Monday, September 7, 2026

What 2K and 4K Actually Mean on a Security Camera Box

What 2K and 4K Mean on a Camera

The Numbers on the Box Are Marketing Labels, Not Standards

For most people the right choice is the frame shape and the upload speed, not the badge. Choose 1080p at a front door and save 4K for a gate or a driveway. A 1080p camera captures 1920 by 1080 pixels, which is 2.07 megapixels, and needs about 2 Mbps of upload where a 4K model can want 15 Mbps.

Only 1080p and 4K have settled meanings, and even 4K bends. Ring lists its 4K doorbell as 2880 by 2880 pixels in a square frame, which is not the 3840 by 2160 shape a television calls 4K.

The Same Pixel Count Can Be Cut Into Very Different Shapes

Shape Beats Label
  • ● 1080p - 1920 x 1080, 2.07 MP
  • ● Ring 1536p - square, taller frame
  • ● Ring 4K - 2880 x 2880, 8.29 MP

Camera makers stopped competing on width once they realised doorbells look at people, not landscapes. A tall frame catches a face and a parcel on the step in one view.

Ring calls its 1536p models head to toe, and that phrase describes a shape rather than a sharpness. The extra pixels went into vertical coverage instead of a wider picture.

So two cameras with similar megapixel counts can frame a doorway completely differently. Reading the pixel dimensions rather than the marketing name is the only way to see which.

Label you see Typical pixel dimensions Megapixels Frame shape What it suits
1080p Full HD 1920 x 1080 2.07 Widescreen 16:9 A hallway or a wide driveway
2K or QHD 2560 x 1440 3.69 Widescreen 16:9 Recognising faces mid range
2K, alternative reading 2304 x 1296 2.99 Widescreen 16:9 The same job, fewer pixels
Head to toe 1536p 1536 x 1536 2.36 Square 1:1 Doorways and parcels on the ground
4K on a doorbell 2880 x 2880 8.29 Square 1:1 Doorways where you want to crop in
4K UHD 3840 x 2160 8.29 Widescreen 16:9 Gardens, gates, long approaches

The two 2K rows are the honest part of this table. Both appear on shelves under the same badge, and one carries about a quarter fewer pixels than the other.

Resolution Is a Bandwidth Decision Before It Is a Picture Decision

What It Costs Upstream
  • ● Ring Battery Doorbell - 2 Mbps
  • ● Ring 4K - 15 Mbps recommended
  • ● Nest Cam max video - 3300 Kbps

Every cloud camera uploads what it records, so resolution spends your slowest connection. That is the constraint most buyers discover after installation.

Ring lists a minimum upload speed of 2 Mbps for its Battery Doorbell and recommends at least 15 Mbps for its 4K devices, calling 20 Mbps or higher ideal. That is a sevenfold jump for one step up the range.

Google publishes per model ceilings instead. Nest Cam battery and wired models are listed at 700 Kbps on high and 1000 Kbps at maximum video quality, while the wired third generation indoor camera and doorbell reach 3300 Kbps.

Multiply that by the number of cameras before you shop. Four cameras at maximum quality on a household with 10 Mbps upstream will fight each other and your video calls.

Night, Lens, and Compression Undo a Higher Number

Where Pixels Are Lost
  • ● Infrared cuts detail after dark
  • ● Wide lenses spread pixels thin
  • ● Bitrate caps flatten motion

Resolution describes the sensor grid, not the light reaching it. After dark, infrared illumination and heavy noise reduction flatten fine detail, which is why a 4K night clip rarely looks four times better than a 1080p one.

The lens spreads those pixels too. A 140 degree field of view, as Ring lists for its 4K doorbell, covers more scene with the same grid, so pixels per metre at the gate fall as the angle widens.

Compression finishes the job. Cameras hold a bitrate ceiling to protect your network, and fast motion in a high resolution stream is exactly where that ceiling shows as blocky smearing.

A useful habit is to judge a camera on a still frame of a moving person at your own distance. That single test tells you more than the box.

The Professional Trade Measures Pixels Per Metre, Not Pixels

Security installers stopped quoting resolution years ago because it says nothing without distance. The measure they use is pixel density, the number of pixels covering one metre of the scene at the point you care about.

That approach is written into an international standard. IEC 62676-4 sets out pixel density levels for detecting, observing, recognising and identifying a person, and a 2025 edition revises those levels for higher resolution digital cameras and the effects of compression and noise.

The practical translation is simple. The same camera identifies a face at the doorstep, recognises a person on the path, and merely detects a shape at the end of the garden.

So the question is never how many pixels the sensor has. It is how many of them land on the spot where you need to tell one person from another.

Higher Resolution Changes What Storage Costs You

Local recording turns pixels into gigabytes, and the maths is unforgiving. Doubling the pixel count roughly doubles the file size at the same settings and length.

That surfaces in two places. A microSD card fills sooner and starts overwriting older clips, and a network video recorder needs more drive space for the same number of retention days.

Retention days are the number to watch rather than card size. A card that held three weeks of 1080p clips may hold barely one week once the same camera records at a higher resolution, and the loss is silent.

Event recording softens the blow, since only motion clips are stored. Google lists typical data use for a Nest Cam on High at 200 GB a month with 24/7 recording, against 10 GB with event recording. Continuous recording is where resolution and retention collide hardest, and where a recorder drive fills faster than anyone expects.

Cloud plans hide the trade behind a monthly fee, so the cost appears as a subscription rather than a card. Our guide to security cameras with no monthly fee covers where that line falls.

Where the Extra Pixels Genuinely Pay Off

There is one case where a big number is worth the bandwidth, and it is digital zoom. Cropping into a recorded clip to read a number plate or a face is only possible if the pixels existed at capture.

Distance is the trigger. A camera watching a gate 15 metres away spends most of its resolution on empty ground, so a 4K sensor leaves something to crop.

At a front door the opposite is true. The subject fills the frame already, and a taller aspect ratio does more for you than four times the pixels.

What the Spec Sheet Should Tell You and Often Does Not

Ask for pixel dimensions rather than a badge, since 2K without numbers is not a specification. Any maker who will not publish the grid is telling you something.

Then look for the field of view in degrees and the aspect ratio together. Those two figures decide how much of your doorway or driveway a given resolution actually covers.

Finally check the network requirement in the support pages rather than the product page. Ring and Google both publish theirs, and the numbers vary by model in ways the marketing never mentions. If you are choosing between camera types first, our comparison of wired and wireless home security cameras covers that decision.

Which Resolution Fits Your Camera Position

Front door with a doorbell camera: A square or tall 1536p frame beats a widescreen 2K one. You want head to toe coverage of a person on the step, and that shape delivers it at a fraction of the bandwidth.

Driveway or gate at distance: This is the case for 4K, provided your upload can hold 15 Mbps. Cropping into a plate or a face after the fact is the only feature that truly needs the pixels.

Hallway or nursery indoors: 1080p is enough, and the saved bandwidth is worth more than the detail. Indoor scenes are close, well lit and small, which is the easiest job a sensor gets.

