Super Triode Connected Amplifier Discussion Notes

Edited and Selected Email Correspondence from the JoeNet, DejaNews and Privately

Last update 21.12.01

Introduction:

The whole of the following text presupposes that one is at least passingly familiar with Hiroshi Uda's detailed English language explanation of the STC topology. One needs to have visited that first before the following Discussion Notes will make much sense at all!

More recently, a second English language explanation of the STC topology has become available. Katsu-san has targeted his explanation at the "valve-challenged" in his article The Super Triode Connection V.1: For the Transistor Generation. Katsu-san clearly is knowledgeable, enthusiastic and speaks English remarkably well, but I still found his article difficult to read and hence his concepts difficult to grasp. Translating from Japanese to English is extraordinarily difficult. I remain filled with admiration for those such as Hiroshi-san and Katsu-san who take on the challenge of translating difficult technical material across this language chasm!

The conversations recorded below commence with a review of a “summary” schematic posted to the JoeNet by Torisawa-san, to aid his valiant efforts to explain the STC circuit to puzzled JoeNet members. My notes open with my own comments on Torisawa-san’s work privately, to my good friend Hugh Dean. Now read on!

STC enthusiast, Katsu-san, includes an alternate description and explanation of the STC circuit (also in English).
STC enthusiast, Katsu-san, includes an alternate description and explanation of the STC circuit (also in English).

Discussion Notes:

Paul Cambie to Hugh Dean, 5/10/00; I’ve redrawn Torisawa-san’s "Figure 1: Practical Circuit of the STC" with minor adjustments (Ref. Figure 1a), including the addition of R2 which should be there.

  1. Putting the B+ and output transformer "upward" gives more of a sense of the T1, T3 SRPP actually being driven in parallel, (sort of), with the output pentode, T2. Jean-Michel seemed to me to be pointing this out, or rather, that the T1, T3 pair remained an SRPP however the remaining explanation of the circuit was structured.

  2. I've included, what I guess is a NFB path, being the connection from the output (T2) pentode’s cathode, back to the g2 (screen) of the input pentode. This connection seems to have been left out of some discussion drawings, but is an integral part of all the STC circuits in Uda-san's text. [Note subsequent discussions which show that this idea is in fact incorrect, at least for AC, as I’ve (in turn) overlooked the Ck in parallel with R2 in Figure 1a and left it out of the picture inadvertently!].

Figure 1a; Practical Circuit of the STC
Figure 1a
Practical Circuit of the STC

The original explanation of the STC circuit, if I understand it correctly, is that triode T1 is a significant NFB path, from the anode of output pentode T2, back to the grid of T2. What if instead, the output pentode T2 were considered as simply being "slaved" off the upper portion of the SRPP? The topology would thereby be using the output pentode as a relatively crude output signal "grunt" supplier, with the lower powered SRPP section providing the more desirable sonic character of the whole thing? T2 is perhaps an "after-burner", or "mother’s-little-helper", so-to-speak?

Such an explanation would leave the T2 cathode to T3 screen path as the NFB path of significance rather than considering T1 to have this role? [This is clearly wrong, at least for AC, as I’ve failed to consider the Ck normally and correctly in parallel with R2 to ground at this point]. The circuit would, if considered this way, confer the significant advantage of using only a low power and relatively low cost triode, to dominate the sonics of a better-than-headphone-powered single-ended transformer coupled circuit, employing relatively low cost but readily available pentodes to provide constant current drive and output grunt.

Yuichi Torisawa, 10/2/00; There's a mistake in my earlier Figure 1. I omitted R2 and Ck. (Ref. Figure 2 below; for my part I hadn’t helped by adding R2 to get Figure 1a, but continuing to leave out Ck).

Jean-Michel Le Cleac'h, 10/2/00;As Yuichi Torisawa wrote previously a bypass capacitor in parallel with the cathode resistor of the output pentode (T2) is missing in his Figure 1. This capacitor discounts any AC feedback between T2 and T3. We should also consider the behaviour of this part of the circuit at DC. Imagine that the mean current through the output valve (T2) rises, (because of ageing, for example).

See Jean-Michel's interviewed at Adnan Arduman's gallery of audiophile friends.
Jean-Michel
Le Cleac'h

The cathode voltage of that valve (T2) will rise too and therefore so will the screen voltage of the input pentode (T3). This will lead to a rise in the current through R1 and the upper triode, (T1), so the mean voltage, (DC), at the cathode of T1 will decrease. And so the DC voltage at the grid of the output valve (T2) will be nearer to the cathode voltage, so the current will increase.

Figure 2; Ck across R2 also added; click here to enlarge image.
Figure 2: Ck across R2 also added

[Actually it will therefore decrease, rather than decrease, by this reasoning, which is a desirable rather than unfortunate effect. It is indeed a DC, not an AC, NFB path. For example, and considering a disturbance to DC conditions, T2 Ik up gives T3 Vg2 up, giving T3 Ia up, giving T3 Va and hence T1 Vg down, decreasing current through R1, giving T2 Vg1 down, countering the original change in T2 Ik].

The inverse demonstration can be done too; a slight decrease of the mean current due too less cathode emission of the output valve will lead to a larger decrease...etc. So if someone builds such an amplifier it would be best to test and set the quiescent current periodically.

[Actually, the mechanism works the other way around, as per the first part of Jean-Michel's explanation. The NFB-for-DC path confers DC stability and obviates the need for re-biasing].

I spent several hours simulating the operation of the different valves of the STC amplifier. With regard to the triode T1, and the way it is used in the 6AK5-6BM8 amplifier, I can see that it acts as a nearly perfect 800kOhm resistor, (in fact V = -30 + [800 x I]), so the non-linearity of T1 cannot explain the overall sound of that particular STC amplifier. Also I can see that the S shape of the transfer curve of the output valve (T2) is in large part compensated for by a kind of inverse S-shape due to the input stage and the way feedback is performed. This leads to an overall transfer curve without any S-shape very near to what we can see with a single-ended triode amplifier. That's a very new thing, as it is very difficult to cancel the “S” shape of the transfer curve of a pentode. This is probably the most important feature of the STC amplifier.

Figure 2a: Hiroshi Uda's 6BM8 STC amplifier; click here to enlarge the image. Figure 2a: Hiroshi Uda's 6BM8 STC amplifier

Torbjørn Lien, 10/2/00; To me it looks like the plain old, (with pentode input, "the real"), Loftin-White amplifier, in the following respects.

Torbjørn Lien at the Arhus 2000 "Triode Festival" audio gatherering.
Torbjørn Lien

The amp uses negative voltage feedback, to stabilise the DC operating points of the amp. When the input pentode (T3) g2 is driven more positive, the input pentode draws more current. This causes the voltage drop across R1 to be increased; and so too is the "negative" bias of top valve T1, and T1 conducts less. The voltage at T2 grid drops. So then the output at T2's cathode drops, and the output valve (T2) draws less current. This counters the original change, completing the DC NFB stabilising loop.

An SRPP, even one that's a little unusual like this one, still inverts phase. I find the application of feedback, as done here, quite interesting and cunning. First something approaching 100% voltage feedback around the output valve, then the "anti-boot-strapping" of the input stage as well, which should induce non-linearities. Puzzling! (And what also is funny is the unusual silence from the anti-feedback crowd on the JoeNet; I've been waiting for the "This simply cannot sound good" or "been there...." or something equally rhetoric!). I believe a quite decent example of this type of amp could be made with an E130L as the output valve, T2.

Tom Ronan; This idea has been around (on the JoeNet), for more than six months, since when Carter Hendricks first posted about it on this list, and until a week ago only he had built it. I don't think any of the anti-feedback fraternity even blinked twice when they looked at it, (if they looked at it), since it wasn't the choke loaded, low Rdc, inter-stage, parafeed, constant current source, boring, only way to get good sound that every one knows is the only way to do things! The sound was so nice and from such odd components that I built one too. Now Carter and I, (and Mr. Torisawa), have heard them. I wanted to use directly heated triodes, so I did, using a 32 pentode as the input and a 47 as the output with excellent results! It's a neat amp and has a powerful and strong tone that is alluring!

Al Marcey; My fellow "6BM8 Club of Japan" member, Hideo Tamura, sent me his STC 6BM8 amp last spring. It is a modified Elekit. I enjoyed it for several weeks and forwarded it to John Atwood for appraisal and measurements. It may become an article in VTV.

Tsugunari Eguchi, 18/2/00; The STC circuit is indeed a little confusing! Mr. Le Cleac'h: the STC is an SRPP? This is a common question. The early portion of the STC is definitely SRPP-constructed. Yes, it's an SRPP; you can determine the resistors' value just as when designing an SRPP. But note that the anode of the upper valve is connected to the anode of the final valve, not to B+ where the impedance is zero. (Indeed this is a fundamental and critical point of difference).

