[time-nuts] Re: Clock display on Linux systems?

[Adam Space]

Thanks for the ideas. The first idea sounds interesting. I recently got a
Raspberry Pi GPS/PPS setup working, which is still having some issues but
overall ok. (In particular, it seems as if there is a few ms lag from the
PPS signal, which I'm not sure what to do about). I've never done anything
with phototransistors though ... Are there any guides for this type of
thing that you know of?

On Mon, Dec 6, 2021 at 3:47 PM Tom Van Baak <tvb at leapsecond.com> wrote:

> Hi Adam,
>
> It would help to know, numerically, what you're goals or requirements
> are. But it sounds like you're not interested in microseconds or
> nanoseconds, maybe just milliseconds?
>
> There are untold number of layers of latency in the system you describe,
> especially since you're using a general purpose computer, a general
> purpose operating system, a general purpose language, general purpose
> display, etc.
>
> But two ideas come to mind.
>
> 1) One way to reduce latency is to have your clock display include
> pixels that flash once a second and monitor that portion of the display
> using a phototransistor. Then compare that pulse to a GPS/1PPS pulse and
> steer your software delays for zero mean. Not only will it reduce the
> latency caused by executing millions of lines of code in the computer
> and inside the display but it will also solve bias and latency issues in
> NTP.
>
> In fact, as long as you can set the clock once, you don't need NTP at
> all. The display will be locked to GPS. The only remaining latency is 1
> ns per foot for the photons to reach your eye and the hundreds of
> milliseconds it takes for your eye and brain to process the image. ;-)
>
> 2) For higher precision the other solution is not use general purpose
> computing as a source of time or display. Instead use logic IC's or
> microcontroller to direct drive LED digits. Do not multiplex the digits
> or segments. In theory this would get you down to the turn-on time for
> an LED, which is likely very sub-microsecond.
>
> Decades ago this would be done with an attractive chain of 7400-series
> TTL or 4000-series CMOS chips with 7-segment LED's. You could make a
> 13-digit display (HH:MM:SS.mmmuuun) clocked at 10 MHz from a GPSDO.
>
> To summarize, the solution(s) you choose very much depend on the
> precision of the timing you require.
>
> /tvb
>
>
> On 12/5/2021 12:53 PM, Adam Space wrote:
> > Most distributions of Linux already have a "clock" application that shows
> > the system time, but I am wondering how to program a more customizable
> > display on a Linux system / Raspberry pi. There are a few solutions that
> > pop up by googling the issue, but these are all insufficient. For
> example,
> > the only solutions in Python (which I would prefer to use if possible,
> but
> > not necessary) I have found are 1) using a library like pygame or tkinter
> > to build a display or 2) doing something more barebones like displaying
> the
> > time and sleep() ing for a second. 1) is terribly inefficient for a
> display
> > of accurate time, because either you have a refresh rate that is low and
> > the clock updates with significant lag, or you have a high refresh rate
> > which eats up large processing power. 2) is also inefficient because the
> > frequency of a local clock may be poor, so long term accuracy can only be
> > sustained with synchronization via NTP for example.
> >
> > Perhaps there are lower level Linux commands that could be used for this?
> > I'm not sure. Any suggestions are welcome.
> >
> > Adam
> >
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com -- To unsubscribe send
> an email to time-nuts-leave at lists.febo.com
> To unsubscribe, go to and follow the instructions there.
>