IT Support & Hardware curriculum

A motherboard's form factor names its size, shape and the layout of its parts, and ATX became the most common design on newer desktop boards. Cases come in their own sizes, from slim desktops through mini, midi and full towers, so check that a case is built for the board's form factor before you buy either part.

The processor mounts in a socket on the motherboard, and the chipset beside it governs how the CPU trades data with memory, drives and expansion cards, and even how much memory the board can take. So a processor is chosen to suit a particular board, never on its own: a chip that looks right on paper still has to match that board's socket and chipset.

Each core is a separate processing unit, and modern CPUs carry many, so different programs can run at the same time on different cores. Multithreading is a different trick: a core works on another stream of instructions during the dead time while data is being fetched. Clock speed is the third lever, since a faster clock processes instructions sooner.

A heat sink and fan carry heat away from the chip, and when cooling cannot keep up, a chip that reaches its temperature limit protects itself by lowering its clock speed. That throttling looks like a machine that starts fast and slows within minutes of heavy load, so check temperatures and airflow before you suspect a failing part.

The firmware's CMOS setup stores the date, time, drive settings and boot sequence, and the motherboard carries a small CMOS battery beside it; a real-time clock chip can keep counting on a coin cell while main power is gone. So a machine that loses its date or settings after being unplugged points first at that coin cell.

Power-on self-test runs first: the firmware initializes and tests the chipset, CPU and memory, and only then brings up the video card. A fault found that early happens before anything can be drawn on screen, so a blank display with the fans running can still be a memory or board problem rather than a monitor problem.

Expansion cards plug into PCIe slots, and a slot performs only as well as the lanes wired to it. A board can have more slot pins than lanes it can drive at once, so a card in a full-length x16 slot that is given fewer lanes will not reach its rated speed. Check the board's lane layout, not just the slot's size.

Before you power a board, look for anything that could short it: a stray screw, nut or scrap of wire near a terminal or the chassis can bridge a connection and do real damage. Shaking the case and listening for a rattle is a quick check, and a short you remove now is one that cannot strike later.

Firmware starts the machine, tests the hardware and then hands control to a boot loader on the first suitable boot device, such as a drive or a USB stick. Boot order and Secure Boot are settings kept in nonvolatile memory: one lists which devices to try and in what order, the other controls which third-party code the firmware will run.

Each DDR generation has doubled transfer rates and lowered its signaling voltage, so a module from one generation is not an electrical match for a board built around another. Memory is specified by capacity, speed, data rate and module type, and the generation is part of that specification, not a detail you can skip.

SO-DIMM stands for small outline dual in-line memory module, the compact module format found in laptops and mini PCs. A replacement has to match both the module type the board accepts and its DDR generation, since a module that is right in one respect can still be wrong in the other.

The memory controller reaches the modules over channels. Field studies of large server fleets found channel failures rare but prolific: a failed channel produced a large share of the errors seen, either transient, from transmission signals falling out of alignment, or permanent, from failed transmission logic on a module.

Error-correcting code memory stores extra check bits with the data; a common scheme corrects any single-bit error and detects two-bit errors in each 64-bit word. Most servers use stronger versions of the idea to avoid silent data corruption, yet ECC is not absolute, since faults that flip several bits in one word can defeat it.

How much memory a system can use is set by the board, not just by how many slots it has: the chipset determines how much memory can be added. More memory lets more work run at once, and when a job needs more than the system holds, software has to work through the data in pieces so it does not exceed system memory.

Memory faults are not always loud: errors on the memory bus can corrupt data silently, with nothing reported. Because a symptom alone rarely names the failed part, isolate it: test a suspect module against a known-good one and change one item at a time, rather than replacing several parts before you have found the fault.

A hard drive stores data on magnetic platters spinning at high speed, read by moving parts, while an SSD keeps it in NAND flash memory with no mechanical moving parts. An SSD reads and writes in microseconds where a spinning disk needs milliseconds, which is why replacing a hard drive with an SSD is such a visible speedup.

RAID 0 stripes data across disks for speed but loses everything if any one disk fails. RAID 1 mirrors the same data onto a second disk, surviving a failure but halving usable space, and RAID 10 stripes and mirrors at once. RAID 5 uses parity, so a lost disk's contents can be rebuilt from the others while giving up less space than mirroring.

