Designing an OTL / Circlotron

[atmasphere]

Years ago Svetlana (the real one) published the grid current territory of the 6AS7G. At the time, Eric Barbour worked for them. He told me later that the current just kept going up as he raised the grid voltage, so he graphed it to the end. I might have one of those datasheets around still, although Svetlana (who are defunct in the US after loosing their name/trademark to New Sensor) published this stuff a good 15 years ago.

Anyway, the tube was pretty linear right to just shy of saturation. By contrast the 6C33 has a grid current character that looks a lot more like a pentode.

 

[banat]

A2 operation mean basicly next:When drive signal on the grid of 6as7 goes positiv(on the top of drive signal envelope)grid become Virtual Anode and take some electrons emision from cathode,rest of electrons(major part) flow thru the grid and become accelerate on the way to Anode(wery simillar acting like g2 of tetrode or pentode tube) ,with typical wery small place betwen the anode and grid backflow is posible from secondary emision from anode,that is curent flow of g1 or A2 class.Grid curent mean some grid power disipation,that is actualy real reason why tube 6as7 with top or botom Grid heatsink is proper for A2 operation and other variant of 6as7 without this heatsink not.6c33c-b have no this grid heatsink so grid maximum curent(power)is limited,all 6c33c grid curent beyond cetrain value overheat the grid structure and produce grid warping(tube failure).but AB class with 6c33c-b is quit OK,special for OTL Circlotron topology.Good Luck to Eweryone

 

[banat]

OTL Multiphase Circlotron,good Concept for OTL Tube Amp but for SS Amp too:OTL Asylum

 

[Miles Prower]

So now, I thought i'd try my hand at building my first tube power amp.

Good for you.

So i was thinking of trying an OTL circlotron, the main thing that's put me off from building a tube power amp in the past is the availability, expense, and designing of an output transformer.

So first, some theory questions. I've designed (very) high power RF (80kW-peak 10%-duty 50MHz) in the past, so i definitely know what happens when the source-transmissionline-load are not completely impedance-matched. But I was also taught that for any voltage source, output impedance should be as low as possible to minimise internal losses in the source and send all the power to the load. This is certainly true of the solid-state amps i've built, lowest possible output-impedance is best.
So which is right? Or is it that at low (audio) frequencies, the wavelength is so long that impedance mismatch doesn't matter so much?

Look at it this way: if your highest audio frequency is 30KHz, and you have a T-line with an 80% velocity factor, your wavelength is 8000M. Unless your speakers are 400M (0.05Y) away from your amp, you won't need to worry about it. In audio, what low source impedance does for you is help keep your speeks under control. That's what's meant by "damping factor".

So, to valves. I can certainly understand that a 400V swing at low current has to be transformed to a lower-voltage higher-current for a speaker-sized load. But for the schematics i've been browsing so far, many people just say 'add more parallel output tubes for lower output impedance', which i certainly understand. But (presuming i can get a spice program for linux and good models) how many is enough, and how low should the output impedance be for (eg) an 8ohm load? exactly 8ohms? lower? and what happens to the output power, distortion, frequency response, all of that, if i have a higher output impedance? (of course, i could just add x more tubes, but i'm on a limited budget here).

Look at it this way: let's say you want 15W into 8R speeks.
P= I^2R
I= sqrt(P / R)= sqrt(15 / 8)= 1.37Arms
I= 1.37sqrt(2)= 1.94Ap

To get that 15W, you're gonna need 1.94Ap. That's going to take a helluvalot of VTs to supply that current, and typical audio types like 6AQ5s, 6V6s, etc. aren't going to be up to that task.

There are, however, types made for pulling the bigamps: TV horizontal defection types. That's what that EL509 is. There are other types as well. There is also a power triode designed for series pass voltage regulators: the 6AS7 and its cousins (6080 (industrial type) and 6082 (26.5V heater version)). This type is a dual triode, has a fairly decent plate characteristic, and is available. With some spec busting, you could easily pull 350mA / cathode from it without impacting service life all that much.

1.94 / 0.35= 5.54

Here, you would need six tubes per phase in parallel to get the required current for the power level. With 6AS7s, that's six holes in the chassis, as opposed to twelve if using singleton triodes or pentodes.

So, to the circuits. I've been googling OTL and circlotrons, and a fair few seem to favour EL509s, which i think i can source a few of. Are there any other suitable tubes with low output-impedance that can be used in OTL designs?

