The Noise Floor

One volume of one journal, in one month, carried the measurement and the explanation — on facing pages.

J. B. Johnson, "Thermal Agitation of Electricity in Conductors," pages 97–109. He had put a vacuum-tube amplifier on a resistor and found a hiss that would not go away, in every conductor he tried, scaling with temperature.

H. Nyquist, "Thermal Agitation of Electric Charge in Conductors," pages 110–113. Four pages. His first footnote is three words long.

Both worked for the same telephone company, on the same problem, from the two ends. Johnson had a number nobody could explain. Nyquist explained it from thermodynamics and statistical mechanics, and the two agreed.

Remember this shape. It happens again in this chain, thirty-seven years later, running the other way.

documented  a measurement → its explanation, same volume

Nyquist's conclusion is worth quoting at its full width, because the last four words are the whole thing.

Not the metal. Not the wire's history, its purity, its manufacture, how much it cost or who made it. He says so outright: "no assumption has been made as to the nature of the two conductors. One may be made of silver and the other of lead, or one may be metallic and the other electrolytic."

And the power, in a band: "The average power, transferred from each conductor to the line in the frequency interval dν … is therefore kT dν." Integrate over a bandwidth B and you have the sentence people quote without the paper: kTB. In volts across a resistor R, his equation (1): E² = 4RkT per unit bandwidth.

So the floor is not a fault in a component. It is a property of being warm. Every resistor above absolute zero has one, the good ones and the bad ones alike, and there is no version you can buy that doesn't.

documented  the hiss → a universal quantity

Nyquist did not measure the floor. He deduced it — and the thing he deduced it from is the second law of thermodynamics.

The argument fits in a paragraph. Wire two resistors together, both at the same temperature. Each one's thermal agitation drives a current through the other, so each heats the other, and at equal temperature the two flows must be exactly equal — otherwise heat moves from one body to another of the same temperature, for free.

Now suppose that were only true overall, and that in some narrow band of frequencies one conductor gave more than it got. Nyquist's move: put a resonant circuit between them, tuned to that band. Now more power flows one way than the other, and heat has moved from one body to another at the same temperature with no work done. The second law says no.

This matters for everything after. The floor is not an unsolved problem. It is not waiting on a better material or a bigger budget. It is the same law that says you cannot un-stir the milk from the tea — and nobody files that under insufficient effort.

documented  the second law → the floor is compulsory

Twenty years on, in the same laboratory's journal, Claude Shannon took that floor and worked out what it costs you in understanding.

Rearranged the way it is usually written: C = B log₂(1 + S/N). And then the sentence that closes the door:

"It is not possible to transmit at a higher rate by any encoding system without a definite positive frequency of errors."

Not "we haven't found a way yet." Any encoding system. No cleverer code, no sharper filter, no more careful listening gets you past it. The noise is in the channel, so it is in the message, and the most you can ever do is trade — more power, more bandwidth, more time, fewer bits.

You cannot filter your way to clarity. That is a theorem, and it is the hinge of this whole zine.

documented  thermal noise → a hard ceiling on communication

Here is the part that gets skipped, and it is the most useful part. Once the floor was proved compulsory, the entire profession changed the question.

Nobody spent the next fifty years hunting for a resistor that doesn't hiss. They read the formula and noticed it has three handles on it, and every one of them is a thing you do to the situation rather than to the signal.

Notice what none of these is. None of them asks the signal to try harder, and none of them asks the receiver to be better than physics. All three change the conditions the listening happens in.

documented  a compulsory floor → design around it

Every one of those three moves, in one instrument, pointed at the sky.

Arno Penzias and Robert Wilson had a twenty-foot horn-reflector antenna and a radiometer built to be quieter than anything else on Earth. A travelling-wave maser — the lowest-noise amplifier then made — cooled by liquid helium to 4.2 K or less. A comparison switch of 0.027 dB loss. And a reference termination that was itself a microwave absorber sitting in liquid helium, twenty litres of it, good for about twenty hours a fill.

They were not looking for the beginning of the universe. They were doing careful radio astronomy, and Wilson states the reason for all that helium plainly in his Nobel lecture:

Read that against Nyquist and it is the same sentence in a different decade. They lowered their own floor on purpose, as far as money and cryogenics allowed, precisely so that anything still left over would have to be real.

Something was still left over.

documented  design around the floor → an instrument that can see past it

You have probably heard this story. It is almost always told wrong, and the correction is better than the version that travels.

The retelling makes the pigeons the punchline — two men chasing a mysterious hiss, cleaning out the birds, and there is the universe. Here is what Wilson actually wrote:

The pigeons were not the answer and they were not the suspense. They were one item on a list of things that did not work, and the sentence keeps going into the part nobody quotes.

