Output inductor in power amps - pro and con

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi Dave,

Yes, the inductor thing and where to put the Zobel can be a point of controversey and maybe even some art.

I always recommend an inductor, even if it is fairly small. Amplifiers with unknown capacitive loads (or, effectively unterminated transmission line cables) can sometimes oscillate in brief bursts that may not be seen, but which may affect the sound. Without some decoupling, as provided by the inductor, it is difficult to get an amplifier to be unconditionally stable for nearly arbitrary loads, especially given the wide variation in operating points of the output stages as they go through an infinite variety of combinations of current and voltage. Even 1uH shunted by about 2 ohms can make a big difference.

The Zobel is very important to the stability of the usual EF output stage, and should be very close to the output stage transistors, with a well-controlled local return to signal ground. It wants to provide good resistive damping to very high frequencies, especially with fast output transistors. I even believe in what I call distributed Zobel networks in the output stage, when a plurality of small Zobels effectively in parallel is physically distributed across the output stage. Then the individual zobels can be smaller and less inductive. Then there is no need for big-wattage, non-inductive power resistors for the Zobel.

I also prefer to have a Zobel on the speaker side of the inductor, and located at or near the amplifier output terminals. We thus end up with a sort of pi output network. This can help the EMI situation and sometimes to some extent the possible speaker cable mis-termination concern. The key to recognizing the possible value of this is to realize that there is often some wiring inductance in the signal path from the output stage to the speaker terminals. Doug may regard this as over-kill, citing the damping effect already there as a result of the resistor across the inductor. The Zobel need not necessarily be big.

Also bear in mind that star grounding can make the return ground path at the speaker terminals be very different at high frequencies than the signal ground present at the output stage. A Zobel at each end mitigates any effects of this difference at high frequencies, damping the line to BOTH grounds.

You can tell that I like Zobels. They tend to help things behave at those high frequencies where there is some magic and layout sensitivity.

Finally, if you really want to be sure, put a Zobel at the speaker - at the far end of the speaker cable. I am not the first to advocate this, and I think some speaker cable manufacturers build this in. This helps make sure that the far end of the speaker cable is at least somewhat terminated at the far end (loudspeakers can look inductive at HF, even when there is a resistive L-pad in the tweeter circuit). While it is certainly true that speaker cables of reasonable length do not act like transmission lines at audio frequencies (and certainly need not be properly terminated at audio frequencies), the transmission line effects that can begin to emerge in the 1 MHz region can in some cases act to de-stabilize an amplifier.

I suspect that there are many amplifiers out there that may sound different merely because they are not unconditionally stable, and they may be issuing subtle parasitic bursts of oscillation in different ways. Similarly, different speaker cables connected to a marginally-stable amplifier may cause the sound to be different by altering the way in which the amplifier is mis-behaving.

Cheers,
Bob

Thanks Bob for your clarifications.
I have your book and it helped me to understand many things a lot better. What I missed was some more LTspice simulations and explanation how to do it. I am learning each day something new. I am retired telecommunication engineer and electronic is just my hobby.
BR Damir
 
... I even believe in what I call distributed Zobel networks in the output stage, when a plurality of small Zobels effectively in parallel is physically distributed across the output stage. Then the individual zobels can be smaller and less inductive...

I also prefer to have a Zobel on the speaker side of the inductor, and located at or near the amplifier output terminals. We thus end up with a sort of pi output network. This can help the EMI situation...

Your explanation in the book made sense to me and I planned to do this.
But it is nice to have confirmation, so thanks to Antonio for the data to quantify the benefits of an outboard capacitor to reduce RFI.
I am somewhat cautious about RFI because I plan to use so-called "phantom zero" compensation which will increase RF return ratio.
A question - I have multiple BJTs in the OPS and it seems reasonable to attach the distributed Zobels directly to the emitters rather than downstream after the emitter resistors. Does that make sense to you?

Best wishes
David
 
It's odd that the output inductor has always been about 1uH, from the days of 2n3055 onwards. You would expect the value to have dropped to 100nH with modern transistors and FETs. Too high a value will be above its self resonance at problem frequencies.

When you talk about RFI... These days the sources are mostly above 1GHz

The inductor is outside the loop, so simplistically it is determined by the speaker impedance and acceptable roll-off, that have not altered, rather than the transistor Ft.
But your point about the self resonance issue makes me think. Do you have any data?

Best wishes
David.


OK. a quick check of typical 1uf inductor has a SR frequency of about 160Mhz. Not a problem I think.
 
Last edited:
It's odd that the output inductor has always been about 1uH, from the days of 2n3055 onwards. You would expect the value to have dropped to 100nH with modern transistors and FETs. Too high a value will be above its self resonance at problem frequencies.