Jean-Michel Le Cleac'h; That's exactly what I wrote about the first stage of the 6BM8 STC amplifier. I only took as an example the original 6BM8 amplifier that is used by the STC creator as a reference. Its first stage is an SRPP, however it is connected at the anode of the upper valve. I don't see why this fact raises so much discussion, after all! I have always been one to advocate the quality of the SRPP stage, (but only well designed SRPP’s, as one can do very bad things also with that topology).

Tsugunari Eguchi, 18/2/00; Torbjorn, the STC is a Loftin-White derivative topology? The 2A3 Loftin-White amplifier is a well-known direct-coupled amplifier. The STC is also kin to the direct coupled amplifier family, but it's not a Loftin-White. The STC employs anode-to-grid feedback. The "modern" version of the Loftin-White may well employ overall feedback, but basically the Loftin-White is a straight-forward circuit . . .

Tokyo based Amateur Audio Amp. Association

The Tokyo based Amateur Audio Amp. Association includes notes on its "amp tour"; "STC on the Road; Our STC (Super Triode Connection) 6BM8 amplifier goes the world!

"We have modified the Elekit-Japan's TU870 6BM8 amplifier kit into STC, and lent it for testing. The amplifier is currently in the United States. We invite you for the voluntary testers of the amplifier who pays for the shipping to the next tester and give us the testing report. We especially appreciate it very much if you could hold a testing gathering with your friends or family and send us the report on the testers' impression."

Hideo Tamura introduces himself and his extensive collection of projects (in English)

Hideo Tamura introduces himself and his extensive collection of projects (in English), and again in English details his special consideration of the Stopping Diode and other special items.

Tsugunari Eguchi details an extensive array of projects, including a 16A8-based STC amplifier

Tsugunari Eguchi details an extensive array of projects, including a 16A8-based STC amplifier.

Torbjørn Lien, 18/2/00; This is incorrect. The Loftin-White circuit was invented in 1929, and contrary to popular belief is not "the same" as "direct coupling" the 2A3 or whatever. (The 2A3 was introduced in 1932 ~ 1933). To recognise and appreciate the revolutionary nature of the Loftin-White circuit, one must view it the context of the technology at the time. That usually involved:

To improve on bandwidth, the Loftin-White used only two stages, directly coupled. Deletion of several gain stages and coupling transformers improved bandwidth. It was only possible to get the required two stage gain with the use of an input pentode. To improve on the stability, or rather, to be able to make this as a direct coupled amplifier at all, the Loftin-White had to find a way to set the amplifier up with a DC feedback loop, that sensed the current in the output valve. Suddenly the amplifier was self-stabilised, and could be fed varying voltages without the usual need for frequent adjustments. (This was done via the cathode resistor, feeding this DC voltage to the g2 of the input stage. The original amplifier also had a hum-bucking circuit, designed to inject some heater-hum in anti-phase into the grid circuit to lower the noise). So, a "true" Loftin-White amplifier is recognised as having a;

(This is admittedly a little simplified, but it needs to be mentioned. I've seen numerous published circuits incorrectly described as being "Loftin-White type" that were just two stage and direct coupled). Regarding now the STC amplifier, it can be seen that it uses the very same technique to maintain DC-stability.

[This is a critical point, from Torbjørn].

Tsugunari Eguchi, 18/2/00; . . . Basically, these answers depend on the definitions of the SRPP and the Loftin-White. Maybe, the STC can indeed be recognised as an SRPP or a Loftin-White if someone wants to. A slope can be ascending and descending at the same time, depending on one's own direction. Additionally, please understand that the super triode connection, (not the STC amplifier), consists of the final valve and the feedback triode, (just like the "Ultralinear" topology consists of the final valve and the output transformer, and the SRPP consists of two valves). They make "a super triode valve", and it can be driven by a constant current, with an input transformer, or by a transconductance driver.

Torbjørn Lien, 18/2/00; I know there are several ways to analyse circuits. Your explanation of it remains, to me, a bit cloudy; this description makes it more difficult to understand, than, I believe, is actually necessary. Also, I don't believe it to be correct. Is it possible to speak of an output valve, and a feedback triode, making a "super triode valve"? I don't think so. This type of feedback is covered in the standard literature as "parallel applied negative voltage feedback". This type of feedback reduces both input and output Z of the stage, making the dynamic Z of Rinput low. Then it is of course convenient to replace the Rload with a much higher, active one, using a valve.

I think this amplifier can be considered as a modern way to realise a two stage Loftin-White -like amplifier, swapping the expensive output triode with a feedbacked pentode in near unity-gain/inverting and low-Z mode. This output stage is, (put very simplistically), approaching the set-up of the old triodes, that, (when loaded), also had a gain near unity, were inverting, and had low-Z. (The comparison ends there).

Jean-Michel Le Cleac'h; I did a lot of experimentation on both the 6AK5 and 6BM8 (T) SRPP stage and on the 6BM8 (T) + 6BM8 (P) stage. What I found is that the 6BM8, with it's 8.2kOhm resistor, is acting like a nearly perfect 800kOhm, (with an added DC voltage bias though). In the description of the 6BM8 STC amplifier it is said that this valve acts as a non linear feedback element. That's wrong.

Tsugunari Eguchi, 18/2/00; Yes, the upper triode has a linear Vak-Ia characteristic, i.e. it acts as a perfect "valve resistor". The point of this recognition is the Vak "stability" of the triode. Triodes with low internal impedance can optimise this merit. But, to assume the upper triode is a "valve resistor" is to easily derive the conclusion that this valve could be replaced with a simple resistor alone. I and other Japanese STC specialists don't favour this view. It can be so replaced with a simple resistor alone, but then the core characteristics of the STC are lost. So, it's essential to understand the circuit combining the feedback triode and the final valve together. The upper triode has u, and amplifies. The cathode resistor is an I-V converter. The final valve has u = 1. The combination of the final valve and the feedback triode enables this characteristic.

Jean-Michel Le Cleac'h; I cannot believe that the sound of the STC amplifier is only due to the 6BM8 (T) + 6BM8 (P) stage. In itself this stage possesses quite bad behaviour. In fact I found that the 6AK5 non-linearity in itself is very important in understanding the behaviour of the amplifier. In the way they are used, when we consider them separately, both the 6AK5 and the 6BM8 (P) have S shape transfer curves, something generally considered bad as a source of H3 distortion. But the interesting thing is that in the way feedback is applied in the 6BM8 amplifier, there is an effective cancellation of the S shape. So the overall amplifier, while largely based on pentodes, possesses an overall transfer function very near to that of a single ended triode amplifier.

[A key point, from Jean-Michel].

Tsugunari Eguchi, 18/2/00; Yes, the SRPP's valves cancel the distortion from each other. I've heard of that being applied in the transponders in satellites, using linearising devices like the upper valve of the SRPP is utilised. I and some other Japanese STC constructors sometimes call the triode the "lineariser". But actually, we use pentodes, transistors and FET's for the first stage. Please let me repeat that an input transformer and constant current source can be utilised as an input stage for the signal. The original first stage, (V-to-I), may not be utilised. You can drive the cathode resistor, (I-to-V), directly, (with a V source). This is the original STC amplifier circuit. Differences in recognition doesn't change the amplifier's sound. Build it first, and think later!

Jean-Michel Le Cleac'h; I'll probably build one, but I disagree deeply with your view; I much prefer to understand how an amplifier operates before building one.

Tsugunari Eguchi, 18/2/00; The STC employs local feedback. There is no output transformer in the feedback loop. If the Olson circuit prior to the feedback age is termed "classical" and the Williamson is considered "modern", then the STC might be called "neo-classical".

Jean-Michel Le Cleac'h; I disagree with the expression "local feedback'. The feedback here just doesn't act on the output stage. The load that the STC stage presents to the input stage is very important to consider. That's the point I raised earlier: there are several schools for describing schematics. Some confusion arises in the paper describing the 6BM8 amplifier when drawing a parallel between the pentode, (with the triode as feedback element), and an op-amp with a non-linear feedback path. Another point of confusion came from the use of the expression "current source" to describe the operation of the 6AK5. In every description I saw of the STC amplifier, everything is described as if the behaviour of the input element, (the 6AK5 in the original STC amplifier), was negligible. It is not, and the feedback is effectively taken from the output of the amplifier, (even if at the primary of the output transformer), to the input stage. That's global feedback, despite the original way it is applied.

Tsugunari Eguchi, 18/2/00; I used the term "local feedback" for feedback without the output transformer primary and secondary being in the loop. Of course the feedback signal taken from the anode of the final valve affects the entire preceding circuit, but we assume it remains "local" insofar as the output transformer itself is not included in the loop. This type of "deep" feedback is sonically okay.

