Everything on a dash cam spec sheet, and what each number is actually for

Every dash cam is sold on its resolution. Resolution on its own tells you almost nothing, because it says how many pixels the camera records and not how much of the world it spreads them across. A lens that takes in 170 degrees and a lens that takes in 125 degrees both produce a 3,840-pixel-wide file; the first one is describing 45 more degrees of scene with the same pixels.

Divide one by the other and you get the figure that actually predicts whether a licence plate stays legible: horizontal pixels per degree of view. Across the 53 cameras here it runs from 13.7 to 27.4, with a median of 22.6. The sharpest is Nexar One; the bluntest is Garmin Dash Cam X110. Both are sold as high-definition cameras.

Two honest caveats. First, almost every manufacturer publishes a diagonal field of view, not a horizontal one. The handful who publish both — BlackVue is the most consistent about it — show the horizontal figure running 20 to 25 degrees narrower. Dividing horizontal pixels by a diagonal angle is therefore not an optical measurement; it is a comparison index that is conservative by roughly the same amount for every camera, which is exactly what makes the ranking hold.

Second, it says nothing about whether those pixels are any good. Sensor size, aperture, HDR and bitrate decide the night result, and a camera with a high figure behind a slow lens loses to one with a lower figure behind a fast one once the sun is down. Read it as “how far away can this camera resolve detail in good light” and read the sections below for the rest.

25+ px/°

Plate-legible at distance

19+ px/°

Plate-legible a few lengths ahead

16+ px/°

Plate-legible close in

under 16 px/°

Establishes the scene, not the plate

The front lenses in this catalog run from 125° on the Thinkware Q200 2CH to 170° on the Rexing V1-4K. That 45-degree spread is the single biggest reason two cameras with the same resolution behave completely differently.

Wide is genuinely better for context. A 170-degree lens catches a car entering from a side street, a cyclist coming off a sidewalk, and most of a parking lot — things a 140-degree lens clips out of frame entirely. If the incidents you worry about happen at intersections or while parked, that coverage is worth buying.

Wide is genuinely worse for detail, for the arithmetic in the section above. It is also worse at the edges in a way the numbers do not capture: a very wide lens bends straight lines and stretches anything near the frame edge, so the plate that is technically in shot at 80 degrees off-centre is smeared anyway.

Rear lenses skew wider still — 112° to 165° here — which is why most rear channels resolve so little. The exceptions are worth knowing about: Miofive S1 Ultra pairs a narrow rear lens with a high-resolution sensor and gets 34.3 pixels per degree out of the rear channel, against a median of 16 across every rear camera here.

29 of the 53 cameras here record a 3,840-pixel-wide file. Not all of them capture one. 2 write a 4K file from a 5-megapixel sensor whose native output is 2,592 pixels wide, with the extra pixels added in processing rather than collected by the lens. That is not a defect and the manufacturers are not always hiding it — one of them puts the word “upscaled” in the product title — but it means the optical detail sits closer to a good 1440p camera than to a native 4K one.

Where a maker upscales, this catalog records the sensor’s real horizontal resolution rather than the file’s, so the pixels-per-degree figure describes what the lens actually resolved. The resolution field says “(upscaled)” on those entries.

The other thing to know is that resolution is a per-channel setting, not a property of the camera. Nearly every multi-channel camera drops the rear or interior channel when the front one is at maximum, and several offer a higher frame rate at a lower resolution. The figures in this catalog are the maximum front-channel mode.

9 cameras here record the front channel at 60fps rather than 30. It sounds like a video-smoothness feature and it is not — it is a plate-legibility feature, and for a specific case it beats resolution.

A car crossing in front of you at an intersection is moving sideways across the frame. At 30fps its plate smears across the exposure of each frame; at 60fps each frame carries half as much motion blur, so there is a decent chance one of them is sharp. For a car you are following, where relative motion is small, the frame rate barely matters and resolution does.

The complication is that HDR and 60fps are usually mutually exclusive, and the menu makes you choose. Cameras here range from 30fps to 60fps on the front channel, and the ones offering the choice put it in the recording-mode menu rather than deciding for you.

