Circuit of the basic amplifier. Heaters to pins 4 and 5. See further into the article for component variations.

I have used this amplifier in a number of projects, the first time back in 1984 in a regenerative receiver. Television valves are used in most of my projects, and this one is no exception. The 6BM8 is a 1950's TV valve initially intended for vertical oscillator (triode) and output (pentode) use. However, it was also given audio ratings and became a very popular valve in Australian TV sets for both applications. In fact, if the vertical scan was getting a bit weak, one could often swap the audio and vertical 6BM8's to restore operation, with the weaker valve being less critical for the audio section.
The 6BM8/ECL82 has a number of derivatives, all the same but with different heater requirements. Europeans will be familiar with the 16A8/PCL82 which has a 16V 300mA heater, common in European TV sets.  There is also a 50BM8/UCL82 with a 50V 100mA heater that was used in portable gramophones of British origin. In some Japanese TV sets was the 11BM8/LCL82 with an 11V 450mA heater. Any of these valves can be used in this circuit provided a suitable heater supply can be arranged.
The circuit shown here is for use from a B+ of 230V at around 40mA and with a 5K to 8R speaker transformer. However, there is a degree of flexibility, and the circuit will happily work on much lower voltages with less power output.

Other supply voltages and output transformers:
To use with other output transformers and/or supply voltages, some components need to be changed in the pentode circuit. These are the cathode resistor (Rk), and the screen grid resistor (Rg2). See the data below:

Triode data:
 

Supply Voltage	170	200
Plate load	220K	220K
Rk	2.7K	2.2K
Vrms output (680K load)	25V	26V
Gain	51	52
Distortion	2.3%	1.6%

Notes on this valve data;
Vp is the voltage between plate and cathode, not necessarily the same as the supply voltage.
g2 voltage is measured between screen grid and cathode. Note that the screen grid resistors where shown for 200V and 272V operation are not bypassed. Lack of bypassing is recommended for stability (i.e. the grid resistors also function as "stoppers" to prevent high frequency oscillation). For 250V operation, it was not stated if the screen was to be bypassed or not. To be effective, the bypass capacitor, if used, needs to be greater than about 8uF
Rk is the cathode resistor value. Where a voltage is given instead, this is the control grid (g1)voltage, which is the same as the voltage developed across Rk when cathode bias is used.
Rk can be calculated by dividing plate to cathode voltage by the sum of plate and screen current.
Vg1 is the negative bias voltage applied to the control grid. Where this is not given, use the value of Rk specified.
ZL is the load impedance; i.e. the impedance of the speaker transformer primary. This is NOT the same as the DC resistance, which is usually much lower.
Ip is the plate current.
Ig2 is screen grid current.
Vrms I/P is the rms voltage required to drive the valve to full output.
Po: Power output.
For the triode data, Supply Voltage is just that, NOT plate voltage.
Vrms O/P is the maximum undistorted output at the plate. The 680K load is the following grid resistor. Note that I recommend 470K as the following grid leak instead of 680K. This is because the 6BM8 can sometimes be prone to grid emission and it is good practice to use a grid resistor of low as possible value. However, the lower the following grid resistor, the lower the gain.
200V operation: Two sets of operating conditions are shown for the pentode at 200V. Both are from different data books, although both Philips.

The output transformer:
If you have a proper valve output transformer such as old stock or a new reproduction, then build the amp as is. Otherwise you will have to use a power transformer or 100V PA line transformer with the correct impedance ratio. Note that imedances are not carved in stone. For example, if the design specifies 5000 ohms, a 7000 ohm transformer will work but just with less power. Even gross mismatches will work if you have no choice, and full power output is not required.