Two smaller repairs while we are here. The famous phrase "white dielectric material" is not in the lecture — Wilson wrote "a white material familiar to all city dwellers," which is funnier. And the pigeons are not in the 1965 paper at all; they arrive thirteen years later, in the Nobel lecture. The story is true. It is just not from where people think, and it does not say what people think.

documented  the obvious explanation → eliminated, floor unmoved

The excess was 3.5 ± 1.0 K at 4080 Mc/s. Their paper is, mostly, a list of everything it wasn't.

And the remainder held. The paper records it in one parenthesis that quietly contains ten months of work: the excess is "isotropic, unpolarized, and free from seasonal variations (July, 1964-April, 1965)."

Isotropic: the same in every direction, so not a place. Unpolarised: so not a machine. No seasonal variation: so not the atmosphere, not the ground, not New York over the horizon.

documented  everything eliminated → a real, isotropic remainder

One volume of one journal, in one issue, carried the explanation and the measurement — on facing pages. Again.

Dicke, Peebles, Roll & Wilkinson, "Cosmic Black-Body Radiation," pages 414–419. A Princeton group had reasoned that a hot early universe should have left a bath of cooled radiation, and were building a receiver to look for it.

Penzias & Wilson, "A Measurement of Excess Antenna Temperature at 4080 Mc/s," pages 419–421. Three pages of things they had ruled out, and one sentence pointing next door: "A possible explanation for the observed excess noise temperature is the one given by Dicke, Peebles, Roll, and Wilkinson (1965) in a companion letter in this issue."

Thirty-seven years after Johnson and Nyquist, the same journal-page adjacency — and this time in reverse. In 1928 the measurement came first and the theory followed it. In 1965 the theory was already set in type when the measurement walked in from a telephone company in New Jersey.

They got the Nobel Prize in 1978 for a hiss they had filed as a fault.

documented  an unaccounted remainder → the oldest light there is

Here is the rung that surprised us while we were checking it. The transfer from Shannon to a room is not ours. It is already in the standard, in the standard's own words.

ANSI/ASA S12.60 caps background noise in a core learning space at 35 dBA and reverberation at 0.6 s. The Acoustical Society's guidance for architects explains why in a sentence that could have come off Shannon's page: you need an "average differential between speech level and background level (called signal to noise ratio) of about 15 dB, which means the A-weighted background noise level should not exceed about 35 dB."

Fifty decibels of teacher, minus thirty-five of room, equals fifteen of margin. A classroom is specified as a channel. And note the profession's instruction about what to do with it: "Rather than boost the instructor's voice through electronic amplification … we have to work to keep the background noise level inside the unoccupied classroom from exceeding 35 dB." Lower the floor. Don't shout.

The same story is in the lights and the offices, and the numbers exist:

Flicker. Wilkins, Nimmo-Smith, Slater & Bedocs (1989) compared fluorescent tubes on magnetic ballasts — pulsating 43–49% at 100 Hz — against high-frequency electronic ballasts under 7%. Double-blind, crossed over. Headaches and eyestrain more than halved. Wilkins' own later note is the honest part: the effect is "attributable to a minority that is particularly affected." The room was not equally bad for everyone. It was catastrophic for some.

Open plan. Kim & de Dear (2013), across a large occupant-survey database, found sound privacy the single largest source of dissatisfaction anywhere in the building: 59% dissatisfied in high-partitioned cubicles, 58% in low, 49% with no partitions — against 18% in private offices. And the thing open plan was sold for did not show up: satisfaction with ease of interaction was no higher than in private offices.

One more thing worth noticing, because this chain has already shown you the good version. Penzias and Wilson's paper is, in large part, a published noise budget: 2.3 K of atmosphere, 0.8 K of ohmic loss, back-lobes under 0.1 K. Reporting the conditions the measurement was made in is simply what a radio astronomer does — otherwise nobody can tell whether the result is about the sky or about the equipment.

Now consider that physical-health research involving Autistic adults almost never reports the lighting or the noise in the room. That is the gap the TRACE framework (Srinivasan, 2026) was written to close, as a reporting standard alongside CONSORT and STROBE: record the sensory context, the pacing, the proximity and touch practices, the access modifications offered and used. Its argument is exactly the radio astronomer's — without them "researchers cannot determine whether outcomes reflect the procedure itself, the conditions under which participation occurred," or who was able to take part at all. One field publishes its noise floor as a matter of routine. The other is still being asked to start.

documented  Shannon's quantity → a building's design spec

This is the joint where a chain like this normally cheats, so we are going to stop and say the disanalogy out loud before we say anything else.

Auditory filtering is not thermal noise. kTB is a floor set by temperature in a conductor. What happens when we try to follow one voice in a hard-surfaced room is a different phenomenon in a different substrate, and none of Nyquist's numbers carry over. There is no k for a nervous system. Anyone who tells you the physics proves something about an Autistic person's hearing is selling you the naturalistic fallacy with a formula attached, and this collection exists partly to refuse that move.

