The official abbreviation for Coordinated Universal Time is UTC. This abbreviation comes as a result of the International Telecommunication Union and the International Astronomical Union wanting to use the same abbreviation in all languages. English speakers originally proposed CUT (for "coordinated universal time"), while French speakers proposed TUC (for "temps universel coordonné"). The compromise that emerged was UTC,[6] which conforms to the pattern for the abbreviations of the variants of Universal Time (UT0, UT1, UT2, UT1R, etc.).[
Having worked on PTP professionally, I know that a few years ago big players used to charge $1,000,000+ to license the software for the algorithm that does the synchronization. Then it would still take many months of work to integrate it into a working product. The fact that this is open source is awesome!
@@tim-w Do you use a cell phone? Tight synchronization is vital to making the radio network work. There are many other applications, but this is probably the one normal people interact with most often. The electrical grid also relies on tight synchronization, but the transition to digital is happening more slowly as they are ultra-conservative in that industry.
@@tim-w Making cell phones networks possible. Making GPS possible. Making banking transactions secure. Proving relativity. Calibrating equipment not to mention more serious applied and pure scientific research.
@@tim-w I currently have better than 1ns timing jitter between my Grandmaster and switch using PTP and Synchronous Ethernet in my home lab and the GM is accurate to about 40ns of UTC. As has been mentioned, timing is critical in communications systems. It is looking likely that 5G will transition in later specifications to a requirement of 5ns timing accuracy. It is fun to play with but there are absolutely real-world applications for sub-ns timing. In the industrial networking space accurate time is needed for synchronising different parts of a production facility and is provided by industrial field bus systems such as EtherCAT. With standard crystals I’ve had better than 20ns of timing jitter between devices on a network. One of the problems is that the GNSS constellations such as GPS, Galileo, GLONASS etc are prone to blocking so it is common to have multiple Grandmasters which may be 100’s of miles apart for resilience. In the UK there are exercises where GNSS is deliberately blocked and these are listed on Ofcom’s website. IT is also possible to have a GNSS firewall where if the GNSS signal is blocked the timing can be sourced from either another remote GNSS Grandmaster or a local Caesium / Rubidium clock. Accurate time is also critical for trading to track when exactly a transaction is placed and also in cars. Many new automotive applications rely on time for position doing and synchronising RADAR systems for spatial awareness.
Oof, this makes me think of how big companies will use open source work, not contribute, and then try to draw away attention from the open source work by rebranding and popularizing their own version that builds on it. I’m thinking about video capture software, that most streamers use. I hope whoever decides to make use of this put in a few of the millions into the work.
Glad to see you and Dr. Byagowi debunking the radioactive hype. Looks like the average Rubidium oscillator consists of 160 picoCuries of radioactivity. For comparison, your average banana measures in at 520 picoCuries. And you *really shouldn't be eating* any Rubidium oscillators you find lying around. During undergrad physics we did a ton of experiments with radioactive button sources and a few using our college's very own neutron source. At the beginning of every lab we were given a safety lecture about the proper way to handle... the bricks of lead we were using for shielding, because those were the real health hazard.
@@anon_y_mousse This issue came up. Person in the class asked if they should be worried about their radiation dose because their significant other's diet was so rich in potassium. Physics professor replied, with a completely straight face, "Well, dosage vs. proximity follows the inverse square law."
They should have more of those WWV type of transmitters all over the US and force auto makers to tune into it for turn signal flashing. No more sitting at red lights watching a bunch of flashing lights sync and unsync.
I have used PTPv2 in wind tunnels, where multiple computers record audio signals from a few hundred microphones. The data is all timestamped, which allows software post-processing to calculate where certain sounds came from. The same effect as a dish that your would physically move. For this to work well, you want nanosecond-scale accuracy. And since wind tunnels are generally large metal structures, running a GPS antenna for each computer is not preferable.
@@jyvben1520 Like a faraday cage, about 20 meters in height. We try to keep the computers close to the microphones (near the ground) and they're generally standing in the wind (300kph). Running long wires in that environment doesn't seem wise.
@@hyperfluff_folf about 50% of aircraft noise comes from the engines. The rest comes from the aircraft itself. Landing gear, flaps, but also small gaps between components. Since noise of landing aircraft is a big factor for airports, manufacturers want to find the sources of the noise. This is generally too complex to simulate with a computer I'm told. You are right that wind is a factor in the propagation of sound. These effects have been modeled, and we apply corrections to our measurements in an attempt to undo these effects. In order to apply the corrections we need to know things like wind speed in the wind tunnel, so these are measured too.
