- Higher power
- Lower distortion
- Can also be used in Single Ended
Schade Feedback or Partial Feedback – Better than ultralinear?
I often repeat to my clients that the sound of tubes by itself does not exist, it is the entire amplifier as a whole that plays, and if you know how to design it properly, more or less you can make any tube sound like another. We are not talking about zero-feedback amplifiers, where it becomes a contest of who distorts in the most “subjectively beautiful” way, but about circuits designed to be HiFi and reproduce the signal well without excessive coloration. Basically, we do not want the “Mona Lisa” to look like a “Jackson Pollock” painting.
In my experience one of the most difficult tubes to tame has definitely been the EL84. The EL84 is a 12 watt dissipation pentode with an internal resistance of 38k. With such a high internal resistance it was also difficult to produce a good sounding push-pull output transformer on low frequencies, because the required primary inductance was really high, and to achieve it you would end up compromising the high-frequency extension. This pentode was very common in vintage radios and in tube guitar amplifiers, such as the famous VOX AC30. It replaced the 6V6 in circuits using Noval tubes. However, in a radio the low-end extension was not important, as the small output transformers used at the time often did not go below 500 Hz by design, and the same applies to guitar amplifiers, which did not need to produce very low frequencies and where distortion was even desirable. In HiFi many classic designs using the EL84 end up having a rather typical, vintage sound, which I mentally associate with the EL84. Especially in push-pull configurations, where Ri doubles, it becomes necessary to apply a lot of NFB if you want to achieve decent damping factors as in this Dynaco 410 clone: roaring, old sounding and in my opinion quite unpleasant. In ultralinear mode the situation does not improve much and, moreover, the EL84 does not even allow you the satisfaction of saying “I will connect it as a triode!”… Connected as a triode, the curves have a very steep slope. Usually in these cases you use a transformer with a high impedance and work in voltage but… NO! The EL84 datasheet recommends not exceeding (static voltage) 300 volts on plate and G2, so either you end up like some who built ridiculous SE triode EL84 amps delivering 0.4 watts or 2 watt push-pulls, then hiding their inability by saying “it is the sound that matters”, or you take another more sensible path.
Even though I had already used it in the past, only today I took the opportunity to talk in depth about a type of circuit called Schadeode or Partial Feedback.
(Type 1)
This circuit topology is defined by the use of two resistors: one feeds the plate signal to the control grid and the other is placed in series with the input signal. Since these two signals are in opposite phase, a differential signal reaches the grid, the result of their mutual subtraction mediated by these resistors, which forms a local feedback loop from plate to grid. This feedback loop provides advantages: it reduces distortion, tube internal resistance and noise, while extending the circuit bandwidth. Since the feedback loop is very short, this system encounters few of the instability problems usually associated with global feedback. The amount of feedback applied depends on the values of R1 and R2. This system implies that the signal generator, meaning the stage that drives the power tube, must have a very low output impedance, typically a cathode follower or an SRPP, whose Zout is negligible compared to the value of R1. If instead you want to apply Schade Feedback to a stage that delivers signal from anode, the circuit would look like this:
(Type 2)
In this circuit variant, the amount of feedback applied depends on the ratio that R1 has relative to R2/R3 and the Ri of the driver triode. Compared to the resistor-based circuit, where it is virtually possible to apply as much NFB as desired (since resistors are perfectly linear), here you are limited by the linearity of the driver stage which, at some point, will be pushed by the signal coming from R1 into saturation or cutoff, causing strong distortion originating precisely from that element.
Let us now look in detail at what this circuit does and why it should be better than the ultralinear connection. As written at the beginning of the page, it has higher power compared to ultralinear. The pentode has the highest efficiency in terms of power transferred to the load (40/50%, in some cases even 60%), but it also shows the highest Ri to the transformer and generally a high distortion rate. A pentode connected as a triode, or a true triode, on the other hand, has the lowest efficiency in terms of power transfer (20/25%), but also a much lower Ri to the load and a low distortion rate. In ultralinear mode, the efficiency is intermediate both in terms of power and Ri to the load and distortion. Then, as if that were not enough, the ultralinear connection is widely used in push-pull circuits, but in single ended use it is limiting and impractical: it produces only a little more power than triode connection but with much more distortion. Despite this, many “distortion lovers” keep using it precisely because they are not fond of high fidelity but of distortion itself, as you can see here. In fact, ultralinear connection in single ended makes no sense, I have already discussed it here.
With Schadeode connection, instead, you can have the power efficiency of a pentode with an Ri to the load and distortion even lower than the same tube connected as a triode, and it can be used effectively also in Single Ended. Wanting to have your cake and eat it too, in this case you can. Incredible? Look at the graphs below, let us first compare the EL84 as a pentode with the same tube connected as a triode…
| Pentode | Triode |
 |
 |
Then we look at the EL84 curves in ultralinear mode, at 20% and 40%, as indicated by the tube datasheet. In this case, as you already know, the curves are valid only if the operating point is set at 300 volts, and it is not possible to move it above or below this voltage without completely changing the graph, as already explained in another article…
 |
 |
As you can see, the EL84 in UL, both at 20 and 40%, does not show impressive curves: you lose power with almost nothing in return, so it is better to leave it connected as a pentode. But let us now see with Schadeode connection, which the uTracer software allows me to simulate, how the curves change. NOTE: all graphs were acquired from a real EL84 inserted in the curve tracer and are not computer simulations.
As you can see, the current/voltage limits are the same as the tube working as a pentode. So, for example, a load line on the left side of the graph can go down to 50 volts with 100 mA of current, transferring the same power that the tube would deliver as a pentode, but the curves have almost the same slope as the EL84 connected as a triode. In a real triode connection, the same load line would stop with tube saturation at 150 volts / 75 mA, therefore transferring much less power. Basically an EL84 as a triode that can deliver the same power as an EL84 as a pentode. But let us see what happens when increasing the amount of local feedback:
The limits are still those of the EL84 connected as a pentode, but the slope of the curves and the gain are greatly reduced, meaning its internal resistance has decreased and now it takes up to -36 volts to bring it into cutoff. In effect we have created a new triode that did not exist before. Nothing to do with ultralinear, and it does not matter much if the curves do not extend upwards like true triodes, because the normal load line will never reach that region. The reduction of the Ri of the virtual tube then becomes useful for increasing the damping factor without using global NFB, but with this kind of feedback, which is even less harmful than conventional feedback. Note for those who want to experiment: the static operating point (voltage/current/grid bias) must be set on the pentode curves, because the virtual triode operation occurs in dynamic but not in static conditions.
Below are some external links on the same topic:
On this site you can find the Alimede project with EL84 driven in Schade Feedback with a 5842 as driver, the very old free single ended project with the 6JZ8 (the damping factor of this circuit remains low because the local feedback rate is low as well). And soon a premium project of a push-pull with EL84 should arrive, where I will try to achieve an acceptable damping factor without using global feedback, with a power of 15 watts. Moreover, to complete the discussion, it should be said that Schadeode can be used with any tube, even a triode, achieving a reduction of its internal resistance.