Rented flat with shared or slow internet: Choose the lowest resolution that identifies a face at your door, then cap quality in the app. Nest cameras adapt bitrate to available bandwidth automatically, so the setting mostly limits the worst case.

Multiple cameras on one connection: Budget bandwidth as a household total, not per camera. Two 1080p cameras and one 4K camera can exceed a typical upstream link during a busy evening.

Reading a Camera Box Without the Marketing Words

Three lines matter and none of them is the badge. Pixel dimensions, aspect ratio, and the recommended upload speed together describe what you will actually see and whether your connection can carry it.

Resolution is the easiest number to print and the least useful in isolation. A well positioned 1080p camera at the right height beats a 4K camera aimed at the sky, every time.

If you are still shortlisting hardware, our roundup of home security cameras sets out which models publish those figures plainly, and confirm current pricing on the official site before you order.

FAQ

Is 4K on a doorbell the same 4K as on a television?

No. Standard 4K video is 3840 by 2160 pixels in a widescreen shape, while Ring lists its 4K doorbell as 2880 by 2880 in a square 1:1 frame. Both land near 8.3 megapixels, but the square version trades width for height at the door.

What resolution is 2K on a security camera?

It is a shorthand for roughly 1440 vertical pixels, usually 2560 by 1440. There is no certification behind the label, so some cameras sold as 2K use 2304 by 1296 or a square sensor of similar pixel count instead.

What limits video quality more, the camera or my internet?

Upload speed, almost always. Ring lists a 2 Mbps minimum for its Battery Doorbell and recommends at least 15 Mbps for its 4K devices, so a higher resolution camera can saturate a slow connection that handled the old one.

Does a taller aspect ratio beat a higher resolution?

For identifying a face at the door, height matters more than width, which is why square and 3:4 sensors have spread. For watching a driveway or a garden, a wide frame at lower resolution usually beats a tall frame at higher resolution.

Will my camera lower its own quality if the network is busy?

Sometimes, and the camera decides. Google notes that Nest cameras adjust video bitrate automatically based on available bandwidth, so a congested network shows up as softer video rather than an error message.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.

What Pa Suction Numbers Tell You About a Robot Vacuum

What Pa Suction Really Means

The Number Is Not a Cleaning Score

Start Here
  • ● Pa measures sealed pressure, not pickup
  • ● Hard floors are served by 8,000 to 11,000 Pa
  • ● No shared test standard exists

For hard floors and thin rugs, about 8,000 Pa is enough. Spend the difference on brush design and navigation instead. For medium or thick carpet with pets in the house, 15,000 Pa and above starts to matter. Flagship models now advertise far more than that, with the Roborock Saros 20 listed at 36,000 Pa and the eufy Omni E28 at 20,000 Pa for $1,099.99 as of September 2026.

Those numbers are real measurements of something. They are just not measurements of how clean your floor ends up.

A pascal rating describes pressure at a sealed intake. Cleaning depends on pressure, airflow, brush contact, and the seal between the machine and your floor, and only one of those four appears on the box.

What a Pascal Measures When the Nozzle Is Sealed

A pascal is a unit of pressure. In vacuum specifications it describes how much lower than room pressure the fan can pull the air inside the machine when the intake is blocked.

Blocking the intake is the important detail. With nothing flowing, the fan reaches its maximum pressure difference, and that peak is the figure printed on the box.

In your living room the intake is never sealed. Air rushes through, pressure drops, and the useful work comes from the combination of pressure and airflow rather than from pressure alone.

This is why two machines with identical pascal ratings can behave differently on the same rug. The one with a wider air path and better floor contact moves more air at the same pressure.

Why the Figures Tripled in Three Years

The specification climbed because it is easy to print and easy to compare. A buyer scanning a listing page can rank 10,000 against 22,000 in a second, and no other robot vacuum spec offers that.

Roborock alone illustrates the spread across its own current range. The Q7 and Q10 series list 10,000 Pa, the Qrevo Slim 11,000 Pa, the Qrevo Curv 18,500 Pa, the Qrevo Edge 2 25,000 Pa, and the Saros 20 reaches 36,000 Pa.

Dreame advertises 35,000 Pa at the top of its X60 range, and eufy quotes 20,000 Pa on the Omni E28 against 8,000 Pa on the older X10 Pro Omni. The whole industry moved in the same direction at the same time.

Some of that gain is genuine engineering in fan and duct design. Some of it is a number chosen because rivals were printing bigger ones, and no outside body checks either claim.

Published Pa Ratings and What They Cost

The Numbers on Sale Now
  • ● Roborock Saros 20 - 36,000 Pa
  • ● eufy Omni E28 - 20,000 Pa at $1,099.99
  • ● iRobot publishes no Pa figure

Figures below come from the manufacturers own product listings, in US dollars, with prices as of September 2026. Confirm current pricing on the official site, since robot vacuum prices move constantly and discounts are frequent.

Model Stated suction Listed price What the number is useful for
Roborock Q7 and Q10 series 10,000 Pa not listed on the range page Hard floors, thin rugs, light debris
Roborock Qrevo Slim 11,000 Pa not listed on the range page Low clearance furniture, hard floors
eufy X10 Pro Omni 8,000 Pa $899.99 Hard floors with occasional rug work
eufy Omni E28 20,000 Pa $1,099.99 Mixed flooring with pets
Roborock Qrevo Edge 2 25,000 Pa not listed on the range page Medium pile carpet, embedded hair
Roborock Saros 20 36,000 Pa not listed on the range page Thick pile, headline specification
iRobot Roomba Max 705 Combo no Pa figure published $1,299.99 Comparison on other specs only

Read the last column before the second. The jump from 8,000 to 20,000 Pa changes what the machine can do on carpet, while the jump from 25,000 to 36,000 Pa mostly changes the marketing.

Notice also how little price tracks suction. The $1,299.99 Roomba prints no pascal figure at all, and the $899.99 eufy prints the lowest one in the table.

The Brand That Refuses to Print a Number

iRobot sells one of the most recognised robot vacuum lines in the world and publishes no pascal ratings. Its product pages describe suction in words, and the specification simply is not there.

That absence is informative. One of the best-known brands in the category decided the number was not worth quoting, which tells you how much weight the industry itself places on it.

It also creates a practical problem. You cannot compare a Roomba with a Roborock on suction, so the comparison has to move to brush type, mopping design, obstacle handling, and dock behaviour.

Our comparison of lidar against camera navigation covers the spec that actually differs between these machines day to day. A robot that maps your rooms well cleans more of the floor than a robot that pulls harder over a smaller area.

Suction Is One of Four Things That Pick Up Dirt

Brush design does the heavy lifting on carpet. A rubber roller that reaches the fibres lifts embedded hair that no amount of static pressure would move on its own.

Airflow carries what the brush loosens. A wide, unobstructed path from nozzle to bin matters more than peak pressure once debris is airborne.

The floor seal decides how much of that pressure reaches the dirt. Ground clearance, side skirts, and wheel height determine whether the machine pulls on your carpet or on the surrounding air.

