Runs entirely in your browser

Test your own connection.

Every request below leaves your device and goes straight to the resolver. Nothing is submitted to PacketPulse’s server, so what you see is the path from where you are sitting — not from where we are.

The address your requests leave through

Checking…

Asking 1.1.1.1 — one of the endpoints measured below — what address it sees.

DNS query round-trip · average

—ms

Over HTTPS, not ICMP

Fastest

—ms

Best round-trip seen

Slowest

—ms

Worst round-trip seen

Jitter · RFC 3550

—ms

Variation between consecutive replies

Query loss

—%

Queries that never came back

Voice quality · MOS

—

Estimated from the figures above

Round-trip, sample by sample

Probe log

Ready. Nothing has been sent yet.

This is not ping. A browser cannot send ICMP — there is no API for it. These figures are HTTPS round-trips to a DNS-over-HTTPS resolver, so they read higher than ping to the same address and will not match PacketPulse’s server-side sweep. Both are true; they measure different layers from different places.

How a browser can measure this at all

Worth understanding before you quote a number from it, because the method determines what the number means.

It leaves from your IP

The page holds no server component. Requests go from your browser straight to 1.1.1.1 or 8.8.8.8 over your own connection and router. Open your browser’s network inspector while a test runs — you will not see a request to PacketPulse.

DNS-over-HTTPS, not ICMP

Browsers cannot open raw sockets, so ping is off the table. Instead each sample is a real DNS query over HTTPS, and the answer is checked — a resolver that replies with nothing useful counts as a failure rather than a fast success.

The first request is thrown away

Opening a connection costs a DNS lookup, a TCP handshake and a TLS negotiation — roughly three extra round-trips. We pay that once, untimed, and measure only what follows. Skip this step and the first sample reads around 138 ms against a 30 ms baseline, dragging the average with it.

Jitter follows RFC 3550, the RTP standard — a smoothed measure of how much consecutive replies differ. A connection that drifts slowly has a wide spread and almost no jitter by this definition, which is the distinction that matters for voice.
“Query loss”, not “packet loss”. No packets are visible from here. What is counted is queries that did not come back inside the timeout — the honest analogue, and a different measurement.
Only two resolvers are offered. A resolver is usable here only if it serves DoH and permits cross-origin requests. Quad9, OpenDNS, AdGuard and others do not, so they are left out rather than listed and broken.
Any HTTPS host can be reached for timing, including your own. The browser cannot read the response, so the result is “it answered, in this long” rather than a status code — enough to tell a reachable endpoint from an unreachable one.
For true ICMP, use the desktop app. A native process can open an unprivileged ICMP socket and measure real echo replies and real packet loss — directly comparable to a server-side sweep.