Three different things go wrong when a singer works too close to a directional microphone, and only two of them can be repaired afterwards. The bass rises, because every pressure-gradient capsule does that at short distances. Plosive consonants arrive as air rather than as sound. Sibilance gets louder and narrower at the same time. The bass and the plosives are fixable in the mix. The third one, the change in how the voice was performed once the singer heard themselves sounding large, is not a technical fault at all and no plugin addresses it.
What is proximity effect actually doing?
It is adding low frequency energy that varies with distance, and the variation is the problem rather than the boost.
Proximity effect exists in every microphone that works on a pressure difference across the diaphragm, which means every cardioid, supercardioid, hypercardioid and figure-of-eight. Omnidirectional pressure capsules do not have it. DPA's technical note puts the ordering plainly: the closer a pattern comes to figure-of-eight, the more proximity effect it shows, because a bidirectional capsule uses both sides of the diaphragm equally.
The manufacturers give real numbers for this. The Neumann U 87 Ai operating instructions state that the low-cut switch compensates the effect so that a uniform frequency response is produced in the cardioid position for a talking distance of 30 to 40 cm, and in figure-8 for 15 to 20 cm. The switch is calibrated for a specific working distance, and that distance is much further than anyone actually sings.
The same manual gives the size of the correction: relative gain at 40 Hz is minus 3 dB flat and minus 18 dB with the low cut engaged. That is 15 dB of designed-in compensation, aimed at a singer standing a foot and a half away.
How much boost are we dealing with at 10 cm?
Enough that the low cut on the microphone was never going to reach it.
A vocal recorded at 8 to 12 cm on a cardioid large diaphragm typically arrives with something in the region of 8 to 12 dB of lift centred somewhere between 100 and 250 Hz, depending on the capsule and the singer's height relative to the microphone. The exact figure matters less than where it sits. Below about 100 Hz there is very little voice, so anything there can be removed without loss. Between 150 and 300 Hz sits the chest weight of most adult voices, and removing that region removes the singer along with the fault.
This is why the reflex fix fails. A 100 Hz high-pass takes out the part that was harmless and leaves the part that is doing the damage.
Why does a fixed EQ curve not solve it?
Because the boost is a function of distance, and the distance changed constantly while the take was recorded.
A singer moves 3 to 6 cm between a quiet verse and a loud chorus without noticing, and at these working distances that swing is worth several decibels of low end. Set a static 4 dB cut at 200 Hz to fix the loud lines and every quiet line goes thin. Set it for the quiet lines and the chorus stays muddy. The take now has two faults instead of one, and the second one was introduced by the repair.
What works is a dynamic cut. A single band around 180 to 220 Hz, Q of roughly 1.2, threshold set so the band engages only on the loudest 20 percent of the performance, 3 to 5 dB of range, attack around 10 ms and release 150 to 250 ms. The band tracks the singer's distance because level and boost move together. Before any of that, clip gain the take by phrase: getting loud and quiet passages within about 4 dB of each other by hand makes every dynamic process downstream behave, and it takes ten minutes on a three-minute song.
Can a plosive be repaired?
Most of them, and the method is surgical rather than tonal.
A plosive is a burst of moving air hitting the diaphragm, not an acoustic event with a pitch. Its energy is almost all below 100 Hz and it lasts 40 to 80 ms. Automate a steep high-pass, 24 or 48 dB per octave at 120 Hz, onto that window only, and pull the clip gain down 4 to 6 dB across the same region. The consonant survives, because the part of a P or a B that the ear recognises is the transient at the top, not the thump at the bottom.
The plosives that cannot be saved are the ones that overloaded something. Neumann note in the same manual that pop-testing a condenser can produce sound pressure levels exceeding 140 dB, and if that hit the preamp hard enough to clip, the damage is at the top of the waveform where the voice lives.
Two things prevent this at source and both are cheap: a gauze pop screen at 8 to 10 cm from the capsule, and angling the microphone 10 to 15 degrees so the breath passes across the grille rather than into it. Shure ship the SM7B with a large close-talk windscreen for exactly this reason, described in the SM7B specification sheet as reducing plosive sounds and giving a warmer tone for close-talk vocals.
Why does close miking make sibilance worse as well?
Because you are inside the region where the high frequency radiation from the mouth is still narrow.
At 30 cm the mouth behaves close enough to a point source across the vocal range. At 10 cm it does not, and the 5 to 9 kHz energy in an S is still a beam. Small movements of the head then produce large changes in how much of that beam lands on the capsule, which is why close-recorded sibilance is both louder and more erratic than sibilance recorded further back.
The practical consequence for the mix is that one de-esser set to one frequency will not hold. Find the actual range first by sweeping a narrow bell, plus 12 dB at Q 6, between 4 and 10 kHz, because an S and a SH on the same singer can sit two octaves apart. Then split the work: 2 to 3 dB before the compressor and 2 to 3 dB after it, rather than 6 dB in one place. A single deep de-esser is what produces the lisp people complain about.
What should change at the next session?
Distance, angle and one recording that nobody makes.
Move the singer to 20 to 25 cm and put the pop screen where their nose would have been. If the voice sounds thin at that distance, the answer is a different microphone rather than a shorter distance, because thinness at 20 cm means the capsule was chosen wrong.
Angle the capsule 10 to 15 degrees off the mouth, aimed at the corner. This costs almost nothing in level and removes most of the plosive energy and some of the sibilant beam at the same time.
Then record 30 seconds of the singer speaking normally before the first take, at the exact distance they will sing from. That clip is the reference for every EQ decision later, because it shows what the microphone does to that voice when nothing dramatic is happening. Almost nobody records it.
Frequently asked
Can I just use a high-pass filter? A fixed high-pass removes the part of the boost that sits below the voice and leaves the part inside it, usually 150 to 300 Hz, and it cannot follow a boost that changes size every time the singer moves.
How close is too close? Neumann specify flat response on the U 87 Ai in cardioid at 30 to 40 cm with the low cut in. Most home vocals arrive recorded three to four times closer than that.
Is a plosive repairable? Usually, with an automated steep high-pass over 60 to 80 ms plus clip gain. A plosive that clipped the preamp is not, because the damage is at the top of the signal.
If you have a take like this
Send me thirty seconds of the raw vocal, unprocessed, and tell me roughly how far the singer was from the microphone and whether there was a pop screen. That is enough to say which of the three faults you actually have and which of them is worth spending time on. Full credit list at /credits, background and contact at /about.