Puzzling audio circuitry

[Aerodyne]

It may be that I've missed something here... or age has finally and perhaps irrevocably caught up with me... but I am rather puzzled by this circuit which shows the two halves of a double triode with the anode of one directly connected to the grid of the other, with the audio output taken from the cathode of the latter triode. My question is this: what is the point of this circuit? It isn't a cathode follower, at least not in the accepted sense (as far as I can see) and if it is a means by which impedance is modified, how does it work?

Can anyone provide a clear and straightforward explanation? Yes, I have searched books and the net and can find nothing about it.

Tony

[wd40addict]

It's the so-called shunt regulated push-pull (SRPP) circuit beloved of Japanese audiophiles. In essence the upper triode acts as current source for the lower. Whether this is actually an improvement over a simple resistor I'll let others decide...

[John10b]

Interesting circuit, it got me thinking and one circuit configuration I vaguely recall, in which the anode of one valve was connected directly to the grid of another valve was in a constant current configuration, I’m not able to elaborate it’s just a distant memory. I did think of a cascade amplifier, but that’s anode to cathode, just can’t recall. I’ll be interested to know what the circuit is used for.
Cheers
John

[Aerodyne]

I'm certainly familiar with the basic idea of the cascode circuit, but this isn't one of them, to my way of thinking. Quite what the point of it is still eludes me; something called a 'shunt' should be in parallel, surely.

Replies so far much appreciated. Perhaps others may shed a little more light on what in my ignorance is a seemingly wasteful use of a valve!

Tony

[Radio Wrangler]

The top triode acts as a constant current source, biassing the anode of the bottom triode. This allows the bottom triode to have high gain - limited by the valve's own mu, there is no anode load resistor to load it.

There's a nice twist that the feed to the next stage comes from the top triode's cathode, this means there is a degree of buffering done by the top valve as well as providing the constant current load.

Of course, this leaves the Miller effect around the bottom triode.

Then there is overall feedback from a purpose-designed transformer winding to the bottom triode's cathode.

Quite elegant and fun.

But it's not the be-all and end-all.

David

[John10b]

Cascode not cascade, of course, I’m blaming my age
Cheers
John

[G6Tanuki]

The top triode is being used as a sort-of constant-current 'load' for the bottom triode in place of the usual anode-resistor.

[Aerodyne]

I noted the negative feedback. I doubt it would make any (audible) difference if the feedback was fed from the secondary in 'normal' fashion. I get the 'buffering' point and can see its use to prevent the following stage loading the driver, but for all that I do wonder about its practical effect. Is it simply audiophoolery, or actually worth the trouble? With my ears, I'd be guessing the latter!

Tony

[G8HQP Dave]

The top valve in an SRPP acts both as an active load and as a (rather modified) cathode follower. The net result is a lowering of output impedance, although not by as much as you might expect. However, this will have little or no effect with a high impedance load so most audio uses of SRPP must be regarded as a popular fad.

The SRPP was invented to drive low impedance loads. There is an optimum load impedance (for lowish distortion) which results in both triodes acting a bit like push-pull, but in this regime the gain is quite low because the optimum load is usually much smaller than the output impedance.

The SRPP is related to two other circuits:
1. the mu-follower
2. the grounded cathode stage with cathode follower and bootstrapped anode load

[turretslug]

It's peripheral to the main subject of discussion, but it seems odd that both chokes are 2H 100mA types- fair enough for the first section of the PSU filter but surely for similar volume and expenditure on iron and copper it would make better sense for the second choke to have rather more inductance and rather lower current rating. Or maybe Hammond were offering BOGOF on a slow-moving item.

[turretslug]

Too late to edit... it's also noteworthy for using a single-ended ultralinear output stage, which can't be very common- though the transformer appears to be a commercial off-the-shelf item.

[Argus25]

It is just another version of cascode. The difference simply occurs due to the point that the grid of the upper valve (or base or gate of a semiconductor) is connected.

In traditional cascode it is just clamped to a fixed voltage, but it can be connected to other points to, so as too add negative feedback and reduce distortion. In the case of the circuit posted, connecting the grid to the lower end of the cathode resistor (which happens to be connected to the plate of the lower valve) adds negative feedback to the upper tube(which is now a "degenerated current source').

The other place to connect the grid of the upper valve, is to ground, but for valves it is harder to meet the DC conditions, but with J-fets, it is perfect, which is how they built the Fetron.

