Almost all modern amplifiers today including the cheapest domestic ghetto blasters have distortion figurers that are magnitudes below what can be audibly discernable. Amplifier distortion figurers are now at the threshold of what can be scientifically measured. An amplifier with a THD of 0.001% is of no discernable difference to one of 0.0001%. Amplifier manufactures today have virtually no influence on these figurers. These extremely low distortion figurers are now governed by the engineering physics of the components that are common to all amplifier manufactures. Quoting comparative distortion figurers with claims of an auditory difference has now become a face of marketing deception.
If there is an actual audible difference between 2 amplifiers (separate from power) then either one or both amplifiers has a design mistake, or one or both amplifiers is faulty. Any modern amplifier designed and made correctly that do not have a design fault or technical fault, cannot be audibly different from any other modern amplifier that is designed and made correctly that does not have a design fault or technical fault, regardless of brand name, marketing hype or cost. Almost all technical specifications of modern amplifiers are asymptotic toward infinity, therefore auditory indiscernible and meaningless by comparison (except for power).
S l e w
Frequency response and distortion figurers of amplifiers are normally measured at 1 Watt. The maximum power of an amplifier is normally measured at 1kHz. At 20kHz most amplifiers can not deliver more than 1/4 of the power quoted at 1kHz as in the graph below. The limitation of power restriction of amplifiers above 10kHz is described as Slew. It is not possible for any electronic circuit or individual component, transistor, FET or mechanical switch to be able to turn on and off instantly at the speed of light. Slew is the highest speed an electronic circuit, transistor or FET can change from on or off.
Volts / micro second is Slew rate. With an amplifier of + - 60V supply rail, the leading edge of a 20kHz sine wave would have to change from 0V to 60V in 1/80,000 of a second ( 12.5uS ). Therefore it has to have a slew rate of at least 0.2V / uS to reach 60 Volts in time. But if the amplifier output can only change at 0.1V / uS the maximum voltage it could reach is 30V which is half way and therefore 1/4 power.
Many amplifiers that use power MOS-FETs have very high slew rates and are easily capable of delivering 20kHz at full power. But many amplifiers that use output transistors are restricted in being able to deliver full power at 20kHz. Slew can be intentionally applied in a driver circuit of an amplifier (dominant pole capacitor) to insure high frequency stability and freedom from internally generated parasitic oscillations.
But there is another unique problem of slew that is caused by the internal limitation of one or both output transistors not being able to turn off as quickly as they can turn on. This problem causes a large quiescent current cross conduction through the output transistors that generates extreme heat. An amplifier with slew limited by output transistor cross conduction can be destroyed by high frequency oscillation caused by a reactive capacitive speaker load or Rf (radio frequency) getting to the input from an external source.
Music power bandwidth
Almost all amplifiers including cheap domestic amplifiers have a frequency capability from 0Hz to 100kHz at low power. The gain of an amplifier should be rolled off below 20Hz to stop un-wanted sub-sonic frequencies getting to the speaker. In a previous era (before solid state technology) when only valve amplifiers existed, the BBC in London did extensive research about how high in frequency response an amplifier needs to be to give faithful music reproduction. They found no evidence that an amplifier which was designed to continue amplifying frequencies above 20kHz was musically any different to an amplifier deliberately designed to restrict frequencies above 20kHz. Amplifiers which were capable of amplifying supersonic frequencies above 20Khz were not only un-stable but were also susceptible to Rf (radio frequency) interference. Hence the statement "The larger the window the more sh*t gets in".
In this era of magnetic recording, the limitation of high frequency response was governed by the tape speed and the high frequency bias oscillator. The high frequency bias oscillator acted as a carrier for the audio signal to be magnetised on to the recording tape. The highest audio frequency that could be transferred on to a recording tape was 1/4 of the bias oscillator frequency. Bias oscillators in the majority of professional recording machines in many of the worlds top recording studios were set at 60kHz which restricted the highest audio frequency that could be recorded at 15kHz. Special mastering recording machines had bias frequencies of 80kHz to 100kHz which enabled 20kHz to be recorded on to a tape. But as a general rule, the higher the bias frequency the less audio energy could be recorded and therefore the higher the signal to noise ratio. Only very expensive tape recording machines were not restricted in this way.
The spectral energy of music is flat to approx 1kHz. The highest notes on any musical instrument is approx 2kHz. Above 2kHz are the harmonics which decrease in energy at approx -6dB per octave. At 10kHz the harmonic energy of music is approx -20dB which is 1/100 of the power compared to the middle sector of the voice range. Therefore very little energy exists above 2kHz at these extreme high frequencies.
By combining all the previous information knowing that no audio information is traditionally recorded above 20kHz and the spectral energy of music decreases at -6dB / octave above 1kHz and the slew limitation of many amplifiers only enable a flat power bandwidth to 1/4 full power demonstrates that slew restriction above 10kHz is almost never a factor in limiting an amplifiers ability to reproduce music with normal use. However in a high power 4 way active system the amplifier driving a tweeter from 6kHz up should not have slew restriction.