Filter and bin condition quietly cap all three. A loaded filter chokes airflow, which is why a machine that cleaned well in month one often disappoints in month six with its spec sheet unchanged.

Peak Suction Is a Short Burst, Not a Setting You Leave On

Advertised figures describe the strongest mode the machine offers. That mode spins the fan at full speed, which drains the battery quicker and is loud.

Most people discover this in week two. The robot gets scheduled for the middle of the day on a quieter setting, and the headline number stops being part of daily life.

Battery limits reinforce it. A large home on maximum power will need a recharge partway through, and the machine will pause, return to the dock, and resume later.

So compare the mode you will actually use. If your routine is a daily quiet pass with a weekly deep clean, the mid setting matters more than the peak, a pattern our notes on self-emptying bases touch on from the maintenance side.

Which Suction Level Fits Your Floors

Match Spec to Floor
  • ● Hard floors - 8,000 to 11,000 Pa
  • ● Pet hair on carpet - 15,000 Pa and up
  • ● Thick pile - suction plus brush design

Hard floors throughout with a couple of thin rugs: 8,000 to 11,000 Pa is plenty. Spend the difference on navigation quality and a dock that fits your space.

Mixed flooring with one shedding pet: 15,000 to 20,000 Pa. This is the band where hair on low and medium pile stops being a repeat job, and where the eufy Omni E28 at $1,099.99 sits.

Thick pile carpet or two or more pets: 20,000 Pa and above, paired with a rubber roller rather than bristles. The brush is doing at least as much work as the fan here, which is the theme of our guide to robot vacuums for pet hair in small apartments.

Hardwood with a fine dust problem: Suction is not your bottleneck. Sealing, filtration, and how often the bin fills matter more, as our notes on robot vacuums for hardwood floors set out.

Homes where a person still vacuums weekly: Buy for daily maintenance rather than deep cleaning. A modest robot that runs every day beats a powerful one that runs on Sundays, and our comparison with a cordless stick vacuum covers how the two split the work.

How to Read a Spec Sheet Without Getting Sold

Treat the pascal figure as a rough tier rather than a score. Under 10,000, around 10,000 to 15,000, and above 20,000 are meaningfully different, and 22,000 against 25,000 is noise.

Then look for the specifications brands print less loudly. Brush material, whether the roller is anti-tangle, ground clearance, bin volume, and filter type all shape the result more than a peak pressure reading.

Check the mopping design separately if you want mopping. Pressure applied to a pad and how the pad is washed have nothing to do with vacuum suction, and many buyers conflate the two.

Finally, ignore any comparison table that ranks brands purely on pascals. Without a shared test method the ranking measures marketing appetite, not cleaning.

What to Compare When Two Models Tie

Start with consumables. Filter, brush, and mop pad prices over three years often differ more between models than the purchase price does.

Look next at how the machine handles failure. What it does when it beaches on a rug edge, loses the map, or finds the dock blocked will shape your experience far more than the fan.

Then check app behaviour and update history. A robot is a connected appliance, and the brand that still ships firmware for a three-year-old model is worth more than one extra suction tier.

For a Roomba, that support now comes from a different owner. iRobot filed for Chapter 11 on December 14, 2025, and on January 23, 2026 became a privately held company wholly owned by Picea, its main contract manufacturer. When it filed, iRobot said it expected no disruption to app functionality or ongoing product support.

If two machines still tie after that, buy the cheaper one and put the difference toward a spare filter set. The pascal gap between them will not be the thing you notice in a year.

FAQ

What does Pa mean on a robot vacuum?

Pascals measure the pressure difference the fan can create when the intake is sealed. It is a measure of pulling force, not of how much dirt the machine collects, because collection also depends on brushes, airflow, and how well the base seals against your floor.

Do I need 20,000 Pa or more?

For hard floors and thin rugs, no. Models around 8,000 to 11,000 Pa handle dust and crumbs on hard flooring, and the higher figures earn their keep on medium and thick pile carpet where hair works its way below the surface.

Are Pa ratings tested the same way across brands?

There is no shared industry test behind the figure. Each brand measures in its own lab under its own conditions, so a 22,000 Pa rating from one company and a 20,000 Pa rating from another are not directly comparable.

Why does iRobot not list a Pa number?

iRobot does not publish pascal figures for Roomba models and describes suction in words instead. That is a marketing choice rather than a sign of weakness, and it does mean you cannot line a Roomba up against a Roborock on this spec.

Does the vacuum run at its maximum Pa the whole time?

Usually not for long. Peak suction runs the fan at full speed, which drains the battery faster and is loud enough that most people schedule it away from working hours, so the advertised figure describes a short burst rather than a whole clean.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.

Sunday, September 6, 2026

What ONVIF and RTSP Mean on a Home Security Camera

ONVIF and RTSP

ONVIF Is a Handshake and RTSP Is the Video

Two different jobs hide behind these acronyms. ONVIF lets a camera and a recorder discover each other and agree on settings, and RTSP carries the video stream itself.

For a home buyer the practical rule is short. If you ever want the camera to record somewhere other than the maker app, look for a registered ONVIF profile and a documented RTSP path, usually on TCP port 554.

Profile T is the one to want now. ONVIF ends support for the older Profile S on 31 March 2027, and Profile T covers virtually the same features with better authentication.

What the Profile Letters Actually Promise

Read the Letter Not the Logo
  • ● S and T - live streaming, T adds H.265
  • ● G - recording and local storage
  • ● A, C, D - access control, not cameras

An ONVIF profile is not a quality rating. It is a fixed set of features that a conformant device and a conformant client must both support, which is what makes two brands work together.

The letters split into two worlds. Some cover video systems and some cover access control, and a camera listing that quotes an access control profile is telling you almost nothing useful.

Profile What it covers Why a homeowner cares
S Video streaming, the original video profile Widest device support today, retiring in 2027
T Video streaming and playback, adds H.265 and analytics events The profile to buy now, replaces S
G Video recording and local storage Needed when the camera or recorder stores footage on site
M Metadata and analytics, video and access control Person and vehicle events passed to other software
A and C Access control systems Door controllers and credentials, not cameras
D Access control peripherals, also covers video Doorbell and entry hardware in mixed systems

Devices can hold more than one profile at once. ONVIF gives the example of a network camera with local storage conforming to both Profile T and Profile G.

That combination is the one to look for if you want footage kept on a card or a recorder rather than in a subscription. Our guide to a home security camera without a monthly fee covers the storage side of that decision.

Why Profile S Is Being Retired in 2027

ONVIF has published an end date for Profile S, and the reason is security rather than age. The authentication mechanisms specified in Profile S are no longer consistent with current cybersecurity recommendations.

The mechanics matter if you are buying this year. After 31 March 2027, manufacturers cannot submit new products, or older products carrying new firmware, for Profile S conformance.

There is an earlier deadline hiding inside that one. The June 2026 version of the ONVIF conformance test tools carries a 9-month validity period, and it is the last version that lets a manufacturer claim Profile S conformance.