Torbjørn Lien, 18/2/00; This is only partially correct, as the input stage is anti-bootstrapped, so a loop, although "minor" has been made, and the feedback cannot be considered purely local.

Tsugunari Eguchi; I used the term "local" to explain that the feedback loop doesn't contain the output transformer. I just don't like the output transformer in a "deep" over all, Williamson-style, secondary to cathode feedback loop. OTL, (Output Transformer-Less), must be the best topology, by my reckoning, but it's difficult to implement and too inefficient. The STC doesn't employ Williamson- feedback through the output transformer, so it's okay. Basically over-all feedback needs phase/frequency "adjusting". This is difficult. In the STC there's no adjusting insofar as its "local", (in my intended sense) NFB. Much easier!

Torbjørn Lien, 19/2/00;Local feedback is only local when no circuit loop is added. For example, series/parallel applied feedback over one valve only. Part of the output transformer in the output valve cathode circuit is indeed local, but regarding the STC, the output signal is allowed to modify the operation of the input stage as well. So we have a, (still minor), loop here. I didn't say this cannot sound good. I think in fact it could, and would like to try it.

Tsugunari Eguchi, 19/2/00; I just wanted you to understand that the Version 1 STC is defined as the triode plus final valve combination. Many people have built an STC, but only a few have built the "ideal" STC, which doesn't contain the "lower pentode". You and Mr. Cleac'h wanted to describe the entirety of the three valves. Actually, this kind of discussion has occurred often in Japan also. Everyone said "Oh, it's an SRPP". Even we ourselves thought so at first, and have referred to the early portion of the STC as the "SRPP Stage". Some said the STC is akin to the Loftin-White like you. Some referred to the triode as a "cathode follower". But gradually we changed our collective view. I just hope you'll see that the combination of the upper triode and the final valve is the core part of STC Version 1 amplifier. Let's quit the "recognition" discussion now!

Figure 2b: The Loftin-White topology, here seen using a 2A3 in a circuit by Jean Hiraga; click here to enlarge image.
Figure 2b:
The Loftin-White topology, here seen using a 2A3 in a circuit attributed to Jean Hiraga, sourced from Bonavolta.

The most significant merit of the STC amplifier is its simplicity. Everyone can build it and enjoy its good sound. Overall feedback needs phase/frequency adjusting. Push-pull circuits are more difficult. Young newcomers don't like difficult circuits these days. I want to promote the joy of building audio equipment by oneself to novices! So, I promote the STC! I've previously written everything needed to understand the STC circuit, so I don't want to repeat this again. Version 1 seems to be the best; this is the version that contains the input pentode, as a V-I converting current source. In summary;

[A key point of difference in explanations of the STC is becoming defined; the Japanese STC fraternity clearly dissociates the driving device from the STC topology's fundamental character. The JoeNet fraternity, in the main, identifies the reverse, holding the drive-device, and indeed its specific characteristics, as completely integral to the operation of the STC topology.

Torbjørn Lien, 19/2/00; Yes, I understand; I find it peculiar though, as this "feedback triode" really is an active resistor, and so belongs to the input stage. It is unusual to name a topology after the output plus half the input stage, but OK! Any input stage, or a different input stage, that reacted beneficially to this kind of anti-bootstrapping, could be used here. The output valve would still have parallel applied negative voltage feedback. The effect of negative voltage feedback, with respect to distortion and output Z, is exactly the same on the output stage, no matter if the feedback is being parallel or series applied. It is then possible to achieve near similar, or maybe better results with part of the output transformer in the cathode circuit. This way the output transformer would have some of it's shortcomings rectified via feedback as well, still without forming a feedback loop. It would be possible to take a good pentode, like the E130L, and a good, split-able, output transformer, like the Lundahl type, then to put half the primary in the anode circuit, and the other half of the primary in the cathode circuit of the output valve. This feedback would give [2(unity gain)] (sic) . . .

Tsugunari Eguchi, 19/2/00; I understand. You've been thinking about this circuit! (GC+CF)/2! (sic; GC = Grounded Cathode, CF = Cathode Follower). So, you've stuck to the feedback concept, I see! Please let me explain.

I don't like the output transformer in Williamson-type secondary-to-cathode feedback. Usually the output Z declines and distortion becomes better. Less noise, wide bandwidth, etc., true! I agree! But I, (and other Japanese STC enthusiasts), think this type of feedback makes the amplifier's transient characteristics worse, (especially when it's deep feedback). Transient characteristics can be "adjusted" but to do this is difficult. The Japanese STC enthusiasts think that the clarity of pulsive sounds such as cymbals or percussions depends on an amplifier's transient response characteristics. So I like the STC, (which is without overall feedback). I don't know if your proposed type of feedback is okay or not okay, as far as transient characteristics are concerned, right now. The proportion of windings of each of the two primaries may affect the answer; but I'm not sure. I'm a builder of cathode follower output topologies, (although I don't recommend them). So, I strongly support your unique idea.

Torbjørn Lien, 19/2/00; . . . As for the input stage, there would now be three choices of supplying it with DC supply voltage, optionally including a feedback process. It could either be;

The sound from a "tri-voiced" amplifier like this could be quite different in it's various settings. With valves of questionable linearity, maybe it would be suitable for a guitar pre-amp?

Tsugunari Eguchi; We live in the CD age. Speakers are inefficient. Music is pulsive. So, we need low output Z! We have depended upon expensive output triodes and overall feedback from the secondary to the cathode for many years to reduce the output impedance, for this reason.

Carter Hendricks I think it misses the point to argue about whether an STC amplifier is like another feedback amplifier. What makes an STC amplifier unique is its implementation, with unusual valves and unsuitable transformers. The circuit, (with Uda's extra tricks), is a way to produce good music from unlikely parts. There may be some way to simulate the deep bass and clear highs of the Uda 6BM8 amplifier. But meanwhile, the proof is in the pudding!

Figure 3; Inverting Op-amp; click here to enlarge image.
Figure 3; Inverting Op-amp

Figure 4; Inverting Op-amp with a Triode; click here to enlarge image.
Figure 4; Inverting Op-amp with a Triode

Figure 5; Ideal STC Circuit; click here to enlarge image.
Figure 5; Ideal STC Circuit

Yuichi Torisawa, Mon, 21 February 2000: The STC imitates an inverting op-amp. Figure 3 shows an inverting op-amp circuit block. Kamijo-san tried to build this schematic with valves. It is shown in Figure 4 and Figure 5.

The schematic of Figure 5 is an ideal STC circuit, but it is difficult to build. Figure 6 is a practical STC circuit. This looks like an SRPP and Loftin-White But the STC's origin is the inverting op-amp. Imagining the inverting op-amp is the best way to reach an understanding of the STC circuit. The transconductance of the final stage and the T1 triode have a dominant effect on how the amplifier sounds. High gm valves sound great in the STC. The T1 triode can be replaced with a resister, but doing so makes the sound dull. Using the T1 triode is one of the STC’s key points.

A 6BM8 based STC amplifier by originator of the topology, Shinichi Kamura.
A 6BM8 based STC amplifier by originator of the topology, prolific Japanese amplifier designer Shinichi-Kamijo, who introduces himself and his projects (in English).

Daniel J. Marshall, 21 February 2000; I built one of these to see what the STC thing is all about. It didn't work well at all, having very low output, (a couple of Watts), with massive distortion. It was one of the worst performing amp’s I have ever encountered. I checked the circuit over very carefully several times, including buzzing it out and making resistance measurements from the valve socket pins. I tried different output valves, including a known good NOS GEC KT88 and a good Tungsol 6550 and tried various things to get it to perform properly, all to no avail. The 6U8 input valve was new and had been tested. The feedback valve was used, but tested quite good; over 100% with no gas, or leakage. All valves were carefully checked, plus the operating voltages seemed to be pretty well what you would expect.

So, I don't quite know what to make of it. I have a lot of electronic bread-boarding experience going back over 40 years and have never had any problem in accurately wiring and checking out anything in the past, plus I have a well-equipped electronic bench to critically evaluate the results. I very carefully checked everything several times and could not find any mis-wiring, or bad components. I did have to decrease the output stage cathode resistor (R2) value in order to get the recommended cathode current. So, I don't know, perhaps this is peculiar to the KT88 version, as others have apparently had good success with other versions. I removed the feedback valve and clip-leaded in an anode resistor for the 6U8 and jumpered it over to the output valve’s grid, and it worked considerably better; (more power, less distortion), though the bandwidth was not good. I didn't try to optimise the modified circuit, and just made a quick test set-up.