SMART is a drive's self-monitoring: it tracks attributes that report the disk's health, and some of them correlate strongly with coming failure. When those values climb, treat the drive as failing and copy the data off before trying any repair, since a current backup is what lets you restore when a computer breaks.

USB flash drives are small, cheap and portable, which is exactly what makes them easy to lose and attractive to attackers. Do not plug in a drive you found, since it can carry malware; encrypt a drive that holds sensitive files; and keep a backup, because a lost unencrypted drive hands its contents to whoever picks it up.

A drive is divided into partitions, and the partition style matters at startup: legacy BIOS boots from the first device it finds with a valid master boot record, while UEFI firmware loads its boot loader from the first MBR- or GPT-formatted boot device. Partitions also let one drive set aside separate areas for system software, working files and backups.

SATA ports were designed for spinning hard drives, and the SATA controller cannot keep up with the speeds a fast SSD can reach, so NVMe was created to talk to storage over PCIe instead. Workstation boards usually carry small M.2 slots for these PCIe SSDs, so an NVMe drive's ceiling is set by the PCIe lanes wired to its slot.

Power in watts is voltage times current, and a rating in watts marks the most a device is designed to handle, not what it normally uses. Size a power supply with headroom above the load rather than matching it exactly, for the same reason a designer picks parts rated well above the power they will actually dissipate.

Every connection between a power supply and the parts it feeds is a potential trouble spot, and loose, broken or corroded connections are among the first things to look for. Before you switch on, check that each cable is fully seated and undamaged; a fault found before power is applied is one that cannot do damage.

Inspect before you switch on: look for stray metal that could cause a short, a discolored or leaking transformer, capacitors with a whitish deposit at their seals, and loose connections. If any part smokes, or you hear boiling or sputtering once power is applied, remove the power at once, because a short you missed is the likely cause.

A surge protector guards against spikes and dips in line power, while a UPS adds a battery that carries equipment through short outages long enough to save work and shut down. Many UPS units protect against surges too, so computers and critical network gear go on the UPS and less sensitive electronics on surge protectors.

Capacitors can hold a charge long after the voltage is removed, and in many circuits that stored charge is enough to cause a lethal shock. So unplugging does not make the inside of a power supply safe to touch; in electronics maintenance, high-voltage capacitors are discharged only with an authorized shorting probe, repeated several times, before any work begins.

A monitor may offer more than one video input, and models differ: in one study's equipment list, one monitor had D-Sub and DVI-D (Digital Visual Interface, digital version), another DisplayPort and HDMI. Identify the port at each end before you choose a cable or adapter, because the computer's outputs and the monitor's inputs do not always share a type, and the plug that fits decides what can connect.

Resolution is how many pixels a screen shows, and refresh rate, in hertz, is how often it redraws them. A higher rate helps only when the whole path can deliver it: one 120 Hz diagnostic monitor relied on two DisplayPort links and a dedicated graphics card, where conventional commercial screens ran at 60 to 70 Hz.

LCD panels come in several modes: twisted nematic (TN) is efficient but has a limited viewing angle, vertical alignment (VA) reaches very high contrast, and in-plane switching (IPS) suits touch panels. OLED pixels make their own light, giving true black and fast response, but parts of the screen that never change, like a taskbar, risk burn-in.

Isolate a display fault by swapping one part at a time for a known-good one. If the defect follows the monitor onto another computer, it lives in the monitor; if it stays with the original computer when a different monitor is attached, suspect the graphics hardware or the cable, then swap the cable alone to tell those two apart.

When a projector is tilted or off to one side of a flat screen, the picture spreads into a trapezoid, the keystone effect. Correct it by moving the projector until it faces the screen squarely, or by digitally reshaping the image to cancel the slant; a curved or uneven surface distorts the picture in ways a simple keystone fix does not cover.

A keyboard is an array of switches that each send a code when pressed, and many wireless keyboards connect over Bluetooth. Bluetooth devices must be discoverable to find each other the first time, then they stay paired; when a wireless keyboard stops responding, check the pairing and that the radio is on before blaming the keyboard.

A laser printer builds each page with a laser, a photoconductive drum, toner and a heated fuser. Defects often point back to one of those parts: a scratched drum leaves streaks, and print quality also drifts with fuser temperature and the condition of the drum.

