Meier Audio has a feature in their amps called “FF” or Frequency Adaptive Feedback. Jan Meier describes it here. His article is detailed yet long and can be hard to understand exactly what it does, and why. Here I give a simpler explanation. FF is based on 3 key concepts.
If my explanation here makes sense, go back and read Meier’s and you’ll get an even deeper understanding.
When it comes to human perception of sound and music, all frequencies are not created equal. The ear is most sensitive to frequencies from around 100 to 3000 Hz. And, most music (at least voices and acoustic music) is concentrated in this range.
Consequently, this is the most critical range for reducing distortion. You might not hear 1% (-40 dB) distortion at 30 Hz, but you can definitely hear it at 2000 Hz.
Analogy: Dolby B
Readers with a few grey hairs remember cassette tapes and Dolby B noise reduction from the 1970s and 80s. Dolby B was brilliant in its simplicity. Tape hiss has a wide frequency spectrum but it’s most noticeable in the treble. If you cut the treble during playback, it reduces hiss but it also dulls the music. So when recording, boost the treble. Then during playback, cut the treble by the same amount you boosted it. You get the same hiss reduction without any reduction in treble, because you’re only cutting exactly what you boosted earlier. The music has flat frequency response and sounds cleaner with higher S/N ratio.
The only drawback to this is that boosting the treble when recording limits the dynamic range. You can only boost it so far, before it reaches peak levels and overloads. Boosting the treble may require you to reduce the overall recording level. However, with most music this not much of a drawback since the energy is focused in the bass. Treble is usually only a small % of the overall energy, so boosting it doesn’t affect the overall level very much.
Solid state amplifiers have a negative feedback loop that reduces distortion and increases stability.
What exactly is negative feedback? A portion of the output signal is inverted, attenuated, and fed back into the input. Imagine what happens when you do this. Because it’s inverted, each distortion tone becomes its mirror-image opposite. As this passes through the amplifier, it opposes the distortion tones that the amp produces. The distortion tones oppose and cancel each other.
Frequency Adaptive Feedback
Combine these 3 ideas and you have Meier Audio’s FF. Start with the musical signal.
- Step 1: boost the critical frequency range (say, 100 Hz to 3000 Hz)
- Alternately, attenuate frequencies outside this range. This can be a better approach since attenuation means no chance of clipping.
- This is the first thing you do when the signal enters the amp.
- Step 2: pass this modified signal through the normal amp / feedback stage
- Step 3: attenuate the critical frequency range
- Do the reverse of what you did in step 1.
- This is the last thing you do before the signal leaves the amp.
In step 2, because the critical frequency range is exaggerated, the feedback loop’s distortion reduction is focused in this range.
In step 3, when you attenuate the critical frequency range back to its original level, this has the side effect of attenuating any residual distortion in that range. This improves the S/N ratio in this frequency range.
In summary, FF does to distortion what Dolby B does to tape hiss. It’s based on the same concept.
Incidentally, the Redbook CD specification has something called “emphasis”, which boosts high frequencies. CD players are expected to attenuate those frequencies on playback. This is akin to Dolby B for digital audio.
Musical Energy vs Frequency
The energy in music (and most other sounds) is not evenly spread across frequencies. Most of the energy is in the bass, and energy drops by about 6 dB per octave into higher frequencies. This is true for most music, from chamber music to rock.
As the overall signal approaches maximum amplitude peak levels, the midrange & treble is forced into extreme amplitude levels because they’re riding on the bass wave, even though they are much smaller, nowhere near peak amplitudes. Any component that is not perfectly linear all the way to max levels, will now increase distortion in the mids and treble. Attenuating the bass lowers the overall signal amplitude, makes the critical midrange & treble frequencies relatively larger, which “unloads” the signal processing to focus on these elements. The signal no longer swings near the max amplitude levels.
Human hearing is most sensitive in the midrange and treble. Since these are at lower levels than the bass, they’re closer to the noise floor. This means recording gives us the opposite of what we really need. We get high S/N ratio in the bass, where we don’t need it, and we get reduced S/N ratio in mids and treble where we need it most.
Concept: boost the midrange & treble when recording, then cut it on playback. Alternately, cut the bass on recording and boost it on playback. Either of these approaches optimizes the S/N ratio by frequency to better match our perception.
Here we’ll play some devil’s advocate.
If distortion is already below audibility, then FF is a solution looking for a problem – what is the point? In fact, the cure could be worse than the disease! FF requires filters on the input and output to shape the frequency response. These filters cause their own distortions (such as phase distortion from analog filters or minimum phase digital filters). The overall effect is a trade-off between the benefits of FF and the drawbacks of having this extra signal processing.
FF actually increases distortion outside the critical frequency range! With FF you will have higher distortion at the extreme low frequencies (because FF attenuates them in the feedback loop). But you’ll have lower distortion in the midrange and treble. FF shapes distortion to match the sensitivity of our hearing: less distortion where our hearing is most sensitive, at the cost of higher distortion at frequencies where we can’t hear it.