Getting it onto the wire · 05 of 11

Ground first

The analyzer decides high or low by comparing each channel against its own ground pin. Leave that pin off and the comparison is against nothing in particular — and the result is not an empty screen, it is a capture full of edges nobody sent.

What "high" is measured against

A logic analyzer has no idea what a volt is in the abstract. Each input is compared against a threshold, and that threshold is referenced to the analyzer's own ground pin. If the pin is on the same copper as the circuit, the comparison means what you think it means. If it is not, the comparison is between two independent things that are free to drift apart.

They will drift apart. Two USB-powered devices on the same laptop are tied together loosely through the machine; a battery-powered target and a mains-powered laptop are not tied together at all. Either way the difference is not zero, it is not constant, and it moves fast enough to cross a logic threshold.

Ground is the wire the other nine are measured against
GND fitted
The ground wire
Channels wired
1
Ground wires needed
1
Edges in the capture
5
One ground is enough and it is not optional. The analyzer decides high or low by comparing each channel against its own ground pin, so that pin has to be on the same copper as the circuit. Fit it first, before any channel, and fit it even when you are only probing one wire.

Why it is worse than a blank screen

If a missing ground gave you a flat trace, nobody would lose an evening to it. The problem is that it gives you the opposite.

Without a reference, the input is comparing against noise, so it crosses the threshold constantly: mains hum, the target's switching regulator, the capacitance of the probe lead itself. The result is a trace that is busy — edges everywhere, at plausible-looking rates — and a decoder that will pick real bytes out of it every so often, because that is what pattern matching does to random data.

So the false conclusion is available and it is the wrong one: the bus looks like it is running and sending rubbish, which points straight at the firmware.

The rule

Fit ground first, before any channel, on every capture, including the ones where you are only probing one wire.

There is nothing more to it than that. It is one hook clip onto a GND pin, a ground pad, the negative rail of a breadboard, or the barrel of a mounting hole that goes to ground. Any of them, as long as it is the same ground the signal is referenced to.

One ground, or several?

One is enough for the kind of work this kit is sold for. The connector has one GND position and it is sufficient for eight channels of microcontroller logic across a hand-sized board.

More than one becomes worth it when the probe leads are long and the edges are fast, because each lead and its return form a loop, and a big loop picks up whatever is nearby. If you are seeing ringing that follows the other channels' activity rather than the signal you are probing, shortening the ground lead and routing it beside the signal lead is the fix — not adding channels.

The one case where ground is not enough

If the target and the host are on separate mains supplies with a real potential between them, connecting the grounds does not just fix the measurement, it makes a current path. On the low-voltage, USB-powered benches this kit is for that is not a concern. On anything mains-referenced — a switching supply's primary side, anything with a live neutral — it is a serious one, and neither this instrument nor any other unisolated one belongs there.

When it does not work

The trace is dense with transitions I cannot account for

Check the ground wire before you read another line of firmware. Without a common reference the threshold gets crossed by the difference between two power supplies, which is noisy, continuous, and looks exactly like a busy bus. This is the single most common first-session failure and it costs people an evening every time.

The decoder found a few real-looking bytes in obvious noise

It will. A decoder does not know the difference between your data and mains hum crossing a threshold — it looks for a pattern and reports what matches. Occasional plausible bytes in a sea of nonsense is evidence of a missing ground, not evidence of a partly working bus.

I grounded to the analyzer's case rather than the GND pin

The case is a moulded plastic shell and is not connected to anything. Ground has to be the GND position on the connector, or the GND hole or pad on whichever breakout board you are using.

Everything is grounded and the trace is still noisy on one channel

Then it is a real signal-quality question rather than a reference question — a long unshielded probe lead next to a switching supply, or a floating input on a pin nothing is driving. Move the ground clip closer to the signal you are probing, and check the pin is actually driven rather than left as a high-impedance input.

Where this goes next

Three attachments in the bag, three component pitches, and the point at which soldering a wire stops being a defeat and becomes the fast option.

What you can grab

Edit this page — content/books/logic-analyzer/ground-first.mdx

Community

Questions about this product

See what other owners have asked, and read their solutions.

Ask a question ↗

Logic Analyzer Kit, 8-Channel 24 MHz

Loading discussions…

Discuss this article

Ask about this page. The answer stays here, on the page it belongs to, for whoever hits the same wall next.

Browse Modules and blocks on the forum