Azelio Boriani said: > The Swiss MicroCrystal claims +/- 1ppm accuracy and +/- 0.09 seconds per day > for their RTC modules. <https://www.microcrystal.com/en/products/ > real-time-clock-rtc-modules/> Note that the slots in that table with 1ppm are for 25C. I looked at the data sheet for the top slot: 2.5ppm over -40 to +85C. There is a graph for the temperature of the crystal that goes to over 100 ppm so I assume there is some trickery to add/drop pulses. That is probably OK if you are driving a watch but may be "interesting" in other applications. -- These are my opinions. I hate spam.

Hi Backing up a bit: Any oscillator ( RTC or not ) is going to age. As it ages, the rate it looses or gains time speeds up. A device that is dead on today might well age a ppm ( or more ….) per year. Oscillators all are going to have a temperature spec. That’s going to apply over some range of temperatures *and* up to some (often unstated) rate of change. You then have a set accuracy. When the device leaves the factory, it’s within tolerance of “dead on”. That might be 0.1 ppm, it could easily be something else. Solder the device into a board via a typical reflow process and it’s not set where the factory put it any more … Past that, you have all of the other nasty little factors. Supply voltage moves things around. Shock / vibration / acceleration move things. On a module that is not hermetic, humidity will get into the act. How much do any of these contribute? That depends ….. could be tenths of a ppm. The accuracy of any oscillator at any point will be driven by the sum of all this “stuff”. The bold print number in that flashy ad likely puts one of the numbers in focus and pretty much ignores the rest ( …. in a constant environment ….). Parsing all the advertising talk often is more than a bit difficult. Do most oscillators do pretty well? Sure they do. This or that goes one way today and the other way tomorrow. Net at the end of the week is some sort of cancelation. If the aging spec is a max, you can bet that most will do better than that limit. However if you have to be *sure* the device will do this or that ….. hmmm ….. If you make wrist watches, the most common solution to this is to set them so they typically go a bit fast. Apparently folks don’t seem to mind arriving a bit early. They get bothered when they show up late. Back in the day, that set target was 10 seconds a month. These days, it’s not quite that much. Fun !! Bob > On Aug 7, 2022, at 1:56 PM, Hal Murray via time-nuts <time-nuts@lists.febo.com> wrote: > > > Azelio Boriani said: >> The Swiss MicroCrystal claims +/- 1ppm accuracy and +/- 0.09 seconds per day >> for their RTC modules. <https://www.microcrystal.com/en/products/ >> real-time-clock-rtc-modules/> > > Note that the slots in that table with 1ppm are for 25C. > > I looked at the data sheet for the top slot: 2.5ppm over -40 to +85C. > > There is a graph for the temperature of the crystal that goes to over 100 ppm > so I assume there is some trickery to add/drop pulses. That is probably OK if > you are driving a watch but may be "interesting" in other applications. > > > -- > These are my opinions. I hate spam. > > > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe send an email to time-nuts-leave@lists.febo.com

On Sun, 7 Aug 2022 15:59:53 -0800, Bob kb8tq via time-nuts wrote: >Hi >Backing up a bit: snip >If you make wrist watches, the most common solution to this is to set them >so they typically go a bit fast. Apparently folks donG��t seem to mind arriving >a bit early. They get bothered when they show up late. Back in the day, >that set target was 10 seconds a month. These days, itG��s not quite that much. It's only a bother in a horse race. A watch that is a few seconds fast is easy to reset; just stop it for those few seconds then restart. If the watch is slow then the hands have to be advanced before restarting. Regards. Bill Beam NL7F

Thanks all. That Dallas Semiconductor model looks a lot better than I've been seeing. The DS RTC on the Beagle or Canaduino boards I've seen off some of your websites were using a lower DS model number, maybe 3107 or some such. The Microcrystal models look good too – I just learned about that company last week, even went to their site, but must not have processed the specs. Maybe I looked at lesser part numbers. Around the same year that Heathkit designed and launched the GC-1000, I discovered that the British magazine *Radio & Electronics World* published a guide to build your own MSF-synced clock (MSF is 60 KHz like WWVB). Their design seems much cleaner than the Heathkit monstrosity – it might be because they use a Zilog Z80 instead of microcontrollers. In 1982 they released the first design guide, and in