Power amplifiers are simply large op-amps with the capacity to drive a speaker, simplified in the right pic. The lower pic shows the external management detail.
(1) A capacitor in series with the input signal only allows AC to pass and blocks any DC from the getting to the amplifier's input from an external source. A small capacitor to ground shorts out any Rf (radio frequencies) picked up by the input lead.
(2) Gain = R1/R2+1 There is no agreed set gain for amplifiers. Gain will typically be between 20:1 to 40:1. A capacitor must be series with R1 R2 to stop the amplifier from being able to have DC gain, to help avoid DC offset at the output.
(3) A Zobel network of various complexity is placed at the output of every solid state amplifier to minimise the possibility of parasitic oscillation, and damp the capacitor reactance of the speaker cable.
(4) A speaker relay with a delayed turn on is used in some amplifiers to avoid turn on and turn off thump into the speaker. Some relay circuits also sense DC offset. Most amplifier failures are the result of a short circuit output transistor. This causes the speaker to be connected across one of the V rails hereby destroying the speaker. The relay circuit should also detect a short circuit output transistor and disconnect the speaker.
The combined small phase shifts within the circuit and internal parameters of the transistors is compounded and fed back through the negative feedback, un-stabilising the amplifier causing 50kHz to 1MHz Parasitic oscillations. Parasitic oscillations are almost impossible to control but are minimised with stabilisation capacitors across the Class A driver and driver transistors, reducing the slew of the amplifier and further reducing the open loop gain above 20kHz, but often with only limited success.
The higher the open loop gain the more responsive, but the more potentially un-stable an amplifier becomes. A stable amplifier may have a low open loop gain of 20,000 with a flat response to 20kHz. But its response above 50kHz may be poor in comparison to an audiophile amplifier with an open loop gain of 100,000 that has a flat response to 100kHz. There are many un-substantiated claims in audiophile marketing that an amplifier with a flat response to 100kHz will sound better than another amplifier with a flat response to only 50kHz. Some audiophile amplifiers are so potentially unstable that the excessive capacitance of a magical speaker cable cause parasitic oscillations within the amp overheating and destroying output transistors.
Zobel network is at the output of all solid state amplifiers. Solid state amplifiers are only stable with a perfect Resistive load that has no capacitive reactance. Anything that causes a slight high frequency rotation of phase at the amplifier output is fed back through the negative feedback path to the -inverting input. The comparator can only function correctly if the phase is exactly correct across all frequencies. If not, the negative feedback becomes positive feedback causing the amplifier to oscillate.
Zobel network design is based on transmission line theory. All wires including printed circuit tracks have specific lengths that randomly co-inside with fractional numbers of Rf (radio frequencies) wavelengths. This can cause slight rotational phase shifts typically between 50kHz to 1Mhz. The Zobel network represents a phase correcting load across the amplifier's output. It attempts to keep the combined small capacitive effects of the amplifiers printed circuit tracks, internal wiring to the speaker terminals, and the speaker cable appearing as a phase coherent resistive load. Only a very few professional amplifier designs have achieved excellent performance stability under most load conditions.
Similar to marriage and helicopters there is no such thing as a naturally stable solid state amplifier. All solid state amplifiers are parasitically unstable once a speaker cable is connected, its only a matter of degree. Cables of differing lengths and styles can cause bursts of parasitic oscillation to randomly appear on different parts of an audio waveform. Changing cable may stop or shift the parasitic to different positions. Parasitic sub-harmonics can be heard within the music.
There is un-intentional incorrect information on Zobel networks on many web sites. A Zobel network in an amplifier does not provide correction for impedance variation of a speaker. Some passive crossovers include a correcting network for impedance variation of a speaker inside the speaker box also described as a Zobel, but it is not the same Zobel as in the amp.
It is essential for the output of an amplifier to be exactly at 0V when there is no signal. The differential input pair have to be exactly equal for DC offset to be zero. DC offset can also be caused by un-equal quiescent current through the output transistors. As transistors gets hot, the HFE (gain) increases, and each transistor will behave slightly differently. This may cause a DC offset 1mV to 1V at the amplifier output. A + or - DC offset of 100mV or greater will cause the speaker voice coil to be biased slightly off centre, either in or out. With the cone biased off centre, the cone will move back and forth non-linearly. The low frequency efficiency of the speaker will be reduced and distorted at higher power.
This problem is often mistaken for a faulty speaker. First, turn the amplifier off. Then turn the amplifier on (without music) and notice if the cone moves slightly, in or out. A small DC offset is often overlooked by most service technicians because it does not effect the power performance or other specifications of an amplifier. Also DC off set may appear only when the amp is hot. DC offset was a major problem in many early solid state amplifiers. The negative feedback R1 R2 must have a 100uF capacitor in series to ground to stop the amp from amplifying any DC that gets to the input from an external source. Also the output of the amp may have a speaker relay with a delay turn on to avoid turn on and off bangs and detect DC offset.