Existing Profile S equipment keeps working. What changes is that a camera bought in 2027 with only a Profile S claim is standing on a certification nobody can renew.

Profile T is the stated replacement and contains virtually all of the Profile S features. Treat a Profile T listing as the safer buy for anything you expect to keep for five years.

Where the RTSP Stream Lives on Your Network

Finding the Stream
  • ● Default is TCP port 554
  • ● Stream path differs by manufacturer
  • ● Never forward 554 to the internet

RTSP is the older and simpler half of the pair. It is a streaming protocol, it usually listens on TCP port 554, and it hands over H.264 or H.265 video to whatever asks correctly.

The address follows a predictable shape, and only the tail differs between manufacturers.

rtsp://username:password@192.168.1.42:554/stream1     # main stream
rtsp://username:password@192.168.1.42:554/stream2     # substream, lower bitrate

The path after the port is the part you cannot guess. TP-Link publishes the exact strings for Tapo cameras in its support articles, and most brands that support RTSP do the same.

Codec choice changes what your network has to carry. H.265 cuts the bandwidth of a stream by up to half compared with H.264, which is the difference between four cameras being watchable over Wi-Fi and not.

One rule has no exceptions. Do not forward port 554 through your router, because that publishes both the video and the camera login to anyone scanning the internet.

Reach it over a VPN back to your home network instead. If the camera lives on a separate network segment, our note on smart home devices on a guest network covers how that changes discovery.

What a Camera Without These Two Gives Up

A camera with no ONVIF profile and no RTSP path is one that only its own app can watch. That is fine until one of three things happens.

The first is a subscription change. Without a local stream you have no route to keep recording when the plan you relied on gets repriced or restructured.

The second is a recorder. Mixing brands on one network video recorder is exactly the problem ONVIF was written to solve, and a closed camera cannot join.

The third is the end of the product. When a brand shuts down its servers, an open stream is the difference between a working camera and a plastic shell, as described in our piece on what happens when a smart home brand shuts down.

None of this makes a closed camera a bad buy. It makes it a rental rather than a purchase, and the price should reflect that.

Which Camera Setup Fits Your House

Match the Profile to the Plan
  • ● One camera, one app - support is optional
  • ● Mixed brands on one recorder - Profile T
  • ● Long retention on site - add Profile G

One camera watching a porch, viewed on a phone: Neither standard is required. Buy on image quality and alert accuracy, and treat ONVIF as a bonus that keeps your options open later.

Three or more cameras from different brands: Insist on a registered Profile T device. This is the case ONVIF exists for, and mixed brands without it turn into three apps and no unified timeline.

Footage kept on site rather than in the cloud: Look for Profile G alongside the streaming profile. Recording and storage are a separate conformance area from streaming, and a Profile T only camera may not expose recordings the way your recorder expects.

A wired system you expect to keep for a decade: Avoid a Profile S only listing from 2026 onward. The certification cannot be renewed after March 2027, so buy the replacement profile now.

Renters who may take the system with them: Prioritise RTSP with a documented path over profile depth. A camera you can point at any software moves house more easily than one tied to a hub.

Four Listing Claims That Do Not Survive Checking

The first is ONVIF compatible. ONVIF states that conformance to a profile is the only thing that ensures compatibility between products, so a compatibility claim without a registration is marketing language.

The second is a profile letter with no version behind it. Ask which profile, then look the model up rather than trusting the box.

The third is RTSP support that turns out to be firmware dependent. Several brands ship RTSP disabled, or removed it in an update, so check the current firmware notes and not an old review.

The fourth is a rating that belongs to a sibling model. Wired and battery versions of the same camera often differ here, because a battery camera streaming continuously would not last a day.

Battery models deserve a second look for that reason. Continuous RTSP and battery power pull against each other, which is part of the tradeoff in our comparison of wired and wireless security cameras.

The Client Side Needs Conformance Too

Half the ONVIF question sits on the recorder. A profile is a contract between a conformant device and a conformant client, so the software you record with has to hold up its end.

That is why a certified camera can still refuse to appear in a piece of free recording software. The camera met the standard and the client never claimed to.

Check the profile letters on both sides before assuming a mismatch is a camera fault. Most recorder projects and NVR makers publish the profiles they implement.

Stream choice is the other setting people miss. Cameras usually expose a high bitrate main stream and a lower bitrate substream, and recording the substream continuously while pulling the main stream for events keeps storage and Wi-Fi load sane.

Analytics travel separately again. If you want person or vehicle events handed to other software rather than reinvented by it, that is Profile M territory, not Profile T.

Check the Conformance Page Before You Buy

ONVIF runs a public list of conformant products, and it is the only place a claim can be confirmed. Searching the exact model number takes under a minute and settles the question the product page leaves open.

Do that check, note the profile letters, then find the manufacturer support article that prints the RTSP path. Two pages, two facts, and you know whether the camera will still be useful when its app is not.

If both come back empty, buy the camera on its merits as a closed appliance. Just do not pay a premium expecting an openness that was never certified.

Resolution claims deserve the same scepticism as protocol claims, since 2K carries no definition at all. Our guide to what 2K and 4K mean on a camera box converts the badges into pixel dimensions and the upload speed each one needs.

FAQ

What is the difference between ONVIF and RTSP?

ONVIF is the handshake that lets a camera and recorder find each other and agree on settings. RTSP is the protocol that carries the actual video once they have agreed.

Is ONVIF Profile S being discontinued?

ONVIF has announced that support for Profile S ends on 31 March 2027. Existing equipment keeps working, but manufacturers can no longer certify new products or new firmware against it after that date.

Does every camera expose RTSP on port 554?

Usually yes, on TCP port 554, which is the default for the protocol. The stream path differs by manufacturer, so check the support page for your exact model rather than guessing.

Can I open port 554 on my router to watch from outside?

No. Port forwarding 554 to the internet exposes the stream and the camera login to anyone scanning for it. Use a VPN back to your home network instead.

Does ONVIF support guarantee my recorder will work with the camera?

Only if a product is registered as conformant. ONVIF states that conformance to a profile is the only thing that ensures compatibility, so a listing that says ONVIF compatible without a registration is a weaker claim.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.

Friday, September 4, 2026

What Lumens and Kelvin Mean on a Smart Bulb Box

What Lumens and Kelvin Mean on a Smart Bulb

Lumens Are Brightness and Kelvin Is Color

The Short Version
  • ● 800 lumens replaces an old 60-watt bulb
  • ● 2700K reads warm, 5000K reads clinical
  • ● Watts now describe cost, not brightness

The short answer is that lumens measure how much light a bulb emits and Kelvin describes the color of that light. Pick lumens by room size and Kelvin by mood. Energy Star lists 800 lumens as the replacement for an old 60-watt incandescent, and 1,600 for a 100-watt. On the color scale, 2700K is the warm yellow of a table lamp and 5000K is the cold white of a hardware aisle.

Smart bulbs did not change either number. They added a dimmer, a schedule and sometimes a Kelvin range you can shift from an app.