Figure 6; Practical STC Circuit; click here to enlarge image.
Figure 6; Practical STC Circuit

Jean-Michel Le Cleac'h, 22 February 2000; The STC output stage shouldn’t be considered just as a new method of connecting the output pentode, and surely not as an AOP (sic) with a non-linear feedback path as it was said. Also to consider the STC output stage, (the pentode and the feedback valve), as a single element is weird to me. If it was so, we could draw characteristic curves for it, (as with the Ultralinear connection for example). But in fact we cannot, because the input valve draws current from the feedback loop. That's why I consider the original 6BM8 STC amplifier as a classical two stage amplifier, the first one being built on the 6AK5 and the triode portion of the 6BM8, and the second stage being the 6BM8 (pentode).

It’s a pity that our Japanese friends don't take our remarks into consideration and prefer to consider that we have still not understood how the design operates. Modifying and improving our collective view about the STC operation, by exchanging our views could lead to improving the operation of such amplifiers, (for example for larger power as with a 6550). The very low output power obtained so far is not surprising to me, as this is something inherent to the actual STC design. From my simulations and measurements, one cannot neglect the choice of the input valve.I built the 6BM8 STC amplifier but with a huge power resistor in place of the transformer, and I use a larger B+. Polarisation is done with a Zener; as the 6BM8 is directly coupled this allows precise voltage measurements and curve tracing.

Jean-Michel Le Cleac'h, 22 February 2000: Due to the way feedback is applied, there is a lot of cancellation of non-linearity between the input valve and the output valve. So therefore I cannot imagine that a more linear input element, such as a transformer will improve anything here. I'll try to put on my web site two French push-pull amplifier schematics from the 1960's that use a same method as the STC to apply feedback between the output valve and the driver.

Yuichi Torisawa, 22 February 2000: Dan, did you build the KT88/6550 schematic? What voltage appears across your KT88 cathode resistor, . . .

Dan Marshall; With the 6U8 (P) cathode resistor set to maximum, the KT88 cathode voltage was about 60V. The pot was a 5kOhm unit, but measured a bit less than 5kOhms. I had used two resistors in series for the 820 Ohm KT88 cathode resistor of which one was a 320 Ohms. Jumpering, (shorting across), the 320 Ohm resistor allowed for adjustment to the recommended 66V on the output cathode. Perhaps a better approach would have been to have added some resistance in series with the pot.

Figure 6b: Hiroshi Uda's KT88 STC amplifier; click here to enlarge the image.
Figure 6b: Hiroshi Uda's KT88 STC amplifier

Yuichi Torisawa, 22 February 2000: . . . and the 6U8A (P) cathode resistor?

Dan Marshall; I didn't record it, but seem to recall a bit over 2V, but am not entirely sure of this value.

Yuichi Torisawa, 22 February 2000; . . . You can remove and jump [sic] the 6U8A (T) and 6AX4GTB. The setting of the variable resister at 6U8A (P)'s cathode is critical. Please try to increase and decrease it.

Dan Marshall; I will reinstall the circuit on the breadboard, add a 2.2kOhm resistor in series with the 5kOhm pot, remove the jumper across the 320 Ohm KT88 cathode resistor and report the results.

Allen Wright, 23 February 2000: The STC circuit is nothing mystic; just a regular amplifier with a common SRPP front end, but with the top of the SRPP returned to the output valve's anode, rather than B+. This unusual connection is a form of bootstrapping. But it's a "reverse" bootstrap, (or "anti-bootstrap" as Torbjørn refers to it), in that the bootstrap acts to make the, (load), impedance of the SRPP triode smaller than if the SRPP was wired conventionally. I tried this load resistor to output valve connection in a test amplifier and had exactly the same experience that Dan had; low power and increased distortion. Bootstrapped loads, (to increase the effective load impedance), are often used in transistor power amplifiers to, (attempt to), linearise the driver stage, and they work quite well when used this way. But conventional theory would say that reverse bootstrapping would increase distortion, and it does per my measurements. Now if someone wanted to try using a valve as an active feedback resistor, then I would suggest using a conventional load resistor, (to B+), on the driver valve, and then also add the SRPP connected triode between the anodes. This should allow the effects to be separated and understood as to their value, or otherwise.

Allen Wright
New Zealand born Allen Wright now lives in Europe where he teaches seminars, produces and films educational videos, and develops and hand wires radical tube prototypes for "VSE" (Vacuum State Electronics).

Hugh R. Dean to Jean-Michel, 23/2/00; I am impressed by the pervasive feedback system of the STC. The concertina operation of the feedback triode worries me and would seem to cost headroom; the total package would appear to 'compress' the waveform at the limits with almost no overload margin, rather like a solid-state amplifier which clips in a dB or so of overload. Your frustration with the Japanese "mystical" approach is justified; the Japanese have a cultural aversion to blunt exchanges; (their language reflects that; it values honorifics and circumlocutions), and there is possibly also a little "pride" in this as well. They are aware that Westerners stereotypically consider them non-innovative - not true, of course, and this prejudicial judgement hurts them.

Hugh Dean to Paul Cambie, 23/2/00; The STC is slowly revealing its secrets. One thing not remarked upon is that as the anode of the output valve goes up, so too must the anode of the feedback triode. But to achieve this, the grid of the output valve must go down, (simple phase reversal). And so too must the cathode of the feedback triode, since it's directly connected. As the anode of the feedback triode goes down, the anode of the feedback triode likewise, but the cathode of the feedback triode must go up. Put simply, we have a concertina, with the anode range of the feedback triode determined by the variations of the anode of the feedback triode, and the cathode range of the feedback triode determined by the variations in the output valve's grid drive. This concertina is thus not symmetrically disposed top and bottom of the feedback triode, and it is forced to generate, (or "suffer", depending on how you view it), a huge Va-k variation on the feedback triode. This Va-k variation is considerably more than that of the output valve itself, and all the more if it's a u valve.

Hugh Dean; designer of the AKSA amplifier, from Melbourne, Australia
Hugh Dean
, designer of the AKSA amplifier, hails from Melbourne, Australia.

[This is the mechanism variously described as "anti-bootstrap" or "reverse bootstrap" by Torbjørn Lien and Allen Wright respectively].

This makes for a highly variable resistor effectively between the anode and grid of the output valve. Now if we view the feedback triode as merely a valve working hard, then it is clear that even moderate swings on the output valve will work the bejesus out of it and take it right to its limits. And when it runs into either cutoff or saturation or both, the effect on the output valve will be draconian, and very, very sudden. Thus two things will result:

  1. Limited power, since very few small triodes can swing the volts of even an average output valve, and the limitations of the small triode will rapidly bear down on the voltage swing available, (so much so you might as well use a 1000V valve in this role!)

  2. Very rapid onset of clip, with almost solid-state sonics at the limit.

In light of this, I just can't see the point. The topology does not permit maximum power from the output valve, since the voltage swing of the feedback triode is the limiting factor, and the nature of the overload is not like a conventional triode; viz graceful, easy, and smooth sounding. It's hard clip, like a "good" solid-state amplifier!

Paul Cambieto Hugh Dean; It seems I was a bit quick to scoff at Torbjorn Lein's terminology of the feedback triode part of the STC circuit as being an "anti-bootstrap" circuit, for it's now pretty apparent that that's a most apt description.

Hugh Dean to Paul Cambie; Yes, a bootstrap takes output to improve operating conditions in an earlier voltage amplification stage. The improvement is usually achieved by bolstering supply voltage to keep a constant current across a resistor. It works really well. But in the STC the situation is reversed; giving a rather interesting effect, and exacerbating clip.

The Loftin-White topology is interesting. Note the resistor from the anode of the SRPP to the cathode of the 2A3; this means the upper triode's Va-k is held almost constant regardless of the signal; in fact as the grid drive to the 2A3 rises, so does the supply voltage of the SRPP – by an almost equal amount, too. This is a genuine bootstrap.

Yuichi Torisawa, 23/2/00; Dan, I think the grid voltage is low but your pot is a 5kOhm unit, which value is indeed correct. Something else is incorrect. Hiroshi Uda said his 5kOhm pot was set to 2kOhm or 3kOhm. Please return the cathode resister of the KT88 to 820 Ohms and adjust the pot by listening to the amplifier’s sound through a speaker. Also check the screen grid voltage of the KT88, between the KT88 and the 6AX4GTB's cathode. It may solve the problem to replace the 12AX7 with a 12AT7 or other alternative type. The way to adjust the bias voltage of the final stage is described in Hiroshi-san's website article as follows;

“3.1 Adjustment of the Bias Voltage of the Final Stage

In an STC circuit, the grid of the final stage valve and the cathode of the voltage-feedback-valve are directly connected. (If the amplifier has however been configured with RC coupling, this paragraph can be skipped). Initially, when experimenting with an STC amplifier, the voltage of this inter-connection point of the voltage amplifier pentode, (or BJT/FET), and the voltage-feedback-valve, will not be particularly high. Generally the point should be set to around 50V to 70V. It is possible to configure the whole STC amplifier by using just "one" B+ power supply. The adjusting process of the bias of the final stage valve is explained for this set-up. The following steps are required to set up an STC amplifier for proper operation.

