Effect chain order: what goes where and why

A DSP chain is not a bag of effects, it is a pipeline. Your mic feeds the first node, the first node feeds the second, and so on down the line. Nothing except the first effect ever hears your raw mic. So reordering two effects is not rearranging furniture: it changes what each effect measures and reacts to. Same effects, same settings, different order, different sound. In the worst case, a broken one.

A rule of thumb for a voice chain

This is a starting order, not a law. It is worth knowing as a default because each position has a reason you can check, and worth leaving when your room or your voice gives you a better one: clean up first, shape after, transform last.

• 1. High-pass. A cut, not a tone move. Rumble is energy even when you cannot hear it as a note, and detectors measure energy, so taking it out first keeps every effect below from reacting to it.
• 2. Cleanup: suppression and the gate. Suppression works on the noise under your voice while you talk; the gate closes the gaps when you stop. Which of the two runs first is a judgment call, and the next paragraph lays out what each way buys.
• 3. Tone: EQ and a de-esser. This is where an EQ adds rather than removes: a lift around the range that carries your consonants, once there is no hiss left to lift with it. A de-esser sits beside it to catch the S before any boost makes it sharper.
• 4. Compressor. Even out the level of a signal that is already clean and already shaped, so it works on your voice rather than on the room.
• 5. Creative effects. Reverb, character, anything decorative. Polish goes on the finished surface, not under it.
• 6. Persona last. A voice changer should receive the final, clean, level signal. Feeding it noise or uneven levels means transforming the noise along with the voice.

The gate and suppression pair, both ways. Put suppression first and the gate inherits a quieter floor: suppression can take a room from around -50 dB down to around -70, so a threshold at -45 seals every gap with room to spare and never shaves the quiet start of a word. Put the gate first and suppression only ever runs on your voice, never on the room, which keeps it from chewing on noise it was not trained for in the pauses, at the cost of a gate that has to judge the raw floor. Neither is wrong. Set the threshold where it works for the order you chose, and listen to the quiet ends of your words either way.

If you remember one phrase, make it this: measure clean, then shape. Every effect that reacts to loudness (the gate, the compressor) makes better decisions when the junk was removed before it, not after.

What breaks when you swap a pair

A default is easier to leave on purpose once you know what each swap actually costs. These two have a specific, repeatable failure rather than a tradeoff, which is what separates them from the gate and suppression pair above.

• Compressor before gate. A compressor with makeup gain raises everything, and everything includes your noise floor. Lift the floor enough and it climbs above the gate threshold you set against the old, quieter floor. The gate opens and stays open, and it looks like the gate is broken when it is doing exactly what you told it to. The advanced tier below does this failure as arithmetic.
• EQ or high-pass last. Rumble is energy even when you cannot hear it as tone, and detectors measure energy. Put the high-pass at the end and every loudness-reactive effect upstream spent the whole chain reacting to garbage: the gate opens on desk thumps, the compressor clamps down on bus rumble that was never going to reach your stream anyway. The high-pass costs nothing at the front and fixes nothing at the back.

Notice that none of these failures come from bad settings. They come from a correct setting being evaluated against a signal it was never tuned for. That is the whole lesson of chain order: an effect's settings are promises about its input, and the order decides whether the input keeps them.

The broken gate, as arithmetic

Here is the compressor-before-gate failure with real numbers. Your room noise sits at -50 dB and you set the gate threshold at -45 dB, giving the floor 5 dB of clearance below the line. Speech crosses the threshold, the room never does, and the gate works.

Now insert a compressor before the gate and dial in +8 dB of makeup gain to restore loudness after the peaks come down. The floor at -50 dB sits far below the compressor's own threshold, so compression itself leaves it untouched; makeup gain, however, is a flat lift applied to every sample that passes through.

The gate compares that floor against the threshold you set when the floor was 5 dB lower:

Same gate, same threshold, same compressor, and every setting is individually defensible. The only bug is the order. Swap the two nodes so the gate runs first and the floor the gate measures is -50 dB again, comfortably under the line; the compressor then applies its makeup to a signal whose gaps are already sealed, so there is nothing in the gaps for it to lift.

The same clearance arithmetic shows why suppression belongs before the gate. Define the gate's working margin as the distance from the floor up to the threshold:

Five decibels of margin is a gate that works until you buy a louder fan. Twenty-five is a gate you stop thinking about.

Fixing order should be one drag

In Patchd the chain is an effects rack on every channel, and the order on screen is the signal flow, top to bottom. Every node here (EQ, suppression, gate, compressor) is free on every channel, and the rack is drag-to-reorder: if your compressor ended up ahead of your gate, the fix is dragging a row one slot, not rewiring anything.

Set the order once, watch the meters agree with the math above, and the chain stays fixed while you tune. You can try it yourself: Patchd is in open beta now, and if you clear the terms your invitation arrives straight away.