The plate to cathode cap

[Lynxtrap]

I'm struggling with some fuzzy high end in the overdrive channel of an amp build.
Looking at different ways to remedy this, I started thinking about the cap between the plate and cathode in some of the ODS circuits (250-330pF in the ones I've seen).

How do these actually work (being connected to the actual cathode and not straight to ground) and how do they affect the gainstage?
I figure there is some kind of local feedback going on?

Speaking of which: has anyone tried the LFB between grid and plate using a cap value that makes it affect only high frequencies?

[ijedouglas]

You are correct, it feeds high frequencies (determined by cap size) back in to the circuit and cancels them out

[Lynxtrap]

You are correct, it feeds high frequencies (determined by cap size) back in to the circuit and cancels them out

Thanks!
Where do you figure it would work best, on the stage that generates most of the overdrive or on the stage before it?

[ijedouglas]

I would try it on both, easy enough to tack in

[martin manning]

...I started thinking about the cap between the plate and cathode in some of the ODS circuits (250-330pF in the ones I've seen).
How do these actually work (being connected to the actual cathode and not straight to ground) and how do they affect the gainstage?
I figure there is some kind of local feedback going on?

The cathode resistor is bypassed so the cap is shunting HF to AC ground.

[Lynxtrap]

The cathode resistor is bypassed so the cap is shunting HF to AC ground.

That makes perfect sense. Would that mean that it doesn't affect the cathode at all, so that there is really no NFB going on?
Is there a straightforward way to calculate the cutoff frequency?
On the other hand, I guess 250-330pf is a good starting point unless Rk and Rp values don't affect the cutoff point in any major way. I figure the effect of Ck in series with the cap can be ignored.

[martin manning]

No feedback going on, and the Ck is five orders of magnitude bigger, so yes it can be ignored. You can try values around 300p. Without looking it up, I suspect it will be be like a LPF with resistance Zo and the cap value paralleled with the Miller capacitance, so about 10kHz.

[ijedouglas]

No feedback going on, and the Ck is five orders of magnitude bigger, so yes it can be ignored. You can try values around 300p. Without looking it up, I suspect it will be be like a LPF with resistance Zo and the cap value paralleled with the Miller capacitance, so about 10kHz.

I was always under the impression that this introduces some negative feedback into the circuit? I guess it's small compared to the hf rolloff to ground though.

[martin manning]

I don't think so. Blencowe has some comments about it, but doesn't mention FB. The roll-off is a combination of the input (Miller) capacitance effect and and the anode bypass cap, but he doesn't give details.

[jelle]

If the cathode cap is 'perfect', as well as the leads/solder joints to ground making a perfect zero ohms path, forming a perfect Thevenin loop, then yes, its shunted to ground.

In practice, the cap is not perfect, nor is the signal path to ground. Its close, but not perfect. Its my understanding that its both a negative feedback loop, as well as a shunt to ground.

The negative feedback works as follows: the 270pF shunt caps place the high frequency content that is not shunted into the cathode caps/ground into the cathodes of the tubes, where they, because the cathode is out of phase with the plate, effectively form negative feedback in an extremely tight closed loop. Pretty nifty, imo.

Other ways for shunting highs in a clipping triode would be 1) placing the shunt cap over the plate resistor, or 2) from plate to grid, or 3) plate to ground, but the response is different.

The problem with the Dumble approach is that the quality and synergy of the components becomes critical for it to work well. Or it becomes buzzy.

[martin manning]

If the cathode cap is 'perfect', as well as the leads/solder joints to ground making a perfect zero ohms path...

The imperfection is the ESR of the cathode cap, of course, so yes that creates a small negative feedback effect.

[martin manning]

How much is that NFB effect? Sweeping ESR from 1 mohm to 32 ohms on a 100k Ra, 1k5 Rk, 4u7 Ck stage with 1p and 270p A to K, it's hard to see any real difference in the HF roll-off with ESR. The change in gain is 0.17 dB at mid band. The 270p snubber cap clearly takes out 3 dB at 10kHz, though.

[jelle]

I have always wondered, and honestly I do not know.

But if the different methods of shunting highs were the same, it would not matter how one places the cap (plate to ground, plate to grid etc.), but it does. As does the quality of the components. And I do not know how to simulate this with programs like Spice either. Spice lies about what sounds good, it only knows about theoretical models, but we are dealing with analog imperfection here.

What I do know that that using the wrong cathode caps, wrong snubber caps, or worse, wrong both cathode *AND* snubber caps, makes the amp buzzy. This is real time empirical data, and the lack of answers does not negate the value of this empirical data.

Maybe the software used to get that graph is like ChatGPT, an approximation at best, as correct as the input it gets and the (limited) knowledge the programmers had how to assemble the information provided. No wonder the use of ChatGPT is wildly contended now. Could you build a great sounding amp using software? Or does it need validation, and fine tuning?

I am sorry I do not have answers, only questions. Why do different types, brands, and even batches of resistors, or even individual resistors and caps sound different? Why does the guitar sound change when reversing the guitar cable? The longer I am in this field, and its been 25 years, the more questions I have. Its wild. But I learned how to fix them in the amp at hand. The way I understand it is that this is what puts the art into the amps, and how Dumble could take a Fender amp, and without even touching the schematic, make it a phenomenal amp.

[martin manning]

Here's another study where the same CC gain stage has a 270p cap connected A to K, A to Ground, and A to B+, along with a reference case where there is no anode bypass cap and no ESR in Ck. Placing a cap from A to G1 is not something that anyone would do without a large resistance in series to limit the NFB. For clarity there are only two levels of ESR shown, 1m and 8 ohms (a typical value). A 68k grid stopper has been added, and although that rolls off more HF, the 270p bypass cap is still worth about 3 dB at 10kHz. The results show that all three cases produce nearly identical results, and as before, the ESR in Ck has a negligible effect.

A circuit simulator does not lie about anything. It's an objective way of thinking and communicating where the assumptions and level of understanding are transparent. The simulator computes the electrical behavior of the circuit using standard techniques, and like any mathematical model, the results depend on its completeness. It is accepted that the model may be incomplete and imperfect, but the qualitative understanding and quantitative assessment of the differences between designs it provides can be used to refine a practical solution. It has no ability to predict the perception of the human psycho-acoustic system, which is unique to each individual. The user has to judge that and decide if it is within the ability of a listener to detect. The Bode plot is a standard way of doing that, but one can also look at phase shift and clipping behavior at relevant frequencies, which is another aspect of the response that a listener may be able to hear. In the second plot it can be seen that the various positions for the 270p anode bypass cap produce nearly identical results for a 1-5V input level sweep at 5kHz. An interesting feature is that the square corners seen in the reference case are rounded-off, indicating a reduction in the higher order harmonics for those clipped waveforms.

[Lynxtrap]

3dB at 10kHz? As a tone shaping option, that's not worth the price of the cap