  1. Set the bias of the first stage voltage amplifier, pentode or BJT/FET, to a low level initially, using VR2 (variable resistor). In this condition, the first stage amplifier is set to a low internal resistance state. Then the greater of the voltage output from the anode of the final stage valve will be carried by the voltage-feedback-valve.

  2. In this condition, the grid of the final stage valve is at lower than normal bias voltage. Also the voltage on the screen grid of the first stage amplifier is lower than normal, and operation of the STC will be incorrect and it'll have distortion.

  3. Now slowly increase the resistance of the VR2 in the cathode (emitter/source) of the first stage amplifier. The operation of the STC will gradually become normal. In this condition, measure the cathode voltage of the final stage valve. Calculate the cathode current using the formula;

    cathode current (Ik) = cathode voltage (Vk)/cathode resistance (Rk)

    Ensure that the operation of the final stage valve is within the valve's data sheet specification limitations.

  4. Then find the most powerful or most favourable operating point of the final valve, within its specification limits.

  5. If the adjustment failed, check and vary the cathode resistors of the voltage-feedback-valve and the final stage valve. Find a proper operational point, by repeating steps (1)~(4) above. In a very few cases this step (5) adjustment was required in the author's early experimental stages. In such cases, the voltage of the cathode of the final stage valve was insufficient. If a BJT/FET is employed as the first stage amplifier, this adjustment will be somewhat critical.

3.2 Issues Associated with the Circuit

The bias voltage of the final stage of the STC V1 circuit has to be set by controlling the cathode voltage. That is to say, the output valve's bias voltage is governed by its cathode current. The circuit is also controlled by the anode (collector/drain) current of the first stage amplifier. In another words, even if the bias of the final stage is set correctly, it is not then automatically assured that the bias of first stage will be correct. Even if the bias of the first stage is set correctly, it is similarly not assured that the bias of the final stage will be correct. Essentially the operation point of the first stage and the operation point of the final stage should be independent. However, the STC V1 circuit is adjusted by only one variable resistor in the first stage, for convenience. So a circuit adjusted for the best point of the final stage doesn't mean the setting of the first stage is at the best point. If the first stage can tolerate a wide latitude of operating voltages, then the first stage can produce proper operation even though its set-up voltages are not precisely suitable.

In most examples of the author's designs, using newly tested valves, the cathode current of the final stage was initially determined and then the cathode voltage, giving the required value of cathode resistance. After doing this design step, the voltage will, in practice, be within plus or minus 10% of this calculated value. The voltage supplied to the 1st stage may be varied from the best point, as its operation is not as critical as that of the final stage. In most cases, the STC V1 amplifier can be adjusted to approximately the correct operating point. In a few worst cases, the author had to change and adjust the cathode resistance of the final stage, to give correct operation of the first stage.

3.2.1 First Stage Using a Pentode

As described above, (3.2 Issues Associated with the Circuit), adjust the cathode current of the final valve to set it within the anode power dissipation limit and the screen grid power dissipation limits. That's all that is required. Check these points and there will be no problem operating the amplifier. In most cases the author adjusted to the position that gave the most powerful or most favourable sound. In this condition, the bias of the final valve will be near the most suitable position.”

Yuichi Torisawa, 25 February 2000: I built two STC amp’s with 6BM8 and 6AN5. They sound good, and vivid. Many of my friends have built many STC amp’s, using different valves. Every one of these amp’s works well. Hiroshi-san, who built one using the KT88, sent his data to me. His amp sounds good. Please refer to the measured data as follows;

Measurement Point

Symbol

Value

B+ power supply before the output transformer;

Vbb

= 413V

1st stage (6U8):

Rk

= 2.8kOhms

1ststage (6U8):

Vk

= 2.1V

Power stage (KT88):

Vk

= 60.0V, (820 Ohms, 73mA)

Power stage (KT88):

Vg2

= 397V to ground, 337V to Vk

Power stage (KT88) voltage drop by g2 LNR (6AX4GTB)

 

= 0.98V

I think that distortion of the STC amp depends on adjustment of the first stage.

Daniel J. Marshall, 25 February 2000: I remounted the STC amp and got it going again. The 820 Ohm KT88 cathode resistor measured 860 Ohm so I paralleled it with a larger resistor which brought it down to approximately 800 Ohms. This allowed the amp to be adjusted to 68V on the KT88 cathode, whereas, before it would go to only 60V. I adjusted it to the recommended 66V while monitoring the distortion on a spectrum analyser and it exhibited less distortion at 66V than at 60V, which you mention above. The other operating voltages were close to those above, but all a bit higher. It seems to be working normally. I found distortion to be fairly low at low power level, (around 1/2W), but to increase substantially as the power is increased. At 2W the 2nd harmonic was down by only around 20dB, which is quite high. The higher harmonics decrease very close to 10dB with each successively higher harmonic, (at the 2W level). At power levels much over 2W the distortion became much worse with the 3rd harmonic being greater than the 2nd. In my first evaluation I was not prepared to accept the fact that it was limited to a couple of watts, and I rather was expecting several times that amount. At 4W it exhibits quite high levels of distortion.

I have been listening to it for a while and it actually sounds quite good at moderate power levels, (with 100+ dB sensitivity speakers), but gets a bit grainy at high power levels. But then, KT88’s sound good in strapped triode with the Allied inter-stage transformer driver which I recently described here and crank out upwards of four times the power.

I did not have both circuits operational at the same time to make an AB comparison, so will reserve final conclusions until later. However, I would say that they are roughly comparable at low power levels, i.e., they both sound very good, but the standard, strapped-triode circuit sounds better at higher power levels, as it does not start coming unglued until it reaches about 6W. Also, it exhibits much lower 3rd and higher order distortion, e.g., at 4W it exhibits -27dB 2nd, -49dB 3rd, -51dB 4th, -58dB 5th and >-80dB 6th and above. This is much lower distortion than the STC circuit at 2W. The STC does seem to have very good bass response. Perhaps I will get both circuits going and A-B compare them for a better comparison. My general opinion at the present is that the STC topology gives up too much power for any advantages it may have. If a KT88 STC would produce 6~8W it would make more sense, but for 2W, it seems a bit of a waste. If I later determine that it sounds appreciably better at low to moderate levels, then "perhaps".

Hugh R. Dean to Dan Marshall, 25/2/00; I was fascinated by your comments on the distortion spectrum. I actually thought this might happen, because of the phase relationships in this strange amplifier, which I believe has virtually no headroom. Your results accord very well with my own predictions. Your testing has been fully and completely objective and you have no axe to grind; any concertina overload function on a small signal valve hooked to the output valve must overlay the sound with its own sonics - and its limitations. It occurs to me that the feedback triode should be capable of swinging at least as much as the output valve, and preferably a little more. To this end, a high voltage triode is really required; and of course it must also have a very low saturation voltage, with enormous linearity across the range. The only valves I know in these categories are hefty triode wired pentodes, and hefty triodes. This seems like a waste of a good valve to me; you might as well use two of them and make a decent push pull amp, but there it is.

Daniel J. Marshall, 25 February 2000: The KT88 in strapped triode mode was with a 6,500 Ohm anode load. The data sheet shows 6,150 Ohms anode resistance for the KT88 in strapped triode mode, so it needs some feedback to attain a sufficiently low damping factor. Perhaps the KT88 is best left to push-pull Ultralinear amplifiers. Since the output transformer does have a screen tap, and it will be very easy to implement, I might try a single ended Ultralinear amplifier with the KT88 using conventional feedback from the anode, just to see how it performs.

[?] You might want to check more than one data sheet for the KT88; one says 670 Ohms, another says 6150 Ohms (maybe it's supposed to say 615 Ohms?). Also, the former says 650 Ohms for the KT90, and the KT90 is supposed to be a drop-in replacement for the KT88. I've also got hardcopy 6550 and KT88 data sheets that show triode ra’s in the 650 Ohm ~700 Ohm range.

Daniel J. Marshall; I thought 6,150 Ohms sounded high, but did not run any tests on it. In fact, I looked it up only after I had already disassembled the circuit. I had presumed perhaps around 1kOhm or so, but didn't measure the amp's output impedance. I did briefly connect a second 8 Ohm load resistor across the existing one and noted that the voltage dropped a relatively small amount, less than you would expect for 6,150 Ohm, but more than from 650 Ohm. But then I was not focusing on damping factor at that moment, so passed by it pretty quickly.

