Explanation of the circuit
Basically, this is a grounded cathode voltage amp followed by an opamp buffer. For full understanding ignore the opamp stage and look at the tubed voltage amp. Just cut (in memory) the wire between the tube and the opamp (+) input. The triode is fixed-biased by the grid-charge current that flows through the 1meg resistor.

To explain this:
The hot cathode always emits some electrons that have enough energy to fly up to the grid. The grid resistor closes the current path between grid and cathode so that the grid discharges across the resistor and following Ohm´s law builds up a negative voltage between grid and cathode = ground. This way of biasing works for many small signal tubes like 6DJ8, ECC88, 6922 but also for 12AT7, 12AX7 etc.. It works for pentodes as well, like DAF96, 1T4. It won’t work for power tubes like 2A3 or so because the bias voltage is in the range of -0.3..-0.8VDC only.

Choosing the value of the grid resistor:
It´s written in the books. It is that value named “max. grid resistor” and often is combined with the type of bias: Automatic or fixed. Choose the value for “fixed”.

Here I have to correct myself:
If the book says “fixed bias” this is for an external negative voltage to be applied to the grid. If the book says “automatic bias” this is for the bias acquired by a cathode resistor. Most books don’t say anything about “grid-leakage bias” !!! If there is a value, use it. If not: Use the value for automatic bias and multiply it by 10. In my circuit the R1 value should be up to 10 megOhm. The circuit works ok for R1 = 1megOhm but one can be sure that there is a small amount of grid current flowing.

Choosing a plate resistor:
Normally you would take the value off the data book, too. But in this case it is not possible because the plate voltage is so low that no curves exist for most of the tubes. Here is a procedure how to get a value without curve-tracing. The plate resistor value is chosen that it will consume 4..6 V from the 12 V DC power. I recommend that this resistor is adjustable if you use any other tube than the 6DJ8 series or want to use other supply voltages. Always adjust the plate resistor to consume 1/2 to 1/3 of the supply voltage. If you try pentodes the plate resistor may consume up to 2/3 of the supply voltage. Start with a high value, up to 200kOhm, and adjust to lower while continuously measuring the voltage between plate and B+ until it fits. Now take the adjustable resistor off the circuit and measure the resistance. You now may choose the standard value nearest to the result of the measurement. An amplified input signal can now be seen at the opamp (+) input if we “repair” the wire we cut above.

The opamp in the prototype is an NJM4456 dual opamp. It is able to feed as much as 70mA into a 150 Ohm load.The NJM is not as “hot” as modern opamps with enormous high gain-bandwidth products. So it is not very likely that you will have problems with noise or oscillation.You may also try other opamps like NE5532 or OPA2134 or so. Make sure that they are “unity gain stable”. That means they will work flawlessly at a gain of 1.That´s the configuration we need. We wire the output to the negative input and feed the signal into the positive input. From the DC point of view this input sits at about 1/2 of the supply voltage, according how exactly you have chosen the plate resistor. In steady-state the output of the opamp follows the input (with a gain of 1 here)and therefore sits at 1/2 of the supply voltage, too.

Side note:
You could use a standard buffer like the BUF634, too. It is more expensive, provides even more current and you need two of those.The result is the same: The output follows the input and sits at 1/2 B+.Ok. Here is another correction:It was reported that the BUF634 would be outside a feedback loop and therefore is nearly uncontrollable. It may work or not…A better way is to put a MOSFET source follower as a buffer behind the opamp.We do not like to feed the headphones with DC, so we need to go with an output cap.Warning: Never go without output cap, otherwise your headphones (and probably your ears) will be ruined immediately.You may use an electrolytic of 220uF for cans above 32 Ohms, like the PortaPro or SportaPro. You should use 470uF for lo-Z headphones.Some people put a small film cap in parallel to the electrolytic to have better high frequency response.

Important: Watch the polarity of the cap. The + side of the cap goes toward the opamp output! Essential are the two resistors after the cap.

The bleeder resistor:
This one, to ground, will allow the cap to charge at power on. This helps to minimize the POP when you connect the headphones.