An inkjet printer fires tiny droplets from nozzles in its printhead, so ink that dries at a nozzle can leave deposits that spoil the output. A receipt printer usually needs no ink at all: thermal paper carries a developer that reacts with heat, so the printer marks the paper with heat alone.

On Windows, printing runs through the Print Spooler service, so it is the first thing to check when jobs stop moving. It is also an attack surface: a critical flaw in it let attackers take control of affected systems, and CISA advised turning the spooler off on domain controllers and other machines that do not print.

What you can upgrade depends on what is replaceable. Adding memory is often the quickest and cheapest upgrade, but some small systems build their storage onto the board as eMMC flash, and not every battery can be removed or serviced by the user, so check before you promise an upgrade.

Keep lithium-ion batteries at room temperature, away from long spells of heat or cold, which can damage them. A swollen battery is damaged and a potential fire hazard: store it away from anything flammable, such as in a bucket of sand or kitty litter, and take it to a battery recycler or household hazardous waste site, never the trash or a recycling bin.

A touchscreen is two systems stacked together: the display that makes the picture and a touch layer that reports where you press. Capacitive sensing reads a light touch as a change in capacitance, while resistive panels need pressure to bring two conductive layers into contact, so a screen with a clear image that ignores touch points at the touch layer.

A keyboard is an array of key switches, each sending a code when pressed, so one dead key points at that key's switch while a whole dead keyboard points at its cable or connection. Loose, broken or corroded connections are classic trouble spots, so reseat and inspect the keyboard's connector before you replace the keyboard itself.

A USB dock or port replicator rides on USB, an external bus built for plug-and-play peripherals that can be connected without shutting down. Each newer USB version is much faster yet fully backward compatible, so a dock still works on an older port or cable but only at that older version's speed.

When a phone or laptop will not connect, check each radio on its own. Keep Bluetooth off when you are not using it, and hidden rather than discoverable; paired devices still find each other in hidden mode. On Wi-Fi, confirm the network name and password before joining, since an impostor access point can use a stronger signal.

A backup is a separate copy you can restore if a device breaks, someone makes a mistake or malware strikes. A sound plan keeps more than one copy, with at least one away from the building, such as in cloud storage, so theft or a disaster cannot take every copy at once, and it tests each backup right after making it to be sure it restores.

Phones and tablets keep their main storage on the board, with eMMC and UFS flash becoming the standard for mobile systems, so you cannot swap in a bigger drive the way you can in a desktop. On models that accept one, a removable microSD card adds room, and that card can be lost or handed on separately from the device.

Two devices talk directly only when their addresses sit on the same network, which the subnet mask defines, and traffic for anywhere else goes to the default gateway. An address outside the local range cuts a device off from its neighbors, while a wrong gateway leaves the local network working and everything beyond it unreachable.

A hub repeats traffic to every port, while a switch sends it only to the port where its destination sits. A router is the gateway between the local network and the internet, and the home box called a router usually holds a switch and a wireless access point too, so a fault in one function can masquerade as another.

VLANs segment one network's traffic into separate groups, for example keeping voice packets apart from data to improve security. The trade-off is design work: VLANs make network addressing more complex and bring their own supporting protocols, some of which are known to be vulnerable.

Ethernet usually runs over copper cable whose twisted pairs act as balanced lines, which reduces the effect of interference. Fiber carries light instead, which brings far greater bandwidth, lower signal loss, immunity to electrical interference, no crosstalk and no need for a common ground between the two ends.

Power over Ethernet sends power and data down the same cable, so a device like an IP camera can run from one network cable without a power outlet beside it. A recorder or switch with PoE ports supplies that power, and adding a PoE switch lets you connect more powered devices than the recorder has ports.

Wi-Fi shares unlicensed bands at 2.4 and 5 GHz, and more recently 6 GHz, each divided into channels, and in the crowded 2.4 GHz band neighboring channels partly overlap, leaving only a few that stay clear of one another. One study of older Wi-Fi devices found that newer networks sharing the band mainly raised delay and jitter, since retransmissions made up for most lost packets.

WEP is no longer considered secure, so use WPA2 with AES at minimum and WPA3 where your equipment supports it. An open network sends traffic in the clear, where sniffing tools can capture passwords. Change the router's default admin password too, because default passwords are easy to find online.

ping tests whether a host can be reached and measures the round-trip time, and traceroute lists the hops along the path. Read silence with care: firewalls can block the ICMP packets these tools rely on, so a missing reply or a gap in the path does not by itself prove a device is down.