Why the Watt Number on the Box Stopped Meaning Anything

Watts measure electricity drawn, not light produced. That distinction did not matter while every bulb was a filament, because a filament converted power to light at a predictable and terrible rate.

LEDs broke the shortcut. A bulb drawing 8 or 9 watts now produces the light an old 60-watt bulb did. Wattage survives as a running-cost figure and says nothing about how bright the room will be.

Packaging still leans on the old habit. Phrases like 60W equivalent or soft white 60 are marketing translations of a lumen figure, printed because shoppers spent decades buying by watts. The lumen number underneath is the one that transfers between brands.

The Brightness Ladder Most Rooms Actually Need

Lumens by Room
  • ● Bedside lamp - 450 lumens is plenty
  • ● Kitchen ceiling - 1,100 to 1,600 lumens
  • ● Garage or workshop - 2,600 lumens

The table pairs the Energy Star equivalence figures with the rooms they suit. Treat it as a starting point rather than a rule, because ceiling height, wall color and the number of fixtures all move the target.

Old wattage Minimum lumens Typical LED draw Where it fits
40 W 450 5 to 6 W Bedside lamp, closet, accent fixture
60 W 800 8 to 9 W Living room lamp, bedroom ceiling
75 W 1,100 11 to 13 W Kitchen ceiling, home office
100 W 1,600 14 to 17 W Open kitchen, basement, large room
150 W 2,600 23 to 28 W Garage, workshop, utility space

Two smaller sources usually beat one bright one. Three 450 lumen bulbs spread across a room produce fewer harsh shadows than a single 1,600 lumen fixture overhead, and dimming them together lands somewhere the single bulb cannot reach.

Lumens also decay. Most LED packages quote a lifetime in hours alongside the brightness, and output drifts down across that span rather than holding flat until failure.

Kelvin Read as a Room Decision Rather Than a Spec

The FTC Lighting Facts panel plots light appearance on a scale running from 2600K to 6600K. Warm sits at the low end and cool at the high end, which feels backwards to anyone who thinks of blue as cold.

In practice the scale collapses into three choices. Around 2700K reads relaxed and flattering, around 3500K reads neutral, and 5000K and above reads alert and slightly clinical.

Match the number to what happens in the room. Kitchens, bathrooms and workbenches reward the cooler end because you are inspecting surfaces, while bedrooms and living rooms reward the warm end because you are winding down.

Mixing Kelvin values within one sightline is the common mistake. A 2700K lamp beside a 4000K ceiling fixture makes both look wrong, and the fix is usually cheaper than it sounds because one bulb has to change, not the room.

What a Smart Bulb Adds to Both Numbers

A smart bulb still ships with a fixed lumen rating, so the app cannot make it brighter than the hardware. What the app changes is the ceiling you actually use, since a dimmed 800 lumen bulb can behave like a 450 lumen one on a schedule.

Color temperature is where the marketing gets slippery. A bulb sold as tunable white lists a Kelvin range such as 2200K to 6500K, while a plain smart bulb lists one number and cannot move off it no matter what the app suggests.

Full color bulbs add red, green and blue emitters. Their white light often measures worse than a dedicated white bulb, which is why many people run color bulbs in accent fixtures and tunable white bulbs where they read.

Heat is the constraint nobody reads for. A smart bulb carries a radio and a driver inside the same housing as the LED, so packaging often warns against enclosed fixtures where that heat has nowhere to go.

Check the fixture before the feature list. A sealed ceiling dome or a tight glass shade shortens the life of a bulb rated in years, and the failure looks like a defect rather than a mismatch.

The dimming path matters too. If the bulb sits behind a wall dimmer, the two dimming systems fight each other, which is a separate problem covered in do smart bulbs work with dimmer switches.

Where the Numbers Hide on the Package

Every general service bulb sold with a medium screw base carries a Lighting Facts panel. It is the small table on the side or the back, and it is regulated rather than written by the marketing team.

The panel lists five things in a fixed order. Brightness in lumens, estimated yearly energy cost, life in years, light appearance plotted on that 2600K to 6600K strip, and energy used in watts.

Smart bulb boxes add their own claims around it. Color counts, app names and voice assistant logos take the large type, while the regulated panel keeps the figures you can compare against a bulb from another brand.

One habit saves most of the guesswork. Photograph the panel of a bulb you already like before it fails, because the replacement decision then becomes matching four printed numbers rather than rebuilding the judgement from scratch.

The Third Number That Never Gets Large Type

Color Rendering
  • ● CRI runs on a scale that ends at 100
  • ● 90 or above for kitchens and mirrors
  • ● Below 80 makes food look gray

The color rendering index scores how faithfully a light source shows color against a reference, on a scale that ends at 100. It is printed in small type when it is printed at all.

Below 80, food and skin start to look drained. At 90 and above, a tomato on a counter looks like a tomato, which is why kitchens and bathroom mirrors are worth the upgrade.

Hallways, closets and garages do not need it. Spending on high CRI in a space where nobody judges color is money that would do more work as a second fixture.

Which Bulb Fits Each Room in a Normal House

Bedroom: 450 to 800 lumens at 2700K, dimmable. The evening matters more than the morning here, and a cooler bulb makes the room feel like an office.

Kitchen: 1,100 to 1,600 lumens at 3500K to 4000K with CRI 90 or above. You are judging color on food and cutting near your fingers, so brightness and fidelity both earn their cost.

Home office: 1,100 lumens at 4000K on the ceiling with a warmer lamp beside the desk. The contrast keeps video calls from lighting you from one direction only.

Living room: 800 lumens per lamp at 2700K, spread across two or three fixtures. This is the room where tunable white pays off, because the same lamps serve both an afternoon and a film.

Garage or utility space: 2,600 lumens at 5000K, fixed. Nothing about the room rewards warmth, and a single bright cool source is the cheapest way to see what you are doing.

What to Check Before You Buy a Four Pack

Read the lumen figure first and ignore the equivalence claim beside it. Two bulbs both labelled 60W equivalent can differ by 100 lumens, and the smaller print is the honest number.

Confirm whether the Kelvin value is fixed or a range. A single number means the app will never change the color, regardless of how the listing photo is styled.

Check the base and the shape against the fixture, since an A19 bulb with a wide smart driver sometimes will not seat in a recessed can. Buying one bulb before buying four costs less than returning a box.

Running cost is the last box to tick, and it is smaller than most people expect at these wattages. The arithmetic sits in do smart bulbs use more electricity, and the question of whether the bulb or the socket should be the smart part is worked through in smart plugs vs smart bulbs.

The Two Numbers That Survive Every Brand Change

Brands, apps and hubs will keep churning. Lumens and Kelvin will not, because they describe physics rather than a product line.

Write down the lumen and Kelvin figures that worked in each room before you replace anything. That short list survives a brand going out of business, and it turns the next purchase into a matching exercise rather than a fresh guess.

FAQ

How many lumens replace a 60-watt bulb?

About 800 lumens. The Energy Star replacement chart lists 450 lumens for a 40-watt bulb, 800 for a 60-watt, 1,100 for a 75-watt, 1,600 for a 100-watt and 2,600 for a 150-watt. Smart bulbs print the same lumen figure, so you can shop across brands with one number instead of guessing from wattage.

