Bits versus bytes
One byte contains 8 bits, so 100 Mbps is a theoretical 12.5 MB/s before overhead. A download manager showing 11 MB/s on that line is consistent with ordinary overhead.
Measure, then isolate
Trivetek is a new independent home-network guide for everyday users. It helps you convert units, record speed tests and separate Wi-Fi problems from ISP problems. It is not an internet provider or a repair service.
This is an independent measurement reference, not an ISP, a speed-test operator or a diagnosis service for a specific line.
Scroll the strip sideways — 6 cards.
One byte contains 8 bits, so 100 Mbps is a theoretical 12.5 MB/s before overhead. A download manager showing 11 MB/s on that line is consistent with ordinary overhead.
Idle latency is measured on a quiet link; loaded latency is measured during a transfer. The gap between them is the queueing signal.
Jitter describes variation in latency rather than the average latency itself, so report it beside the mean, not instead of it.
Packet loss is the share of sent packets that do not arrive. Sustained loss forces retransmissions and reduces effective throughput.
The FCC fixed-broadband methodology uses three concurrent TCP connections for download and upload throughput measurements.
A single run cannot separate device, Wi-Fi, router and ISP effects. Only repeated comparisons with one changed variable can.
Six steps from units to isolation
Divide megabits per second by 8 to get megabytes per second, then subtract overhead separately.
Network speed is measured in megabits per second (Mbps), while file sizes and disk transfers are usually shown in megabytes (MB). One byte contains 8 bits, so 100 Mbps corresponds to a theoretical 12.5 MB/s before protocol and retransmission overhead. A browser download that shows 11 MB/s on a 100 Mbps line is consistent with ordinary overhead, not proof of a fault by itself.
To estimate an ideal transfer time, multiply the file size in megabytes by 8, divide by the speed in Mbps, and treat the result as a floor. An ideal 250 MB transfer at 100 Mbps takes 20 seconds by 250 × 8 ÷ 100, before overhead. Real transfers add TCP headers, acknowledgements and retransmissions, so the observed time is longer.
Keep the two conversions separate in your notes: unit conversion explains the gap between the advertised number and the file manager, while overhead explains the gap between the ideal time and the clock. Mixing them produces a false diagnosis of a slow line.
| Reading | Unit | What it answers | How to use the reading |
|---|---|---|---|
| Download throughput | Mbps | How fast data arrives | Baseline and repeat runs |
| Upload throughput | Mbps | How fast data leaves | Baseline and repeat runs |
| Idle latency | ms | Round-trip delay on a quiet link | Quiet-link baseline |
| Loaded latency | ms | Delay while a transfer runs | Change during a transfer |
| Jitter | ms | How much latency varies | Spread beside the mean |
| Packet loss | % | Share of packets that never arrive | Percentage over the same run |
| Ideal 250 MB transfer at 100 Mbps | s | Floor before overhead | 20 s before overhead |
| Theoretical rate at 100 Mbps | MB/s | Bit/byte conversion | 12.5 MB/s before overhead |
| FCC throughput method | TCP connections | Published measurement method | 3 concurrent connections |
| FCC idle-latency sampling | packets/hour | Published measurement method | ~2,000 packets/hour |
Each panel pairs a unit with the question it answers, so the six readings stay distinct.