Theoretically, there are lots of high powered TV HDs that would work. Unforch, during the mid-1970s and the "CB craze", all too many of these very types went into illegal, 1000W, CB "linears". Being illegal rigs, they all too often weren't designed right, and burned up tubes very fast. It's not so easy to get these types, nor is a future supply guaranteed. 6AS7s are your best bet these days.

My own inclination would be to build the Circlotron back end with power MOSFETs, and the front end hollow state.

If you really want some decent advice about hollow state Circlotrons, search up "atmasphere's" posts.

 

[Alastair E]

The old 'Allan Kimmell' Sweep-Tube OTL is a fair design...

I re-designed this to use 6C33C dropping the screen-drive of Pentodes of the Kimmell design for a more normal grid drive to the 6C33C..

A Pair of 6C33c will happily drive 25W into 8 ohm, with peaks over 50W....

Here you go, its a complicated beastie, but worth it--

Sweep Tube OTL Monoblock © 2004

 

[SemperFi]

Good for you.

Look at it this way: let's say you want 15W into 8R speeks.
P= I^2R
I= sqrt(P / R)= sqrt(15 / 8)= 1.37Arms
I= 1.37sqrt(2)= 1.94Ap

There are, however, types made for pulling the bigamps: the 6AS7... With some spec busting, you could easily pull 350mA / cathode from it without impacting service life all that much.

1.94 / 0.35= 5.54

...With 6AS7s, that's six holes in the chassis...

.. 6AS7s are your best bet these days.

I get about 550mA out of each triode section without cathode follower driving the grids. With a CF driving the grids I get about 750mA. I've had my first OTL running for almost 10years now, not every day but still many hours, and the tubes are still fine

So in a real world OTL you 'only' need 1.94/0.55=3.5 tubes, or four holes in the chassis.

 

[atmasphere]

I've built a small Circlotron OTL that has 4 6AS7Gs per channel. I did not set it up for class A; I used a 200V B+ and limited the idle current to about 25 ma per section. The amp makes 25 watts into 8 ohms at clipping, although it can do over 50 watts into higher impedances.

The 6AS7G will work nicely in Alan Kimmel's design also. **They are a lot less trouble than 6C33s!!** For example, your socket's life will be measured in decades rather than in 1000s of hours (6C33 sockets are usually pretty baked after about 5000-7000 hours).

 

[ilimzn]

I hear Ralph's statement that experimentally, at least in the Atma-Sphere, the current though the tubes only cuts off at the bottom of the sinusoid if clipping has been reached (i.e. full power). But I still don't understand WHY this can happen, since it appears to contradict the standard argument that implies class AB beyond a rather lower power. Is the class A2 issue playing a role here? I can see how that means the maximum plate current can be bigger at the top of the sinusoid than might otherwise be achieved, but I don't see how it could help with preventing a cut-off down at the bottom of the sinusoid.
Can anyone shed light on this, with some mathematics of what it happening?

This is because you are looking at it from a standpoint of an ideal amplification device, where current increases in one half of the output stage by the same amount as it decreases in the other. The 6AS7 is actually a rather non-linear device, which surprisingly gives it some very desirable characteristics when used in an OTL. In essence, it leads to something close to what is known as hyperbolic class B - because the tubes compress one side of the sinusoidal waveform, it is very difficult to get them to cut off, even without going into class A2/AB2. Because of this it is possible to set a considerably lower idle current than half the maximum, and yet still have both sides conduct all the time. Weather this is truly class A operation, is a whole different discussion, as it depends on what definition of class A you use (the one about percentage of cycle conduction, or the one about cut off)

 

[atmasphere]

The 6AS7 is actually a rather non-linear device, which surprisingly gives it some very desirable characteristics when used in an OTL.

Through most of its range the 6AS7 is pretty linear like any triode, but at the extremes of operation it *is* a little odd. I think your explanation is the most succinct/accurate that I have seen. I would like to add that the 6AS7G, in addition to these traits, is also very economical, making good power for the number of tubes used, filament current used and overall cost of parts.