[?] Maybe you need to try 2kOhm or so for strapped triode...

Daniel J. Marshall; I was operating the STC at 3,500 Ohms load, which is the closest I could get to 2,500 Ohms with the existing primary and secondary taps, and the speaker impedance. I will measure the output impedance when I get that circuit operational again and calculate the anode resistance. Thanks for pointing that out. Since I had only passed by that aspect quickly, I should have not been commenting on it. I had a Sylvania manual close by and it calls for 4kOhm for the 6550 in push-pull triode configuration, which would suggest that perhaps the 650 Ohm value may indeed be correct.

Yuichi Torisawa, 28 February 2000: I discussed with my friends why your STC was low power. We suspect that there might be oscillation. The 6U8A is apt to self-oscillate, because of its pin connection. Pin #1 is the anode of the triode section. Pin #2 is the grid of the pentode section. They are close each other. If you try the STC using the KT88 again, please connect the grid of the pentode section to ground with a 100pF capacitor. If the STC still is low in power, please disconnect the line from the anode of the 12AX7 to the anode of the KT88, and connect the anode of the 12AX7 to B+; i.e. from the anode of the 12AX7 to a B+ point before the output transformer. This will change the STC into a conventional SRPP, temporarily, for the purpose of testing.

If the temporary SRPP still has low power, something else in the first stage is incorrect. Please measure the voltages at the following points;

  1. g2 of the 6U8A pentode section

  2. Anode of the 6U8A pentode section

  3. Anode of the 6U8A triode section

  4. Cathode of the 12AX7

  5. Anode of the 12AX7

I'm advised that 5~6W may be expected from an STC using a KT88.

Dan Marshall;

1.

g2 of the 6U8A pentode section

66V

 

2.

Anode of the 6U8A pentode section

35.4V

 

3.

Anode of the 6U8A triode section

35.2V

Hmmmm, a bit less than the
anode of the pentode section

4.

Cathode of the 12AX7

39.5V

 

5.

Anode of the 12AX7
with approx. 420V B+

440V

 

Yuichi Torisawa, 29 February 2000; The bias of the KT88 is -26.5V. (39.5V - 66V = -26.5V). This is too deep. A proper bias is about -16V. Increasing the cathode resister of the 6U8A pentode decreases the bias. But you have disassembled your STC amp test circuit already?

Dan Marshall; The GEC data for the KT88 calls for –48V bias at the grid and a screen voltage of 440V. There is no way –16V can be correct, unless the output valve is totally exhausted, or perhaps the screen bypass capacitor is very leaky, causing a low screen voltage. I checked the screen bypass capacitor for leakage before installing it and it was very low, around 1mA. Even -26.5V seems low to me, for the Vak = 400V or so, at which it was operating. A bias voltage of -26.5V resulted in an anode current of approximately 80mA, which would indicate that it was the correct value for that particular output valve. I was using a good, tested Tungsol 6550 when the measurements were made, but they are pretty much interchangeable with the KT88. I have a little stereo single-ended 6BQ5 chassis. Perhaps I will convert one channel to STC topology and compare them next time.

Yuichi Torisawa, 2nd March 2000: I do not have the GEC data and cannot find the voltage of the bias at 330V, (anode-cathode), and 330V, (g2-cathode). But -16V seems to be less than a proper bias. The KT88 is famous as a loose quality controlled valve. The Ohms of the schematic could be a rare case [sic]. I will build an STC using a KT88 myself in a few months time and check its performance.

Carter Hendricks, 28 February 2000; How can Tom and I have good luck building and listening to our STC amp’s while friends have bad luck even looking at these unusual circuits?

Daniel J. Marshall, 29th February 2000; I wouldn't say I had bad luck, rather that I had greater expectations than a mere 2W output from a KT88.

Carter Hendricks, 28 February 2000; I think I know how these amp’s work, and I know how they sound. This amplifier sounds powerful. The first thing one notices is the strong, fast bass.

Daniel J. Marshall, 29 February 2000; Yes, I commented that I thought they have very good bass response.

Carter Hendricks, 28 February 2000; But it also has great piano tone: the lower notes convey the rich wooden structure of the piano, the high notes are percussive and clear.

Daniel J. Marshall, 29 February 2000: Yes, I also commented that it sounded quite good.

Carter Hendricks, 28 February 2000: Too clear, perhaps. My second Uda amplifier is less a test mule and more carefully finished. It is prettier, and it has higher resolution. I even thought that there was some higher frequency glare to this new amp, but that turned out to be my CD player. With a phono stage and LP's most of the edge went away.1 Still, this is no warm and fuzzy amp.

Daniel J. Marshall, 29 February 2000: Agreed, mine sounded clear and accurate.

Carter Hendricks, 28 February 2000: The STC amp uses 100% anode-to-grid feedback in the output stage. This provides exceptional drive for quite unexceptional transformers. The only reason to use the STC topology is to be able to use little, inexpensive transformers. Compared with what the Tezukuri guys are using, my US$30 Hammonds are gross overkill.

Daniel J. Marshall, 29 February 2000: Well, I had started breadboarding another project using the SV811-10’s, so diverted my attention to the STC just to see what it was all about. Talk about overkill; I was using an 11lb Hammond output transformer.

Carter Hendricks, 28 February 2000: All of the gain of the output valve is applied to the STC. Though Hiroshi Uda has tried triodes, pentodes have higher µ to available to contribute to the STC. They provide the most STC effect to these small transformers. But remember that all of the output valve gain is "gone." Power is voltage and current. The STC pentode valves are, (usually), direct coupled but operate conventionally. They have ordinary bias limitations. The "book" details the 6BM8 at 200V anode, 200V screen, bias = -16V, anode current = 35mA. The amplifier valve has a µ of say 70. But with output bias = -16V little of this u can be used. (Although we do have an amplifier with great sensitivity). So our voltage swing is limited by the low bias points of power pentodes, the same valves which provide the high µ which makes the STC work well. But the µ of the output valve is all used up by the STC. These amp’s do not produce Watts of power. For example, with this 6BM8 amplifier, played really loud, with my ~90dB speakers,2 I measured ~2V peaks on my DVOM, corresponding to 1W. Played much louder, the sky falls in. So let's select a very powerful valve, like the KT88; 400V, 80mA but, oops! grid = -16V. More current, yes, but the same voltage constraints, the same µ = -1, so something like double the power; 2W. Not exactly what Dan had in mind.

Daniel J. Marshall, 29 February 2000: Right, 2W is what I found to be the maximum power with the KT88. I was driving 100dB speakers and 2W can go pretty loud. Yuichi claims that others have gotten 5W~6W from STC KT88’s, but I could not get that much.

Carter Hendricks, 28 February 2000:The STC amp’s come from Japan, where triodes and transformers are expensive and where many valve enthusiasts are a bit obsessive. The STC movement is a breath of fresh air. Cheap TV valves, cheap little transformers, wonderful music. Now I still think that my Uda is a great little amp. Played loud, the STC amp’s sound wonderful.

Daniel J. Marshall, 29 February 2000:I thought that mine sounded best at moderate volume where distortion is acceptably low, but I did not do critical comparisons with other amp’s.

Carter Hendricks, 28 February 2000: Two photographs of my little Uda amp may be seen at my website [here reposted by permission].

I've been constructing most things from a sheet of 1/8” aluminium, with the power supply and output transformers hung underneath. I've had good luck with a Sakuma/Sato ground bus system. Yet while the amps sound powerful, they are not so (technically) powerful. These are sensitive amps that provide good bandwidth through the speakers, but if forced too far they collapse, like, well, pentodes. The resolution of the circuit is enough to reveal the tone of these TV valves. More problematic, the amps are good enough to remind me how much I like US$2k CD players.There's a bind here. Still, Tom and I think that an STC amp is the best crazy cheap amp one can build,3 especially from new, (not scrounged), parts.

Figure 7: Carter Hendrick's STC amplifier, Feb. 2000.
Figure 7: Carter Hendrick's STC amplifier, Feb. 2000.

1I think the best path for further development is the selection of valves with better tone, and the use of other inexpensive transformers with better mechanical construction than the Hammonds.

2 I've used my Uda’s with Diatone’s, the Goodmans, and a pair of nice Neat bookshelf speakers. The Neat’s have crossovers and tweeters but the Uda’s did surprisingly well; one reason I thought the STC amp’s had much more power.

3 Yes, I prefer the quality of the music from a well-built amplifier using a direct heated triode and Magnequest transformers!

Figure 8: Carter Hendrick's STC amplifier, Feb. 2000.
Figure 8: Carter Hendrick's STC amplifier, Feb. 2000.

