aioscilloscope / NovaOS vs stock / Siglent SDS800X HD

NovaOS vs stock on the Siglent SDS800X HD

NovaOS runs on our SDS814X HD from a USB stick, in RAM, beside Siglent's firmware, which stays installed and untouched. What is measured so far is below, each result with its date and method; Siglent's firmware on the same unit is measured the same way next, and the list after the table says what that run adds.

The family page: Siglent SDS800X HD.

CH1Measured

Each row: what Siglent's firmware gave, what NovaOS gave, the date of each, and how we measured it, with the raw data where we kept it. Every figure here was measured within a day of this page's build (2026-10-17). "Not yet measured" and "not tested" mean we have no such measurement from that day.

WhatSiglent's firmwareNovaOSMethod
Waveforms a second, by the scope's own acquisition count (:ACQuire:NUMACq?), 1 us/div, 10k pointsnot yet measured725.4 a second (median of nine 10 s windows) 2026-10-17how
Whole 10k-point records a second over SCPI (:WAVeform:DATA?)not yet measured13.61 a second (0.272 MB/s) 2026-10-17how
Answer time of :TRIGger:STATus?, 50 reads back to backnot yet measuredp50 1.85 ms, p99 6.06 ms, max 6.76 ms 2026-10-17how
A screenshot over SCPI (:PRINt? PNG)not yet measureda 1024x600 PNG in 0.509 s, 3 of 3 tries 2026-10-17how
SCPI round trip, 500 *OPT? queries back to back on one connectionnot yet measuredp50 1.69 ms, p95 4.02 ms, p99 5.40 ms, max 6.11 ms 2026-10-17how
Sustained *OPT? queries a second over 10 snot yet measured453.3 a second 2026-10-17how
Remote screen (VNC on port 5900), frames a second over 10 snot yet measured3.79 frames a second at 1024x600 2026-10-17how

CH2Not yet compared

Each measured the same way on both systems, Siglent's firmware on the same model:

CH3How we measured

Enough to repeat each measurement on your own scope: the units, the scene, the cabling, the arithmetic, the exact commands, what we know is uncertain, and the raw data. Times are UTC.

NovaOS against Siglent's firmware on the SDS814X HD, read from the wire

Units

  • NovaOS: an SDS814X HD running NovaOS from a USB stick, in RAM (its build of 2026-10-17); the scope's own flash is never written.
  • Siglent's firmware: the same SDS814X HD booted on its own flash (firmware 4.8.12.1.1.6.5, the *IDN? firmware field), no options; not measured yet.

When

NovaOS: 2026-10-17, 13:31 to 13:34 UTC, three runs back to back. Siglent's firmware: not measured yet.

Scene

  • The CAL output (a 0 to 2.93 V, 1 kHz square) on CH1 at 1 V/div; 1 us/div, 10k points, 1 GSa/s; edge trigger on CH1 at 1.45 V, AUTO; the other channels off; nothing else connected.
  • One client at a time: SCPI on the raw port 5025, the remote screen on VNC port 5900. The setup is read before and after each run and was the same.

Equipment and cabling

  • A PC on the same network; no other instrument.

Arithmetic

  • Round trip: the time from sending a query to the last byte of its answer, on one open connection; p50, p95 and p99 are the percentiles of 500 such times.
  • Queries a second: answers received in 10 s, divided by 10.
  • Waveforms a second: the scope's own acquisition count read at the start and the end of a 10 s window, the difference divided by the time between the two answers; three windows a run.
  • Records a second: 20 whole 10,000-point records pulled one after another, divided by the time they took.
  • Remote screen: incremental frame updates received in 10 s, divided by the time between the first and the last.
  • A screenshot: the time until the whole image has arrived.
  • Each figure is the median of the three runs (for waveforms a second, of the nine windows); the range follows it where given.

Tool

Our bench script for this comparison, the same commands on either firmware; each step is plain SCPI or VNC over the network, so any client that times its own requests can repeat it.

*OPT? sent 500 times back to back on one connection to port 5025, each round trip timed; the error queue (:SYSTem:ERRor?) read before and after.
*OPT? sent as fast as answers come for 10 s.
:PRINt? PNG sent once a run.
:WAVeform:DATA? for 20 whole records of channel 1 in 16-bit words; the :WAVeform: settings read first and put back after.
A VNC client on port 5900 asking for incremental updates for 10 s.
:TRIGger:STATus? sent 50 times back to back, each timed.
:ACQuire:NUMACq? read at the start and the end of three 10 s windows.

Known limits

  • One unit. Siglent's firmware has not been measured yet on the day: every stock cell says so until it is.

Raw data