This thread has some discussion on this: Need a Zobel Network (or RC shunt) in tube amplifier?
The output inductor is also a Zobel link for matching with the speaker transmission line (Wire).
There seems to be at least one commercial amp with provisions to bypass the inductor. It has two sets of output speaker terminals, one being inductor free.
It is interesting the effort some manufacturers go to in order to avoid all these useful and sonically transparent components
I've experience with two commercially successful solid state amps that lack the RC across the output and one that also lacked the series inductor to speaker.
Dynaco ST120 1966 original with 60636 Output transistors (200 khz ft) got away with only 11 turns wire wound 2" diameter around the output capacitor (3300 uf) series the output capacitor. When production changed to TIP3055 outputs (about 1970?) with 2 mhz ft, the service bulletin had a 1000 ohm series .01 uf cap installed across the output banana jacks - thus a zobel was added.
Allen S100 organ amp was extremely commercially successful built in 1980, and only had a ferrite bead on the output, after mj4502mj802 single pair output transistors. the feedback is kind of minimal. Sound source is limited to about 7 khz max, 32 hz min. Allen ran the output through 4' zip cord to a power up silence relay, then back in same cord, then out 150' of multiwire cable to their own (exclusively) speakers. After 30 years service and a change of location, the silence relay started needing contact cleaning every 8 weeks to maintain continuity, so I bypassed it. Speakers were also changed to two way KLH23 on 15' of 16 ga zip cord. (The allen driver surrounds had rotted). Immediately the amp started picking up sports talk (AM) radio. I added 11 turns wire around a 10 ohm 3 watt parallel wirewound resistor to the output, right inside the speaker screw terminal, and the interference went away. Sounds great again (after ~20 capacitors changed).
Dynaco ST120 1966 original with 60636 Output transistors (200 khz ft) got away with only 11 turns wire wound 2" diameter around the output capacitor (3300 uf) series the output capacitor. When production changed to TIP3055 outputs (about 1970?) with 2 mhz ft, the service bulletin had a 1000 ohm series .01 uf cap installed across the output banana jacks - thus a zobel was added.
Allen S100 organ amp was extremely commercially successful built in 1980, and only had a ferrite bead on the output, after mj4502mj802 single pair output transistors. the feedback is kind of minimal. Sound source is limited to about 7 khz max, 32 hz min. Allen ran the output through 4' zip cord to a power up silence relay, then back in same cord, then out 150' of multiwire cable to their own (exclusively) speakers. After 30 years service and a change of location, the silence relay started needing contact cleaning every 8 weeks to maintain continuity, so I bypassed it. Speakers were also changed to two way KLH23 on 15' of 16 ga zip cord. (The allen driver surrounds had rotted). Immediately the amp started picking up sports talk (AM) radio. I added 11 turns wire around a 10 ohm 3 watt parallel wirewound resistor to the output, right inside the speaker screw terminal, and the interference went away. Sounds great again (after ~20 capacitors changed).
Last edited:
It depends on the circuit. A complimentary emitter follower output probably doesn't need a Zobel network but a complementary feedback pair, also known as a Sziklai pair, probably does. Quasi-comp outputs used a CFP on the negative side so they need a Zobel network. The concept is that a CFP can oscillate all by itself so there has to be " frequency compensation" to stabilize the feedback between the two transistors. The resistor loads the collector in a low-Q way that a cap alone does not. Years ago we tested a Quasi-comp module that buzzed when you pushed the speaker cone in one direction but not the other direction, because the module did not have a Zobel network and pushing the cone generated current that either the CFP or EF had to absorb, the EF being stable and the CFP not.
The R+L series network adds the resistor value to the high frequency loading of the global feedback so that the speaker wire etc does not impair the phase margin and stability. The final cap puts an upper limit to the HF impedance and is not part of the Zobel network. All of this is intended to create a predictable high frequency load because speakers and speak cables are unpredictable at RF frequencies.
Something that needs to part of this discussion is a resistor in series with any feedback network phase lead cap. If a (small) cap is connected directly between the speaker output and the negative input, a path for RF is created that often causes RF rectification. No cable, including a speaker cable should connect directly (or via cap) to a fast semiconductor or a tube grid. Best practice is a resistor about 10% of the circuit audio impedance ( and 10x the RF impedance).
The R+L series network adds the resistor value to the high frequency loading of the global feedback so that the speaker wire etc does not impair the phase margin and stability. The final cap puts an upper limit to the HF impedance and is not part of the Zobel network. All of this is intended to create a predictable high frequency load because speakers and speak cables are unpredictable at RF frequencies.
Something that needs to part of this discussion is a resistor in series with any feedback network phase lead cap. If a (small) cap is connected directly between the speaker output and the negative input, a path for RF is created that often causes RF rectification. No cable, including a speaker cable should connect directly (or via cap) to a fast semiconductor or a tube grid. Best practice is a resistor about 10% of the circuit audio impedance ( and 10x the RF impedance).
Last edited:
since coming across this thread, neither the exiting posts or new posts have expressed any problems when using a zobel.
in this thread vssa squarewave response it appears the zobel is causing a problem.
in this thread vssa squarewave response it appears the zobel is causing a problem.
In many ways a valve amp is a whole 'nother beast from a solid state amp, partly because of the output transformer and its leakage inductance. Nevertheless, a great many of them oscillated with no load. I'm guessing that they would not do that if there was a reasonable Zobel network across the output. I do seem to recall that the McIntosh amplifiers did not oscillate under no-load conditions, but I believe that they did not employ a Zobel. Very interestingly, on the MC275 I think they took the feedback from a separate secondary coil. I believe that their legendary output transformer was pentafilar wound.
Cheers,
Bob
Your comment about the phase lead capacitor in the feedback path is spot-on. A. N. Thiele warned about this in the early '70s. For this reason, I never resort to a lead network in the feedback path.
However, in fairness, a very small capacitance can be used across the feedback resistor if it forms a proper capacitive voltage divider with the capacitance seen at the IPS feedback node. This will largely eliminate the pole at that node that is otherwise there. But this isn't really the same as creating a phase lead. To do this right, the capacitance would have to be on the order of the feedback node stray capacitance divided by the closed loop gain - a very small capacitance indeed. I've never tried it.
Your suggestion of a lead capacitor with a resistor in series with it is a better one if one is going to use a feedback lead capacitor. To get some lead in the loop (and eliminate an RHP zero), I will sometimes put some resistance in series with the Miller compensation capacitor.
Cheers,
Bob
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.