Meteor showers on a phone: the one night stacking works against you

The shower has been circled on the calendar for weeks. You go out, tripod up, phone aimed northeast, four hundred eight second frames running until three in the morning. Back home you launch the stack and get exactly what the software promises: clean sky, pinpoint stars, noise all but gone. Not a single meteor, though you counted eleven with your own eyes. Nothing has gone wrong with the program. This is the average doing its job on an event that lasts less than a second.

Half a second of glowing gas, a hundred kilometres up

The thing you are trying to photograph is usually smaller than a grain of rice. It hits the atmosphere somewhere between 11 and 72 kilometres per second: the IMO puts the Perseids at 59 and files them under swift, while the Geminids come in at 35 and count as medium. Friction, the hand-rubbing kind, has little to do with it. The grain compresses the air ahead of itself until that air glows, its own material vaporises, and the stripped atoms excite the air along the path. What you see is that column of gas, lit up as a rule between 80 and 120 kilometres of altitude.

That altitude produces the number which governs everything else. An object a hundred kilometres away, moving at 59 kilometres per second across your line of sight, covers a little over half a radian each second, call it thirty degrees. That is the perpendicular case, the best one. The IMO notes that a meteor is usually visible for under a second. Put the two together and you understand why a Perseid draws ten or fifteen degrees of sky and still feels instantaneous.

Long exposures make it worse

On deep sky the rule is familiar: more integration time, more photons, less relative noise. A meteor turns that on its head. The light it deposits on any single pixel arrives during the milliseconds its trail crosses that pixel, and not one instant longer. Sky background, meanwhile, keeps charging for the whole exposure. Double the shutter time and the meteor’s signal stays exactly the same while the background doubles and its noise grows by the square root of two, so the trail’s contrast gets worse with every passing second.

The conclusion is counterintuitive enough to stick: for meteors, short exposures do better. The brake comes from the other side, and it is called dead time, because between one frame and the next the phone is writing a file and during those moments it is looking at nothing. Eight seconds of exposure with a second and a half of writing is an eighty-four per cent duty cycle: across three hours of work, twenty-eight minutes of sky the sensor never saw. The upper limit is set by Earth’s rotation, and we have already counted those seconds lens by lens.

Averaging deletes the exception, and that is its job

Now the sore point. Stacking adds frames and divides by how many there are: whatever appears in all of them survives intact, whatever appears once gets divided by three hundred. Stars are in every frame. A meteor is in one. Across three hundred shots its brightness drops to a three-hundredth, which is another way of saying gone.

Live stacking is doing precisely the job it was built for. Its criterion is repetition: keep what returns identical frame after frame, damp everything else, and that is exactly how random noise dissolves. Under that rule a meteor is indistinguishable from a cosmic ray hit, something that shows up once and needs removing.

The fix has a dull name and simple behaviour: maximum combination. For every pixel you keep the highest value found anywhere in the sequence instead of the average. Every meteor survives, laid over the same sky. So do the satellites, the aircraft and the hot pixels, because maximum is as blunt an instrument as its name suggests.

Hence the serious workflow, which runs over the same sequence twice: average for the clean background and pinpoint stars, maximum to cut out only the trails you have chosen to keep. Then the two images are blended. It is a composite in the full sense of the word, and the caption says so.

How many you will actually get

The ZHR you read in headlines is a laboratory figure. The IMO defines it as the hourly count for an ideal observer with the radiant at the zenith under a sky of limiting magnitude 6.5: the Perseids are worth 100, the Geminids 150. Neither number will ever reach your internal storage, and there are three cuts along the way.

The first is geometric. Observed rates scale with the sine of the radiant’s altitude, so with the radiant at forty-five degrees you keep barely seventy per cent. The second is your sky: the IMO correction divides by the shower’s population index raised to the difference between 6.5 and your local limiting magnitude, and with an index around 2.2 under a 5.5 sky that more than halves the count again. The third is framing. A 24 millimetre equivalent lens covers roughly seventy-four degrees by fifty-three, a little over a sixth of the celestial sphere, and meteors outside the frame do not exist.

Thirty meteors an hour overhead therefore become five inside the frame, and of those five the sensor holds on to the brightest. A well organised Perseid night yields a handful of good trails out of four hundred exposures. That streaked sky you saw going round social media is almost always a composite of several nights, often several cameras.

Where to point, and when

The radiant is pure perspective: the particles travel on parallel paths, and their tracks appear to diverge from a point the way rails appear to meet at the horizon. Aiming at it is the instinctive move and the worst one. Meteors there come almost straight at you, foreshortening squashes the trails, and you are left with two degree dashes. Ninety degrees away the tracks are gloriously long, but the density per square degree falls off. The compromise sits thirty or forty degrees to one side, with the framing well above forty degrees of altitude, where the atmosphere absorbs less.

The timing is not up for negotiation. What matters is the radiant’s altitude, and for nearly every shower it climbs through the second half of the night. The Perseid radiant sits at declination +58, so from most of Europe it never sets, yet in the evening it skims low across the northern horizon and only culminates towards dawn, which is precisely where that sine starts working for you. Then there is the Moon, which on its own outweighs everything else. A half lit disc still high during the peak strips out the faint meteors as efficiently as a street lamp planted ten metres from your tripod.

The boring part that decides the night

Wide angle, always: a tele narrows the field and the odds collapse in proportion. Focus locked manually on the stars and never touched again. White balance fixed, or three hundred frames drifting in colour temperature will ruin the average and the maximum alike. Six to ten seconds depending on the lens, ISO high enough to keep the sky background above the read noise, and the interval between frames trimmed to whatever minimum the app allows.

Then the logistics, the real enemy of shower nights. Four hundred RAW files at twenty-odd megabytes each come to eight gigabytes, which is why JPEG remains a defensible choice on a night whose purpose is counting trails. Three hours of continuous shooting will flatten a battery long before dawn, so bring a power bank and a short cable, and keep an eye on the dew, which always picks the moment the sequence is going well.

Meteor, satellite or aircraft

The first pass through the files holds a different surprise: satellites outnumber meteors, and by a wide margin. Three signatures separate them. A meteor lives in one frame only, its trail brightening, peaking and dying out. A satellite turns up in consecutive frames, always in line, with flat steady brightness, and under a maximum stack it rebuilds a straight line across half the picture. An aircraft gives itself away: the trail is dashed by blinking anti-collision lights, and often carries two colours.

For one night the trade reverses. Everything you have learned says to add, to average, to keep what repeats; a meteor shower asks the opposite, which is to save the single frame where something happened and let the other three hundred and ninety-nine go. Do that, then go back to stacking. The Milky Way in the background is still waiting for its six hundred exposures.