Gain Structure, and Where Your Levels Actually Go Wrong

You have a track that came back noisier than it sounded in the room, or smaller, and the tab you have open is a preamp. That is the usual order of events and it is the wrong one. The preamp is one of five things putting noise into that recording, and on most sessions it is not in the top three.

Gain structure is the unglamorous business of deciding how much amplification happens at each point in the chain, and where the level lands when it reaches the converter. It costs nothing and it is entirely under your control. I have sat through a lot of preamp demos that were really level demos: somebody sets up a comparison, the new box is quieter and more open, and what actually happened is that the gain finally sat in the right place. The improvement in the room was real. It just did not come out of the circuit, and it is worth knowing where it did come from before you pay for it.

Why the New Box Sounded Better

Take the demo seriously, because in most of them something genuinely changed. Four things change when a new preamp goes into a chain, and the circuit is the least of them.

The two boxes were not at the same level. Matched knob positions are not matched outputs, and the louder one wins. Half a decibel reads as more detail and more weight, and nobody in the room hears it as a volume comparison, because volume was not the thing anybody thought they were adjusting.

The gain got set on purpose, probably for the first time on that source. Nobody plugs in a new preamp and leaves the knob where the old one sat. They set it while listening, during a real take, with their whole attention on the level. If the old chain had been running 10 dB hot into the converter, the new one is not, and the take that comes back still has its transients on it.

Everything downstream stopped being overdriven. This is the one most often mistaken for a preamp's character. A track recorded 12 dB hot drives every analog modeled plugin it meets 12 dB past the point that plugin was calibrated for, and then does the same to the bus. Record the same performance at a sane level and the whole mix arrives less squashed, which is precisely what people describe when they describe a better preamp.

The amplification moved to a different part of the range. A preamp with more clean gain in hand supplies 60 dB without going to the top of its travel, which is where inexpensive designs get noisy and gritty fastest. That difference is real and worth money, but it is a difference in where the gain sits rather than in how good the circuit is, and an inline booster, covered below, buys the same effect for a fraction of the price.

All four are level. Three cost nothing and the fourth costs far less than the box. The rest of this article is the evidence, and the one case where a better preamp genuinely is the answer.

What You Are Actually Hearing

Noise arrives from a fixed set of places, in the order the signal meets them, and they are not equal contributors.

The source. A fingerpicked guitar three feet away is a small signal, and nothing downstream makes it bigger without making everything around it bigger too.

The room. An ordinary domestic room sits between 35 and 45 dBA with nothing happening in it. Refrigerator, forced air, traffic, the computer's fans. A good studio room gets into the twenties, and that is construction rather than equipment.

The microphone's self noise. A condenser has electronics inside it and they hiss. The spec is equivalent noise level in dBA: a quiet modern large diaphragm condenser is under 5, a typical one is 12 to 18, a cheap one can be north of 20. A passive ribbon or dynamic has no electronics, so its noise is negligible. Its problem is output.

The preamp. Specified as equivalent input noise, usually at 60 dB of gain with a 150 ohm source. The floor is thermal noise from that source resistance, about minus 131 dBu unweighted across the audio band, and nothing can sit below it. A very good preamp lands two or three decibels above that, around minus 128. Published figures at minus 131 or lower are A weighted, which is a different measurement rather than a better circuit. A cheap one lands between minus 120 and minus 125.

The converter. A current 24 bit converter gives you 110 to 120 dB of range. In a music chain it is almost never what you are hearing.

Put numbers against a normal case. A condenser with 15 dBA of self noise, in a 40 dBA room, needing 35 dB of gain for a vocal. The room is 25 dB louder than the microphone. At an ordinary large diaphragm sensitivity of 20 to 25 millivolts per pascal, that microphone's own hiss refers back to the preamp input at somewhere around minus 109 to minus 111 dBu, against the minus 122 a mediocre preamp contributes at that setting. Two noise sources 12 dB apart do not add up to anything: the quieter one moves the total by about a quarter of a decibel.

So in that session a better preamp is inaudible. What you are hearing is your room, then your microphone, and a purchase cannot move either unless the purchase is a microphone or a building. The same arithmetic disposes of the rest of the shopping list. A cable has no noise floor worth measuring, and what a cable can and cannot do is settled elsewhere. A better converter solves the one item that was not a problem.

There is exactly one case where the preamp genuinely is the dominant noise source: a low output microphone, a quiet source, and 55 to 70 dB of gain. That case is the last section of this article.

Where the noise is from	The figure	Can a purchase move it
The room	35 to 45 dBA in an ordinary domestic room, into the twenties in a good studio room	Construction, not equipment
The microphone’s self noise	Under 5 dBA on a quiet modern large diaphragm condenser, 12 to 18 on a typical one, north of 20 on a cheap one. Negligible on a passive ribbon or dynamic	Only by buying a different microphone
The preamp	Around minus 128 dBu on a very good one, minus 120 to minus 125 on a cheap one, against a thermal floor of about minus 131	Yes, and only at 55 to 70 dB of gain
The converter	110 to 120 dB of range on a current 24 bit converter	In a music chain it is almost never what you are hearing

Recording Hot Was Correct, Once

Pushing levels as high as they will go was right for about forty years, for two reasons that both stopped being true. On tape the noise floor was the medium. Two inch multitrack gave you a signal to noise ratio somewhere in the sixties per track and narrower formats gave less, the hiss was audible, and it accumulated on every bounce. Every decibel you put on the tape was a decibel you gained on it. And the ceiling was forgiving: tape does not clip, it compresses, so approaching saturation rounds the transients and thickens the low end. People liked it. Whole genres are built on it. Level bought a cleaner track and a better sound at once.

