Hz and an impedance of 32 ohms, the coupling capacitor value must equal 250 μF. Two of these capacitors were they high quality film capacitors would fill a large cabinet, leaving no room for the circuitry and batteries! The alternative to using film capacitors is either using electrolytic or a blend of  electrolytic and film. Working on the assumption that the best capacitor is no capacitor, this circuit is definitely worth a second look.
     
A Transformer Coupled Example
    Transformers help tubes work into low impedance loads and they also protect the load from the high voltage on the other side of the transformer. Both these tasks seem ideally suited to a tube headphone amplifier. Even 300 ohm headphones are awfully low for most tubes and even a small DC offset at the output could damage the headphone's drive element. A good transformer could leverage the even Grado's 32 ohms up to a manageable 3200 ohms and offer zero DC offset.
   The problem is finding a good transformer. Even the best transformer in the world isn't all that good. Aside from bandwidth issues, a transformer is plagued with a large size and heavy weight, leakage inductance, primary and secondary DCR, saturation, eddy currents within its core, proximity hum pickup, physical noise generation, and sometimes untamable high frequency resonances. Add to this list that headphone impedance spread of 32 to 300 ohms makes finding a transformer with useful secondary taps difficult. Let us also add to this list that array of useful output transformers has shrunk since the tube glory days; I have a few old catalogs that list a few ideal transformers (out of hundreds and hundreds that were not suitable) that are no longer in production. The capacitor does not look so bad now does it?
    Assuming that a good transformer can be found, the design of the headphone amplifier is really no different from the design of power amplifier, a smaller in scale, but not fundamentally different.

    The single-ended amplifier's general of 2rp for a primary impedance is a good starting point. Given that the spread of useful rp's begins at 1k and ends at 4k, we need a transformer whose primary impedance falls between 2k to 8k and centers on 4k. A primary of 4k coupling into a secondary of 300 ohms gives us a winding ratio of 3.65:1, as dictated by the formula:
    Ratio = √(primary / secondary).
We need to know this ratio because it is unlikely that we will find a transformer specified for a 300 ohm secondary load. In other words, a transformer that has a primary impedance of 8k and a secondary of 600 ohms has the required winding ratio. The ratio also gives the current magnification and the voltage division from primary to secondary. For example, a 36.5 volt swing and a 1 mA current swing on the primary equals a 10 volt and 3.65 mA swing into the 300 ohm load impedance. In addition, the ratio indirectly gives us the output impedance, as the ratio squared is the impedance ratio of the transformer. Thus, for example, an rp of 2k becomes an output impedance of 150 ohms.
    With secondary of 300 ohms how do we drive 32 ohm headphones? If the secondary winding does not have multiple taps, we don't try, in this case, our best bet is to find a transformer with an 11:1 winding ratio and preload the secondary with a 100 ohm resistor. This resistor will give the transformer's limited inductance something to bite on when the load is 300 ohms. But if the secondary has multiple taps, the going gets easier. A center-tap will yield a secondary impedance of 75 ohms; a .707 tap, 150 ohms; third-tap, 33 ohms. In other words, a center-tap increases the winding ratio by 2.
   The next question is where to place the transformer. Should it go at the plate or the cathode? Should it be used directly or in a para-feed arrangement? The plate loaded amplifier gives us gain; cathode loading, a much lower output impedance and less distortion, but at the cost of no gain and a high input voltage swing; and a para-feed arrangement, a potentially lower noise output and no primary current.