• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Beyond SET and ParaFeed.... Complementary Current Triode

Having participated in some efforts recently to DC compensate the SET transformer, so the air gap could be eliminated, I have worked up some new circuits that are analogs to SET design.

These provide the same single triode output tube gain, but provide either twice the power, or even 4 times the power (see next post), while preserving zero DC flux in the output transformer. This means that an ordinary P-P type transformer can be used, which will provide better bandwidth (ie. better Bass, and the option of NFB if wanted), and the cost and weight are less with more power output.

This uses the technique of complementary current to drive the other winding of a P-P transformer. Your first impression will be that this is just some sneaky version of P-P, but a closer look is required. The pentode is used as a programmable current source. With no plate feedback effect ( as with a triode output ), its high impedance output has no say or control of the output voltage. The triode is calling the shots here.

See figure. The constant current source in the "tail" guarantees that the currents in each half of the circuit add up to a constant. This means that for every change in current the triode commands to one winding, an equal but opposite change in current occurs in the other pentode/ transformer half. Since the other winding section is inverted in phase with respect to the output, this just means an exact doubling of the current change as far as the output is concerned. So the output current changes will simply be doubled by this exact duplication. Like having double the gm and wattage rating of the triode.

(using P = I squared R might lead one to think this would quadruple the output power. But, since the triode now sees a higher impedance, 2x, due to the pentode's assistance, we would use .7 times the turns on each half of the xfmr to get the same impedance back...., so power ends up being just doubled. Not bad though, twice the power output from your favorite triode. )

The pentode grid bias gets adjusted for equal DC idle current in the xfmr halves, so no DC is present flux wise in the xfmr. The 1 Ohm resistors are for setting bias. The pentode could be replaced with an N channel Mosfet, since it is just acting as a programmable current source. (You will notice that the pentode, or Mosfet, are operating in grounded grid mode, so the same current goes in as goes out, so no sonic effects on the signal) The triode's cathode sees 1/gm of the pentode (or Mosfet) as a cathode resistance to ground, so the Mosfet will reduce this effect dramatically.

One can still use a partial cathode feedback winding, or distributed loading, for the triode if desired, just put the winding in series with the triode's cathode to CCS connection point.



  • cct1.gif
    6.4 KB · Views: 3,399
alternate circuit

Here is an alternative circuit that accomplishes complementary current triode operation for a single winding type primary transformer.

See next figure. This uses a setup that looks like it's halfway between Mu follower and SRPP. The LEDs provide a constant DC voltage reference, C1 holding a constant voltage during operation.

With a constant voltage effectively across R1 and R2, the IR drops must sum to a constant. R1 and R2 are nominally equal in value, so the triode current and Mosfet current must therefore sum to a constant. (R1 and R2 are relatively small value resistors here, being used as current sense resistors.) This is just what we want, complementary currents. The triode controls the output and the Mosfet acts as a programmable complementary current source. (could use a pentode in place of the Mosfet too) Notice that a Mu follower provides a near constant current load, and the SRPP provides a voltage follower output, but here we use a varying current source on top.

In reality, the Mosfet (or pentode) needs some gate to source (grid to cathode) drive signal for operation, so R2 gets bumped up in value a little to provide this. R2 would be adjusted under large signal conditions into a dummy load to get zero DC current thru the transformer. This would be done using a milliammeter in the xfmr primary to ground connection and tweaking R2 during large sinewave signal (drive to triode) operation for DC null.

The previous circuit, in the last post, has an advantage over this circuit in that its instantaneous current draw on the B+ supply is constant. (all circuits here are class A, which at least have constant average current draw).

(couldn't get the figure accepted, even though its only 8.4 KB. I will try to post it separately in the next post.)

figure 2 hopefully

no luck, let me try figure 3 instead

no luck again

Well, figure 3 uses the figure 2 design on each end of the transformer primary. This is like what is called H bridge operation.
However, the triode drive is kept on only one side of the xfmr, the other side gets a pentode on the bottom with an AC grounded grid (like in fig. 1), and both sides have their cathodes connected to a CCS like in the first figure.

I'll have to try re-scanning the figures I guess.

figure 3 finally

This design provides 4x the power of the triode in normal SET mode. It also has constant instantaneous current draw on the power supplies like figure 1, so should be very easy on power supply filtering requirements.

