[morzh]

Your basic flow is this:

http://en.wikipedia.org/wiki/File:VacRect2E.png

This above is a simple full-wave tube rectifier.
It is not filtered.

http://www.radiomuseum.org/tubes/tube_5y3.html
This is now filtered.

Now, this above will still have pulsations which are

Vpulse=I/(2*f*C)

Where Vpulse is the ripple voltage, f - Frequency (60Hz) and C - capacitance in Farads (say 12uF is 0.000012F).
Which at say 50mA of current will yield 34V of ripple.
One would say why not increase the cap 100-fold and get rid of ripple.
Valid suggestion but with tube rectifiers the capacitance is limited by what the tube can handle.

The best way of getting rid of the ripple is adding the second stage of the filtering. Most typical is LC. The inductance L is either a separate choke or the field coil used as such. And then there is the second cap (say 8uF cap).
The L will resist the change of current and so your current will become triangular and not suqare-wave type. The capacitor will smooth the voltage. Good thing is there does not have to be a small limit on the second cap size as the current is limited by the choke.

Choke also has a parasitic parameter, the resistance which is quite large - it is from hundreds to thousands of Ohms. This brings another degree of filtering due to the time constant RC. It is not very effective in your case as the time constant has to be larger than your ripple's period, so the output cap has to become say 80uF to have some good effect with 400 Ohm resistor.

Problem with purely resistive filtering is the resistor needs to be large and will dissipate lots of power whereas the choke only dissipate the power due to its parasitic resistance and if it were pure inductance it would not dissipate anything at all. Things being equal the capacitor increase and not resistor is desirable, but then you should see if you overload your tube when your resistor is small and the second cap is large.

Another problem with increasing the resistor is it will drop the voltage, and if you were dropping say 30V before you threw out your field coil, and you decided to increase the resistor twice, you will be dropping 60V. Not only this will increase the dissipated power 4 times, but it will also drop the Vbb by the same 30V which might adversely affect the operation of the tubes, especially of the audio output stage which typically works at full rectified Vbb voltage.

What would I do.
If I had extra space, I would rather install a separate choke of comparable inductance and parasitic resistance (which are always listed in datasheets) to your field coil.
And then increased the output capacitance to further attenuate the ripple.