DHCP hands out IP addresses automatically as leases; a device keeps its address by renewing, and an address it stops renewing goes back to the pool when the lease expires. A static address is typed in by hand, so it has to fit the network's range, mask and gateway, or the device cannot reach anything.

A port number tells a host which service a connection is for, so firewalls can allow or block services by port. Some older services send everything in cleartext, passwords included, with Telnet and SMB version 1 the classic cases, which is why guidance favors their encrypted successors and why an exposed remote desktop port is a common ransomware entry point.

A hypervisor runs virtual machines, keeping them isolated while they share the host's CPU, memory, storage and network. A Type 1 hypervisor installs directly on the hardware; a Type 2 runs on top of an ordinary host operating system, which then has to be trusted and kept patched along with it.

Hardware-assisted virtualization gives the hypervisor a more privileged processor mode and hardware page tables, which cuts the code it needs and strengthens isolation. Plan virtual machines against the physical host: NIST recommends never provisioning beyond a key physical resource such as CPU cores, and keeping memory over-commit, which lets total VM memory exceed the host's RAM, off by default.

Desktop virtualization delivers each user's desktop from shared infrastructure, so people can reach their own environment from almost any device, wherever the internet is available. It simplifies managing applications and desktops, and in one hospital rollout, outside access ran over a VPN to keep the connection secure.

The cloud models differ in who manages what. With IaaS you get processing, storage and networks and run your own operating systems and software; with PaaS you deploy your own applications on the provider's platform; with SaaS you simply use the provider's applications, often through a web browser, and manage none of the layers underneath.

Work in order: record the user's description, find the cause before you change anything, fix it, then prove the whole system works and write down what you found and did. Verifying matters because a technician takes nothing for granted; one fault you fixed can hide a second that is still there.

Change one thing at a time and test after each change. If you swap several parts at once you cannot tell which one mattered, and there is no sense risking a new part before you have found the fault. Isolate a suspect from the rest of the system when you test it, so a neighboring part cannot mask the result.

Before any fix that could touch data, make sure a current backup exists and that it restores. A backup is what lets you recover when a computer breaks, a step goes wrong or malware strikes, and testing it right after you make it is the only way to know it will work when you need it.

Reproduce a fault before you fix it, so you can tell when it is truly gone. Intermittent faults are the hardest: simply opening the case can make a heat-related fault vanish, so electronics bench technicians warm a suspect board gently until the fault returns, then cool suspect parts one at a time with a nonconductive coolant spray.

Support work is layered: network support specialists often work under administrators who handle the more complex tasks. Part of the job is telling team members and managers about major problems or recurring complaints, with the customer's own description written down so whoever picks up the issue starts from facts.

A few common complaints map to a few causes. No boot can be a fault caught in the firmware's early hardware tests or no valid boot device; a blank screen with fans running may be memory or the board, not the monitor; and a machine that slows after minutes of heavy work may be throttling from heat.

A static shock you can feel takes thousands of volts, but sensitive chips can be damaged by tens of volts, so harm happens without any sensation and may only show as an early failure later. Ground yourself with a corded wrist strap, keep parts in protective packaging until needed, and hold them by the body, never the pins.

Lifting and moving equipment by hand, especially repeatedly, puts you at risk of back and muscle injuries, which is why NIOSH built an equation to judge lifting tasks. Keep the bench orderly too: a screw or scrap of wire left inside a case can short a board, so track every fastener you remove.

Old electronics are not regular trash. Lithium-ion batteries go to battery recyclers or hazardous-waste collection, because in a garbage or recycling load they can start fires; CRT glass contains lead and counts as hazardous waste when it is thrown away rather than recycled; and drives should be wiped, then destroyed, before a machine leaves your hands.

A useful ticket records the user's own description of the problem, each defect you found, what you changed and the result. Writing down every fix as you make it keeps the trail clear for whoever reads the ticket next, and it lets you prove the repair worked instead of assuming it.

Keep an inventory that records each device's make, model, serial number and location, along with what information it handles. That record answers the first questions before any repair: exactly what the device is, where it lives, and how much depends on it.

Explain solutions in words a nontechnical user can follow, and listen closely, asking questions when the description is unclear. Many people who call support are already frustrated, so patience is part of the work, and walking them through steps succeeds only if they understand each one.

