What it actually records

One bit, per channel, per tick

Every sample interval, the analyzer looks at each of its eight inputs and asks one question: is this above the threshold or below it. The answer is a bit. It stores the bit and moves on.

That is not a simplified description. There is no analogue path, no voltage stored anywhere, and no setting that changes it. A capture is a rectangle of bits — eight wide, as many rows deep as the capture ran — and everything either application shows you is drawn from that rectangle.

What the trade buys

Throwing the voltage away is what makes the rest possible. A bit per channel is a byte for all eight, so a full-rate capture is a byte every 41.7 nanoseconds and nothing more, which is cheap enough to stream continuously over a USB cable for as long as you leave it running. That is why you get eight channels at once, and why a capture can be seconds long rather than a screen wide.

It is also what makes a decoder possible. A protocol decoder does not want a waveform; it wants to know which side of the threshold each wire was on at each instant, because that is what the protocol was defined in terms of. Handing it bits rather than voltages is handing it exactly the thing it needs.

The two things it cannot do

It cannot tell you about signal quality. A rising edge that overshoots to 4.5 V and rings for 200 ns is recorded as a rising edge, because every one of those excursions was on the same side of the threshold. If you are chasing reflections, slow rise times or a driver that cannot hold a line, the trace here will look perfect the entire time.

It cannot tell you about a level that is wrong but stable. A 3.3 V rail loaded down to 1.6 V is doing something obviously wrong, and this instrument records it as a wire that is low and stays low. The capture is not misleading — it is answering the question it was asked, correctly — but the question was never "what voltage is this".

Both of those are meter and oscilloscope work. It is worth knowing the boundary early, because the alternative is an hour spent looking for a setting that does not exist.

What the threshold is, and is not

There is one, it is fixed, and none of the three boards in this kit does anything to a signal on its way in — no divider, no shifter, no clamp. So a channel either crosses the threshold or it does not, and the practical rule that follows is short:

• 3.3 V and 5 V microcontroller logic is what the kit is for, and it is what it is good at.
• 1.8 V logic is the case to test rather than assume. Capture a signal you can independently verify is toggling, and check the trace toggles with it.
• Motor supplies, unregulated rails and anything above 5 V stay off the channels entirely. There is no protection here worth relying on, and this is the one mistake that ends with buying another analyzer.

The resolution floor

Because samples are taken on a fixed grid, every edge in a capture is recorded at the tick it was noticed on rather than the instant it happened, and every duration is a count of ticks. At the top rate a tick is 41.7 ns.

That is the floor under every timing number this instrument will ever give you, and it does not go away because the readout prints more digits. Logic 2 has the figure that makes it concrete, at the point you first reach for a measurement.