Hacker News Reader

Comment by willis936 | Hacker News Reader

10 ns is the quoted accuracy, but it’s not easy to verify that and it is very easy to mess up something at that timescale. If something is cheap, I’d be willing to bet their PPS output is not going to be 10 ns accurate. Most applications won’t tell the difference between 10 us and 10 ns accurate, but when you need it, you’ll know it.

View on HN

The jitter can be a really important parameter.

for example in pdf: https://processors.wiki.ti.com/images/f/f1/TI_GPS_PPS_Timing...

My masters thesis is on jitter measurement techniques in 25 Gbps baseband signals :)

Jitter can be tracked and filtered in PPS signals. The related parameter, wander, is interesting to me. Phase noise above 10 Hz is jitter. Phase noise below 10 Hz is wander.

Why?

¯\_(ツ)_/¯

Allen variation is the typical metric for wander.

Fascinating! Do you know of any good textbooks or application notes that go into jitter measurement techniques?

I found this note from Keysight, is this is a good place to start?

http://literature.cdn.keysight.com/litweb/pdf/5988-6254EN.pd...

That note is interesting, but a bit dense and most of it is standard specific implementation details that aren't too important without first understanding why the standards were written the way they were. That question isn't usually well documented, so a bottom-up understanding of the testing is valuable imo.

If you want to become a time-nut, then check out time-nuts[0]. They know their stuff and will be the most valuable resource on topics related to GPS timing: wander, allen deviation, and jitter of various sources (quartz oscillator vs rubidium, etc.)

I will say that the big test and measurement (T&M) players (Keysight, Tektronix, Rhode&Schwartz) have been the main drivers of T&M research. It's a strange field where academic work and commercial interests line up, so all of the academics work for commercial companies. They'll happily share their results and preferred approaches (except for the secret parts) in white papers and app notes. I don't think there is a good centralized repository of such documents, so it may require some hunting. Knowing keywords might be helpful. If you're mainly interested in jitter then hot things to search for are "jitter", "jitter decomposition", "TIE", "wander", and "dual-Dirac" (just off the top of my head, I'm sure the rabbit hole goes deeper if you start digging into more recent work).

I was in T&M for 7 years. I'm not currently but likely will be in the future. This[1] is the bible I learned on. It's $120 new and not too easy to find free (most of the people who would find it interesting can afford it, ie not in curriculum). Most of my learning was through osmosis over the first few years, then I taught myself the rest. I only used the book as a reference after the first year, but I wouldn't say I knew it all until maybe 5 years in (I was a student at the time too). When I flipped through it towards the end of my time I found it was a good teaching source. I suffer from the curse of knowledge, but I think if you have EE chops then this book will make perfect sense and even help build the intuition necessary to be a good test engineer.

The book is all practical T&M techniques for high-speed SERDES, so wander is outside of the scope of the book. There are other bits in the book besides jitter too (lots of tasty stuff on encodings and impedance).

0. http://www.leapsecond.com/time-nuts.htm

1. https://books.google.com/books/about/An_Engineer_s_Guide_to_...

Thanks so much! Time-nuts looks like a gem. I was starting to learn about serdes, so more the reason to check out [1].

I know lots of applications that could take advantage of 10 picosecond accuracy, if only it was feasible without spending hundreds of thousands of dollars.

You can get multiple devices and compare the output. Doesn't give you absolute accuracy of course. But you can set one as "master" and diff to that.

I dunno.

Back in the day when "selective availability" was turned on they added "brown noise" with strong low frequency components that would defeat attempts to use temporal averaging to approve GPS accuracy.

I remember setting a receiver up outside and looking at the tracks go in circles like a spirograph.

Selective availability is off since almost 2 decades.

That's how long ago it was.

Yes. But what does location accuracy have in common with PPS accuracy besides that you need a fix?

Everything! :D GPS measures the time difference between multiple satellites that broadcast their location and time. You need very accurate timing information to have accurate location. Timing error that manifests itself in PPS would also manifest itself in location data.

Reply on news.ycombinator.com