Direct coupled SET proposal

I’ve had a few 6CB5As kicking around for a while waiting for a project. The 6CB5A has been documented by Thomas Mayer and Ale Moglia, among others, as a great option for triode strapping. Thing is, I like trying new things when I build and repeating a cap coupled formula for a two stage single-ended amp just wasn’t making it to my short list of projects (which I don’t quite have time for anyways).

A recent discussion reminded me of an idea to use the shunt cascode topology to direct couple to a single-ended output. It required some extra power supply rails, including a fairly large negative rail. These requirements aren’t anything too unusual; you see them with Morgan Jone’s Crystal Palace amplifier or any kind of fixed bias scheme (in a way).

Anyways, the more recent discussion reminded me of a thread discussing a novel way of direct coupling two stages by stacking the power supplies. This is kind of similar to the Free Lunch AKA Monkey on a Stick arrangement. Applying this idea to my original shunt cascode brainstorm lead me to this:

We have a shunt cascode input stage. The output resistor (R2) idles at about 75V across it. This feeds a MOSFET source follower, which will have just a couple volts less on its output. So we have a very low impedance output at around 75V above ground and we want to direct couple that to the next stage. This is where I think it gets exciting.

We raise the cathode of the output stage so that it is positive relative to the grid (at 75V) by floating the output tube power supply (V2). The voltage we float it at is roughly equal to the target bias voltage plus half the target output swing. In other words, we raise it by twice the bias voltage for A1 or twice plus a bit for A2.

The output tube anode is connected normally and the cathode returns to the point where the output supply (V2) floats on the bias supply (V1). Our input stage is powered by another supply floating on the bias supply (V3). Our input and grid driving circuits are all referenced to ground and direct coupled. We can set the output bias by adjusting the current through R2.

Here’s a more fleshed out version:

It looks like a lot in the schematic, but I’ve already got shunt cascode and grid driver circuits on small PCBs. The power supplies don’t need to be anything exotic in this case as the input has decent PSRR already. The higher current output could use simple CLC filters as well.

Will I build it? I hope eventually. By summer I hope to have the workshop basically finished. I’m already enjoying having all the tools and parts in one (heated) space!

Stacked SMPS PSU for tubes

I used a 48V switch mode power supply in the El Estudiante headphone amp and am pleasantly surprised with the quiet background and relative simplicity. When it comes to higher voltages though, you will not find many AC/DC switch mode power supplies at vendors like Mouser or Digikey. While a beefy low voltage DC supply could feed a DC/DC booster (see Millet’s 10W booster project and various eBay listings), I’d like to find something that is more easily repeated by others with parts from major PSU manufacturers.

I came across the idea of stacking low voltage SMPS supplies in a couple places and the idea intrigued me as a scalable and affordable approach to creating B+ (see here and here). The Meanwell EPS-15 48 are regulated and isolated AC/DC supplies that sell for about $8. Stacking six of these would supply 300mA at about 300V.

This schematic shows the general outline of what I’d like to try with a single-ended amplifier. The V1 supplies produce the anode to cathode voltage for the output tube. These are referenced about 100V above ground by the V2 supplies. The input tube’s B+ is a combination of the V2 and V3 supplies. All in all, this would call for seven 48V supplies, plus another low voltage supply for heaters. The final cost would be somewhat less than a traditional transformer and CLC filter, but there are more interesting reasons to try this.

All the supply gymnastics make it very easy to direct couple the two stages. In this example, Q1 is a gyrator load and Q2 sets the reference voltage. This would also allow us to drive the output tube into A2 operation. My choice for output tube here would be a 6V6: a really sweet sounding triode that otherwise doesn’t produce much in A1 operation. Power would still be low (around 2W), but that would be plenty for headphones or enough for high-efficiency speaker systems.

Direct coupled stages

If you spend enough time haunting DIY tube amp websites and books, you will inevitably come across the theme of direct-coupled tube circuits. N-type and p-type transistor sandwiches make direct coupled circuits almost trivial. Tubes, which are “n-type” only, are not quite so simple to marry anode to grid. And yet the siren sings, drawing in the adventurous tube spirits.

“Why should we want to direct couple in the first place?” you ask, your socratic gland tingling. Many solid state amplifiers take advantage of the direct coupling to increase the levels of negative feedback. With tubes, we’re often more interested in maximizing the inherent linearity of triodes in open loop Class A amplification (but for a good counter example, see Jones’s Crystal Palace in Valve Amplifiers 4th ed).

Ostensibly, eliminating a coupling capacitor or transformer leaves less in the signal path between input and output, making whatever you are building more transparent (if you consider caps to be a significant source of coloration). Eliminating a coupling capacitor also removes a potential source for blocking distortion (if you are prone to driving amps to clipping, though a cap-bypassed cathode resistor can still cause you problems). In my opinion, the most compelling reason to direct couple is that it makes A2 (positive grid bias) operation a possibility.

Following are some (mostly untested) scratch-pad ideas and notes for “simple” direct-coupled SET amplifiers.

Fig 1: Simply using an abnormally large cathode resistor under the output tube raises its cathode above the anode voltage of the driving stage. This dissipates a lot of extra power in the output section and doesn’t really contribute anything to A2 operation. Still, a fun party trick.

resistor load dc

Fig 2: Using a resistor divider to lower the dc voltage seen by the output tube’s grid. This reduces the gain of the first stage and probably still requires you raise the cathode of the output stage (see Jones for good reading on this).

level shift dc

Fig 3: The Free Lunch style of choke loading the driving stage is as nifty as it is temperamental (in my experience). You are still dissipating power in the cathode of the output tube. See also Loftin-White variations discussed at TubeCAD.

choke loaded dc

Fig 4: Currently simmering on my back burner, a MOSFET gyrator sets a reliable voltage on the grid of the output tube and its low output impedance enables A2 operation. Rather than raising the cathode by dissipating power in a bias resistor, the cathode is raised by a separate power supply (must be rock solid). Additional stacked supplies provide B+ for the output tube and driving stage.

mosfet dc

It should be pointed out that direct coupling will almost always require some extra calculating, measuring, and adjusting of whatever you build (you get a glimpse of this with the El Estudiante cathode resistor trial and error). You’re also likely to pigeon hole a direct coupled circuit to very specific tubes, not to mention bias points (which must be maintained). But despite these warnings, once you’ve heard the legend of the circuit without caps, it may already be too late.