There is a threshold curve for statistically significant audibly detectable frequency response differences, this plot has the advantage of being on the web:
http://www.provide.net/~djcarlst/abx_crit.htm
from that set of curves it is pssible to argue that some speaker impedance curves would give response variations above the audible thershold difference for some speakers and 2uH added series inductance (some speaker impedance curves are measured in Stereophile equipment reports also on the web)
http://www.provide.net/~djcarlst/abx_crit.htm
from that set of curves it is pssible to argue that some speaker impedance curves would give response variations above the audible thershold difference for some speakers and 2uH added series inductance (some speaker impedance curves are measured in Stereophile equipment reports also on the web)
Let s not get into a long discusion on this subject - we already covered it on another thread. To summarize:-
1. Unless you know a few tricks like JC does, its unlikley that you will be able to design an amp that will operate under all speaker load conditions without an output inductor
2. The output inductor (1-3uH typicallywith parallel a 2-4 Ohm resistor) is designed to isolate capacitive loads from the output stage where such loads cause excessive phase shift resulting on oscillation. It canm also help to reduce the ingress of RF picked up by the speaker cable.
3. If you are using your circuit with a well defined load (e.g. active speaker) you can getaway with it.
4. Don't forget that a typical speaker load along with the associated cable has many times the inductance of a typical output inductor of 1-3uH i.e. it is completly swamped by the wire and speaker load inductances
5. Conclusion: Mortals should use an output coil of 1-3uH in parallel with a resistor of between 2-4 Ohms.
1. Unless you know a few tricks like JC does, its unlikley that you will be able to design an amp that will operate under all speaker load conditions without an output inductor
2. The output inductor (1-3uH typicallywith parallel a 2-4 Ohm resistor) is designed to isolate capacitive loads from the output stage where such loads cause excessive phase shift resulting on oscillation. It canm also help to reduce the ingress of RF picked up by the speaker cable.
3. If you are using your circuit with a well defined load (e.g. active speaker) you can getaway with it.
4. Don't forget that a typical speaker load along with the associated cable has many times the inductance of a typical output inductor of 1-3uH i.e. it is completly swamped by the wire and speaker load inductances
5. Conclusion: Mortals should use an output coil of 1-3uH in parallel with a resistor of between 2-4 Ohms.
Once I made a marginally stable amp. At the output binding post, if I put 2.2uF cap across +/-, the whole amp oscillates. But when I put a 30cm cheap speaker cable (ordinary twin cable), it doesn't oscillate when the 2.2uF cap is placed after this 30cm cable.
Seems the inductance of 30cm cheap cable is enough to prevent that amp from oscillating, but I don't know what is the inductance value of 30cm speaker cable.
Bonsai said:
I agree absolutelly with you. For the 5 reasons that you mentioned above exactly, especially the nr. 5. We haven't the luxury of testing our projects with a big variety of expensive hi-end loudspeakers. The only that we are in place to do are simulations with dummy loads. I use the method of the 8Ù resistive dummy load paralleled with a 2ìF cap, because the same practice used from Bascom H. King in "Audio" magazine. In practice now, in the graph of my previous post nr. 6, it appears the damping of oscillation from an inductor of 0,8ìÇ. As for the coil i preffer always thick wire of 2mm diameter - it has realy enormous dimensions - for lowering as much as possible the resistive part ( a common practice that i know from Crest Audio and Peavey amplifiers - the 0,8ìH value used in Peavey amplifiers) and usually for lowering the length of coil i modify it in multilayer and i place it in vertical direction on the seperate output protection PCB just before the output relay ( far away from the main amplifier PCB, exactly as in Crest amplifiers ). Finally from measurements with or without the coil in output by injecting squares in input from my Hameg function generator of 10MHz i haven't seen - from my own eyes by looking at the screen of my Hameg DSO of 50MHz - any difference in the shape of output waveform at least in the audible spectrum. Maybe my Hameg instruments are not so accurate in comparisson with HP instruments or Tektronix - the latter are of course of China origin today.