Also, the triode sees 4x the load impedance due to the assistance of all the complementary currents, so the primary winding can have its turns ratio cut in half for the same Zo load as in SET mode. (of course, the actual # turns will be QUITE a bit lower yet than even 1/2 that on the SET xfmr, since we have no air gap in this xfmr, fewer turns required to get same primary inductance)

Don :)


  • cct3.gif
    7 KB · Views: 3,198
Hi Don,

These are interesting circuits, and remind me of John Broskies clever doodling at tubecad.com.

They are clearly all designed to remove or cancel standing DC current in the transformer. Admirable, but a bit complicated. Why not just use parafeed? It gives excellent results, and uses an audio optimized choke which is expensive, but it's much simpler and should remove the 'sonic overlay' of silicon which your mosfet technique would doubtless add......

How about an attempt to rid Class AB of the nasty switching artefacts? Or a tube VA followed by a Class AB output stage?


pros and cons

Hi Hugh,

Well, the original start of the DC compensation thread was motivated, I think, to find something simpler or cheaper than parafeed. The figure 1 ( CCT1 ) schematic is reasonably simple I think. The choke for parafeed is expensive and heavy and also parafeed has some limitations due to the capacitor in the circuit. Aside from the usual paranoia over capacitors, the main concern is resonances between the inductors (2 of them actually, including the xfmr) and the C, which complicates design and may cause limitations to bandwidth, especially if NFB is contemplated.

CCT1 not only avoids all this, but doubles the power output from the triode. A lot more triodes would be useable for CCT operation than SET, with the doubled power. The current source makes for an accurate complementer for the current signal by simple algebra, or Kirchoffs Laws. The use of a cascode Mosfet or transistor at the top of the CCS tail can remove any variable junction capacitance issues. (of course, one could use a tube CCS instead) The pentode or Mosfet side are operated in grounded grid where the same current comes out as goes in, so has to be pretty transparent. So I feel pretty confident on the sound quality of CCT1.

CCT2 and CCT3 (figures 2 and 3) on the other hand are a little more risque on transparency. I will have to listen to them to see how they perform. But getting 4x power out of your favorite triode has to be a strong attraction.

On the Class AB and switching spikes during crossover problem, see here:
The Hawksford Error Correction idea was developed for exactly this reason (although for SS designs originally) and is well regarded in the SS community. No reason it can't be used for tube designs too.

"Tube VA followed by a Class AB output stage"

Using the CCT2 circuit as a driver for a class AB SS output stage should work pretty good. Complementary drive currents would preserve the triode sound yet produce symmetrical drive for the N and P channels. Would want Hawksford EC in the SS output stage to clean it up. Get that SET sound with 300 Watt gusto.

Another interesting approach are Hybrid output stages. More than just class AB SS, these use SS and Tube in the output stage itself.
(Easier to drive the tubes too.)

Fig. 1:
Does LTP mean anything to you? It's PP. Surely you've seen the cheapa$$ amp schems which use just one driver tube for PP output with a common cathode resistor or CCS to drive the other? Same thing, just better cathode balance, but unbalanced devices so you'll get lots more 2nd H.

Since the dynamic resistance of a pentode plate is towards infinity, a single one at idle opposite the triode is true balanced-DC SET. Let me put it this way: if there is some delta I in the pentode plate, it will affect the effective load resistance the triode sees; thus, it must be either helping or fighting the triode. Zero dI means zero interaction means lone triode.

Fig. 2:
Score one for the SS guys: all a mu stage was ever designed to do was drive the output with something stiffer than the triode, which gets the red carpet and velvet rope. I'm assuming your resistor divider is designed to share the load somewhat, so this isn't strictly true; however, the FET is in the signal path regardless, doing its share of the oomph. If you don't count SS as an amplification device, then yes it is tubed; however, if you are a hardlining tubophile, this schematic will be burned in effigy.

Fig. 3:
Same deal, twice as bad: H bridge is *NEVER* the answer*. You get double the voltage drop in any direction. Note that output impedance is *in series*, so Zo is DOUBLE a single stage.
CT transformers were invented to solve this problem by allowing both tubes to sit on the ground while bumping away with only one series voltage drop at any given time. (Half bridge is certainly feasable, but lacking P-type tubes, you need a coupling transformer, and that just gets ugly.)

