LM386 headphone amp, i need to lower gain

[amptramp]

If you put a resistor in series with the output, do you want the Zobel network to be on the amp side or the speaker side?  It is there to counteract the rise in impedance of the speaker / earphone at high frequencies and its effect on the feedback loop and it is quite the load on the amp at high frequencies.  It may not be necessary with a series resistor since the load is isolated somewhat from the output and the feedback path.  At the very least, it should be recalculated for its intended use.  You may find lower power consumption with the change to the 47 nF + 10 ohm network.

[Dormammu]

amptramp
In this particular case - I think this whole impedance mumba-jumbo thing doesn't matter. It's just performance control, not audiophilliac  puzzles.   

[Rob Strand]

If you put a resistor in series with the output, do you want the Zobel network to be on the amp side or the speaker side?  It is there to counteract the rise in impedance of the speaker / earphone at high frequencies and its effect on the feedback loop and it is quite the load on the amp at high frequencies.  It may not be necessary with a series resistor since the load is isolated somewhat from the output and the feedback path.  At the very least, it should be recalculated for its intended use.  You may find lower power consumption with the change to the 47 nF + 10 ohm network.

It's a valid point.

Treating the LM386 as a black box (ie. unknown innards) it would be best to keep the Zobel on the output of the LM386.   The bonus here is it will probably help damp parasitic oscillations.   The series resistor just makes the load look nice and resistive but the everything is still happy with the Zobel on the LM386 output.

[PRR]

...The international standard for headphone amplifiers is actually 120 ohm however ....

I believe that is a "testing" standard. Like testing speakers on an open baffle. It is reproducable, that's all.

Although it may be a fair approximation of 150r/200r broadcast console line amps. (Even if they are "zero-Z", you like some series resistance to protect against goofy or shorted loads.)

And IIRC the 120r standard came before the Sony Walkman and universal use of low-Z phones/buds.

Mr Sony was not stupid. He may have deliberately designed to tolerate various sources and to be a design that other makers would tend to copy (not fragment the market).

[Rob Strand]

I believe that is a "testing" standard. Like testing speakers on an open baffle. It is reproducable, that's all.

Form what I remember it originated as a recommendation for headphone amplifiers   I'm fairly certain about that because the standard also specified maximum voltage levels.   The main intent was to stop excessive output levels on low impedance headphones.

If you consider headphones ranging from 16 ohm to 600ohm then the geometric mean impedance is sqrt(16 * 600) = 98 ohm.   If you consider the ranges as 32 ohm to 600ohm the geometric mean is sqrt(32*600) = 138 ohm.  All in the 120 ohm ball-park.  The significance of the geometric mean impedance is an amplifier with that source impedance will have the least variation in power when *different* impedances are plugged into the output.
[EDIT: 
The standard actually states:
This output is designed to produce, as far as possible, a constant sound pressure level in the
headphones for a given setting of the volume control, irrespective of the impedance of the
headphones over the range 8 ohm to 2000 ohm.

Note that geometric impedance = sqrt(8*2000) = 126 ohm
]

Because these standards have some degree of harmonization it's likely the resistance value spills over to the test standards.

Speaker designers tend to allow for some degree of cable impedance when they consider the damping of the bass response.

I'm not aware of ever reading what allowance headphone designers make for amplifier impedance in the *design*.   (Presenting test results is a different matter.)

Something I've noticed about the IEC *Audio* standards in general is they are not mandatory.   In many instances you can use a different test configuration to their recommendations.  In this case the standards only say all you need to do is document this configuration in the test results!  Crazy stuff for a standard.

[Rob Strand]

Here you go,

The standard is about audio interfaces
IEC 61938 from 1996
https://store.standards.org.au/product/iec-61938-2018

It is referenced on this page,
https://nwavguy.blogspot.com/2011/02/headphone-amp-impedance.html

Here's an extract from the posts at the bottom,

FWIW, John Woodgate is involved in some of these Audio standards.  I used to post with him on the old
usenet groups sci.electronics.design.  He's a smart dude with a lot of experience.
http://www.woodjohn.uk/

"
http://gilmore2.chem.northwestern.edu/faqs.htm

Is an amplifier's damping factor important to headphone performance?
With loudspeakers, the lower the amplifier's output impedance, the higher the damping factor into the rated load. Damping factor is given as the ratio of loudspeaker impedance to the amplifier's output impedance. As the theory goes, the higher the damping factor, the better the amplifier's ability to control a loudspeaker's low frequency response (when the motional reactance of the system is at maximum), because the low output impedance of the amplifier allows any back-emf generated by the loudspeaker to be absorbed by the amplifier. That theory has been discharged by members of the audio community as unsubstantiated.
However, even if the theory were correct for loudspeakers, its applicability to headphones is suspect. John Woodgate, a contributor to The Loudspeaker and Headphone Handbook (1988), had the following to say about the effect of damping factor on headphone performance:

Headphone transducers are resistance-controlled, not mass-controlled like loudspeaker drivers above the main resonance. In any case 'damping factor' is largely nonsense - most of the resistance in the circuit is the voice-coil resistance and reducing the amplifier source impedance to infinitesimal proportions has an exactly corresponding effect on damping - infinitesimal.

