassuring. The remanence adds harmonic distortion, though possibly very little when the AC and DC are balanced through the cores. Perhaps the sound of the Dynaco output transformer is making up for something missing in the system, like an inherent analytical quality. Or, maybe I am a cheapskate. Or, maybe I am still hearing something of greatness that was always present in the original ST70, observable above whatever else the ST70 was doing wrong.
As a starting place, I recommend you finding an old crusty ST70, at a bargain price, with guaranteed-good output transformers, stripping everything to bare metal, cleaning the metal with steel wool, replacing the output tube sockets, the input jacks, adding speaker binding posts, a high reliability toggle power switch, use all new screws and washers, a 3-prong power cord with ground-lift switch for the chassis, then start with just the upgraded main PCB and an upgraded power transformer. Purchase any tightly match quad of 6550 or KT88 from a reputable tube monger. Make output transformer connection decisions based on the amount of power you will need (triode versus ultralinear). Pound-for-pound, you’ll have one of the best amps money can buy. From there, you can try the expensive output tubes, upgrade to the regulation boards for the input tubes, etc.
An Experiment – A Virtual Tertiary Winding?
First, we are talking specifically of tubes with a 4th element, the screen grid (G2), which could be a tetrode, beam tetrode or pentode. We are not dealing with triodes. Blumlein outlines a connection where the screen grids (G2) are fed a lower DC voltage than the plate, which is desirable for a number of reasons, mostly for tube longevity. Vacuum tube designers, not amplifier designers, never intended for the screen grids to be fed the same voltage as the plate. With a standard ultralinear output transformer, like these Lundahl or the original Dynaco pieces, the screen (G2) received the same DC voltage as the plate. In a perfect world, the screen grid would be fed a very pure and stable DC voltage, regardless of the rest of the circuit. The screens aren’t there to amplify, really. They are there to “screen,” which is why they are called “screens.” Think “shields” as in Star Trek, then go look up “screen grids.” Eventually, because the tubes were used contrary to what they would prefer, the engineers issued recommended voltages, with built-in safety margins, for their use with standard ultralinear output transformers where the anode and G2 would receive the same DC voltage.
The essential element of Ultralinear operation is local AC feedback. A portion of the amplified audio signal, determined by the design of the output transformer, is taken from the output transformer and injected back into the screen grid (G2) of the output tube. This is the most effective form of feedback in tube audio, along with degenerative cathode feedback. The feedback loop is the shortest possible, so that it suffers from none of the sky-is-falling shortcomings attributed to loop-feedback, or global-feedback. Some ultralinear transformers have 40% taps. Others have some other percentage. This is the percentage of the amplified AC voltage, present on the plate of the output tube, that will be fed back into the screen grid (G2). But, while the screen is fed 40% of the AC signal, it is usually fed 100% of the DC voltage. By the way, why do triode-connected-pentodes have lower output, lower impedance and lower distortion than standard pentode connection? I assume that 100% feedback occurs when you connect the plate to the screen grid (G2). This lowers the gain of the tube, lowers the internal resistance of the tube, and lower distortion, just as feedback should. Feedback can work. If you don’t believe it, stop using all modern devices (cars, phones, computers, etc..).
As I was saying, the screen grid should receive a lower voltage than the plate, as was the intention of the engineers at the tube manufacturers. How do you lower screen grid (G2) voltage, making the tube happy, and still use ultralinear operation? There were a few transformers with a special winding, called a tertiary winding. This special winding would be wound quasi-parallel with the winding for the plates (anode), and “pick up” a percentage of the plate winding. Like the normal ultralinear transformer, this tertiary winding could be wound to supply any percentage of the amplified AC signal, greater or lesser than 100%, but would be something like 40%. This special winding, instead of getting the same DC voltage as the plates, would be connected to a separate (lower) DC voltage for the screens (G2), one that was regulated, preferably. One such transformer was made by Acrosound, the TO-350.
Why weren’t more transformers made like this? Because it is much more difficult to wind. The ultimate expression of transformer chaos might be the design for the McIntosh MI350/MC3500, that operated the screen grids (G2) on a tertiary winding, fed by lower DC voltage while the screen grids (G2) of the output tubes were receiving lower DC voltage, just as suggested in pentode operation of the tube. Other tricks were going on, like cross-coupling, and cathode feedback, in addition to a tertiary winding for the screens. The MI350/MC3500 was very expensive and difficult to produce. So, due to the realities of manufacturing and costs, we have rarely seen appropriate output transformers that allowed for the connection.