@Richard Cranium Yep, where PTP is used you sync after the fact. In general the strategy in measurement systems is to directly store your measurements on disk, unmodified. Wind tunnels are expensive to rent, so you want to have your data to pass through as little lines code as possible. Every line of code is a possible bug, and you cannot recover from bugs in your acquisition code.
The T-shirt was well Co-ordinated also... ;-) My first job as a sysadmin 30 years ago was setting up and managing xntp on our network, so this takes me way back. Back then you had to ask permission to have our servers "ping" off of a university's stratum 1 clock over in the next province. It's fascinating to learn how NTP is not accurate enough these days. But wow, a Stratum one clock for so cheap sounds amazing. And no, it did not impress the ladies back then either. 😆
I did an internship at a telecom company, every piece of lab equipment (network switches, oscilloscope etc.) was modded to be able to use the same clock signal via bare (coax) cabling.
I've run Stratum 1 servers from RPi 3s for a few years. While testing code for NTPsec, I had a group of 4 fed from the same antenna with carefully crafted cables to ensure that signal path was the same for each device. This was used to experiment with convergence differences between different NTP solutions. Also, my first professional presentation was on time synchronization using xntpd as one of the options. I also have code that is part of that project for adjusting the time in the kernel to reduce drift, basically adjusting the value of each 'tick' to account for the variance in the systems base oscillator.
We use ptp to coordinate audio and video over a network in the Audio Visual field. It allows us to get video walls synced up for making sure they are frame accurate. We actually don't care about the exact time just that all the device are synchronized. It also means we have to use somewhat specialized switches that support QoS and non-blocking.
@@PsRohrbaugh honestly for video we end up using the Netgear av line switches. They come in a good range of sizes and configurations preset that's make manufacture specs for encoders and decoders. faster and easier deployment that way.
Jeff, On that scope you can use Measurements to measure time between the two pulses so you don't have to count ticks. You can also use Cursors to manually get a time delta by placing one cursor on one pulse and the other on the other pulse. Measurements and Cursors are very powerful tools in the scope and it's well worth the time to learn how to use them. Best wishes.
Got many CM4's on order, in the meanwhile, the sun is out today so the master clock in the garden (sundial) is all we have! PS, black for signal and yellow for ground???
About 20 years ago, I read the Linux NTP README and discovered that I could buy an obsolete GPS receiver for $25 and a serial cable for another $10 to turn pretty much any Linux box into an accurate NTP server. The hardest part was making a battery eliminator, so I didn't need to keep replacing the AA cells in the GPS receiver. Today you could probably use an old cellphone and a USB OTG cable for next to nothing.
@@StringerNews1 afaik there are GPS/modem modules with serial interface for around 30$ to be used for Arduino, drones and other microcontroller-based stuff
@@bobbydazzler6990 why are you asking me this? I've never found a 1 Hz tone useful for anything I do, certainly not a replacement for the standard 10 MHz tone carried over a coax cable. I was talking about making an inexpensive NTP server, as was the OP. If you really need to count one second intervals, you could use the phone's clock or something. What exactly is your use case?
@@bobbydazzler6990 sorry, but I'm not your parents. You're not entitled to my time. You're acting like I owe you something, and that's simply not true. If you want strangers to respond to you, try asking nicely, and learn how to hold a conversation.
2:40 Radio Clocks "...will sync-up after a few minutes..." Only locally, from a 'continental' perspective. More-distant locations will have to wait until the wee hours (e.g. 2:00 AM local), when the VLF signal can reach their location.
Indeed. I currently live in the next state over from where the WWVB transmitter is, and my "atomic clocks" that use WWVB typically need several tries to sync up. The best case is they start at midnight, and finish at 12:04, but it's not unusual to have them trying again at 0100, 0200 ... all the way to 4 AM. When I lived in Chicago, these types of clocks almost never synced up. Too much electrical noise in the city.