Daniel J. Marshall, , 29 February 2000: I built up the other side of the amplifier breadboard, using the bridged Allied inter-stage transformer with 6EA7 driver. I really like the sound of this circuit, it is very clean and clear sounding. Nothing ever seems to get muddled. I have a couple type 45 valves and wanted to try one out, since I had heard so many good comments about them. It sounds quite excellent, very clean and clear without being overly bright or harsh, with everything clearly delineated and never muddled sounding. It behaves very well up to 2W before it goes into excessive distortion and when it does it does so gracefully. The output transformer was pinched from a Japanese receiver. It is a bit larger than the cheap single ended transformers used in low-end consoles, but not large by any means. Frequency response is flat to 20kHz and to around 20Hz~25Hz on the low end. Bass response is good, except is a bit weak on very low notes, around the vent tuning, (32Hz), where the speaker impedance is quite low.

I have a pair of Heath EA-1/AA-191, (single ended 6BQ5), amp’s which I upgraded recently and thought they sounded better than they had any right to sound, but in comparison to the 45 amp, they are not as clean and clear in the upper end. I have a 6BQ5 single ended chassis out of a Silvertone console which sounds fairly good. Perhaps I will modify one channel to the STC configuration and give it another chance. Then I can easily make a more meaningful comparison, with the other, (conventional), channel.

Step Hydro, 29 February 2000; Has anyone compared the STC and the Darling in listening tests? With the Darling it is plain from our discussions that the transformers will be quite limiting. It seems that the STC excels at making the best of cheap transformers however.

John Levreault, 3 March 2000: The following are some observations and comments based on exchanges with Carter Hendricks. I've never built an STC, so my comments are just theoretical observations.

  1. Grounded-grid Stage; At first I was confused by the assertion that the voltage feedback valve was running in a grounded-grid mode. However, now I understand that the action of the op-amp and voltage feedback valve will force the voltage on the (-) input of the op-amp to be equal to the voltage on the (+) input. After all, that's the way op-amp’s work. Referring to Uda's Figure 2, since the (+) input is at ground, (or some other DC voltage chosen for convenience), then the voltage on the (-) input will also be at ground. Hence the correct usage of the "grounded grid" term. The cathode resistor is there for biasing purposes and will create only a small disruption in the "grounded-gridness" of the stage. In practice, the output valve replaces the op-amp.

  2. Gain; Uda-san states that the gain of the compound stage comprised of the output valve and the feedback valve is equal to (u - 1), where u is that of the feedback valve. Furthermore, the gain of the input pentode will equal its transconductance (gm) x R1, which may be less than 1. This is because the "top" of R1 is held at AC ground by virtue of the NFB of the voltage feedback valve.

  3. Output Valve; In addition, Uda refers to the fact that the transconductance of the output valve is absorbed by the feedback. This is true. The output valve does not contribute to the overall gain of the amplifier. In addition, one very important effect occurs by virtue of the feedback path. The output impedance of the output pentode drops substantially. This may be why the STC topology works so well with inexpensive transformers. Generally speaking, cheap transformers don't have much in the way of low frequency extension when used "normally". However, the low frequency extension of any transformer will improve when driven from a relatively low impedance, so the STC may indeed improve the low frequency response of such a transformer. Uda-san reminds us that this topology is intended to achieve optimum performance from inexpensive valves and iron.

  4. Constant Current Source, (a.k.a. the Input Pentode); Proper biasing of the input pentode in critical to the operation of an STC amplifier and may be where that KT88 example is failing. The first problem is driving the grid of the output valve. Note that you only need around -15V of bias. If one were to ground the output valve cathode, then we'd need to swing between –30V and ground on the grid of the output valve. Therefore, if we want a direct-coupled driver, we need to elevate the output valve cathode with a cathode resistor so that the driver can operate off a positive anode voltage.

The second problem is that, with a direct-coupled driver, the quiescent grid voltage of the output valve will be equal to the quiescent anode voltage of the driver, in this case the input pentode. The higher the driver anode voltage, the larger, (and lossier), the cathode resistor on the output valve. For best amplifier "efficiency", the anode voltage of the driver will be as low as possible. However, very few valves like to work with low voltage, say 30V, on their anode, but the 6AK5 happens to be one of them. Pentodes in general are more capable of operating at low anode voltages than triodes, although the 6DJ8/6922 is an exception. In order for a pentode like the 6AK5 to operate properly, (i.e. linearly), it needs some screen voltage. The 6AK5 happens to operate with around +50V on the screen.

This screen voltage is derived from the cathode voltage of the output valve. That means that, no matter what screen voltage is chosen for the input pentode, its anode will, (typically), swing from 30V below its screen to a voltage the same as that on its screen. Remember that the input pentode is direct coupled to the output valve. If you choose to use a different input pentode, then the cathode resistor of the output valve must be chosen to develop enough screen voltage for the input pentode. Can you use a 6AC7? Sure. But you'll probably need at least 100V on its screen. That means a rather large cathode resistor on the output valve, which will dissipate lots of heat and require a larger power supply voltage. But it can be done. I think the 6AK5 was very carefully chosen.

Rick Francis, 22 August 2000; I recently decided to build an STC amplifier, or rather make one out of a stereo single-ended 6BQ5 amp I pulled out of an ugly Magnavox console a couple days ago. Hiroshi Uda's article explains what is special about the STC design: I think it is an excellent introduction. A couple of things to keep in mind are:

  1. the STC doesn't claim to give state of the art results, but instead it is a neat trick to make the most out of small cheap iron, and

  2. one of the things you give up is measured output power.

A little console amp seems perfect for the job, and from this single-ended 6BQ5 amp I'll go with the Version 1 6BM8 design. The little output transformers can only be happier with the lower current of the 6BM8. I kludged a 7-pin socket where the amp had a 9-pin socket for a 6EU7, and drilled another 7-pin socket, so now I'm set for the 6AK5 input valves. The power supply voltage is close enough.

I bought the whole ugly console for US$15 at a garage sale just days ago, and the amp came out with two screws and disconnecting spade clips and phono jacks. I put the rest at the curb and someone took it home within 24 hours, probably to recycle the large woofers! These big consoles usually don't sell at estate sales (gee, I wonder why not?), so if you can wait till the end, you'll get it for free from the people they hire to drag the remains to the landfill! At least that's how it works here.  I hooked up the stock amp last night, and it worked well, no hum, but had that odd beamy-phasey left-right sound that I think results from grossly mismatched left-right parts. So I cut out the parts to redo it as an STC.

I mention all this to inspire cheapskates, as I expect my total investment to stop with the original US$15. If I really like the results, I can build something nice later.

Yuichi Torisawa; The super triode connection was devised from anode to grid NFB. Normally, a feedback resister is put in for anode to grid NFB. In the super triode connection, this resister is replaced with a triode valve. Another feature is the use of a constant current source in the first stage. A FET or pentode etc. is used in the first stage. They have high internal resistance. Therefore the super triode connection has high anode to grid NFB. It makes the internal resistance of the power pentode in the final stage very low.

The sound is wide, flat, vivid and has strong low frequency content. A large powerful power supply is important for the super triode connection. Using a large power transformer and many capacitors in the power supply is important. Provide at least 100mA and 2,000µF for each power pentode. Allowing for more than 100mA is not due to the idling current of the power pentode, but for the signal current to drive the speakers. High momentary current is readily available from the super triode connection. Kamijo-san devised this circuit, and it is revealed at his website. Unfortunately, the site is not fully interpreted into English, but the schematics in there are useful. Hiroshi Uda's explanation of the STC is in English.

Rick Francis, 26 August 2000; I hope to start wiring the STC today. I mounted binding posts and input jacks, and I think I figured out how to mount almost as much capacitance as the schematic calls for: 100µF for the first filter, then 200µF then 600µF. I can manage 100µF, 200µF and then 480µF, while still keeping the tiny chassis, by using TSHA's.

Aaron Bohnen, 15 Dec. 2001; I've been thinking a little recently about something I saw on one of the ubiquitous cool-looking Japanese schematics, (can't remember which one exactly - seemed like it was a regular feature on a few of them) - a stopper diode on the B+ supply to the outputs of an single-ended amp. I wasn't really reading carefully but I even seem to remember a mention of this recently in this group. Gary Kaufman maybe?

Anyway, since current flows only forward from the B+ to the anodes, through the valves, etc. I can't see what purpose the stopper diodes have. Even so it seems they are used and if I remember correctly someone here said they even thought they improved the amp! So I'd love to know what's going on with these? If they really do have a positive effect I'd love to know why/how.

Dan Marshall, 15 Dec. 2001; These amp's are known as UDA amp's, after their designer [Hiroshi Uda in turn attributes the topology to its correct originator, Shinichi Kamijo]. They were a topic of discussion here some year, or so, ago. In addition to the stopper diodes, the design used a triode coupled from the anode of the output valve back to the anode of the driver valve. This was called the "lineariser" and provided the negative feedback.