Is 2700K or 5000K better for a house?

Neither is better and they answer different questions. 2700K is the warm yellow most people associate with a lamp in a living room, while 5000K is the blue-white of a workshop or a garage. Rooms you relax in read better warm, and rooms where you inspect things read better cool.

Can a smart bulb change its color temperature after you install it?

Only if the bulb is sold as tunable white or full color. A fixed 2700K smart bulb can dim and schedule but cannot change its Kelvin value, and the packaging will state a single number rather than a range such as 2200K to 6500K. Check that range before assuming an app can shift it.

Does dimming a bulb to 50 percent cut the lumens in half?

Roughly, but not exactly. Dimming an LED lowers output without warming the light the way a filament does, unless the bulb advertises warm dim or dim to warm. Half brightness on an 800 lumen bulb also does not feel like half as bright, because the eye responds to light on a compressed scale.

What is CRI and does it matter for a smart bulb?

It matters most where you look at color closely, such as a kitchen, a bathroom mirror or a craft table. The color rendering index runs to 100, and bulbs at 90 or above keep skin and food looking normal. In a hallway or a closet a lower number is rarely noticeable.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.

Thursday, September 3, 2026

What Dedicated Backhaul Means on a Mesh WiFi System

Mesh Backhaul

Backhaul Is the Conversation Between Your Nodes

Backhaul is the link your mesh units use to talk to each other, and it is a separate job from serving your phone. If you can run a cable, choose wired backhaul, because it is the only arrangement with no speed penalty at all.

Wireless backhaul comes in two forms. A dedicated radio carries node traffic on its own, while a shared radio does that work alongside your devices and gives up roughly 50% of throughput at every hop.

The band count on the box is how you tell the two apart. ASUS publishes up to 7,800 Mbps for the tri-band ZenWiFi ET9, and TP-Link publishes up to 33 Gbps for the quad-band Deco BE95.

Part of each of those totals belongs to a radio your laptop may never be allowed to touch. That gap is what the rest of this article is about.

Why a Shared Band Halves Your Speed at the Second Node

The Halving Rule
  • ● One radio receives then resends
  • ● Each wireless hop costs about half
  • ● Two hops leave roughly a quarter

A radio cannot receive and transmit at the same instant on the same channel. That single physical constraint produces the whole problem.

When a dual-band system uses its 5 GHz radio for both your devices and the hop to the next unit, every packet crosses that radio twice. The result is close to half the throughput at the satellite, before walls and interference take their own cut.

Add a second hop and you are down to roughly a quarter. This is why a three-unit system strung out in a line often performs worse than a two-unit system placed sensibly.

The halving is not a defect and no firmware update removes it. It is the reason the dedicated backhaul band exists at all.

Dedicated, Shared and Wired Are Three Different Arrangements

Shared wireless backhaul is the default on dual-band systems. Nothing is reserved, the router allocates airtime between your devices and the next node, and both slow down together under load.

Dedicated wireless backhaul reserves a radio, usually the 6 GHz one on tri-band hardware. ASUS describes exactly this on the ZenWiFi ET9, where the 6 GHz band can be set to carry mesh backhaul traffic only.

Wired backhaul runs the link over Ethernet instead. The wireless radios stop carrying node traffic altogether, and ASUS notes that cabling the units reserves all three bands for your devices.

There is a hybrid too. TP-Link markets Backhaul Aggregation on its Deco line, which blends wired and wireless links between units rather than choosing one.

What the Band Count on the Box Actually Tells You

Reading the Spec Sheet
  • ● Dual-band shares the 5 GHz radio
  • ● Tri-band can reserve one radio
  • ● Headline Gbps counts every radio

Band count is the most useful number on a mesh box and the most commonly misread. It tells you how many radios exist, not how many you get to use.

Dual-band means 2.4 GHz and 5 GHz. There is no spare radio, so wireless backhaul is shared by definition.

Tri-band adds a third radio, usually 6 GHz on Wi-Fi 6E and Wi-Fi 7 hardware, which can be reserved. Quad-band adds a fourth, typically a second 5 GHz radio.

The headline speed counts all of them added together. That is the catch worth understanding before you compare two price tags.

System Bands Vendor headline speed Vendor coverage What the backhaul can do
TP-Link Deco BE25 Dual-band Up to 5 Gbps 6,600 sq ft as a 3-pack Shared with your devices
TP-Link Deco BE67 Dual-band Up to 14 Gbps 8,100 sq ft as a 3-pack Shared with your devices
ASUS ZenWiFi ET9 Tri-band Up to 7,800 Mbps 5,700 sq ft as a 2-pack 6 GHz can be reserved for backhaul
NETGEAR Orbi RBE873 Tri-band Up to 21 Gbps Up to 9,000 sq ft Third radio available for node traffic
TP-Link Deco BE95 Quad-band Up to 33 Gbps 9,100 sq ft as a 3-pack Fourth radio available for node traffic

Figures are the vendor published specifications as of Sep 2026 and describe theoretical maximums across all radios combined. Confirm current pricing and specifications on the official site before buying, since model line-ups change every year.

The Coverage Number Assumes a Backhaul You May Not Get

Coverage claims and backhaul quality are the same story told twice. NETGEAR quotes up to 9,000 sq ft for the RBE873 and up to 6,000 sq ft for the RBE872, and an add-on RBE870B satellite is quoted at a further 3,000 sq ft.

Those figures assume the satellite still has a usable link back to the main unit. Push a node to the edge of its range and the coverage circle it draws is filled with a connection that has already halved.

Square footage also assumes an open plan. Brick, plaster with metal lath, foil-backed insulation and a concrete stairwell all shorten the real number, and they shorten the backhaul link first because it is the longest link in the system.

Treat published coverage as an upper bound. Our comparison of mesh systems and single WiFi 6 routers covers when the extra nodes earn their place at all.

The 6 GHz Band Is Fast and Short Ranged at the Same Time

Reserving 6 GHz for backhaul sounds like an obvious win, and in a compact home it is. The band is wide, clean and almost empty of neighbours.

It is also the shortest ranged band in the system. High frequencies attenuate faster through walls, so the reserved radio is the first one to fail across a long house.

Good firmware handles this by falling back to a shared 5 GHz link when the 6 GHz hop degrades. Your speed drops, quietly, and the system keeps working rather than dropping the node.

This is the practical reason node placement matters more on a 6 GHz backhaul than on an older 5 GHz one. The premium band only pays for itself while the nodes can hear each other clearly.

An Ethernet Cable Beats Every Wireless Backhaul

If a cable can reach, run the cable. A wired backhaul removes the halving entirely, frees every radio for your devices, and behaves the same at three in the afternoon as at midnight.

Existing wiring often gets you there. Coaxial runs can carry MoCA adapters, and a single Cat 5e run left behind by a previous owner is enough for a two-node system.

Powerline is the fallback rather than the answer. It works, it is usually better than a two-hop wireless chain, and it is unpredictable across different circuits in the same house.