Fotios
Bonsai said:
Pretty much. But there is one factor you didn't mention. Speaker cable is a mis-matched transmission line. This means that at certain very high frequencies (MHz) it can look quite inductive and at others it can look quite capacitive. IOW it is not quite right to treat a speaker cable as if it is a pure inductance between amp and speaker.
I agree that it is prudent to use an inductor unless you know what you are doing. Designing an amp, especially a high feedback amp, to be adequately stable in to real loads is extremely difficult.
PS: I've measured straight wire to have an inductance of about 5nH/cm.
PPS: A loop of wire, a circuit, has additional inductance related to the size of the enclosed area of that circuit.
While I have no opinion whether coils are audible or not, I try to keep a bit of open mind in such issues and I don't mind exercises of hypothetical reasoning. So, if we assume the ouput coil is audible, can we imagine an explanation that might explain it? So let's speculate a bit instead of just dismissing it by adding inductances etc. If for no other reason, so at least just for the fun of it. 
Tradebarm has already brought up the things that came to my mind too:
1. The coil and the speaker wire have different physical locations, so the explanation might lie there rather than in the inductance per se. As Traderbam said, what about the magnetic field of the coil affecting the small signal circuitry? Since the current dependes on the speaker impedance, the output current may have quite a different spectrum than the output voltage, so if the current induces a voltage in the feedback loop, for instance, perhaps that might distort at least the frequecy response of the amp? What happens if we add the coil just at the output connector instead of on the PCB, for instance? Would John still hear it?
Then, what about the other way around, ie. the coil picking up a field? In a class AB amp, the supply currents are heavily distored. We already know this can affect the circuit by inductive coupling with a bad PCB layout. Well, a coil is much better at picking up magnetic fields than a PCB track is. On the other hand, if this is the case, I suppose it should show up clearly already when testing with a single sine wave.
2. Assume it has something to do with high frequencies, like RFI, slight damped oscillation tendencies of the amp under certain load conditions or maybe back EMF from the speakers can cause fast rising edges with a HF spectrum? Whatever the reason for the HF, maybe we have to think transmission lines here, so we have to consider impedance matching of inner wiring, connectors speaker cable etc. Then perhaps the coil could make thing worse sometimes?
Yes, I know this is mostly very speculative, but if a phenomenon possibly exists, it usually requires quite a bit of speculation and creativity to figure out if it actually exists and why.
Tradebarm has already brought up the things that came to my mind too:
1. The coil and the speaker wire have different physical locations, so the explanation might lie there rather than in the inductance per se. As Traderbam said, what about the magnetic field of the coil affecting the small signal circuitry? Since the current dependes on the speaker impedance, the output current may have quite a different spectrum than the output voltage, so if the current induces a voltage in the feedback loop, for instance, perhaps that might distort at least the frequecy response of the amp? What happens if we add the coil just at the output connector instead of on the PCB, for instance? Would John still hear it?
Then, what about the other way around, ie. the coil picking up a field? In a class AB amp, the supply currents are heavily distored. We already know this can affect the circuit by inductive coupling with a bad PCB layout. Well, a coil is much better at picking up magnetic fields than a PCB track is. On the other hand, if this is the case, I suppose it should show up clearly already when testing with a single sine wave.
2. Assume it has something to do with high frequencies, like RFI, slight damped oscillation tendencies of the amp under certain load conditions or maybe back EMF from the speakers can cause fast rising edges with a HF spectrum? Whatever the reason for the HF, maybe we have to think transmission lines here, so we have to consider impedance matching of inner wiring, connectors speaker cable etc. Then perhaps the coil could make thing worse sometimes?
Yes, I know this is mostly very speculative, but if a phenomenon possibly exists, it usually requires quite a bit of speculation and creativity to figure out if it actually exists and why.