Sony’s STARVIS family dominates this category and the part number is worth reading. 29 of the 53 cameras here name a STARVIS 2 sensor — usually the 8-megapixel IMX678 for a 4K channel or the 5-megapixel IMX675 for 1440p. STARVIS 2 roughly doubles low-light sensitivity over the original generation and widens dynamic range, which is the difference between a headlight being a white blob and being a car.

12 cameras here do not publish a sensor part number at all. That is not automatically a bad camera, but it does mean a 4K claim cannot be checked against anything — you cannot tell whether an 8-megapixel sensor is behind it or a 5-megapixel one being upscaled. When a maker publishes the part number, the specification is verifiable; when they do not, you are taking the marketing at its word.

Sensor size matters as much as the part number and is published even less often. A 1/1.8-inch sensor collects substantially more light than a 1/2.8-inch one at the same pixel count, which is why the IMX678 cameras here generally beat the IMX675 ones at night despite both being STARVIS 2.

The f-number is how much light the lens lets through, and on an unlit road it does more for the result than the pixel count does. The spread here runs from f/1.3 to f/1.9, which sounds narrow and is not: f/1.3 passes roughly twice as much light as f/1.9.

The practical consequence is that the ranking by pixels per degree and the ranking by night performance are different rankings. A 4K camera at f/1.8 and a 1440p camera at f/1.6 will trade places depending on how dark the street is. Where two cameras use the same sensor and differ only in aperture — which happens within single manufacturers’ own ranges — the faster lens is the better night camera, full stop.

A fast aperture also has a cost nobody mentions: less depth of field, so a camera focused for the road ahead is slightly softer on something very close. On a dash cam focused at infinity this matters almost never, which is why every maker chases the fastest lens they can afford.

High dynamic range combines differently-exposed captures so that a scene containing both a bright area and a dark one comes out with detail in each. On a dash cam the scene that needs it is the mouth of a tunnel, a low winter sun, or a car in shadow against a bright sky — not a dark road, where there is nothing bright to balance against.

That distinction is the one people get wrong. HDR does not make a dark road brighter; aperture and sensor size do that. What HDR fixes is the plate that is invisible because it is in the shadow of the car above it while the sky behind is blown out.

Most multi-channel cameras apply HDR to the front channel only, and the ones that run it on every channel independently say so — it is worth checking, because a rear camera pointed at headlights is exactly where the dynamic range problem lives.

A camera can record 4K and still throw away the detail, because the file has to be compressed and the bitrate decides how hard. Compression artefacts land worst on exactly the content a dash cam exists to capture: small high-contrast detail moving across the frame, which is a description of a licence plate.

Very few makers publish a bitrate, which is why this catalog has the field and rarely fills it. The proxies you do get are the codec and the maximum card size. H.265 packs the same quality into roughly half the file of H.264, which is why some 4K cameras give a usable loop on a 256GB card and others do not.

If a camera lets you choose a bitrate in the menu, the high setting is nearly always the right one. The storage it costs is cheap next to a plate you cannot read.

A dash cam only needs stored power for one job: closing the file it is writing when the car cuts power. There are two ways to do it, and they age completely differently.

A lithium cell holds a lot of energy and hates heat. A car parked in summer sun can reach well over 140°F inside, and a pouch cell held at that temperature swells, distorts the case, and eventually stops the camera — the single most common way a windshield-mounted camera dies. A supercapacitor stores far less energy, enough only to finish the file, and tolerates a much wider temperature range without degrading.

46 of the 53 cameras here use a supercapacitor. The published temperature ceilings run from 140°F on the Vantrue Nexus 4 Pro to 176°F on the Rexing V1-4K. 9 cameras publish no operating range at all, which is worth weighing: a supercapacitor camera whose maker will not state a ceiling is asking you to take the claim on trust.

The practical consequence for a battery-powered camera is not that it will fail tomorrow. It is that it is a consumable in a hot climate and a durable good in a temperate one, and the price does not reflect the difference.