[Synchrodyne]

As I understand it, the cascode circuit as developed and named by Wallman et al cascaded a grounded-cathode triode input stage with grounded-grid triode output stage. I think the original used AC-coupling, and that RCA was the first to use direct coupling. But either way, the grid of the output valve was at signal ground, although not at DC ground.

The SRPP circuit shown in the original post does not seem to have a grounded-grid output triode. Rather its grid sees the same signal voltage as is present at the input stage anode. So it appears not to meet the requirements for a cascode, at least as originally defined.


Cheers,

[Argus25]

The SRPP circuit shown in the original post does not seem to have a grounded-grid output triode. Rather its grid sees the same signal voltage as is present at the input stage anode. So it appears not to meet the requirements for a cascode, at least as originally defined.

Normally though the AC signal is extracted from the anode circuit of the upper valve, and if it were, then in the original post circuit you could ground the grid of the upper valve to AC signals with a capacitor to ground. But since they decided not to, the AC output is available from the cathode of the upper valve instead, allowing grounding of the plate to AC signals. Many circuits are variations on the theme of another.

[Synchrodyne]

The attached might help somewhat to place the circuit at interest.

[kalee20]

If you move the connection of the coupling capacitor to the LOWER end of the 560Ω resistor, then the upper triode and 560Ω resistor becomes just a 2-terminal bit of circuitry, a constant-current load for the lower triode. This allows it to give the full gain equal to the μ of the valve, and at low distortion. Just as Radio Wrangler and others say.

However, the grid resistor of the following stage rather spoils this by reducing the load from near infinity to 270kΩ.

Taking the output from the upper end of the 560Ω resistor gets a bit of this back, in a way which I can partly see, but I'm missing the essential paper and pencil to work it out right now!

I wouldn't, however, call it a cascode - the upper valve isn't operating as a common-grid stage.

[Radio Wrangler]

The cascode circuit takes its output from the anode of the upper valve, and the anode of the lower valve has very little signal voltage on it as it is loaded by the grounded grid valve above it.

This pretty much kills the Miller effect of the two valves and replaces the Miller pole in the response with TWO poles, but both at appreciably higher frequencies. So the usual cascode circuit doesn't increase the gain at low frequencies, but it extends what gain there is to much higher frequencies. It is a broad-banding technique.

The circuit in the original post is very different to the cascode. In this circuit, the full output voltage is developed at the anode of the bottom triode and the Miller effect is in full force. The gain of that bottom valve is increased to an amount set by its own anode impedance as all other loads are buffered by the top valve. This increase in gain acts to strengthen the Miller effect and reduce the frequency of the Miller pole.

So the posted circuit offers an increase in LF gain at the price of a reduction in bandwidth compared to a single triode stage. It offers a reduced output impedance for driving subsequent stages, but the current availability is still ultimately limited by the quiescent current in the triode string.

What it shares with the cascode is the passage of HT current through two triodes in series. What it does, though, is quite different.

It's an interesting circuit and rather clever, but it's rare because its advantages lie in a direction which isn't often needed and its disadvantages remain significant.

David

[Maarten]

Philips used this output stage topology in several radio and TV models from the early 1950's up to about 1971 (B/W TV), with pentodes instead of triodes. They developed a range of 800 ohm loudspeakers for it.

[Aerodyne]

Thanks to all who have contributed. Armed with the magic term 'SRPP' I searched again on the net and came up with this detailed explanation (see link below). It doesn't quite meet my requirement of simplicity but it does cover the ground, offering equivalent circuit descriptions and the development of the original idea.

So, from this thread I have again learned something I'd never heard of before, which goes to prove that one is never too old to learn (or would that be better to say that one can't teach an old dog new tricks?). Either way, it has proved an interesting if winding path to follow.
I won't, however, be including it in any future amplifier build.

http://www.valvewizard.co.uk/SRPP_Blencowe.pdf

Tony

[G8HQP Dave]

Originally Posted by Argus25

It is just another version of cascode.

No. Not related to the cascode, apart from the fact that the two valves are wired in series - which is not an essential element of the cascode anyway. In the cascode the upper valve raises the output impedance; in the SRPP it reduces the output impedance. In the cascode the upper valve greatly reduces the voltage gain of the lower valve; in the SRPP the upper valve raises somewhat the voltage gain of the lower valve. In some senses you could say that the SRPP is the opposite of a cascode!