 

[cnpope]

This is because you are looking at it from a standpoint of an ideal amplification device, where current increases in one half of the output stage by the same amount as it decreases in the other. The 6AS7 is actually a rather non-linear device, which surprisingly gives it some very desirable characteristics when used in an OTL. In essence, it leads to something close to what is known as hyperbolic class B - because the tubes compress one side of the sinusoidal waveform, it is very difficult to get them to cut off, even without going into class A2/AB2. Because of this it is possible to set a considerably lower idle current than half the maximum, and yet still have both sides conduct all the time. Weather this is truly class A operation, is a whole different discussion, as it depends on what definition of class A you use (the one about percentage of cycle conduction, or the one about cut off)

I agree with you, that the non-linearities play a role here. But what still puzzles me is this: Let's say the plate voltage is 150V, and the idling bias current is set to about 70mA. This means a grid voltage of about -65V, I think. If the amp is operating in Class A2 mode then at the peak of the sinusoid the grid voltage is going positive. Maybe just to be conservative, we say it reaches 0V. That means the voltage excursion is 65V between the midpoint of the sinusoid and the peak. Therefore at the bottom of the sinusoid, the grid voltage is about -130V. Now, the curves on the 6AS7 data sheet end long before -130V of grid voltage is reached, but it looks pretty clear that the plate current will be pretty well zero, long before the grid goes that much negative. Thus I am still puzzled about how it can be Class A.

One thing I have neglected in the calculation above is that plate current is determined by the voltage difference between grid and cathode, and the cathode of the 6AS7 is not at a fixed voltage, but is itself rising and falling with the signal. I haven't quite figured out how to factor this into the calculation, but it somehow doesn't seem like this would be enough of an effect to make all the difference.

I also assumed that the waveform from the driver stage would be a sinewave, and so if midpoint to peak took it from -65V to 0V, then midpoint to trough would take it down to -130V. I guess we can assume the driver is supplying pretty much an ideal sinewave, right?

If anyone can shed light here, I'd be most grateful!
Thanks,
Chris

 

[Jeb-D.]

6c33c-b have no this grid heatsink so grid maximum curent(power)is limited,all 6c33c grid curent beyond cetrain value overheat the grid structure and produce grid warping(tube failure)

There is fairly popular A2 6c33c SE design that has been floating around which claims 30Wrms. This would lead me to believe it's grid can tolerate some abuse (but how much is the question). The amp does use an EL34 as a driver though

This thread has got me wondering if I could squeeze any worthwhile gains out of my OTL by inserting a direct coupled CF driver stage. Perhaps something like an ECC99.

 

[banat]

There is fairly popular A2 6c33c SE design that has been floating around which claims 30Wrms. This would lead me to believe it's grid can tolerate some abuse (but how much is the question). The amp does use an EL34 as a driver though

This thread has got me wondering if I could squeeze any worthwhile gains out of my OTL by inserting a direct coupled CF driver stage. Perhaps something like an ECC99.

6c33c-b with some A2 design,but which one?Circlotron,Futterman,inverted Futterman,there is so many design.One design which I am Invented and first used on this World is OTL Multiphase balanced bridge HI FI Amplifier.In these OTL Multiphase output stage Concept I use DC coupled driver stage(inspired with Mr Karsten- Atmasphere driver stage design),each output 6c33c-b tube have asociated separate driver/bias configured of 1/2 of 6h8c(or 1/2 of 6sn7gt)per phase,extra suplyed with bipolar +-350v line(non regulated,but wery beefy PSU).Why DC coupled-Beacose output tubes behavior under this driver/bias topology is controlled up to maximum.On the first line -output tube INSTANT OVERLOAD RECOVERY,second -No EXPENCIVE Coupling capacitor( finaly newer perfect designed but wery expencive,similar like OPT) and third,most important- Quality of Reproduced Sound.In this my Multiphase Concept each driver tube have` easy`job beacose drive only one output stage tube grid(not the bank of parallel conected tubes(grids).And each Output Tubes ` look` in own Independent PSU DC line,Result is: NO AC(from driver) or DC(from output stage PSU) interaction betwen Output power Tube(or interaction betwen the same output Tubes),and must admit -more complex PSU.About driver stage againeople dont understand next,in A2 or even in AB2 class Driver stage `pull` some power from Output Tube grid(not push) so lowest internal driver stage impedance is here Crucial for stable output stage operation.On these days I experiment with 6c45p-e(triode) for CF- DC coupled driver stages, with diferent PSU/bias topology-have Excelent Sonic results.(EL34 for driver-No Way).Good Luck

 

[banat]

Forgot to say this:To avoid curent overdrive of the output power tube grid(6c33c-b) here is very simple solution:Grid stopper Resistor,470-820 ohm/2w value(placed near as posible to tube)