Remember when all those Chinese 6550/KT88 were blowing up back in the ‘80s? Many went supercritical not because the plate dissipation was surpassed, but because the screen grids were operating at much higher voltage than those dicey tubes could handle. If the relative spacing of the screen grid to the other grids were incorrect, or if the tube were somewhat gassy, ultralinear operation seemed to push those tubes over the edge. But true pentode operation of those flaky tubes usually worked. In pentode connection/operation, where ultralinear wasn’t an option, the screen grid (G2) voltage was always lower. You will see this on many classic amplifier designs of the ‘40s, ‘50s and ‘60s.
Screen grids have a large control over the operating parameters of the tube. Really, it is better to call the screen grid “G2,” and the control grid “G1,” without giving them names. Calling G1 the “control grid” infers that G2 (the screen grid) has no control over the tube. Not true. That’s why Eimac wrote “Care and Feeding of Power Grid Tubes” and they spend a great deal of time on screen grid voltage and proper connections. No, a kt88 isn’t a 50,000 watt tube, but the principles are the same. From Eimac: “The additional grid serves as a shield, or screen, between the input circuit and the output circuits of the tetrode, and is called a “screen grid.” In addition to serving as a shield, the screen is the accelerating element attracting the electrons from the cathode.” Besides, there are amplifiers that use the screen grid (G2) like a control grid (G1), where both grids are tied together and driven by the signal to be amplified.
Back in the gold old days of tube audio, a simple dropping-resistor was used to derive the lower voltage needed for true pentode (not ultralinear) connection/operation. This was never ideal in a class AB amplifier because the power supply voltage would sag, causing the screen grid (G2) voltage to change. Just modifying an old amp by applying a regulated DC supply from a separate regulated power supply would give much better performance. However, with the K&K mods, the tubes are running class A, so the supply voltage remains relatively constant, so this isn’t as much of a concern in this application. A byproduct of Class A operation is that the power supply has a constant drain, and the power supply voltage does not fluctuate very much, one of the reasons Class A operation is preferred.
As an experiment, to use the ultralinear output transformers, but to lower DC voltage, I connected an 80V zener diode in series with the screen grid (G2) tap of the output transformer, then bypassed the zener diode with a film capacitor so that the AC signal bypassed the zener. The screen grids received the same AC signal as the normal ultralinear connections, while receiving lower DC voltage. The AC signal traveled through the film capacitor, bypassing the zener. The DC voltage was partially blocked by the zener, and couldn’t travel through the capacitor. So, DC had one path, and AC had another path, but they both started from the same transformer, and went to the same screen grid (G2). I used a .47uf capacitor of 200 volts (the voltage the capacitor “sees” will be the voltage drop across the zener diode, not the plate voltage). I was worried I would hear some kind of noise generated by the zeners, but heard nothing on two relatively high efficiency speakers with zero signal.
A note: I do not think this method would work with class AB, since the tube would be operating in cutoff condition, not conducting any current. When operating in Class A, the tube draws a constant current. The zener must have current passing through it to regulate. Without current draw, the zener does not work properly. So, this application of the zener would only work with a tetrode/pentode that is operating in class A.
A thought occurred to me. Can you generate a virtual ultralinear AC voltage with a voltage divider network, running from the plate to the screen grid (G2), while feeding the screen grid (G2) a lower DC voltage from the power supply, dropped to a lower DC voltage via the zener? IDK. Just a thought. I am not an engineer.
You might ask “why try this.” Good question. My friend Marion wanted trouble-free operation. After biasing the KT120 at a moderate 55ma per tube, and dropping the DC voltage to the screen grids (G2) with the zener diode, I believe I have built an amplifier with very good sound, relatively high output for a class A amplifier, and with longer tube life. If you were driving high efficiency horns, I’d say “connect it triode.” It does sound slightly better in triode, but life is about making pragmatic choices, especially when you need more power and reliability.
Copy editor: Laurence A. Borden
- ← Previous page
- (Page 3 of 3)
