[time-nuts] GPS for Nixie Clock
johnswenson1 at comcast.net
Sat Jul 16 03:08:45 EDT 2016
Yes, I was planning on using a high speed photo diode to actually
measure the turn on time of the digits. I hadn't thought of the turn OFF
time, do I want the old digit to be turned off before the new one lights
up or for them to be overlapping? I have been thinking about what
threshold to use, 50% intensity is probably about as good as any other.
It might turn out that different digits turn on differently, so I will
have to calibrate each one separately.
On 7/15/2016 4:57 PM, Chris Albertson wrote:
> If you are going for the sawtooth correction then you also might want
> to add some kind of forward correction for the delay in the tubes and
> the drivers. Your MOSFET gates the nixie tube itself have capacitance
> and switch times that will delay the switch of the display and of
> course the digital processing in the FPGA takes some number of
> nanoseconds. I think you might need some way to actually measure all
> of these as any estimate might be your single largest source of error.
> I don't know how to measure it. Perhaps a pair of phototransistors
> one aimed at a PPS LED and one at the nixie tube. This unknown delay
> is likely larger than the sawtooth correction. at this level you
> might have to define when a digital is actually "on" as there is
> likely some thermal constant and the numbers don't light up instantly.
> I'd bet the turn on time is larger than the sawtooth correction.
> What is "on"? 50% brightness?
> It gets hard when you start caring about tiny increments of time. I
> have a mechanical clock, about 14 inches in diameter that is slaved to
> NTP. The designer took a big short cut. Time is kept internally at
> the hundreds of microseconds level and the pulse goes off to the
> stepper motor at the correct time well at least at the 100+
> microsecond level but the hands don't move instantly because (1)
> slight gear backlash and (2) they have mass. I can actually SEE the
> delay with my eyes. The designer must have forgotten that a "move"
> command requires some milliseconds to execute (I'm thinking about
> 100ms or more). I don't care but it's fun to think the actual display
> is 10,000 times less accurate then the internal timekeeping. You
> don't want this to happen to happen nixie clock
> BTW I did not build my mechanical NTP clock. I got a free broken
> clock and had to fix it, cut and soldered a few traces, fixed some
> cracked parts and learned how it works in the process.
> Finding which PPS to use is easy, you can do that by eye. Compare the
> serial data stream to the time on your NTP sync'd computer. A full
> second off problem is easy to see.
> On Fri, Jul 15, 2016 at 3:53 PM, John Swenson <johnswenson1 at comcast.net> wrote:
>> Yep, that is theory. The fun part is going to be getting the right edge for
>> the new PPS. Half the time it will the one before the PPS from the GPS and
>> half the time it will be the one after. From the sawtooth data I should be
>> able to figure out which is which to align it to the new LO.
>> John S.
>> On 7/15/2016 3:17 PM, Bob Camp wrote:
>>> If you are going to go “full boat” then you probably should get the
>>> sawtooth correction out of
>>> the GPS and feed that into your control loop. You will need something you
>>> can run out at the
>>> “few hundred seconds” sort of time constant.
>> time-nuts mailing list -- time-nuts at febo.com
>> To unsubscribe, go to
>> and follow the instructions there.
More information about the time-nuts