(*Well okay, to be fair: full bridge is useful when limited supply voltage is a design requirement; half bridge requires capacitor coupling or a split supply. Still, it's only practical where low voltage drops are attainable, as with good solid modern SS.)

Technically, if the bias is set properly, the SS will disappear (as far as voltage drop, being it saturates in the 1-5V range), leaving only the tubes' (30-200V) voltage drop as the limiting factor. Now it will look like half bridge, which looks like PP with the CT arranged out.

...Thanks for the thoughts, but I'm afraid you'll have to continue looking for that ever-elusive *original* tube circuit. ;)

Hi Tim,

"Fig. 1"
Yes, I've seen those designs. They usually have two identical pentode outputs and global NFB to set the voltage gain. Here we have a triode using its local internal feedback to set the voltage gain, but just on one side. This preserves the signal quality of a SET design. And yes, the even harmonics will be preserved, unlike typical P-P.

And yes, the pentode complementary current variation does affect the load impedance seen by the triode. It "exactly" doubles the load impedance. The key here is that it "exactly" doubles the load impedance, thus preserving signal fidelity with the SET design. One can also look at it as just doubling the Gm of the triode too. Still the same sound as SET, just louder.

Using just a CCS to balance DC in the transformer is quite acceptable too as far as sound quality goes, but it wastes twice as much power, why not get something useful for it. Here we get "exactly" twice the output for the same power. The CCS only design gets 25% max theoretical power efficiency, the complementary current design gets 50% max theoretical efficiency, just like in a normal class A P-P design.

"Fig. 2"
Just use a tube in place of the Mosfet, as I suggested for an alternative. Then it just looks like a cross between SRPP and Mu Follower. Yes, both devices provide equal drive to the load, I'm not trying to hide anything.

"Fig. 3"
Yes, H bridge does double the output impedance. And doubling the number of output drives halves the impedance. So ends up the same. Thats why I corrected myself in the next post on the turns ratio.

With two devices stacked up, one normally has to double the B+ voltages, like for SRPP or Mu Follower. Here we can drop back to normal B+ supply levels for similar power output due to the H bridge, so not so silly. Why use a dangerous +700V when +350V will do?

The two sides of the bridge are being operated with complementary currents as used in the Fig. 1 design (due to the CCS), so the single triode is again setting voltage gain by its local feedback. That's how it preserves the SET sound. And even harmonics.

Don :)
maybe out for the day!

I read in the paper where MIT just held a conference for time travelers on Saturday. I'm thinking of attending tomorrow. So I might be out for the day. But I guess I could put it off for a day or two. Let's see, how many megawatts per day will that cost to get there? Warming up the tubes in the Tardis.


Sch3mat1c said:
Fig. 1: Does LTP mean anything to you? It's PP...
Fig. 2: ... If you don't count SS as an amplification device, then yes it is tubed; however, if you are a hardlining tubophile, this schematic will be burned in effigy.
Fig. 3: Same deal, twice as bad... full bridge is useful when limited supply voltage is a design requirement... it's only practical where low voltage drops are attainable, as with good solid modern SS.
...Thanks for the thoughts, but I'm afraid you'll have to continue looking for that ever-elusive *original* tube circuit. ;)

Well, that would be my comment too ;)
Seeing fig 1 I first thought, OK, PP amp, what's new?
The forementioned John Broskie's pages are a good place to revisit from time to time, I'd say ;)
this was posted on last Saturday

Hi John,
A triode output with a CCS above it to cancel DC to the xfmr does indeed waste power. Attaining only 25% efficiency at best.

The CCT circuits however operate similarly to a P-P Class A amplifier since the current source is driven in a "similar" inverted manner (at least visually on a scope). This allows 50% maximum theoretical efficiency. Putting a scope to the CCT circuits, one would be hard pressed to see any difference from a typical P-P Class A circuit operation. The actual diference is in just the small distortion components.

The CCT circuits preserve the signal waveform in exact replica form to a SET design, preserving even harmonic distortion components. The true P-P designs cancel out the even distortion components by means of exact symmetry in the inverted drives. The CCT drives are not "exactly" symmetrical in this way, but one would need a distortion analyzer to see the difference in the signals between CCT and P-P. (too subtle to see on a scope)

Hi Piergiorgio,
The Murray amplifier is indeed interesting, thanks for bringing this up. It would appear to be a SRPP input stage driving a triode output with an active triode totem- pole load. The top triode is getting an inverted drive signal via the pentode. The pentode in turn is acting like a tube op-amp in inverting input mode. It drives the top triode in such a way as to null the input current from the output signal with the input current from the drive signal to the bottom triode (they are inverted with respect to each other), insuring accurate matching of the final output with the SRPP output.

