[time-nuts] Extended length Phase Noise Measurements with TimeLab

Tom Van Baak tvb at LeapSecond.com
Fri May 3 17:16:36 EDT 2013


For long measurement runs especially, I almost always do my data collection separately from my data analysis and plotting. That way there is no TimeLab CPU or RAM limit. You can start/stop/exit TimeLab or Stable32 or Plotter or Excel at will without worrying about losing raw data.

The data collection tool I use does a file checkpoint every minute, so if there is an app or OS or PC or cable or switch or TSC glitch or TSC crash or lab power failure I don't lose hours or days or weeks of data. It also means I can use old cheap PC's that run at 1% busy (no fan) to log tens of data streams instead of having to have an instance(s) of TimeLab peg my CPU and turn my laptop into a hair dryer.

I collect TSC 5110A and 5120A data from serial (1 Hz) and LAN (1 kHz) ports. I have never, ever, had problems with serial data. Serial ports may be old fashion but they are stateless and so more reliable for data logging than stateful USB or LAN connections.

I do have issues with long-term reliability of the TSC 5120 LAN port data stream. The TCP timeout bug was introduced in one of the very early firmware upgrades and no one has been able to track it down. I've spent hours snooping packets. The pre-production TSC5120/BSD firmware used to output 1 kHz phase samples evenly about every millisecond over ethernet. At some point before production the firmware "got smart" and buffered packets into chunks as large as 1024 or 5120 or 8192 samples each. My hunch is this optimization correlates with the rare TCP hangs that users see. I've tried multiple PC's and multiple OS's with the same result so I'm pretty sure it's the TSC firmware, and not me or my PC's or even TSC hardware to blame.

I can't help you with the "measurement  parameters exceeded" problem. I know the TSC boxes have limits on dynamic frequency variations. To debug, or at least to get around this, I would run a high-res frequency counter (like a 53132A or CNT-90) in parallel with the TSC analyzer, since those counters have no dynamic range limits.


----- Original Message ----- 
From: "Tom Knox" <actast at hotmail.com>
To: "Time-Nuts" <time-nuts at febo.com>
Sent: Friday, May 03, 2013 11:21 AM
Subject: [time-nuts] Extended length Phase Noise Measurements with TimeLab

> Recently I was looking at a co-workers Lab View measurements taken with a 5120A. She is studying vibration on Phase Noise. I asked how Lab View compared to TimeLab and she said in many ways she prefers TimeLab but ran into a problem. The research she is currently conducting requires collecting data in extremely long measurement periods beyond the TimeLab limits. Is she missing something? If not is measurement length an easy fix? One other problem during her project was related to the 5120A, occasionally the vibration must hit a resonance and the 5120A stops measurement saying measurement  parameters were exceeded. It would be nice to be able to adjust or disable that feature. Thanks in advance for your ideas.
> Thomas Knox

More information about the time-nuts mailing list