A SATA drive needs two cables: a data cable that runs to the drive controller on the motherboard and a power cable from the supply. Check both before blaming the drive, since a missing data link leaves it invisible to the system and missing power leaves it dead.

A graphics card belongs in a PCIe slot wired with enough lanes, since a card short of lanes runs below its rating. Heavy cards also throttle when they reach their temperature or power limit, so airflow and power delivery matter as much as the slot you choose.

Installation media works only if the firmware boots from it. Firmware searches for a boot device such as a USB drive and runs its loader, following a boot order kept in settings, so put the installer ahead of the internal disk in that order or the old system will simply start again.

Size a virtualization host by the virtual machines it must carry. The processor time guaranteed to each VM cannot add up to more than the host offers, so cores and clock capacity cap how many VMs fit, and generous RAM matters because NIST recommends keeping memory over-commit off by default.

A storage server is built around its disks: mirroring with RAID 1 or 10 survives a failed drive, and RAID 5 saves space by using parity. A single box is still a single location, though, so keep a copy somewhere else as well to protect against theft or a natural disaster.

A thin client does little computing of its own; it gives you an interface, often just a web browser, to applications that run elsewhere, as with cloud software or desktop virtualization. That makes the network and the servers behind it the real dependency, so plan for what happens when either is down.

A router has two passwords that matter: the Wi-Fi password and the admin password that controls every setting. Default admin credentials can be printed on the device or easy to find online, so change them first, then check the maker's site for newer firmware, which fixes flaws and security holes.

Every device on a Wi-Fi network joins by the same network name, or SSID, and a factory-default name can reveal the router model to an attacker, so rename it. Many routers broadcast 2.4 GHz (longer reach, slower) and 5 GHz (faster, shorter range, often less crowded), and can add a guest network with its own name and password.

A router's DHCP server hands out addresses automatically, and its client list shows which devices are connected. That suits phones, but a device others must reach at a known address, such as a camera you forward a port to, needs one that does not change: either set by hand on the device or kept for it by a DHCP reservation on the router.

Port forwarding tells the router to pass inbound traffic on a chosen port from its public address to one device inside, so remote clients can reach it. That deliberately opens a path the router would otherwise refuse. UPnP lets devices set up such forwards on their own, which is why CISA's home-network tips advise disabling it unless you need it.

Many routers' built-in firewalls add access controls and web filtering you can tailor, which is where content filtering usually lives. Priority is the other half: when a household streams, games, video-calls and uses work VPNs at once, everything lags, so dedicate the 5 GHz network to the most important uses or wire that device in.

Before buying hardware, move the router to a central spot, since placement alone changes coverage. A range extender or a mesh system can then raise signal strength in far rooms, but a direct Ethernet cable still gives a device the highest speeds and takes it off the crowded Wi-Fi, so wire anything that sits near the router.

DSL rides existing copper phone lines, cable uses TV coax and its speed varies with traffic load, and fiber sends light through hair-thin glass. In FCC measurements fiber had the lowest median latency, 7 to 14 ms, cable sat in the middle and DSL was highest, and cable and fiber held their advertised speeds more consistently than DSL.

A phone can act as a mobile hotspot, sharing its cellular connection with a laptop or another device that has no cellular radio of its own. Check the data plan first, because heavy use can run past data caps into overage charges, and mobile broadband is typically slower than wired or fixed wireless service.

Geostationary satellites orbit about 22,300 miles up, so each request makes a long trip and the high latency can impair highly interactive uses such as voice calls. Non-geostationary constellations orbit lower for low latency but need many satellites, handing your link from one to the next. Service needs a clear line of sight to the provider's satellite, and extreme weather can disrupt it.

The modem speaks your provider's medium, such as coax for cable service or a dish for satellite. The router sits behind it: your devices reach the internet through it, and it is usually where the Wi-Fi, the built-in firewall and the list of DHCP clients live. Knowing which box does which job tells you where to look when one part fails.

As a 2013 NIST guide, withdrawn in March 2026, describes it, a small resolver on your device asks a resolving name server when you type a web address, and that server answers from its cache or queries the DNS hierarchy for you. Each answer is a typed resource record: one type maps a host name to an IP address, while others point to a domain's mail servers and its name servers.