Neither survives the move to a converter. The noise floor no longer moves when you change level, because it belongs to the room and the microphone and both sit upstream of the gain control. And the ceiling is a wall. There is no number above zero dBFS, nothing musical happens on the way to it, and a converter that runs out does not compress. The habit carried forward into a chain where one of its benefits evaporated and the other turned into a penalty. That is most of what I hear when somebody sends me a track that is noisy and thin at the same time. Thin usually means the transient is gone, and transients go at the top of the range, not the bottom.

Where Minus Eighteen Comes From, and What Sits Above It

The target most people have heard and few have been given a reason for: peaks around minus 18 dBFS on the way in.

When digital recording arrived, somebody had to decide what analog level corresponded to what digital level. Professional analog gear runs at a nominal operating level with 18 to 24 dB above it before the circuit stops being linear. Two calibration standards ended up in common use, one landing that nominal level at minus 20 dBFS and one at minus 18, which is why both numbers get quoted and why the argument between them is not worth having. Nominal sits roughly 18 to 20 dB down from the wall, and the gap above it is where the transient lives.

The peak you did not predict is the first claim on that gap. A singer who leans in on the fourth take, a snare hit harder than the three before it. Crest factor, the distance between average and peak, runs 15 to 20 dB on percussive sources and 6 to 10 dB on something sustained or already compressed. Set by average and you clip the first kind. Set by the loudest thing that has happened so far and you clip the moment somebody plays better.

The plugins are the modern claim, and the one people miss. Every analog modeled plugin has a calibration point built in, almost always in that same minus 18 to minus 20 region, because it was modeled from a box that operated there. Feed one a track averaging minus 6 and you are driving the model 12 dB into a saturation region the original hardware reached on its worst day. Then you wonder why everything sounds flat and crowded before you have touched a fader. Summing has a claim too: twenty tracks peaking at minus 3 do not sum to minus 3.

The objection is always the same. Does recording that far down cost me noise? No. On a 24 bit converter you have well over 100 dB of range below the peak, and your noise floor was set by a room and a microphone sitting 30 dB or more above the converter's. Recording quieter costs nothing you could measure. That worry is a 16 bit worry, and before that a tape worry.

The Clip Light Is Not the Meter You Want

A clip light watches one node in the circuit, usually a few decibels below the rails, and it is a peak detector with a hold time. It answers one question, whether you have hit the wall, and it answers it after the fact. It cannot tell you where you are: nothing in the light distinguishes sitting 25 dB below clipping from sitting 2 dB below it. The preamp's clip point and the converter's are not the same place either, so you can be clean at one and already over at the other.

Meter at the converter's input, reading peak, ideally on something that shows average and peak together, because the distance between those two is the number you are really setting. Then set the gain during the loudest section of an actual take rather than a chord played to check, watch it for a full pass, and leave it alone. A gain knob moved mid session is a decision you will be undoing in the mix.

A Condenser and a Ribbon Are Two Different Problems

Most people set both the same way. They should not, because what goes wrong is at opposite ends.

A condenser is a top end problem. Output is high, 10 to 30 millivolts per pascal, so it wants 20 to 40 dB of gain and the preamp's noise is nowhere near the conversation. What it has instead is internal electronics the source can overload before your preamp is involved at all. This is a common misdiagnosis in home recording: a drum overhead or a close trumpet distorts, the engineer turns the preamp down, nothing improves, because the distortion happened inside the microphone.

The test takes ten seconds. Turn the gain down 10 dB. If the distortion is quieter but still there, it is upstream of the gain control, which means the microphone, and the fix is the pad on the microphone body or a different microphone. If it cleans up, the gain was simply too high.

A ribbon is a bottom end problem. A passive ribbon is foil and a transformer, with nothing inside to overload and an SPL limit that is mechanical rather than electrical. It will sit in front of a guitar cabinet all day. What it gives you is an output 15 to 20 dB below a condenser, so 55 to 70 dB of gain on anything quiet, which is the one case where the preamp's own noise is the loudest thing in the chain.

So set it from the other direction. With the session quiet, bring the gain up to where you think it needs to be and listen to what you are amplifying. If the room arrives first, that preamp is fine on that microphone and you are done. If hiss arrives before the room does, you have found the preamp's limit, and no amount of careful metering gets you past it. While the gain is up there, check phantom power is off before anything gets plugged or unplugged, because 48 volts across a shorting cable can stretch a passive ribbon permanently.

Boosters and Pads Fix Opposite Ends

Both are gain structure tools and both get used for the wrong problem.

