DHT Push-Pull

When building a Push-Pull (PP) output stage using DHTs, not implementing a proper DC filament regulator solution like the one developed by Rod Coleman, it’s a wasted opportunity in my opinion. The benefits already proven in the SET topology are available in the PP stage as well.

Whilst working on different PP stages over the past years, I came up with a series of options and configurations for the output stage using the Rod Coleman regulators. There are some advantages and disadvantages on each one of them. The purpose of this article is to present and review each one of them.

As we can see on the diagram above, there are some interesting topologies for consideration:

1. Two regulators with fixed bias (Circuit 1): Each DHT valve has its own fixed bias at the grid (circuit omitted for simplicity) and each filament has its own Rod Coleman regulator.
2. Two regulators with mixed bias (Circuit 2): Each DHT valve has its own fixed bias plus a filament resistor to provide a level of cathode bias. Each filament has its own Rod Coleman regulator.
3. Two regulators with mixed bias abd potentiometer (Circuit 3): Each DHT valve has its own fixed bias plus a shared filament pot to provide a level of cathode bias. Each filament has its own Rod Coleman regulator.
4. One regulator with fixed bias (Circuit 4): Each DHT valve has its own fixed bias at the grid (circuit omitted for simplicity) and both filament share a  Rod Coleman regulator via a series configuration. This circuit is similar to circuit 1 but with one filament regulator instead.
5.  One regulator with mixed bias (Circuit 5): Each DHT valve has its own fixed bias plus a filament resistor to provide a level of cathode bias and both filament share a  Rod Coleman regulator via a series configuration. This circuit is similar to circuit 2 but with one filament regulator instead.
6. One regulator with mixed bias abd potentiometer (Circuit 6): Each DHT valve has its own fixed bias plus a shared filament pot to provide a level of cathode bias. Each filament has its own Rod Coleman regulator. This circuit is similar to circuit 3 but with one filament regulator instead.

So what are the key differences between the topologies presented? There are three things that define each topology:

1. Connection of the filament regulator (series or parallel)
2. Number of regulators
3. Bias (fixed or mixed)

For particular reasons the parallel connection is not shown as described later.

Connection of the filament regulator

Series connection

1. Topologies 1,2 and 3 use two regulators. Each valve has its own regulator. This helps in accommodating the filament resistance differences between valves unmatched. With separate regulators we can adjust the current of each filament independently.
2. Topologies 4,5 and 6 use one regulator. The filaments are connected in series. This reduce the number of regulators needed for the output stage (from 4 down to 2). However, there are some disadvantages to this configuration:
   1. The raw supply now needs to be of higher voltage to accomodate two filaments in series plus the regulators. Despite that the power requirements of the filament circuit remain the same, it may be more difficult to obtain the transformer at the rated voltage needed. In addition to this, capacitors rated at higher voltage (e.g. 35V instead of 16V) are more expensive.

Parallel connection

One topology that I discarded after initial analysis is the parallel connection of the DHT filaments to reduce the number of regulators from 2 to 1 per channel (like in series mode described above). In this mode, one filament end is grounded an the other is shared with the other DHT and connected to the regulator. In this way, both filaments are connected in parallel. The regulator has to handle twice the current as before. This makes the raw supply easier as voltage is not increased, however the regulator transistors have to  dissipate more heat given the current is now doubled. This creates all sorts of implementation challenges (i.e. heat dissipation), in particular for the transmitting valves.

With parallel connection of the filaments, the AC signal in the cathode are in opposite phase. The regulator will present a high impedance (I.e. Gyrator and CCS) to the filament so the cathode signal will cancel nearly each other. The geometry of the filaments will determine how close this gets. Rod makes some interesting notes around this configuration:

“The important thing is that the currents flowing in opposition may well degrade the sound. Possibly, it will degrade it very badly! I would try it out, and listen – but a good outcome is not guaranteed!”

“The music signal current in each of the PP sides is in anti-phase, like the cathode current (in fact, it is the cathode current!). Since any improvement in filament supply quality can usually be heard, I would suspect the parallel connection sounds worse than individually regulated DHTs.

Series might  be better, but I would check by listening for a long time.” (Rod Coleman)

 

Bias (fixed or mixed)

The addition of some cathode feedback provides DC stability though negative feedback. Few valves like 4P1L can be implemented in filament bias with reasonable raw supplies. With exception of some high-mu power transmitting DHTs, the DHTs used in the output stage have low-mu. The addition of several ohms as part of the filament resistor has minimum impact in the output impedance. Remember that the filament resistor is reflected as (μ+1) times in the anode. This is the main disadvantage (or price we have to pay) of the mixed bias implemented in topologies 2, 3, 5 and 6.  Some 2-5V would be ideal

The addition of a potentiometer as shown in topologies 3 and 6 proves quite handy to balance the currents on the valves in particular if they are not matched.