I built one using a GEC KT88 valve. While it sounded OK, it would only hack a couple Watts before going into severe distortion, and I was not willing to settle for a couple Watts from a GEC KT88 valve when you can get several times that amount using a conventional single-ended circuit. I tried shorting out the stopper diodes and couldn't discern any difference, either in bench measurements using a spectrum analyzer, or in listening tests. On the other hand, someone, Grover Gardner, as I recall, swore by the stopper diode principle and said he had started using them in all his designs. He also liked the sound of the UDA design, lineariser and all. So, as Al Marcey commented, I guess you will just have to try it for yourself and see whether you perceive any improvement. As for me, I abandoned the UDA circuit in favor of more conventional circuits.

Grover Gardner, 15 Dec. 2001; I did try the stopper diodes in a few of my single-ended projects, and while there was a definite difference in sound, in the end I did not choose to implement them. But there was a difference, mainly a sharpening of the image and an extra "crispness" to the signature. Certainly worth trying to see if it improves your system.

Dan Marshall, 15 Dec. 2001; . . . The output valve anode/driver anode feedback arrangements does work well though, especially with small (read scrounged) output transformers exhibiting excessive leakage inductance, though a resistor seems to work better here for me than does the "lineariser" valve.

The entire UDA concept is oriented toward getting good sound from small, inexpensive output transformers and, to some degree, does seem to work, if you have very sensitive speakers and can settle for a low power amplifier. However, some of the cheapie 6BQ5 pentode feedback amp's, like the little Heaths, also sound quite nice and put out more power at lower distortion than does the UDA design. If the output transformer has a lot of inductive leakage, the anode to anode feedback will provide a very stable amp, even with a lot of feedback, as it doesn't have to cope with the phase shift associated with output transformer leakage inductance.

Aaron Bohnen, 15 Dec. 2001; Hiroshi Uda addresses the stopper diode in his article on the STC topology. I'll shamelessly extract the portion dealing with the stopper diode from Hiroshi Uda's article. The article is all about STC amp's but we've talked those over quite a bit, so no comments on that right now. Here's the bit about the stopper diodes:

1.5.1 Stopping Diode (SD)

The SD may potentially be applied in both a voltage amplifier stage and in a power amplifier stage. The following discussion is in regard to the power amplifier application; experiments have not yet revealed any effectiveness of the SD in voltage amplifier applications.

(1) Effect of the SD

Inserting a diode between the B+ terminal of the output transformer and the power supply itself, in the typical STC amplifier, "stiffens" the sound of low frequency music content.

This phenomenon implies the existence of some inter-relationship between this type of amplifier and the loudspeaker system. The SD becomes more effective when a small output transformer is used, rather than a large one.

The effect of the SD is not merely specific to the STC amplifier however. It is also effective in other non-NFB or local-NFB amplifiers. In the case of overall loop NFB amplifiers, the effect does not appear at all or is reduced. Supposedly, the reason is that a loop-NFB amplifier acts to amend the shape of signal waveforms passing through, by the NFB action.

(2) Supposed Operation of the Diode in an Ideal Power Amplifier

The action of the diode appears similar to the operation of a "diode-switch", which acts as a signal path when forward biased with some DC current.

In an ideal amplifier, the final stage tube acts as a signal source and the output transformer acts as a load. The power supply should ideally have zero internal impedance. These three circuit elements constitute a DC loop, although the power supply itself is not involved with the signal path, as such.

Just the signal source and the load on the DC loop should share the signal power. The DC loop is essential however, to convey the signal power. There will be some impedance mismatching between the signal source and the load. The mismatch means that power is partially "pushed" through the output transformer, with a remaining portion absorbed and dissipated by the signal source, i.e. the final stage tube.

(3) Approach to an Ideal Power Amplifier

The ideal power amplifier would be supplied pure DC power ONLY from the power supply. The SD acts to make the power supply act as a pure power supply, i.e. never handling or affecting the signal. In another words, the diode causes the signal power to be restricted to the signal power source and the load. By virtue of such assumed operation, the diode is termed a "Stopping Diode". Further, the B+ power supply should contain enough energy storage capacity to support the amplifier's transient power requirements. It should contain about 5 or more times the "uF" of capacitance as the nominal B+ current in "mA". For example, an amplifier that requires about 200mA DC at no-signal, should preferably have a power supply capacitance of (5 x 200 = ) 1,000uF or more. This requirement for large capacitance yields a power supply with less internal DC resistance; that is to say, one providing close to the performance of an ideal power supply.

(4) SD Effectiveness is Related to Tube Type

Through listening tests, the effectiveness of the SD varies; pentode --> beam tetrode --> triode. This tendency implies a relation with the internal resistance of each type of power tube.

(5) SD Effectiveness for the Screen Grid

While the screen grid is normally connected to B+, the screen grid current is affected by the input signal. Inserting an SD in series with the screen grid produces a detectable sonic improvement, heard in listening tests. The operation of the SD here is similar to that of a plate circuit SD.

(6) SD Effectiveness for the Push-Pull Amplifier

When an SD is used with a center-tapped output transformer for a push-pull amplifier, inserted between the center-tap and the B+ power supply, the sound of low frequency music content "stiffens". The assumed principle of operation is that the SD pushes back or shuts off unbalanced signal power which appears on the center tap of the output transformer. In some cases trialed, the low frequency content of music signals became excessively boosted. The effect appears to be dependent on the inductance of the output transformer and the extent of the unbalance in the signal components.

(7) SD Effect Generating Elements

As a general rule, all kind of diodes, such as silicon diodes, diode-connected tubes, and diode-connected transistors etc. can be used for the SD. However, any element high in internal resistance, (for example more than 50 Ohms in the dynamic state), is unsuitable as an SD for the plate circuit. The internal resistance limits the output power, and "loosens" the sound of low frequency music content, etc. From this viewpoint, a power silicon diode such as the 1000V, 1A, 1N4007 is most suitable. A paralleled 6CA4/EZ81 or diode connection of the 6AS7G, 6AS7GA, 6080, 5998, 5998A would be considered second. Most general rectifier tubes or damper tubes are not suitable, except where inserted into the screen grid circuit.

Aaron Bohnen, 15 Dec. 2001; So, reading all of that and looking at the schematics a little closer I'm still a little puzzled but it does look like maybe a way to reduce inter-channel crosstalk at the B+ supply level? Possibly this locks down the bass a little better - something like the improvement we sometimes hear between monoblock and stereo amp's (i.e. having independent B+ supplies, maybe hence lower crosstalk), possibly leading to the "tighter bass" observation?

I'm working on an EL34 single-ended amp wired up in triode right now so I think I'll give the B+ stopper diode a try. I'm still very interested in hearing any thoughts on WHY it might be good. My theory about the inter-channel crosstalk at the B+ supply is still pretty half-baked.

Phil,15 Dec. 2001; I would say try it, and if you notice a difference in the sound then try a 1 or 2 Ohm resistor instead. It may be that the diode's ability to pass current in only one direction has nothing to do with any improvement in sound; especially if this ability is never used. In that case it must be some other characteristic of the diode that is responsible for the change in sound. The diode's resistance is the only other thing I can think of. On the other hand, if replacing the diode with a small resistor results in a totally different change in sound, or no change at all, then I'm with you as to why it works, namely clueless.

Guido Tent; Did you try them yourself? What is the result according to your ears?

Al Marcy; I thought it improved one of my output circuits. A little. A friend in Japan sent a 6BM8 amp with the Uda STC circuit and the 'lineariser' diodes. I liked the sound of my 6BM8 amp better, but, I had much bigger output transformers and a killer power supply. Tamsan's amp was very fine for its tiny size, a modified Elicit.

Richard Jones, 17 Dec. 2001; Could it be to prevent any back EMF getting into the power supply, i.e. if the voltage across the output transformer rises above the B+ voltage?

Acknowledgments:

My sincere thanks to Al Marcey, Dan Marshall, Tsugunari Eguchi, Yuichi Torisawa, Jean-Michel Le Cleac'h, Hiroshi Uda, and Hugh Dean for their generous permission to use their material in this fashion. My sincere and particular thanks to David Dlugos, for preparation of the various schematics from the original ASCII text drawings, and Carter Hendricks for his photo’s and incisive and valuable advice regarding content.

My apologies in advance to all those whose materials I have used without clearly and directly stated permission to do so. If you see your material here, and would prefer it withdrawn, I should be happy to oblige immediately!

David Dlugos
My heartfelt thanks to software and transmission-line loudspeaker specialist, Canadian David Dlugos, who prepared the various schematics from the original ASCII text drawings seen in this article.

Home