“24-hour parking mode” is on nearly every box in this category and means at least four distinct behaviours. The question that actually matters is whether the camera records the approach or only the impact, because a clip that starts when the bumper is already touching yours tells you very little about who was driving.

Radar-buffered 3 of 53 cameras

Radar notices movement near the car and wakes the camera before anything touches it, so the approach is recorded without leaving the camera running all night.

Buffered 35 of 53 cameras

Keeps a rolling buffer, so the seconds before an impact are saved as well as the impact itself. It is the most useful parking mode and the hungriest.

Time-lapse / motion 15 of 53 cameras

Records continuously at a reduced frame rate, or only once motion or an impact is detected. You get context, at one to three frames a second.

None 0 of 53 cameras

No parking recording. The camera stops when the engine does.

The second question is how long it lasts. A buffered mode keeps the camera awake and drains a car battery in a day or two; owners of one radar-equipped camera here measure about ten percent of a battery overnight even in its energy-saving mode. The newest answer is a deep-sleep mode that wakes on impact and draws almost nothing, which turns standby from hours into days or weeks — at the cost of missing the first moment of the event.

A dash cam plugged into a 12V socket records while the engine runs. That is all. Any form of parking coverage needs power when the ignition is off, which means one of three things: a hardwire kit tapped into the fuse box, an OBD-II power cable, or a dedicated battery pack.

Only 7 of the 53 cameras here include the cable. Everyone else sells it separately, which is the cost people leave out of the comparison — and kits are not interchangeable even within one manufacturer’s range, because the connector changed from micro-USB to USB-C partway through.

A hardwire kit has three leads: constant power, a switched accessory feed so the camera knows when the engine stopped, and a ground. Getting the first two the wrong way round is the single most common install fault, and it produces exactly the symptoms people blame on the camera — parking mode engaging while driving, or never engaging at all. A cheap test light finds the right fuses in five minutes.

The kit’s low-voltage cutoff is what stands between parking mode and a car that will not start. Set it as high as your battery tolerates; a cutoff that lets a lead-acid battery sag toward 11.8 volts repeatedly will shorten its life even when the car does start.

Loop recording rewrites the same flash cells continuously, every hour the camera runs. A standard memory card is rated for occasional photo storage and wears out under that duty cycle in months. High-endurance cards use different flash designed specifically for surveillance video, and they are the single most important accessory purchase.

The reason this matters more than it sounds is that a failing card presents as a broken camera. Recordings that stop after ten seconds, double beeping, files that will not play back, a camera demanding a format on a card that worked yesterday, weeks of footage that silently never recorded — all of these are card symptoms, and all of them get blamed on the camera first.

Card ceilings here run from 128GB to 1024GB, and 1 camera uses built-in eMMC storage with no removable card at all — which removes the failure mode entirely and also removes the card you would otherwise hand to an insurer.

Two habits are worth forming. Format the card in the camera rather than on a computer, and open the most recent files once a month to confirm the timestamps are current. The worst version of a card failure is the silent one you discover the day you need the footage.

52 of the 53 cameras here have a GPS receiver built in; the rest take an optional module, usually integrated into the windshield mount, and a handful have none. What it records is a location and speed track alongside the video, stamped into the file or a companion log.

That track is the difference between a clip that shows a collision and a clip that shows a collision at 34mph in a 35 zone at a specific intersection. It is also the feature most likely to be half-installed: on several cameras the GPS lives in the mount rather than the body, so hardwiring the power cable directly into the camera instead of through the mount silently disables it.

Higher-rate receivers (10Hz rather than 1Hz) log ten positions a second instead of one, which matters for reconstructing a short, fast event. Most cameras do not publish the rate.

Almost every camera here creates its own Wi-Fi network that a phone joins to pull clips. The band decides whether that is a thirty-second job or a coffee break: 9 of these cameras are 2.4GHz only, and moving a few minutes of 4K footage over 2.4GHz is genuinely slow enough that owners give up and use a card reader.

5GHz — and Wi-Fi 6 on the newest models — is roughly an order of magnitude faster in practice, and it is the feature that makes a 4K camera usable rather than theoretical. If a camera records 4K and has a 2.4GHz radio, the resolution is mostly decorative unless you are happy pulling the card.