Using a power transformer:
The impedance ratio is the square of the turns ratio. So, assuming we have a 240V to 12V transformer, the turns ratio is 20. Now square this and we get 400.
Multiply by the speaker impedance (we'll use 8R) and we get 3200R as the impedance the valve will see. You can work out endless possibilities here with the range of off the shelf transformers available. For this amplifier, the secondary current rating of the transformer can be anything from 150mA upwards. Generally, power transformers work fairly well as they usually have sufficient primary inductance. The more inductance the better; something above about 10H if possible. One 2851 type transformer I measured had 15H. A real single ended valve output transformer with 7K primary I measured had 10H.
It is recommended by some people to rearrange the laminations into E-I format to prevent core saturation when there is DC flowing through the primary, as in the case of a single ended class A amplifier. With only 30mA or less I haven't found this to be significant. On the contrary, I have found results to be worse because where the inductance was already insufficient, it is now reduced further. This results in a very trebly sound with hardly any bass.

There are a number of things to try if you find your PA line transformer isn't performing correctly. Symptoms include excessive treble or instability.
1) Connect in ultra linear or triode mode. This works well for the MM1900.
2) Increase value of capacitor across the 4.7K in the feedback network.
3) Rearrange laminations (only if sufficient inductance).
4) Try a transformer with a larger core; the M1120 generally works better than a MM1900.

Triode and Ultra linear modes.
Operating an output pentode in triode or ultra linear is nothing new, and is mentioned here in its own right; not just as a cure for poorly performing speaker transformers. The trade-off is reduced sensitivity and power output. However, sound quality is improved. In triode mode, the 6BM8 amplifier has a sensitivity of about 300mV, and power output dropped by about two thirds. Ultra linear operation is a compromise between pentode and triode modes; i.e., the best of both worlds.
To modify the circuit for triode operation, remove the g2 resistor and connect screen to the plate. This gives 100% screen grid feedback.
For ultra linear operation, take the B+ end of the screen resistor and connect it to a tapping on the primary. Obviously, ultra linear cannot be used with a single untapped winding. The position of the tapping will determine the amount of screen feedback With the M1100 transformers, the 2.5K tap is ideal, assuming the 5K connection is to the plate, and the common to the B+. The closer the tap is to the plate, the more like a triode it will behave. Conversely, the closer to the B+ the tapping is, the more like a pentode it will be.

Unstable & Faulty 6BM8's.
The 6BM8 is a moderately high gain valve, the triode having a mu of 70, which is a little more than a 12AT7 triode. The pentode has a mutual conductance of 7500umhos, compared to a 6V6's 4100umhos. This is why there is a 47K resistor in series with the pentode grid. It must be mounted right at the pin of the socket with the shortest possible lead length on the grid side. With the internal grid to cathode capacitance, it acts as a "stopper" like the screen resistor. The centre shield of the valve socket must also be earthed.
Over the years, I have found two 6BM8's which appeared to be faulty. These oscillated at certain settings of the volume control. They can actually be used if necessary, by such measures as increasing the feedback capacitor (try .0022uF) and/or removing the bypass capacitor across the pentode cathode resistor. Severe cases may warrant triode or ultra linear operation.
Grid emission is also something I have encountered with some 6BM8's. If the current of the output stage seems too high even though the bias resistor is correct, measure the voltage between cathode and grid. The grid should be negative to the cathode by a certain amount; if this amount is less than the equivalent positive value across the bias resistor, it indicates grid emission (use a DMM or other high impedance meter). If this is the case, try reducing the grid resistor.  Gain will decrease slightly, however, and the grid coupling capacitor should be increased to .1uF to retain low frequency response.
If it is necessary to go below 220K, the valve has a real problem.
Note that the measured bias voltage across the cathode resistor may be less than the data specifies. For one thing, the valve might not have 100% emission. Also, the actual supply voltage to the screen grid and plate may be less than specified if the voltage drop across the speaker transformer primary and cathode resistors has not been taken into account. 1% accuracy is not needed here! The key thing to check is that the pentode has the correct bias resistor for the conditions selected.
I do not recommend fixed bias be used as the valve could go into thermal runaway and be ruined.