Can someone help me understand this topology?

[Wigwam Jones]

I happened across this page - it shows a 12AU7 tube being used in an output position, without any kind of input tube stage. Naturally, I am sure output is measured in milliwatts, but it looks very cool.

12AU7ƒAƒ“ƒv

I find the final schematic fascinating, but I do not understand how it works. I am not used to seeing CRD diodes, so I'm not sure what's going on here.

Also, the design calls for a pair of Sansui ST-24 input transformers per channel. I see they are quite popular (and apparently inexpensive) in Japan for hobby tube amplifiers, but I can't seem to locate anyplace in the US to purchase them, or a suitable replacement for them. Any thoughts on that?

Thanks!

 

[sreten]

Hi,

The ST24 is a 1K:2K transformer, (ratio 0.7:1) but its
main purpose it to act as a phase splitter as I see it.
I can't see why standard 600R types would not work.
(In fact I think loads of small transformers would work.)

CRD's are just self current biasing FET current sources.

rgds, sreten.

 

[rayma]

I am not used to seeing CRD diodes, so I'm not sure what's going on here.

The current diodes are in parallel and add their currents, just like Zener diodes in series add their voltages.
The higher the impedance of the current source, the better the CMRR. These take the place of the usual
resistor connected to a negative supply.

 

[Wigwam Jones]

Thanks! Fascinatingly simple, it seems. No resistors or capacitors in the signal. Does this not seem unusual?

 

[Wigwam Jones]

Hi,

The ST24 is a 1K:2K transformer, (ratio 0.7:1) but its
main purpose it to act as a phase splitter as I see it.
I can't see why standard 600R types would not work.
(In fact I think loads of small transformers would work.)

CRD's are just self current biasing FET current sources.

rgds, sreten.

What would be a good choice as a 600 ohm CT input transformer in that case? From what I could tell, the Sansui are very cheap in Japan.

 

[rayma]

No resistors or capacitors in the signal. Does this not seem unusual?

This topology has also been used for power amplifiers, using a voltage regulator like the LM317
connected as a current source, but it must run class A at all times.
http://www.tubecad.com/2011/08/blog0210.htm

 

[sreten]

Hi,

No feedback, gain will wander, and depend on the tubes.
No method of balancing mismatched tubes for gain
or standing current in the two halves.

Sometimes things can be too simple. As this seems.

rgds, sreten.

 

[Wigwam Jones]

This topology has also been used for power amplifiers, using a voltage regulator like the LM317
connected as a current source, but it must run class A at all times.
Cathode Bias with a Counstant Current Source

Let me ask you this then. Given the topology and referring to your link, would a common cathode triode work in this application, say maybe a 6j6?

 

[JonSnell Electronic]

The valves cannot bias because the centre tap of transformer should be at 0V and NOT on the cathodes. The triodes will be at zero grid voltage with no negative with respect to the cathodes.

 

[Wigwam Jones]

The valves cannot bias because the centre tap of transformer should be at 0V and NOT on the cathodes. The triodes will be at zero grid voltage with no negative with respect to the cathodes.

I'm afraid that's a bit beyond me, but I understand parts of it, I think. Are you saying that based on the schematic shown, the amplifier cannot function?

 

[john_tracy]

My line stage uses this topology. Substitute a Lundahl LL1690 input trannie, Russian 6N30s, Lundahl LL1689AM line out and a cascode MOSFET CCS (needs negative supply). My phono pre and power mono-blocks also use the same topology. Only caps in AC signal loops are RIAA caps in phono pre and they are shunt, not series connected. Sound is very expressive, dynamic and crystal clear. This circuit is very sensitive to the quality of the cathode CCS. Cascode MOSFET CCS will out perform and sound better than parallel JFETs but require a neg. supply (simple Schottky bridge with a choke input and cap works well). For currents of 10mA or less a cascode BJT CCS like those found in M. Jones 4th ed. will sound even better. For output stages with lots of current I've found a current sink using a single big power MOSFET controlled by a TL431 chip to be stable and has worked well for me (PPP KT120s, 300mA @~70V). Tried parallel CCSs for this and all ended up oscillating. Surprisingly, to me anyway, given the PSRR of a differential stage is that going the extra mile and using CCS fed active shunt regs for the B+ such as the Salas design makes for a significant sonic gain. All this leaves you with a nice clean AC signal loop that's isolated from the PS and ground.

 

[BinaryMike]

If you look at the curves on the datasheet, you will see that zero grid bias is about right for designed cathode current in this setup. So it should work just fine, but understand that you're working in class A2 here. The grids go positive under drive so it's important to provide a very low driving impedance to keep distortion under control. This output stage might be an ideal candidate for direct op-amp drive, if you're interested in that sort of thing.

 

[rayma]

Let me ask you this then. Given the topology and referring to your link,
would a common cathode triode work in this application, say maybe a 6j6?

Sure, if the DC conditions are properly set.

 

[rayma]

Are you saying that based on the schematic shown, the amplifier cannot function?

If needed, negative grid bias can be had by using cathode resistors,
or a floating bias supply in the transformer-to-cathodes connection.

 

[BinaryMike]

A bit more explanation: In this case, the constant current source is actually adjusting the triode's overall cathode-anode voltage to settle on the 30mA (36mA?) operating point.

 

[sreten]

Hi,

The constant current source is sinking current from both valves.
So they will operate with 15mA(18?) in an ideal world. It isn't.

rgds, sreten.

 

[Wigwam Jones]

From what I understand (about every third word, but I am trying!), it seems the overall consensus is that this is not a good topology, or that it's good on paper but impossible to exist in the real world. Is that about it?

Taking it a step further - assuming I could find substitutes for the unobtainium Sansui ST-24 input transformers in the USA, and I built this thing, I could expect it to sound bad? It seems so simple that it might be worth building just to see, but if ya'll think it's right from Satan's buttcheeks, I guess I'll pass. Any thoughts?

 

[sreten]

Hi,

I will look good in a sim, but is not a real world circuit.
It will have serious real world problems you don't want.

rgds, sreten.

 

[BinaryMike]

It might sound interesting, in the same sense that low-feedback SET amps often do. They can add an euphonic coloration that's mostly due to 2nd harmonic distortion and relatively low damping factor. As sreten points out, triode matching within a 12AU7 is basically a roll of the dice, so the amount of even harmonic in this design will be highly variable. Tube rolling fans would love it.

 

[45]

The valves cannot bias because the centre tap of transformer should be at 0V and NOT on the cathodes. The triodes will be at zero grid voltage with no negative with respect to the cathodes.

It is biased at 0V. Each tube will run at some 110V/16 mA which requires 0V bias. Of course it means that it will be driven into positive grid. So one needs some power drive and not just a common preamp.