Re: Re: Re: Function of Output Inductor
Well put, Glen.
PS: When people say they hear a difference between a coil and no coil, I wonder what kind of difference. Maybe an amp without coil is perceived as "crystal clear" because it is just ringing.
G.Kleinschmidt said:
You've got that the wrong way round: It's the compromised designs that generally achieve a high degree of capacitive load stability without a load isolating inductor.
Well put, Glen.
PS: When people say they hear a difference between a coil and no coil, I wonder what kind of difference. Maybe an amp without coil is perceived as "crystal clear" because it is just ringing.
Any output stage will tend to act as a current source with 90 degree lag at RF. This is because the gains of active devices start to roll-off and parasitic capacitances become no longer negligible. A proper RF load is required in order to achieve predictable RF gain/phase and load-independent stability.
A 2.2uF film capacitor (crossover type) is no longer a capacitor above 500Khz or so, which is the typical resonant frequency for a part having 30mm to 40mm lead spacing. It exhibits a very low impedance dip at that frequency and then becomes inductive. Most amplifiers won't oscillate with that load. That's why physically smaller capacitors can become very handy for testing (but never parallel them, as this causes new resonant modes, apart from self resonance, to appear). Internal (and external) amplifier wiring will also contribute to lower the effective resonant frequency of the load (to 100Khz or below), and a LR output network (2uH, 4ohm) will add considerable damping to that resonance, particularly above 250Khz (where it's mostly required).
A piece of speaker wire (like two 2.5mm^2 wires running parallel and close together) exhibits less than 1uH per meter.
This is not vodoo, it's just modelling everything as the L, R and C that it contains. It's a very healthy practice that helps to produce very reliable circuits without doing a million of empirical tests (what you usually do when you don't actually understand what's going on).
A 2.2uF film capacitor (crossover type) is no longer a capacitor above 500Khz or so, which is the typical resonant frequency for a part having 30mm to 40mm lead spacing. It exhibits a very low impedance dip at that frequency and then becomes inductive. Most amplifiers won't oscillate with that load. That's why physically smaller capacitors can become very handy for testing (but never parallel them, as this causes new resonant modes, apart from self resonance, to appear). Internal (and external) amplifier wiring will also contribute to lower the effective resonant frequency of the load (to 100Khz or below), and a LR output network (2uH, 4ohm) will add considerable damping to that resonance, particularly above 250Khz (where it's mostly required).
A piece of speaker wire (like two 2.5mm^2 wires running parallel and close together) exhibits less than 1uH per meter.
This is not vodoo, it's just modelling everything as the L, R and C that it contains. It's a very healthy practice that helps to produce very reliable circuits without doing a million of empirical tests (what you usually do when you don't actually understand what's going on).
I think nobody disputes the basic circuit theory. However, alla models have assumptions that must hold for them to be valid. The questions we should ask are thus rather if we are perhaps failing to see that som assumption does not hold. That is most probably the case if there is something to explain at all here.
G.Kleinschmidt said:
An audio power output stage with a single pole response at "RF" is an object of fantasy.
I meant 90 degree or worse
(for those claiming unconditional stability without specific RF loads and RF isolation).Anyway, with some effort and the right Sanken LAPT and Toshiba parts (or a pair of IRF540 for that matter) you can shift the second and further poles quite high. An unity-gain bandwidth product above 10Mhz is not an object of fantasy (maybe an object of luxury?
) My first output stage made with Toshiba 2SC4793/2SA1837 and Sanken 2SC3624/2SA1295 was happily self-oscillating at 30Mhz... Things have changed a lot since 2N3055 
PMA has it right, high speed output devices and improved compensation as allowed removing coils from being absolutely necessary. The only real compromise can be a little bit of slew rate limiting from the fastest that the amp can produce. Still, 100V/us or more is enough for me, at least, I have done up to 700V/us with coil and a similar circuit however. That was in the 'old days' about 25 years ago.