What the research says, carefully: Anna Remington and Jake Fairnie, in "A sound advantage: increased auditory capacity in autism" (Cognition, 2017), find that Autistic adults have a larger auditory perceptual capacity — and that this single fact predicts both results at once. We detect additional sounds others miss, which shows up as superior performance on the task we were asked to do and as greater distraction by the sounds nobody asked us to process.

That is not a deficit account and it is not a superpower account. It is a capacity account, and it means the sound reaching us was already received. "Tuning it out" is not a filter placed before the input; it is work done after, on material that has already arrived.

And acoustics has now gone and asked the question directly. María del Carmen Rosas-Pérez's 2026 doctoral thesis Disabling Acoustics surveyed 311 noise-sensitive people alongside interviews with twelve Autistic adults. Asked whether they can filter out irrelevant background sound, none of the Autistic participants said they find it easy — 96% could not do it or struggled, against 77% of the neurotypical respondents (χ²=25.08, p<.001). Read that second number carefully: everyone in this survey was already noise-sensitive, so it is a difference measured inside a sensitive sample, not a claim about the general population.

Here is where the analogy actually breaks, and the break is worth more than the resemblance. Thermal noise is stationary: kTB is the same in the tenth hour as in the first, and a receiver that runs all day is no worse off at dusk. Rosas-Pérez found the opposite in people — no habituation, but an accumulation effect, with fatigue and distress building over the short and long term. So the floor in a room is crueller than the floor in a circuit, not merely similar to it. A resistor never gets tired of being warm.

We mark this contested because it is: perceptual-capacity accounts are actively argued over, replication is ongoing, and no one study settles a population. We are not going to pretend otherwise to make a stronger ending.

contested  a floor in a channel → a capacity in a person

"Just tune it out." "It's not that loud." "You'll get used to it."

Those are not our inventions. The first two are what Rosas-Pérez's interviewees reported being told, over and over, since childhood — a pattern her participants' accounts led her to name sensory gaslighting. One of them put the whole argument in a sentence:

And "you'll get used to it" has an evidence base worth looking at. The idea that reducing exposure makes sensitivity worse — the reasoning behind telling people not to wear ear protection — traces to a study of five participants, none of them noise-sensitive, listening to two pure tones over two weeks. That is the foundation under advice given to people whose daily lives are nothing like it. Careful measurement, generalised past what it can carry, and handed back to us as a character note.

Hold that next to the instrument at Holmdel. Penzias and Wilson had the quietest receiver on the planet and it still had a floor, and not one person in that building thought the floor was a character flaw. They cooled it. They shielded it. They narrowed the band. Then they measured what was left and published it.

The move engineering made in 1928 was to stop demanding that the receiver be better than physics, and start changing the conditions the listening happened in. That move is what an accommodation is. Noise-cancelling headphones are shielding. A quiet room is a cooled load. A shorter meeting is a narrower band. None of it is special treatment; all of it is standard practice, applied to people.

And we mark it leap, because it is one. The physics does not prove this. Thermal noise entails nothing whatsoever about who deserves a quiet room — that is a moral claim and it has to stand on its own feet. What the physics does is narrower and still worth having: it removes an excuse. It makes "just tune it out" visibly the wrong kind of request — one that no receiver ever built has been able to honour.

leap  change the conditions, not the receiver → accommodation

This zine is an argument, and an argument is the least useful thing in the room if you are the one in the room.

So: everything above says the floor is set by the environment, therefore the environment is the unit of repair. The work of writing down how to actually do that has been done, by us and by people we learn from, and it is free.

Sensory Access at Work is the one-page version for a workplace — ear protection as normal kit, cutting HVAC and open-plan chatter, and the rest of the senses nobody budgets for. Designing Sensory-Safe Spaces is the longer pattern for a room you can change. Neuroception and Sensory Load collects what our complex sensory experiences are actually like, in our own words rather than an observer's.

And there is now an instrument. The third study in Disabling Acoustics builds AcoustInA, an Acoustical Inclusivity Assessment tool, tested by thirteen experts across twenty-one real spaces — which means the question how disabling is this room, and by how much has stopped being rhetorical and started being something you can go and score.

And the frame that reaches furthest: Sensory Trauma, the name Autism Wellbeing gave to something Autistic people had been describing all along — that the events which harm us need not be the extreme ones, that they arrive through taking a shower or going shopping, and that our responses are proportionate to our genuine, lived experience even when they are mislabelled as something else.

Which is the last thing this chain wants to say. The hiss at Holmdel had been arriving for thirteen billion years before anyone built an instrument quiet enough to notice it, and when they did they assumed for ten months it was a fault in the equipment. Being persistently told your reading is instrument error does not make it instrument error.