April 1983 they updated it as the "Rewbichron 2". See here: https://worldradiohistory.com/Radio_and_Electronics-World.htm The antennas alone make WWVB seem like a better choice than WWV – little ferrites vs. a giant loop. I wish I understood the disciplining better. I still don't grok PLLs. I don't understand what exactly the Heathkit does to its oscillator, and how long the benefits last without a sync. The descriptions of GPSDO make it sound like it's not real disciplining, just software correction. If disciplining is just software then I assume I can find code. I wonder if the high accuracy watch movements all use aged crystals. Cheers, Joe On Sun, Aug 7, 2022, 17:21 Bob kb8tq via time-nuts <time-nuts@lists.febo.com> wrote: > Hi > > Backing up a bit: > > Any oscillator ( RTC or not ) is going to age. As it ages, the rate it > looses > or gains time speeds up. A device that is dead on today might well age > a ppm ( or more ….) per year. > > Oscillators all are going to have a temperature spec. That’s going to apply > over some range of temperatures *and* up to some (often unstated) rate > of change. > > You then have a set accuracy. When the device leaves the factory, it’s > within tolerance of “dead on”. That might be 0.1 ppm, it could easily > be something else. Solder the device into a board via a typical reflow > process and it’s not set where the factory put it any more … > > Past that, you have all of the other nasty little factors. Supply voltage > moves things around. Shock / vibration / acceleration move things. > On a module that is not hermetic, humidity will get into the act. How much > do any of these contribute? That depends ….. could be tenths of a ppm. > > The accuracy of any oscillator at any point will be driven by the sum of > all this “stuff”. The bold print number in that flashy ad likely puts one > of > the numbers in focus and pretty much ignores the rest ( …. in a constant > environment ….). Parsing all the advertising talk often is more than a bit > difficult. > > Do most oscillators do pretty well? Sure they do. This or that goes one > way today and the other way tomorrow. Net at the end of the week is some > sort of cancelation. If the aging spec is a max, you can bet that most will > do better than that limit. However if you have to be *sure* the device > will > do this or that ….. hmmm ….. > > If you make wrist watches, the most common solution to this is to set them > so they typically go a bit fast. Apparently folks don’t seem to mind > arriving > a bit early. They get bothered when they show up late. Back in the day, > that set target was 10 seconds a month. These days, it’s not quite that > much. > > Fun !! > > Bob > > > On Aug 7, 2022, at 1:56 PM, Hal Murray via time-nuts < > time-nuts@lists.febo.com> wrote: > > > > > > Azelio Boriani said: > >> The Swiss MicroCrystal claims +/- 1ppm accuracy and +/- 0.09 seconds > per day > >> for their RTC modules. <https://www.microcrystal.com/en/products/ > >> real-time-clock-rtc-modules/> > > > > Note that the slots in that table with 1ppm are for 25C. > > > > I looked at the data sheet for the top slot: 2.5ppm over -40 to +85C. > > > > There is a graph for the temperature of the crystal that goes to over > 100 ppm > > so I assume there is some trickery to add/drop pulses. That is probably > OK if > > you are driving a watch but may be "interesting" in other applications. > > > > > > -- > > These are my opinions. I hate spam. > > > > > > _______________________________________________ > > time-nuts mailing list -- time-nuts@lists.febo.com > > To unsubscribe send an email to time-nuts-leave@lists.febo.com > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe send an email to time-nuts-leave@lists.febo.com

Hi There are a lot of plots out there looking at WWVB based devices. Since there are a range of devices, there also is a range of performance. Accuracy in the tenths to hundredths of a second is not uncommon depending on the wall clock or module you are looking at. Propagation is a biggie with WWVB. The height of the ionosphere varies over the day. That gives you a significant swing in “time of arrival” of the signal. This limits what you can get without going a bit crazy. The same changes also impact signal levels and make reception a bit difficult in many areas on a 24 hour a day basis. The stock answer to all of this today is a GPS based device. Getting something under a hundred nanoseconds out of one is pretty much a slam dunk. Bob > On Aug 7, 2022, at 5:53 PM, Joe Duarte via time-nuts <time-nuts@lists.febo.com> wrote: > > Thanks all. That Dallas Semiconductor model looks a lot better than I've > been seeing. The DS RTC on the Beagle or Canaduino boards I've seen off > some of your