Sharing a folder lets other machines on the network open it, so scope it narrowly: share a dedicated folder rather than a whole drive, password-protect what you share, and allow sharing only on home or work networks. On public Wi-Fi, an open share is a way in for anyone else connected to the same network.

A client VPN extends a host into a private network by tunneling through public networks, which is why an organization may require one to reach its data. That is a different job from a consumer VPN, which CISA notes mostly shifts your risk from your internet provider to the VPN provider rather than removing it.

A proxy sits between clients and servers: it receives each web request, forwards it on, and relays the reply back. A forward proxy does this for clients going out, so it can allow listed sites and refuse the rest, and a browser pointed at a wrong or missing proxy can fail to load pages even when the network itself is fine.

In a 2013 NIST guide, withdrawn in March 2026, a DHCP server assigns an IP address to a device that asks, and it can also hand the device the addresses of the DNS resolvers it should use. The address is not permanent: it goes back when the device returns it, and some DHCP servers also add the device's name to DNS and remove it again at that point.

One common setup pattern: an unconfigured device broadcasts a temporary Wi-Fi network of its own, and when your phone joins it a setup page, a captive portal, opens so you can configure the device. Setup is also the moment to replace any default username and password, since defaults can be found online or printed on the device itself.

Segmentation keeps one compromised device from reaching everything else: on a well-separated network an intruder's foothold stays contained. At home the simple version is a guest network with its own name and password, which keeps a visitor's infected phone off your main network; larger networks use VLANs for the same isolation.

Firmware is the software inside a device, and on many gadgets you must check for updates yourself on the maker's website or keep its phone app current. Old devices you no longer use are a weak point because their security goes out of date, so disconnect them from the network rather than leave them running.

A liquid crystal panel makes no light of its own; a backlight shines through it, and many LCDs of the 2000s were lit by a CCFL tube driven by a high-voltage inverter. So an image that is faintly visible under a strong light but otherwise dark points at the backlight or inverter, not the panel. OLED screens emit their own light and need neither.

A lithium-ion pack holds energy even when the laptop is off or the battery seems used up, and a damaged or crushed cell is a fire hazard. Not every battery is meant to be removed by the user; when one comes out, cover its terminals with non-conductive tape, bag it on its own, and take it to a battery recycler rather than the trash.

Thermal paste fills the microscopic air gaps between the processor and its heat sink so heat crosses that joint easily, while the fans carry it away. When either path fails, temperatures climb, and a hot processor throttles, cutting its clock speed and sometimes destabilizing the system, so slowness under load can be a cooling fault.

In the classic BIOS startup sequence, the firmware walks its list of boot devices, such as a hard drive, optical drive or USB drive, and boots the first one with a valid boot record; that record's loader then starts the operating system. A no-boot-device error means the search found nothing usable, while a drive that is found but fails to start points at the loader or the OS.

Mirroring (RAID 1) and parity (RAID 5) let an array rebuild a failed disk's data from the others, so it keeps running degraded; RAID 6 keeps two sets of parity, and striped RAID 0 keeps none, so one failure there loses everything. A degraded mirror or single-parity array has already spent its redundancy, so until the rebuild finishes the data depends entirely on the disks that remain.

Before treating slowness as a drive fault, rule out the processor: malware and cryptomining can monopolize it and make the whole system, storage included, feel slow. Know the media too: NIST's 2014 media-sanitization guide, since withdrawn, noted that flash drives spread writes over spare cells with wear leveling and that a device may not clearly report its kind, so check first.

Received signal strength (RSSI) is a useful first look, but it swings with movement and surroundings and weakens through walls, so take several readings rather than one. Check the neighbors too: if many other networks crowd the 2.4 GHz band, the 5 GHz band is often clearer, though its signal covers a shorter range.

In the original 802.11 design the client, not the network, decides when to move to another access point, and it decides without seeing the whole network. Devices often cling to a weak access point even when a nearby one would give a steadier link, a pattern called the sticky problem, so a device can perform poorly right beside a strong access point.

Wi-Fi shares the unlicensed 2.4 and 5 GHz bands with other gear. One study reported microwave ovens cutting throughput by 28 to 49 percent within 1 to 3 m and Bluetooth speakers by 5 to 9, though another found no throughput effect at all; analog cordless phones and analog wireless video cameras, which transmit continuously, can cut it by 90 percent or more.