Cabling also fixes a problem people blame on mesh generally. Devices that keep dropping and reconnecting are often sitting behind a weak backhaul rather than a weak client link, as our guide to reconnecting smart devices after a router change explains.

What Backhaul Changes for Smart Home Devices

Most smart home gear runs on 2.4 GHz and asks for almost no bandwidth. A doorbell chime, a plug state change and a sensor trip are a few hundred bytes each, so backhaul speed is irrelevant to them.

Backhaul stability is not. When a node loses its link to the main unit, everything joined to that node disappears from your network at once, and the outage lands on the whole room rather than one device.

That is why weak backhaul reads as a smart home fault rather than a WiFi fault. Automations fire late, a camera clip fails to upload, and a lock reports offline in the app while your phone still shows full bars on the same node.

Cameras are the exception on bandwidth. A pair of 2K streams uploading at once is real traffic crossing every hop, and it competes with your devices directly on a shared backhaul.

The practical rule is to put anything that streams on the main unit, and let the satellites carry sensors and switches. It costs nothing and it removes the single most common source of intermittent smart home failures.

Where People Lose the Backhaul They Paid For

The most common mistake is placing satellites at the dead zone. Nodes belong between the router and the dead zone, roughly halfway, because a node has to hear the router before it can help anything.

The second is chaining. Many systems let node three link through node two, and each additional wireless hop repeats the halving.

The third is burying the main unit in a media cabinet. Metal shelving and a closed door attenuate the backhaul link at both ends of the house at once.

The fourth is ignoring the app. Most systems report backhaul quality or connection strength per node, and that screen answers in ten seconds what a speed test only hints at.

Which Backhaul Setup Fits Your Home

Pick by Building
  • ● Wired if any cable run exists
  • ● Tri-band for long or brick houses
  • ● Dual-band for small flats

Flat or small house under about 1,200 sq ft: A dual-band two-pack is enough. One hop over a shared radio across a short distance costs you little, and the money is better spent on a newer standard than an extra radio.

Long or multi-storey house with solid walls: Buy tri-band and place the satellite within clear range of the main unit. This is the case the dedicated radio was designed for, and it is where the halving hurts most.

Any home with an existing Ethernet or coaxial run: Use wired backhaul and buy the cheaper dual-band system. A cable makes the third radio close to redundant, which is a rare chance to spend less and get more.

Home with 40 or more connected devices: Choose tri-band, then check the per-node device counts rather than the headline speed. Airtime is the constraint here, and our look at how many devices one router can handle covers where the ceiling actually sits.

Renting, with no drilling allowed: Tri-band wireless, placed conservatively. Powerline backhaul is worth trying before you assume a wireless hop is the only option.

Three Checks Before You Choose a System

Count the radios rather than reading the speed. A dual-band system quoted at 14 Gbps still has no radio to spare, and a tri-band system quoted lower may serve your devices better.

Look for the word dedicated or a backhaul setting in the app, not just the band count. Some tri-band systems put all three bands in the client pool by default and leave the reservation to you.

Measure at the far node before you decide the system failed. A speed test beside the router tells you about your internet connection, and a speed test in the back bedroom tells you about your backhaul.

If you are still choosing between a mesh system and a simple extender, our comparison of mesh WiFi and WiFi extenders covers the same halving problem from the other direction.

The Radio You Never See

Backhaul is invisible in daily use and it decides most of what people notice about a mesh system. Speed at the far end, video calls that survive a walk upstairs, and cameras that stop dropping all trace back to it.

The decision is short. Cable it if you can, buy a reserved radio if you cannot, and place the nodes so they can hear each other either way.

Everything else on the box is a maximum measured in conditions your house does not have.

FAQ

What does dedicated backhaul mean on a mesh WiFi system?

Backhaul is the traffic between your mesh units, not between your phone and a unit. Dedicated backhaul reserves one radio for that job, so the link between nodes does not compete with the laptops and speakers in the room.

Does backhaul really halve my speed at the second node?

On a shared wireless backhaul each hop roughly halves throughput, because one radio receives and resends the same data. Dedicated and wired backhaul avoid that penalty, which is why the far end of a house feels so different between systems.

Is Ethernet backhaul better than a dedicated wireless band?

Wired backhaul wins in almost every case. ASUS states that connecting mesh units by Ethernet reserves all three WiFi bands for your devices, which is a better outcome than any wireless arrangement can produce.

Does a higher band count mean faster WiFi for my devices?

Not on its own. Vendors count every radio in the headline figure, so a quad-band system such as the Deco BE95, published at up to 33 Gbps, includes a radio your devices may never be allowed to use.

Can node placement change how good my backhaul is?

Yes, and it is the cheapest upgrade available. Moving a satellite so it sits within clear range of the main unit does more for backhaul quality than paying for a higher tier system that is placed badly.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.

Wednesday, September 2, 2026

SACC vs ASHRAE BTU on a Portable AC

SACC vs ASHRAE

Size the Room With SACC and Ignore the Bigger Number

If a listing shows two BTU figures, the smaller one is the honest one. SACC is the Department of Energy rating, and it typically lands 25% to 45% below the older ASHRAE figure printed on the same box.

TCL states plainly that a model previously sold as 14,000 BTU under ASHRAE is now considered a 10,000 BTU SACC unit. Nothing about the hardware changed. Only the test did.

So size your room against the SACC number, and treat any ASHRAE figure as marketing history rather than capacity. A unit picked on the ASHRAE number will run continuously in the heat and never reach the thermostat setting.

Two Ratings Measured in Two Different Rooms

Two Test Rooms
  • ● ASHRAE - one test point at 80F dry bulb
  • ● SACC - 83F weighted 0.8 plus 95F weighted 0.2
  • ● Only SACC subtracts single hose infiltration

The gap between the two numbers is not a rounding difference or a manufacturer taking liberties. It comes from two test procedures that put the machine in two different situations.

The older ASHRAE method uses a single condition: 80F dry bulb with a 67F wet bulb, which works out to roughly 51% relative humidity. That is a mild, stable summer afternoon.

The DOE method uses two test points and averages them with fixed weights. The formula is SACC equals ACC95 times 0.2 plus ACC83 times 0.8, so eighty percent of the score comes from an 83F room and twenty percent from a 95F room.

That weighting is why SACC tracks a season rather than a moment. A machine that holds capacity as the outdoor temperature climbs keeps more of its rating than one that fades.

What Changed on October 1 2017

The DOE test procedure for portable air conditioners became mandatory on October 1, 2017, under 10 CFR Parts 429 and 430. From that date, capacity claims on covered units had to be stated in SACC.

Manufacturers did not all move at once. TCL says its own transition began with products manufactured in 2020 and that by 2022 every TCL portable used SACC instead of ASHRAE.

That staggered rollout is why the confusion persists years later. Older stock, third party marketplace listings, and product pages that were never rewritten still lead with an ASHRAE figure.

The practical test is simple. If a listing shows a big BTU number with no rating name next to it, assume nothing and look for the EnergyGuide label or the specification sheet.

