Lo-Fi: every effect side by side

What it shows. A steady 1 kHz note through Lo-Fi, drawn sample by sample (44,100 samples a second) with nothing averaged, at each tested Sampling Freq setting. Use the slider to change the setting.

How to read it. A smooth wave means the effect is passing every sample. Flat steps mean it is holding one sample and repeating it: the longer each step, the lower the effect's own sample rate, and the less the result looks like the note that went in. The flat top and bottom at about ±0.45 are the effect's clipping ceiling at its default Word Length.

What it found.

• The steps are a whole number of samples long at every setting, and at 8 of 8 settings above 0 the step is exactly the setting plus one: setting 15 holds 16 samples (0.36 ms, a rate of 2,756 Hz), setting 31 holds 32 samples (0.73 ms, a rate of 1,378 Hz), setting 47 holds 48 samples (1.09 ms, a rate of 919 Hz), setting 63 holds 64 samples (1.45 ms, a rate of 689 Hz), setting 79 holds 80 samples (1.81 ms, a rate of 551 Hz), setting 95 holds 96 samples (2.18 ms, a rate of 459 Hz), setting 111 holds 112 samples (2.54 ms, a rate of 394 Hz), setting 127 holds 128 samples (2.90 ms, a rate of 345 Hz).
• That is 44,100 Hz divided by (setting + 1), which matches Yamaha's value table.
• At setting 0 nothing is held.

What it shows. A single note sliding up from 20 Hz to 20 kHz over three seconds through Lo-Fi, at each Sampling Freq setting. For every moment of the slide the chart marks the frequencies present in the output: cyan where it is the note itself, violet for its harmonics, amber for anything else.

How to read it. The frequency being played runs left to right, what comes out runs upward, both on musical (log) scales. A clean effect would draw only the cyan diagonal. Violet lines parallel to it are ordinary harmonics. Amber lines that fall as the note rises are images: copies reflected off the reduced sampling rate, which is why a rising melody through Lo-Fi is accompanied by tones that descend. The dotted curves are where images of the note must fall if the rate is 44.1 kHz ÷ (setting + 1). Drag the slider through the settings.

What it found.

• At Sampling Freq 0 (the full 44.1 kHz) there are no falling lines, but 294 harmonic points: Lo-Fi adds odd harmonics of its own before any rate reduction.
• At the reduced rates, between 86% and 90% of the stronger amber points (within 30 dB of the loudest, while the note is still below half the rate) sit within two-thirds of a semitone of an image of the note or of its 3rd or 5th harmonic at the predicted rate.
• The rate is what the division says, and nothing filters the images out afterwards.
• By Sampling Freq 127 (345 Hz) the note can only be followed to about 172 Hz before images take over, and only 54 of the marked points are the note at all.
• Only lines within 42 dB of the loudest at each moment are marked, read from 93 ms slices.

What it shows. A 1 kHz tone rising smoothly from −69 to −4 dBFS through Lo-Fi, with Sampling Freq at its full rate so that only Word Length is acting: everything in the output that is not the 1 kHz note (harmonics, aliases and noise together), relative to the note. One line per Word Length setting.

How to read it. Input level left to right; lower is cleaner. A line falling steadily to the right is a fixed noise floor being left behind as the note gets louder. A sudden jump upward is the moment the note hits the ceiling.

What it found.

• Until the ceiling every setting follows the bypass line down to about −59 dB (bypass reaches −62 dB): no quantisation noise is added, which is what reducing word length from the bottom would do.
• Past the ceiling the residue jumps to between −24 and −9 dB at −6 dBFS in, the level of a hard-clipped wave.
• The lower the Word Length, the earlier the jump: the setting is a ceiling on the output (−4 dBFS at 15, −0 dBFS at 31, +3 dBFS at 47, +6 dBFS at 63, +7 dBFS at 79, +7 dBFS at 95, +8 dBFS at 111, +8 dBFS at 127, where 0 dBFS is the level a full-scale tone returns at through bypass), above +24 dB of plain gain.
• At Word Length 0 nothing comes out at all.
• A shorter word does not make quiet sounds grainy here; it takes headroom away from the top.

What it shows. Every pair of controls that was recorded at all four corners, each control at its low and its high test setting, for the Lo-Fi effects. For each pair: what the second control does to the tone with the first one low, what it does with the first one high, and how far apart those two are as a share of the larger.

How to read it. One dot is one pair; hover for which. At the left edge the second control does the same thing wherever the first one is, so the two simply add. At the right edge it does something at one end of the first control and nothing, or the opposite, at the other. The same pairs are drawn as curves on each effect's sheet above, under two controls together.

What it found.

• Of 36 pairs in which the second control changes the tone, 26 are ones where the first control changes what the second does.
• The pair furthest apart is Flanger Feedback with Flanger Delay on Attack Lo-Fi.
• Tone only is compared, smoothed over about two thirds of an octave, so a change in timing alone does not show here.
• On an effect that moves by itself part of any gap is the movement caught at a different moment.