Test equipment and 1PPS inputs

[japlytic]

You've most likely seen test equipment with an external reference clock input, but I think a 1PPS input (or support for it) should also be included considering it is easier to route a 1PPS signal over long distances than a 10 MHz signal.

When routing the 1PPS signal and serial port data from a GPS receiver on the rooftop, I prefer fibre optic cables over coaxial or twisted pair since fibre optic cables are immune to ground loops, electrical safety issues and EMI, along with using a power supply with a high degree of isolation.

[stj]

Re: Test equipment and 1PPS inputs

how is that usefull?
1 pulse per second means your going to have to start looking at how clean the leading edge is, and what PLL is going to sync with it?

[Curious.George]

Re: Test equipment and 1PPS inputs

how is that usefull?
1 pulse per second means your going to have to start looking at how clean the leading edge is, and what PLL is going to sync with it?

Depends more on how consistent the edge is and how tightly you want to lock to it (and over what period). Obviously, a higher frequency reference gives you more opportunities to lock in a shorter period of time.

I've long been using a software PLL to frequency lock (not PHASE lock!) to the AC mains in most of the products I design. I bandpass filter the AC line frequency to eliminate glitches, noise, etc. (this is relatively easy to do algorithmically; a fair bit more difficult with analog circuitry!) and capture the "(local CPU) time" of each transition. Then, slowly (once locked) use this to tweek the local time-of-day clock -- taking advantage of the long-term frequency stability of the AC mains and relying on the stability of the CPU clock for the short term.

(Obviously doesn't work well for devices that see lots of short-term environmental changes that can screw up the CPUs short term stability.)

NTP (and PTP) use similar algorithms.

[redwire]

Re: Test equipment and 1PPS inputs

Utility/Power generating industry I worked in uses master GPS 1PPS sync pulse for all the datarecorders. Regulations demand you have data which is time-synchronized to the grid (phase accurate), so they can track outages back to root cause. Otherwise, you never know which breaker tripped first.
Example is Arbiter Systems 1092C GPS clock popular in big substations etc.

The problem was long cables cause delays and soft rise/fall times when sending the 1PPS pulse all over the plant. So you get 10's usec error from that, even with proper termination resistors.


For S/W PLL, as I have seem those algorithms fail. They are also used in pipeline industry, where 1PPS is used to sync interrupters for cathodic protection.
The S/W bug was to only move the PLL if the sync pulse was within a time window, say 100msec. Noise, interference would sometimes knock the S/W PLL outside that window and it would never re-synchonize. Caused a lot of grief, having to reboot equipment to get it to synchronize.

[Curious.George]

Re: Test equipment and 1PPS inputs

Utility/Power generating industry I worked in uses master GPS 1PPS sync pulse for all the datarecorders. Regulations demand you have data which is time-synchronized to the grid (phase accurate), so they can track outages back to root cause. Otherwise, you never know which breaker tripped first.

The problem was long cables cause delays and soft rise/fall times when sending the 1PPS pulse all over the plant. So you get 10's usec error from that, even with proper termination resistors.

Termination won't compensate for propagation delays through copper (~1ns/ft). Hence the need for a GPS receiver at every client *or* an active time distribution protocol (e.g., PTP)

For S/W PLL, as I have seem those algorithms fail. They are also used in pipeline industry, where 1PPS is used to sync interrupters for cathodic protection.
The S/W bug was to only move the PLL if the sync pulse was within a time window, say 100msec. Noise, interference would sometimes knock the S/W PLL outside that window and it would never re-synchonize. Caused a lot of grief, having to reboot equipment to get it to synchronize.

Not a very smart algorithm, then! :>

For the line frequency clock (very different from a PPS signal!) to be effective, it must, by its very nature, expect the "reference" to drift out of that window with some regularity (because short term stability of the mains frequency isn't well controlled). Also, it has to accommodate the reference disappearing for variable lengths of time (i.e., power outages and/or folks just turning devices OFF!) So, you have to know when to open the window and "hunt" for the reference as well as whether or not to rely on things that look like they might be the reference.

PTP is now used rather extensively (and inexpensively) for wired connections between "smart devices". I use a variant of it to ensure I have precise time synchronization throughout my network (because I use the network's idea of time to synchronize audio, video, and the ordering of "events and actions" in a distributed whole). With it, I can synchronize the clocks on nodes to better than a microsecond (because the network traffic is traveling down the same conductors as the "clock information", you can actively measure the propagation delays involved and compensate for them).

[japlytic]

Re: Test equipment and 1PPS inputs

Although not a piece of test equipment (but a piece of digital communications equipment), a Tandberg MT5600 DVB-T modulator can have the internal reference oscillator disciplined by a 1PPS input.