Reading Both Numbers on the Same Unit

Retailers who publish conversion tables give you a usable map between the two scales. The figures below are the ones commonly listed, along with the room sizes those tables recommend.

ASHRAE BTU DOE SACC BTU Drop Room size the tables claim BTU per square foot
8,000 5,200 35% Up to 250 sq ft About 21
10,000 6,500 35% Up to 350 sq ft About 19
12,000 7,800 35% Up to 450 sq ft About 17
14,000 9,100 35% Up to 600 sq ft About 15
14,000 (TCL example) 10,000 29% Manufacturer stated Model specific

The conversion tables apply a median 35% difference, while the published range across models runs from 25% to 45%. The TCL example sits at the shallow end of that range.

Look at the last column rather than the middle one. As the units get larger, the recommended room grows faster than the capacity does, so the implied BTU per square foot slides from about 21 down to about 15.

Window air conditioner sizing has long worked from roughly 20 BTU per square foot. A chart that stretches 9,100 SACC across 600 square feet is quietly assuming a shaded, well insulated, single occupancy room.

Single Hose Units Lose the Same Air Twice

The Hose Penalty
  • ● One hose exhausts room air outdoors
  • ● Replacement air arrives hot through gaps
  • ● SACC counts that penalty, ASHRAE does not

The largest structural reason the two numbers diverge is the exhaust hose, and it is the part shoppers most often miss.

A single hose unit blows hot condenser air out the window. That air has to be replaced, and it is replaced by outdoor air pulled in through door gaps, window frames, and every other leak in the room.

So the machine cools the room and then partially undoes the work by importing warm air behind itself. The room sits at slight negative pressure the whole time it runs.

The SACC procedure accounts for that infiltration penalty. The ASHRAE procedure does not, which is exactly why single hose units lost the most capacity when the rating changed.

Dual hose designs draw their condenser air from outdoors and send it back outdoors, so the room is not used as the air source. They still lose some rating in the conversion, but less of it.

This is also why sealing the window kit matters more on a portable than on a window unit. Foam gaskets around the hose panel are not cosmetic, they are the difference between the rated capacity and something well under it.

How Many Square Feet a SACC Rating Actually Covers

Square footage alone is a weak predictor, which is why two rooms of the same size can need different machines.

Ceiling height changes the volume you are cooling. A 300 square foot room with a 9 foot ceiling holds a quarter more air than the same footprint at 8 feet, and sizing charts assume the lower number.

Sun exposure moves the requirement further. A west facing room takes a heat load through the afternoon that a north facing room never sees, and glass area matters more than wall area.

Then add the occupants and the equipment. People, a desktop machine, and a kitchen all put heat into the room continuously, and none of it appears in a square footage chart.

Insulation and air sealing decide how much of that load stays. In an older building with single glazing, a unit sized off the generous end of a retailer chart will run without stopping on the hottest days.

The way out is to size from the conservative end. Take the room area, multiply by roughly 20 BTU per square foot, then add capacity for height, sun, and occupancy rather than subtracting for a hopeful chart.

Which Rating Fits the Room You Are Cooling

Size by SACC
  • ● Bedroom under 250 sq ft - 5,200 SACC and up
  • ● Open living space - 9,100 SACC and up
  • ● Top floor or west facing - add a size

Small bedroom under 250 square feet: Look for 5,200 SACC or better. This is the one case where the retailer charts and the 20 BTU per square foot rule roughly agree, so the smallest units are honestly sized.

Open plan living area over 400 square feet: Start at 9,100 SACC and treat anything below that as a supplement rather than a solution. An open kitchen adds a heat source the chart never counted.

Top floor or west facing room: Add a size beyond what the footprint suggests. Our guide on cooling an upstairs bedroom covers the stack effect that makes these rooms behave larger than they measure.

Rental where a window unit is not allowed: A portable is the right call, and here the dual hose premium pays for itself because you cannot reliably seal an old rental window. The comparison in portable versus window air conditioners sets out the tradeoff.

Dry climate under about 30% humidity: Check an evaporative cooler before spending on compressor capacity at all. Our portable AC and evaporative cooler comparison explains where that swap works and where it backfires.

Three Listing Traps That Hide the Weaker Number

The first trap is the unlabelled BTU figure. A headline that reads 14,000 BTU with no rating name is almost always ASHRAE, because no manufacturer leads with the smaller number by choice.

The second trap is a specification sheet that lists SACC in a footnote while the product title keeps the old figure. Marketplace listings inherit titles from years of previous versions, and nobody rewrites them.

The third trap is a room size recommendation carried over from the ASHRAE era. The capacity claim gets corrected, the coverage claim does not, and you end up with an honest BTU number attached to an optimistic square footage.

There is a fourth, smaller trap worth knowing. Some listings quote a heating BTU alongside the cooling figure on reverse cycle models, and the heating number is often larger. It says nothing about cooling capacity.

Buying From the Number That Was Actually Tested

The rating change was a correction, not a downgrade. The machines did not get weaker in 2017, the test stopped flattering them.

Find the SACC figure, size the room at roughly 20 BTU per square foot, then add for ceiling height, afternoon sun, and the people and equipment in the space. If only an ASHRAE number is shown, cut it by about a third before you compare anything.

And if the room is large or the window seals badly, weigh a dual hose model. It is the one design difference that changes how much of the rating survives contact with a real room.

FAQ

What does SACC mean on a portable air conditioner?

SACC stands for seasonally adjusted cooling capacity. It is the Department of Energy figure calculated as ACC95 times 0.2 plus ACC83 times 0.8, so it blends a hot day and a warm day instead of reporting one ideal test result.

Why does one portable AC list two different BTU numbers?

Because they measure different things. The ASHRAE figure comes from a single 80F test point, while SACC averages two temperatures and subtracts the heat that a single hose unit pulls back into the room. The DOE number typically lands 25% to 45% lower.

What is 14,000 BTU ASHRAE in SACC?

Roughly 9,100 SACC on the conversion tables retailers publish. TCL says a model once sold as 14,000 BTU under ASHRAE is now a 10,000 BTU SACC unit, so the exact landing point depends on the model rather than a fixed formula.

Is the BTU number on a portable AC always SACC now?

Yes, if the label follows the DOE rule that became mandatory on October 1, 2017. Units still advertising only an ASHRAE figure are usually older stock or listings that were never updated, and the comparison is not apples to apples.

How many square feet will 9,100 SACC BTU cool?

Retailer charts pair 9,100 SACC with rooms up to 600 square feet, which is about 15 BTU per square foot. Window unit sizing has long used roughly 20 BTU per square foot, so treat the generous end of those charts as a best case.

Sources

About the author. Jay Lim runs HomeVersus as an independent, one-person publication. Articles are researched against official documentation, pricing pages and regulators rather than hands-on lab testing. How we research · Report an error


Some links may be affiliate links. We may earn a commission at no extra cost to you.

This article was written with AI assistance. It is researched and fact-checked, not based on personal hands-on testing unless explicitly stated.