The Zobel network provides a stable, resistive, load to the amp and its output devices to prevent them oscillating at very high frequency. Without it an amp could suddenly destroy itself when the wrong combination of speaker load and cable is connected. It also makes it easier to ensure stability when there is no load connected. Unless you can prove by analysis that your amp design doesn't need it, then include it. 10R and 100nF are reliable values for most situations. The 10R resistor needs to be rated for some power as it will see non-negligible currents if there is significant top octave signal. Also called a Baucherot cell, see here: https://en.wikipedia.org/wiki/Boucherot_cell
Another way to look at it is it ensures the phase and gain margins you calculated are valid (they are calculated assuming resistive load without phase difference in output current and voltage). You also need a series inductor to isolate the amp output from highly capacitive loads.
The Zobel is there to ensure the OPS remains stable at HF with little or no load. It’s not there to ensure loop gain stability. Localised stability and loop stability are not the same thing.
With modern high fT output devices, the Zobel and base stoppers are more important than ever.
If there is Miller cap on the VAS, the open loop gain is attenuated by Miller effect. You don’t need the RC at output. The 10R put extra load at high frequency and it reduces the your gain/phase margin, not improves.
Nope, that is not how you compensate emitter followers.
Here is how to compensate emitter followers.
https://nathaliebeimler.com/tech/audio_amplifier_output_stage_comparison.html
Jxdking, if you had read Self and Cordell and designed a few amps you would not be confusing loop stability with localised OPS instability.
The link you put up does not represent a real OPS because none of the parasitic elements present in a real OPS are included in the model. And I have yet to see a real world speaker load that’s is purely resistive.
The link you put up does not represent a real OPS because none of the parasitic elements present in a real OPS are included in the model. And I have yet to see a real world speaker load that’s is purely resistive.
I’m afraid this is not correct. You get OPS HF pole migration (means reduced gain and phase margin) with any load, but especially capacitive loads. The whole idea with pole splitting aka Miller comp is to ensure the loop gain and phase are adequate under worst case conditions. The Zobel is there to address a separate issue.
I you decide to omit an output coupling inductor, you will be throwing away HF loop gain in order to keep the amp stable with capacitive loads, or you will have to run it with low loop gain or no feedback.
Last edited:
Both my OEM's (HK680-Sonance 260) have the resistor at ground. My design (Wolverine) has the cap @ ground.
Both the OEM's and DIY's are 10R/47-100nF combos. Inductors are all 1-2uH.
My Sonance has a 2-layer inductor (how nice !) , but they took NFB from the P-channel output (fixed that).
Oh , a bootstrapped badger/wolverine - hard to screw these designs up. layout can give you close to PPM - regardless !
All you need is C8=47-68pF (600-800K UG) for comp. nothing else needed.
OS