The known conflict: the camera’s 2.4GHz radio shares spectrum with Android Auto and Bluetooth, and in some cars enabling one drops the other. The usual workaround is to leave the camera’s Wi-Fi off and switch it on only when you want a clip.

13 cameras here support LTE, either built in or through a module, and 13 of them gate features behind a paid plan — remote live view, cloud clip backup, push alerts when a parked car is hit, GPS tracking. Those are genuinely useful and they are not included in the price on the box.

The pattern worth understanding is that LTE is usually a module you buy separately and a plan you pay for monthly, on top of a camera that already costs more than its non-connected sibling. On several cameras here, none of the marketed cloud features work at all without both.

The alternative that costs nothing: several cameras will join an existing Wi-Fi network — a car’s built-in hotspot, or a phone’s — and upload over that instead. It only works while the network is present, which is fine for uploading a drive and useless for watching a parked car.

Your dashboard is reflected in your windshield. On a bright day that reflection is in the same frame as the road, and it is the most common reason daytime dash cam footage is unusable. A circular polarising filter blocks light arriving at the angle a windshield reflection arrives at, and removes most of it.

It costs about thirty dollars, screws onto the lens, and has to be rotated once to the right angle. 10 of the 53 cameras here include one in the box; one builds the polariser into the lens assembly so there is nothing to buy, fit or align. Everyone else sells it as an accessory, which is why most cameras on the road do not have one.

The cost is about a stop of light, so a CPL makes night footage slightly darker. Most owners leave it on anyway, because the daytime improvement is dramatic and the night penalty is small.

A cabin camera without infrared is close to useless after dark. The interior of a car with the dome light off has essentially no light in it, and no sensor can expose what is not there. 11 cameras here carry infrared LEDs, which illuminate the cabin at a wavelength people cannot see and turn it into usable monochrome without distracting the person sitting in it.

Interior channels are almost always the lowest-resolution camera in a multi-channel system, typically 1080p and occasionally less. That is a defensible choice: the job is to identify who was in the car and what they did, at two feet, not to read anything at distance.

Audio is the part that catches people out. Recording audio inside a vehicle is governed by state consent law, not by camera law, and the rules differ substantially between states. Every camera here lets you turn the microphone off; none of them help you work out whether you should.

17 of the 53 cameras here have no screen at all. That is a design choice with two real benefits: nothing glows in your peripheral vision at night, and nothing announces from outside the car that there is a camera worth breaking a window for. The cost is that aiming the camera and checking a clip both require a phone.

Mount type matters more than it looks. Adhesive mounts hold better in heat and are effectively single-use — peeling one off to move cars usually means buying a new pad. Suction mounts make a camera portable between vehicles and are the ones that let go on a hot windshield. Magnetic click mounts are the best of both and are rare: one manufacturer here makes taking the camera indoors a one-motion job, which is the only theft defence people reliably keep doing.

On placement, high and central behind the rearview mirror is both the best optical position and the one that keeps you inside most state windshield-obstruction rules.

There is no federal rule about dash cams. What does exist is state law about obstructing a windshield, and it was written for things like air fresheners and parking permits rather than cameras. The shapes it takes:

• States that prohibit windshield attachments outright, with narrow exceptions. California and Minnesota are the ones usually cited. Mounting is still possible in a defined area rather than anywhere.
• States that define a permitted area — typically a small square in one of the lower corners, or a band behind the rearview mirror. This is the most common form.
• States with a general obstruction standard and no measurements, so it comes down to an officer’s judgement about your view of the road.

Mounting high and central, tucked behind the mirror, satisfies nearly every version of this and is also where you want the camera optically. The genuinely risky area is a suction mount low on the driver’s side.

Audio is a separate question with separate law. Recording conversations inside the cabin engages state wiretapping and consent statutes, some of which require every party to consent. It is the part of dash cam ownership most likely to cause a real problem and the part nobody reads about.

This is general information rather than legal advice, and it changes. Check your own state’s vehicle code and its consent statute before you install or before you leave the microphone on.