Side note:
You should connect / disconnect the headphones always if the amp is up and running. You won´t notice POPs and this helps to preserve the sound qualities of the headphones. The bleeder resistor should be about 10 times higher than the value of the headphone impedance plus the output resistor value.

The output resistor:
The output resistor is chosen between 22 and 150 Ohm depending on which final gain the amp should have.The output resistor forms a a voltage divider together with the headphone impedance.As a rule of thumb choose the low value for high-Z headphones (>=250 Ohm) and the high value for lo-Z headphones (32 Ohm). Changing the value of the output resistor has _some_ effect on the sound and the stability of the opamp stage.The NJM likes load impedance of 100 Ohm and higher best.

Feeding the heater:
This is done by an LM317T voltage regulator wired as constant current source. You only need one for two channels because the tube is a double triode, having two triode systems inside. Check the tube data: The 6DJ8 is fired with 300mA constant current or 6.3VDC constant voltage. We choose constant current. This is easy to achieve with the LM317 and an additional resistor. The current flowing is given by the formula:
I = 1.25 V / R, or solved for R:
R = 1.25V / I (desired_heater_current) = 1.25V / 0.3 A = 4.2 Ohm

Construction detailsI am the least talented case builder under the sun so I always prefer readily available boxes.This one is an Altoids “Winter Green” tin. I guess it works with any of those Altoid tins.I don´t know if it works with Fisherman´s friends, too. I will test it if someone sends me a box of Penguin mints (hint, hint).

So here´s the look into the inside:
The tube socket was a standard bakelite chassis mount type. I had to “Dremel”-away some material to get it fit into the box.All resistors are standard 1/4 watt metal film types.The circuit diagram does not show the small Alps pot serving as a volume control. The value is 10kOhm but it may be as high as 100kOhm as well.You can leave it out if you have a volume control with your source.I use paralleled resistors to achieve the resistor value for the current source LM317T. You may use one resistor as well but be sure that it can consume at least 1 watt. The configuration of my amp is not bullet-proof. Because I use 1/4 watt resistors in parallel. They get really hot !The inside of the tin is covered with cardboard. This is really necessary because the LM317T has the output at heatsink and would be shorted to ground if touching bare metal.The layout itself is a bit messy, after all. I placed the components according to space optimization. I would be rather happy if somebody could provide a PCB layout.Building this amp was fun, but sometimes a drag because of the small enclosure.

Power supply
As you see 12VDC are necessary to power this amp. It consumes a little more than 300mA. You may use a batt pack of 10..12 NiCd or NiMh rechargables giving you about 5 hrs. of continuous listening if you choose 2000mAh cells. Or power it off your car battery if you go for a camping trip. Use a wall-wart at home, preferably a regulated and well filtered one. Try solar cells if you live in a sunny place.

How does it sound ?
Oh, well. After you had such a hard job to build this amp it is very unlikely that it sounds bad. The sound is mainly determined by the tube. It works at a very low plate voltage so that it probably produces a lot of distortion because the curves are far from being linear. There is a way to measure distortion using a PC-based tool. But I did not try. My PC soundcard is a lousy onboard chip with ugly noise from the computer. ps. So it really is a task someone else should perform. I think the distortion is of mainly even order, as characteristic for triodes. I experienced the sound being very warm, easy to listen to. Not that sterile as transistor amps and not as aggressive as some opamp based amps. I think the NJM and the 6DJ8 are good companions giving a nice and unfatiguing listening experience. This amp works fine with a large bandwidth of headphones. Maximum usable voltage is 4V(peak) and up to 70mA current which leads to a minimum load impedance of 53 Ohm. This is the absolute maximum and the amp will distort heavily. But I admit that I did not listen at that level. My oscilloscope and a dummy load did that for me.

Side note:
A 53 Ohm load impedance consists of the 22 Ohm output resistor in series with an about 30 Ohm dummy load. According to Ohms law this means the dummy only gets about 2 V(peak) of audio. Imagine this going into your Sennheiser PX200 and you will know what LOUD means.