However, the source impedance affects the *frequency response* of a loudspeaker because the motional impedance varies with frequency, and thus so does the voltage drop across the source impedance. This means that the source impedance (including the cable) should be less than about one-twentieth (not one two-hundredth or less!) of the rated impedance of the loudspeaker, to give a *worst-possible change* in frequency response from true voltage-drive of 0.5 dB.

The motional impedance of headphone transducers varies very little (or should vary very little - someone can always do it wrong!) with frequency, so the source impedance can be high with no ill effect.

The IEC 61938 international standard specifies that headphones should be driven by a 120 ohm source - regardless of the impedance of the headphones themselves. If the headphones were designed to this standard, then an amplifier's high output impedance should have little effect on the sound of the headphones. In general, headphones with a flat impedance curve over the audio range will not be affected by high output impedance. For example, in May 1995, Stereo Review published a review of the Grado SR125 headphones. The impedance curve of the SR125s, which have a nominal impedance of 32 ohms, varied from 31 to 36 ohms over the entire 20Hz to 20kHz spectrum. Not all headphones may be as well behaved as the Grados, but neither do they usually have the roller-coaster impedance runs of a loudspeaker. Tube amplifiers (with their high output impedances), it should be noted, have very low damping factors."

[PRR]

I sit in awe of your superior library.

[Rob Strand]

It's something I've worked on in the past.  I remembered John Woodgate wrote some stuff on it and I was just lucky to get a hit on his name.

[Vivek]

Rob, your research is phenomenal !

Respects !

[amptramp]

This source:

https://audioxpress.com/article/voice-coil-lab-notes-improved-zobel-network

implies that the resistor in a Zobel network for a speaker is 1.25 times the voice coil resistance and this would include any series resistor you add and the value for C would be the part of the inductance of the load that is not shunted by any other resistance divided by the square of the Zobel resistor.  This would give much higher values for the Zobel resistor and lower values for the Zobel capacitor.  This would in turn result in much lower power consumption from the amplifier.  This assumes the Zobel network is connected to the amplifier output rather than the speaker.  You would have to determine the inductance of the headphones to select the proper capacitor value.

If you all remember the General Electric Transistor Manual, they designed a number of amps and always used 100 nF and 10 ohms as the Zobel network regardless of any other amp characteristic.  This may be OK where you have an amplifier that has to be used with any 8-ohm speaker they decide to drive with it, but for the case of a headphone amplifier, the 47 nF / 10 ohm values hog way too much output power.  If you are operating from a battery, Zobel network dissipation is the major power requirement and with the correct network, you could extend battery life dramatically and reduce dissipation in the LM386.

[PRR]

> the 47 nF / 10 ohm values hog way too much output power.

Only above 159kHz !?? A simple low-pass in front should take RF way down. A 50kHz low-pass is traditional and almost essential(*) in guitar amplification.

And the article seems to focus on flattening loudspeaker impedance. The 0.1u/10r on the NatSemi notes is about swamping load inductance so it does not reflect-in as capacitance to the NFB amplifier's core and spoil the stability.

(*) you "don't need it" until you play a truck stop or around police/taxi or AM broadcast transmitters; then you NEED it.

[anotherjim]

I don't think I've seen a Zobel on this kind of phones output but this is just a snippet of the first thing I grabbed from my device. So I'm not saying it's never done to fit a Zobel. Anyway. +/-15v rails on this and the unfamiliar (to me) BA15218 is yet another "audio" dual opamp.

If you want to read the rather sparse datasheet.
https://datasheet.octopart.com/BA15218-Rohm-datasheet-14139886.pdf

[Rob Strand]

I don't think I've seen a Zobel on this kind of phones output but this is just a snippet of the first thing I grabbed from my device. So I'm not saying it's never done to fit a Zobel.

There's differences between the opamp case and the power amp case.   It's all to do with feedback loop stability.

No Load:
- Some power amp don't like operating into an open circuit at high frequencies.   A Zobel network ensures
  the load looks low and resistive at high frequencies.
- The inductive load of a speaker creates a similar no-load condition at high frequencies.
   For amplifier stability the values of the Zobel network depend on the amp more than the speaker.
- opamps are different:  opamps are always stable into no load

In general you should use a Zobel network on a power amplifier.   If the datasheet shows a Zobel you should use it.
To go against that you would need a lot of supporting evidence - which is almost impossible to create since you know
nothing about the innards of a power amplifier chip.

Capacitive loads:
That's a whole different story.   

- Capacitive loads stuff up all amplifiers with feedback: power amps and opamps.
- For power amps a common way to "disconnect" the capacitive load is to add an LCR network.
Then to ensure the amp is still stable with the "disconnected" load you need
to place a Zobel network across power amp output.

This pic should also show cable capacitance.
https://www.globalspec.com/reference/22138/203279/output-networks


- For opamps the simplest way to handle capacitive loads is to add a series resistor to the
  output.   In general, the value of the *minimum* resistor depends on output impedance
  of the opamp, the capacitive load, and what frequency the opamp rolls-off.  If you add
  a large valued resistor (say 1k in most cases) it will pretty much work for all opamps,
  except some low power units.

  If the resistor value isn't large enough the capacitive load can introduce phase shift
  into the feedback loop and cause the opamp to go unstable.

  That's why you see a 470 ohm to 1k series resistor on the output of an effects pedal.
  Here, the capacitive load is the cable capacitance.

- Beyond that, there's a whole lot of techniques,
  https://www.analog.com/en/analog-dialogue/articles/ask-the-applications-engineer-25.html