A captive portal pairs a web server with a DNS redirect, so a newly joined device lands on one page whatever address it asks for. Smart devices use the trick for first-time setup, and attack tools use it too: they stand up a fake portal so a visitor believes they are being asked for the Wi-Fi password and types it in.

On Android, Google Play Protect scans apps to detect and block malicious ones. CISA's guidance for high-risk mobile users is to keep it enabled, review its scan results regularly, and use caution with third-party app stores or sideloading, which means installing an app from a source other than the official store.

Mobile device management lets an organization secure, monitor and support enrolled devices, pushing apps, data, configuration and security settings over the air. It can also send a remote wipe that deletes a device's data, which is worth making clear to anyone enrolling a personal phone.

Both major phone platforms list in their settings which apps can reach sensitive data such as location, camera and microphone. CISA's guidance is to revoke any permission that is unnecessary or excessive for what the app actually does, and to avoid granting those three unless truly required.

Updates patch security holes and fix bugs, and attackers keep targeting a hole for months or years after its fix ships, so turn on automatic updates. Install them only through the device's own updater or the vendor's site, never from a link in an email, and avoid updating over untrusted networks such as hotel Wi-Fi.

Backups can live on external media or with a cloud service, and either protects you when data is lost, corrupted, infected or stolen. Encrypt the backup so it stays confidential, and make copies regularly; a backup app can automate the schedule so protection does not depend on someone remembering.

Laser printing runs six steps: charge, expose, develop, transfer, fuse and clean. That order maps defects to parts: a scratched photoconductor drum leaves streaks, a dirty drum or a wrong fuser temperature leaves stray specks, and since fusing bonds toner with heat and pressure, toner that rubs off points at the fuser.

Color printers build images from cyan, magenta, yellow and black, and getting the same color out of different printers and scanners is hard enough that it often takes calibration. To calibrate an instrument is to find how its output varies, by special measurements or by comparison with a known standard, which is what makes two devices' results comparable.

A sheet passes the stages of the paper path in a fixed order, transfer from the drum and then the fuser, which bonds the toner with heat and pressure, and a double-sided job adds a pass for the second side. Noting exactly where a sheet stops tells you which stage to inspect, and double-sided printing is also linked to stray toner marks.

Direct heat-printing label and receipt printers put down no ink or toner: they heat coated stock whose leuco dye and acidic developer react to turn the heated spots dark. That is why only the matching coated media will print, and because warmth alone triggers the reaction, labels and receipts stored near heat can darken where nothing was printed.

Each USB generation raised the ceiling: USB 1.x reaches 12 Mbps, USB 2.0 480 Mbps, USB 3.0 5 Gbps and USB4 40 Gbps. Most devices still carry USB 2.0 and USB 3.0 ports, so check which one a fast external drive is plugged into before blaming the drive for slow transfers.

Thunderbolt 3 and later run over the USB-C connector, so a port that looks identical may or may not carry it. One dock's port list, for example, rates its Thunderbolt 3 USB-C port at 40 Gbps beside a plain USB-C port of the same shape rated at 5 Gbps. Check the port's specification before promising Thunderbolt speed.

A USB Power Delivery charger, as a 2019 DOE rule describes the standard, can move between five, nine, fifteen and twenty volts on request from the device over a digital protocol, and it allows more current at the top level than at the lower ones. That negotiation is how one charger can serve a phone, a tablet and a laptop with different input needs.

USB-C names the connector, not what the port behind it supports. One USB connection can carry several functions at once, and power over it is negotiated, but whether a given port offers Thunderbolt speed depends on how that port was built, so check its specification; a look-alike cable can also quietly limit what you get.

An adapter can add a missing connector, such as an Ethernet adaptor for a laptop that has no Ethernet port, but it can only pass on what the port behind it supplies. Data still moves at that port's USB generation speed, and a basic USB port supplies just 5 volts at up to 500 mA, which limits what an adapter can power.

A serial link joins a terminal device to a communications device, and it works only if both ends agree on the standard, the transmission speed, the bits per character, and whether parity and stop bits frame each character. When a serial console shows garbage or nothing, compare those settings on both ends first.

Three IPv4 blocks are set aside for private networks: every address beginning with 10, the addresses from 172.16 up to the end of 172.31, and every address beginning with 192.168. They are not routed on the public internet, so a router translates them to its public address. Watch the middle block's edges, and note that 169.254 addresses are link-local, a separate range.