websites were using a lower DS model number, maybe 3107 or > some such. The Microcrystal models look good too – I just learned about > that company last week, even went to their site, but must not have > processed the specs. Maybe I looked at lesser part numbers. > > Around the same year that Heathkit designed and launched the GC-1000, I > discovered that the British magazine *Radio & Electronics World* published > a guide to build your own MSF-synced clock (MSF is 60 KHz like WWVB). Their > design seems much cleaner than the Heathkit monstrosity – it might be > because they use a Zilog Z80 instead of microcontrollers. In 1982 they > released the first design guide, and in April 1983 they updated it as the > "Rewbichron 2". See here: > https://worldradiohistory.com/Radio_and_Electronics-World.htm > > The antennas alone make WWVB seem like a better choice than WWV – little > ferrites vs. a giant loop. > > I wish I understood the disciplining better. I still don't grok PLLs. I > don't understand what exactly the Heathkit does to its oscillator, and how > long the benefits last without a sync. The descriptions of GPSDO make it > sound like it's not real disciplining, just software correction. If > disciplining is just software then I assume I can find code. I wonder if > the high accuracy watch movements all use aged crystals. > > Cheers, > > Joe > > On Sun, Aug 7, 2022, 17:21 Bob kb8tq via time-nuts <time-nuts@lists.febo.com> > wrote: > >> Hi >> >> Backing up a bit: >> >> Any oscillator ( RTC or not ) is going to age. As it ages, the rate it >> looses >> or gains time speeds up. A device that is dead on today might well age >> a ppm ( or more ….) per year. >> >> Oscillators all are going to have a temperature spec. That’s going to apply >> over some range of temperatures *and* up to some (often unstated) rate >> of change. >> >> You then have a set accuracy. When the device leaves the factory, it’s >> within tolerance of “dead on”. That might be 0.1 ppm, it could easily >> be something else. Solder the device into a board via a typical reflow >> process and it’s not set where the factory put it any more … >> >> Past that, you have all of the other nasty little factors. Supply voltage >> moves things around. Shock / vibration / acceleration move things. >> On a module that is not hermetic, humidity will get into the act. How much >> do any of these contribute? That depends ….. could be tenths of a ppm. >> >> The accuracy of any oscillator at any point will be driven by the sum of >> all this “stuff”. The bold print number in that flashy ad likely puts one >> of >> the numbers in focus and pretty much ignores the rest ( …. in a constant >> environment ….). Parsing all the advertising talk often is more than a bit >> difficult. >> >> Do most oscillators do pretty well? Sure they do. This or that goes one >> way today and the other way tomorrow. Net at the end of the week is some >> sort of cancelation. If the aging spec is a max, you can bet that most will >> do better than that limit. However if you have to be *sure* the device >> will >> do this or that ….. hmmm ….. >> >> If you make wrist watches, the most common solution to this is to set them >> so they typically go a bit fast. Apparently folks don’t seem to mind >> arriving >> a bit early. They get bothered when they show up late. Back in the day, >> that set target was 10 seconds a month. These days, it’s not quite that >> much. >> >> Fun !! >> >> Bob >> >>> On Aug 7, 2022, at 1:56 PM, Hal Murray via time-nuts < >> time-nuts@lists.febo.com> wrote: >>> >>> >>> Azelio Boriani said: >>>> The Swiss MicroCrystal claims +/- 1ppm accuracy and +/- 0.09 seconds >> per day >>>> for their RTC modules. <https://www.microcrystal.com/en/products/ >>>> real-time-clock-rtc-modules/> >>> >>> Note that the slots in that table with 1ppm are for 25C. >>> >>> I looked at the data sheet for the top slot: 2.5ppm over -40 to +85C. >>> >>> There is a graph for the temperature of the crystal that goes to over >> 100 ppm >>> so I assume there is some trickery to add/drop pulses. That is probably >> OK if >>> you are driving a watch but may be "interesting" in other applications. >>> >>> >>> -- >>> These are my opinions. I hate spam. >>> >>> >>> _______________________________________________ >>> time-nuts mailing list -- time-nuts@lists.febo.com >>> To unsubscribe send an email to time-nuts-leave@lists.febo.com >> _______________________________________________ >> time-nuts mailing list -- time-nuts@lists.febo.com >> To unsubscribe send an email to time-nuts-leave@lists.febo.com > _______________________________________________ > time-nuts mailing list -- time-nuts@lists.febo.com > To unsubscribe send an email to time-nuts-leave@lists.febo.com