I'm hard pressed to say whether this is P-P or SET like emulation or what. The difference between the CCT and P-P designs is already rather subtle, its in just the distortion components. I guess the final arbiter would be to put a distortion analyzer on the output and see if it has even harmonic components or not. But regardless of this "SET emulation" issue, its an interesting design in its own right.

I would say it looks more related to output error correction schemes actually. Yet it does have a certain similarity to the CCT2 (figure 2) design in deriving the drive signal for the top device from the bottom output device. The CCT2 derives the drive from the output of the bottom device, whereas the Murray design really derives the top drive from the input to the bottom device, but in a feedback kind of way to get the final output to match it, rather than the top drive signal itself.

I recall a SS design in Wireless World mag. a few years back that used the top output device to insert an error correction for the bottom device in a similar manner. I believe it attained very low distortion all-together (both even and odd harmonics greatly reduced). So this may really be a difference from both SET and P-P design, but subtle in detail.

Hi ilimzn,
Perhaps you would be happier with the name "SET emulation" or assymetric drive P-P. Obviously, as I originally pointed out, it does look like P-P at first sight. The differences are subtle indeed. Conventional P-P cancels out even harmonics from the output by means of its exactly symmetrical inverted drive signals. People spend a lot of time argueing the merits of various inverter stages.

CCT circuits on the other hand preserve the even harmonic structure of a SET design. A subtle detail not immediately obvious, even on an oscilloscope. But regardless as to weather one wishes to call it P-P with assymetric drive, or "SET emulation", or CCT, the sound will be different than normal P-P. It should SOUND just like a SET amplifier, and people can hear the difference between normal P-P and SET.

Don :)
Kevlar Balloon!

Hi Tim,
I'm quite aware of the non-linear 3/2 power law for pentode current versus drive signal. But if you will notice, the CCT1 and CCT3 designs uses the bottom pentode (or Mosfet) in the grounded grid configuration. The same current comes out as goes in the cathode except for screen current siphoned off in the case of the pentode.

As long as the plate voltage does not drop to near the screen voltage, the screen current is a near constant percent of the plate current due to interception, so has low distortion. But is a problem at higher voltage swings. The Mosfet does not have this problem, and is really a better part to use here. One might consider a constant current source for the screen supplies, would prevent distortion, but not sure how the tube would lke it, probably would oscillate at HF.

The Mosfet's very high Gm also is an advantage in the long tail/CCS operation too, since it basically pins the cathode voltage of the triode, preventing any impedance modulation in the triode cathode circuit. I would use the best part for the job, but some are very insistant on using "glass" everywhere.

In the CCT2 and CCT3 circuits, the main criteria for the accuracy of the upper devices is again high Gm, so I would prefer the Mosfet there too. But screen current is not an issue in CCT2 and CCT3 (top devices) since we are using their cathode currents. So using a lower Gm pentode will just mean a little more bumping up of R2's value for drive signal. The bottom pentode (in CCT3) would however be better with a Mosfet to avoid the screen current distortion (as in CCT1).

One COULD...I suppose, use a 2nd triode in place of the bottom pentode in CCT1 and CCT3. This will still work in grounded grid operation to convey the complementary current and will avoid screen current distortion. The penulty would be higher drive voltage on its cathode to get the required current for the CCS to be happy. This would produce some impedance modulation in the main triode's cathode circuit. Don't know off hand how this would effect the operation of the main triode as far as sound quality. Would just have to try it and listen I guess, maybe compare spectrum analyzer results too.

Don :)
Cascode in place of pentodes

Just had another thought, could use a cascode of triodes in place of the bottom pentodes to cure the screen current distortion problem. So looks like an all tube version is still looking do-able.

A very practical solution would be to use a grounded gate Mosfet with a triode cascode above it in place of the bottom pentode. Then the mosfet can be a small signal, low voltage type and the tube can handle the high voltage output. And no extra B+ required.
Can feel the wind picking up in the sails, no credible show stoppers yet.

Don :)