In a 2013 NIST guide, withdrawn in March 2026, DNS answers are cached for their time to live, so after an entry changes a client can keep using the old address until its cached copy expires or is cleared. That guide says devices get two or more resolvers for fault tolerance and warns of malware that swaps a system's resolver for a malicious one, so check which servers a device uses.

Bandwidth is how much data a link moves per second, measured in Mbps; latency is how long a packet takes to cross it. They fail differently: a download can be fast while calls feel sluggish, because satellite links or a crowded Wi-Fi network add delay. Higher speed tiers do tend to have lower latency, since each packet takes less time to send.

A packet counts as lost when it never arrives, or in FCC measurements when it takes longer than three seconds. Industry and ITU standards put about 1 percent loss as the point where highly interactive uses such as voice calls degrade noticeably, so a ping test that drops a few packets in every hundred points to a real fault even when speed tests look fine.

MAC filtering admits only devices whose hardware address is on the router's list, which keeps unknown hardware off but does nothing to separate the devices it lets in. A guest network gives that separation: visitors join with their own network name and password, so malware on a guest's phone stays off your main network and devices.

Typical signs are a slow or crashing computer, a browser home page that changes or redirects, new toolbars or add-ons, and system tools that are suddenly disabled. Sustained high processor use and heat with no obvious cause can mean someone is using the machine to mine cryptocurrency. First stop signing in to accounts from it, then update the security software and run a scan.

Security software can detect, quarantine and remove known malware, but it relies on recognizing known code, so brand-new threats can slip past. After a suspected infection, stop using accounts from the device, update protection and scan, then change passwords and turn on two-factor authentication in case credentials were taken. Network segmentation limits how far it spreads.

A changed home page, unexpected redirects and toolbars or add-ons you did not install are signs of malware in the browser. Some sites mine cryptocurrency through JavaScript only while the page is open, while persistent miners keep running after you leave. Get software by typing the vendor's address, since scammers place bogus software ads on search engines.

Lockout policies commonly allow three to five bad attempts in a set period, so repeated guessing at one account locks it quickly. Attackers who know this spray one password across many accounts instead, staying under the threshold, so an absence of lockouts is not proof that no one is trying; CISA's first listed defense is multifactor authentication.

Encrypting the whole disk makes a lost laptop useless to anyone without the password or PIN. The same property cuts both ways: lose every way to unlock it and the data stays locked to you too, so a recovery key has to be stored safely before trouble, and any copy kept outside the device is also a way for someone else to read the data.

Not every attack comes over the network: someone can read a screen over your shoulder, which an inexpensive privacy screen protector blocks, or steal the device and gain unrestricted access to its data and linked cloud accounts. Encrypting storage and requiring logins for cloud apps limits what a thief gets.

Safe Mode starts Windows with only a few services running, so a fault that disappears there likely comes from something that normally loads at startup. The trade-off is that security tools may not be running either: the Snatch ransomware reboots machines into Safe Mode precisely to encrypt files while antivirus is out of the way.

Startup is a chain: the firmware finds a boot device, runs its boot loader, and the loader starts the operating system. Work out which link broke before reaching for a tool. Firmware damage has its own recovery path: many systems let someone at the machine restore a known-good BIOS, and some detect corruption and restore themselves from a second firmware copy stored apart.

Rolling back returns to the version that worked before an update, which fits a fault that began right after updating. NIST's 2011 BIOS guidelines show the catch for firmware: an older version can carry known vulnerabilities, so systems that authenticate updates may require a special rollback process, such as someone physically present at the machine.

Firmware searches its boot list, which can include USB drives, and boots the first device it finds with a valid boot record. A flash drive or external disk that carries one, ahead of the internal disk in that list, gets booted instead, so the machine starts the wrong system or stalls; unplug attached storage and retry before assuming the internal drive has failed.

A voltmeter is always connected across the two points being compared, in parallel with the circuit, which lets you read a supply's output while it runs. An ammeter measures current, so it is placed in series, in the current's path. Matching the meter's mode and connection to the quantity is the first step of any bench measurement.

Manufacturers commonly supply utilities that update the BIOS from the normal operating system, either directly or by scheduling the update for the next reboot. If an update goes wrong, many systems let someone at the machine restore a known-good version, some restore themselves from a second firmware copy stored apart, and authenticated update schemes may demand a special process before an older version goes back on.