When you talk about RF ingress, back in the 70s that meant amateur HF and CB. These days the sources are mostly above 1GHz

Hi davidsrsb,

I seem to recall higher values in the early 70's on several amps, perhaps as high as 4uH, but values declining over the years to 2uH or below. I used 0.5uH on my MOSFET power amplifier in 1983.

For me, less than 0.5uH of isolation gets scary and seems insufficiently effective.

Cheers,
Bob
 
David... look the post #7 Damir

Yes. I noticed that Antonio's post #15 seems to be inconsistent with your #7.
If I understand Antonio's chart correctly then the outboard capacitor circuit results in a much lower input into the feedback network. This is makes sense to me.
I don't understand your result where a low value shunt added across a source makes no difference. I suspect there is some issue, perhaps with how the cable source is handled, but I have not yet checked and did not want to comment prematurely.

Best wishes
David

Is current source a reasonable model?
I have no RF model experience. How should cable capacitance and inductance be handled?
 
Last edited:
Yes. I noticed that Antonio's post #15 seems to be inconsistent with your #7.
If I understand Antonio's chart correctly then the outboard capacitor circuit results in a much lower input into the feedback network. This is makes sense to me.
I don't understand your result where a low value shunt added across a source makes no difference. I suspect there is some issue, perhaps with how the cable source is handled, but I have not yet checked and did not want to comment prematurely.

Best wishes
David

Antonio simulation shows output voltage before an inductor(.ac sim) only. Certainly it shows an influence of the cap connected after an inductor influence, but the same influence shows on the voltage after an inductor. I was trying to show the difference between output voltage(or in my case, impedance) after and before a inductor, actually a speaker line garbage suppression by showing V(out)/I(I1)/V(vout)/I(I1) in dB. Could be that my result are not good enough, I did not include any cabel resistance/capacitance, but I think it was not importand.
Sorry but it not easy for me to express my self in English technically correct.
BR Damir
 
It is reported more than once that application of output inductor may lead to higher distortion and less (!) stability with some amps. I remember that some John Linsley Hood amps disliked it and Alexander current feedback amp deliberately avoided using it because he noted that harmonics are more pronounced. Also, Rod Elliott's P3A with CFP output stage dislikes output inductor.
 
It is reported more than once that application of output inductor may lead to higher distortion and less (!) stability with some amps. I remember that some John Linsley Hood amps disliked it and Alexander current feedback amp deliberately avoided using it because he noted that harmonics are more pronounced. Also, Rod Elliott's P3A with CFP output stage dislikes output inductor.

Its not that JLH amp disliked output inductor, just JLH used output resistor instead and thus decremented a dumping factor.
 
Dadod,

JLH designed many amps. The amp that showed some instability with output inductor was one of his MOSFET amps.

Do you mean this amp? I never used it with output inductor so I can't tell.
 

Attachments

  • JLH80W.jpg
    JLH80W.jpg
    67.7 KB · Views: 314
As a big believer in the devil is in the details I really enjoy threads like this, thanks.
I think that one needs to determine for themselves what additional circuitry if any is needed and why. In my case I became interested in trying to better understand why emitter followers can oscillate and found "Why Circuits Oscilliate Spuriously, by Dennis Feucht" to be a good read (in addition found that you can model negative impedances such as -R in LTSpice). In an example amplifier I am looking at, such oscillations can take place when there is a load capacitance between 50p and 4000p (when parasitic inductances are added especially to the output transistor base). At which point combinations of rc's and L's and base stoppers can be used to tame such oscillations (and remove resulting resonances). Dave, not 100% certain but having an R immediately on the emitter followed by the rc would be my first guess at the better solution.
I have seen mentioned numerous times that simulating a nearly full scale sinewave can bring on such oscillations, and that an even better approach is to add a small square wave pulse train to the sinewave. What I have found is that that changing the rise/fall times of the pulse can also make a big difference, especially if these times are made longer, closer to the equivalent bandwidth.

Bonsai, for the case described above I dont see any difference in the main loops phase margin, although there can a difference in gain margin especially if you run the ac sweep with various dc inputs (bias the amp output closer to a rail).

Thanks
-Antonio
 
Last edited:
Hi davidsrsb,

I seem to recall higher values in the early 70's on several amps, perhaps as high as 4uH, but values declining over the years to 2uH or below. I used 0.5uH on my MOSFET power amplifier in 1983.

For me, less than 0.5uH of isolation gets scary and seems insufficiently effective.

Cheers,
Bob
1uH is 0.12 Ohms at 20 kHz so I suppose it is already low enough not to disturb the frequency response across the audio band. There is no point in trying to push for even lower values.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.