An inline booster, the Cloudlifter and the FET boxes that followed it, takes phantom power and puts 20 to 25 dB of its own low noise gain right at the microphone. It does not improve your preamp. It moves a chunk of the amplification upstream so the preamp supplies 40 dB instead of 65, which is the part of its range where it is quiet and well behaved. A passive ribbon or a low output dynamic into a preamp that goes gritty at the top of its travel is exactly the case it solves, at a fraction of what a preamp costs. Two conditions: it needs the phantom for itself and blocks it from continuing, so it is no use to a condenser or an active ribbon, and it presents its own fixed load, which changes how some ribbons and older dynamics behave. If your preamp already delivers 70 dB of quiet gain, the booster is a stage doing nothing.

A pad does the opposite job and lives in two places, which is where the confusion is. The pad on the microphone body attenuates ahead of the microphone's own electronics, and it is the only one that helps when a capsule is overloading. The pad on the preamp attenuates at the preamp input, which helps when a very high output microphone meets a very loud source, or when somebody puts a line level signal into a microphone input. Engaging the preamp pad when the microphone is the thing distorting fixes nothing and throws away 20 dB of signal to noise on the way.

Going the other way there is a legitimate move: pad the input 20 dB and turn the gain up 20 dB on a preamp whose sound you want, and you get more of its input stage at the same output level. That is deliberately bad gain structure bought on purpose, which is a different thing from arriving at it by accident.

Tool	What it does	The problem it is for
An inline booster	Puts 20 to 25 dB of its own low noise gain right at the microphone, so the preamp supplies 40 dB instead of 65	A passive ribbon or a low output dynamic into a preamp that goes gritty at the top of its travel. No use to a condenser or an active ribbon
The pad on the microphone body	Attenuates ahead of the microphone’s own electronics	A capsule that is overloading. This is the only one that helps
The pad on the preamp	Attenuates at the preamp input	A very high output microphone on a very loud source, or a line level signal in a microphone input

What Sixty Decibels of Gain Actually Costs

At 60 to 70 dB of gain the preamp's equivalent input noise is the noise floor of the recording, so this is where the money is real. A good preamp at minus 128 dBu sits within three decibels of the thermal limit, which leaves almost nothing to buy. An inexpensive one at minus 120 to minus 122 is 6 to 8 dB worse, and there is usually a second effect on top of the specification: many affordable designs reach their gain figure by cascading stages, so the last 10 dB of travel is noisier than the first 50 and the hiss rises faster than the knob suggests. Some also lose headroom at high gain settings, which is exactly when a ribbon in front of a loud amplifier finds it.

Six to eight decibels is the difference between a usable ribbon track and an afternoon of cleanup. So yes, on a Coles 4038 or a Beyerdynamic M160 on a quiet acoustic source, or an SM7B in front of a soft speaker, a better preamp does what people claim. The SM7B is the clearest example in the market: a quiet microphone, and most of the people who bought a preamp to fix it needed 25 dB of gain moved upstream instead.

I spent four years building microphone preamps, from 1997, as the first employee of the company that made them, and the whole reason those boxes existed was that last stretch of gain. On a condenser at 35 dB, the preamp in a competent modern interface and one costing ten times as much both contribute noise far enough under the microphone's own that the microphone sets the floor either way. The expensive one earns its money in three specific places: high gain on low output microphones, headroom on loud sources, and a voice you went looking for on purpose. Wanting a record to sound better in general is not one of them.

If you do compare two, compare them properly. Same source, same position, output levels matched to within a fraction of a decibel, switching inside one sitting. Level matching is the whole test. Without it you are comparing volumes and putting a price tag on the result. The order I use for any listening applies here too.

The Short Version

Before buying anything, find out which noise you have. Record thirty seconds of your room with nothing playing, at the gain you actually use. If what you hear is the room, buy nothing: treat the room or move the microphone closer. If it is hiss and you are running less than 45 dB of gain, the microphone is the noisy part. If it is hiss and you are running more than 55 dB into a passive microphone, the preamp genuinely is your floor, and that is the one case where the purchase is justified. Even then, try 25 dB of inline boost first. It costs a fraction of a preamp and solves the same problem by moving the gain rather than improving it.

Set peaks around minus 18 dBFS during the loudest part of a real take, then stop touching it. Meter at the converter, not at the clip light. Use the microphone's pad for a distorting capsule and the preamp's pad for a hot input, never the second for the first.

And if a box you are auditioning sounds better, match the two to within a fraction of a decibel before you believe it.

Do that and one of two things happens. The noise drops and the transients come back, in which case you have just saved the cost of a preamp. Or it does not, in which case you know exactly which link is the problem and you can buy the right box instead of the famous one.

---

Or Bring the Session In

The listening room at Free River Studios, 215 E. Lewis St #301 in Livingston, sits on the third floor of a working recording studio, so the monitoring and the levels were dealt with properly before anyone brought a record into it. Bring the mix you have been fighting with and a record you know cold, and play them back to back. An hour of that tells you whether the problem is your chain or your room. It is free and it is by appointment.