It seems that filament bias became very popular recently. There is a long thread on the DIYaudio discussion forum about preamps with the 26 triode and filament bias seems to be used a lot there. So I thought it would be a good idea to write a few articles about this biasing method. This one starts with the basic introduction of the concept.
I came up with the idea more than 12 years ago when I spent a lot of effort to minimize the capacitors in the signal path. I do not claim that this is an original idea. I have seen similar schemes in old textbooks especially with battery DHTs.
The most conventional and in my opinion also one of the best biasing methods is good old cathode bias. In this scheme the plate current is returned to ground through a dropping resistor which is placed between cathode and ground. In case of directly heated triodes, cathode and filament are the same electrode. A point needs to be chosen as the cathode connection. In case of AC heated filaments this can be the center tap of the filament transformer secondary or the wiper of a hum pot between the filament terminals. Filament bias is only relevant for DC heated triodes. In this case many people still put a hum pot across the filament and use the wiper as connection point. I'm of the opinion that in case you heat with DC, it should be as clean as possible so that no hum bucking is necessary and no ugly pot distrubs the signal path. I commonly choose the negative filament as the cathode point. From there you connect your cathode resistor to ground.
If the full amplification shall be utilized or if the triode is transformer coupled, it is advisable to bypass that cathode resistor with a capacitor so that the cathode has a low AC impedance to ground. Otherwise the internal resistance of the tube would be increased by roughly the cathode resistor value multiplied by the amplification factor of the tube. This would move the internal resistance into a region which is not usable with transformer coupling in most cases. In case of directly heated filaments the bypass cap will also shunt any noise voltage which might be present between the B+ and filament supplies. Such noise voltage can build up if transformers without electrostatic shields are used. Obviously this cap is in the signal path and has a great influence on the sound of a gain stage. So how can we eliminate it?
In the case of cathode bias, the plate current through the cathode resistor generates a voltage drop which elevates the cathode to a positive potential, which in turn translates to the grid being negative with respect to the cathode. The beauty of this scheme is that it lets the tube find it's own operating point. As the tube ages and emission drops, the bias voltage decreases which counteracts the aging effect somewhat. Also in case of some fault the cathode resistor acts as a safety mechanism. In order to eliminate the cthode bypass cap we also have to eliminate the cathode resistor or reduce it's resistance to a very low value compared to the plate resistance of the tube. One way to do that is fixed bias which feeds the grid with a bias voltage from a separate supply. Obviously that additional supply is in the signal path, but we wanted to minimze that. Filament bias utilizes the filament current in addition to the plate current to generate the bias voltage through a cathode resistor which now can be much smaller.
The picture above shows conventional bias vs. cathode bias. The difference is minor, the negative end of the filament supply moves to ground rather than the negative filament terminal of the tube. Now the filament current flows through the cathode resistor. With filament currents typically in the range of one to several amperes, this means we can develop 10s of volts for the bias with resistances in the 10s of Ohms. This is about a factor of 100 below the typical plate resistances and won't hurt much if left unbypassed.
Now the filament and B+ supply are referenced to the same ground, no danger that any electrostatic noise voltage builds up between the two. Of course the filament supply needs to be very clean and well filtered since everything present on top of the filament voltage will be amplified. In the schematics above the filament voltages are supplied through chokes to the triodes. This is to isolate the filaments from the supply and capacitance therein. Instead of the choke the filament bias voltage can also be supplied through a constant current source which seems quite popular lately. I still favour the passive way with good iron though.
This scheme works nicely and is proven in various configurations. It simplifies the signal path and ties filament and b+ supply nicely together. Of course the self biasing and safety aspect of cathode bias are lost. This basically acts like a fixed bias stage now. There is another disadvantage: The cathode resistor needs to dissipate a lot of power. This calculates as filament current times bias voltage. So in case of a 26 DHT we will dissipate about 10-15W in each filament bias resistor, depending on the operating point which is chosen. Worse in a 801A or 10Y: Here we get 20-50W depending on the operating point. This is some hefty power dissipation which requires serious resistors with proper heatsinking. I still use filament bias, especially with triodes like the 26 or the UX201A which only needs 0.25A filament current. With these the heat dissipation is managable. With other triodes I mostly use a scheme called ultrapath which is another way to get the cathode bypass out of the signal path. In some cases I even combine them both.
I will go a bit more into deatils in future articles and also show some actual implementation in a new preamp which I plan to build. There will also be an article about a further development beyond filament bias which I named DirectPath. This eliminates the last remaining cap in such a stage, the last B+ filter cap. Stay tuned!