As someone who went through the hassle of researching and waiting until I could snag a good deal on a used enterprise GPS clock, I'm stoked to see how accessible accurate NTP / PTP sources are becoming for the home lab community now. (Yeah, you're not the first, and neither was I, but hey, it's a cool club, and we're happy to welcome you!) I want to warn anyone who's eyeing old enterprise gear, make note of how long it's been end of life, check whether you can get firmware updates without a[n expensive] support contract, and verify that it was updated recently enough to handle the GPS Week Number rollover on April 6, 2019. The oldest and most affordable stuff out there cannot, so it's basically useless, and the gear that's new enough to still work is not as affordable as it should (IMHO) be. Caveat emptor, and whether you go refurbed enterprise or DIY, happy timekeeping!
This is fantastic for me. I have equipment that uses GPS for timing. Last year some of the GPS units stopped working and the equipment defaulted to NTP, which was not accurate enough. I then discovered they could use PTP, but the cost of a server was too much for me. Now maybe I can implement PTP.
But only if the network switches and other gear actually support PTP. If they don't, then you won't have a way to get the PTP signal across LAN segments so that the clients can talk to the server(s).
This video is a godsend. I am currently working on a project where positioning is determined based on Time Difference of Arrival between two signals. Even though the project is in very early stage and we don't need perfectly synced devices right now, but eventually we'll need it. Else nothing will work. Thanks for making a video on this topic. It'll definitely help me immensely.
Jeff.. look at the timing features etc for the upcoming WiFi 7. 802.11be includes full QoS and time services so all your wireless devices can also sync.
"Maybe i'll be the first homelabber with a time server in my rack" Sorry to break it to you, but that prize was handed out a long time ago, even on a Pi. Running GPS based stratum 1 servers on RPIs has been common for long.
What a strange coincidence. I saw recently that VMware 7.0 supported PTP as a time sync source, which led me down the rabbit hole of all this stuff. Very cool video!
You tease us so much. Most of us will not get our hands on ANY sort of RPi this year (or next, the way things are going). So you may as well do projects with, or educate us on, the hadron collider.
I love videos like this, though this issue in time syncing doesn't affect me personally. I am fascinated by it because the lengths people go to, to perfect solutions like this is impressive and greatly underrated by the majority. I can't help but and geek out about videos like this. Keep up the great content Jeff!
This is a great post.... I just did something exactly like this... My Pi Hat uses multiple GPS sats to give me +/- 400ns time /date and I use NTP to get it to my networked servers...
Yet another reason to ditch "Daylight Savings Time", with its twice/year shenanigans!
Před 2 lety+13
I always get hyped up from these videos to have extra precise network time on my LAN for no apparent reason; your presentation is great. :D Is there such thing as a database of NICs to see which support PTP protocol? Eg. can I check somehow if an onboard LAN on a motherboard supports this or not? Because having a CM4 PTP server is just one part of the equation. I guess the big investment is going to be on the PC side.
Unfortunately, it seems like IEEE1588 support is often buried in tech specs for NICs :( And not all that support it really seem to have full driver support, either. It seems like Intel is the best bet, even many older NICs support it (e.g. i210 I think).
Před 2 lety+2
@@JeffGeerling so this will take some "time" (ba dum tss) to get mainstream unfortunately. :(
PPT also works with software time stamping. We have a Raspberry Pi 4 with a GPS Hat (with PPS) running as a PPT Server. We use this to time sync our web server which is located in the basement where we haven't any change to get a GPS signal. Using software instead of hardware time stamping reduce the accuracy to the micro second range. But that's at least two magnitudes better than using NTP from a time server via internet
@@ChrisA4A4 True! And I didn't even touch on that ability. And from some further investigation, it looks like it's _possible_ the chip in the Pi 4 could have 'unofficial' support for hardware timestamping, too.
At work we're looking at using our Cisco ASR routers as PTP GM clocks, with GPS reference. It looks like it's a pretty straightforward setup, so I'm sure there's more than a few CCIE types who would have this set up in their home labs. Granted a couch change raspberry pie is a lot cheaper than a $100K router, but businesses retire these things all the time, and it's not uncommon for equipment to be redirected from the e-waste bin to employees' home labs. (EDIT: Not that I'm speaking from experience.)
I can guarantee you are not the first with a PTP time server in their home lab. I don't remember who it was, but I remember seeing that someone had multiple atomic clocks in his homelab. This guy even did a test with him going up a mountain with the atomic clock and getting down and comparing two clocks.
A few years back I had a Raspberry with GPS as the only time source, because that RPi was used outdoors without any network connectivity. I remember extremely good how I struggled to get GPSd and NTPd talking to each other.
For test bench use, a GPS Disciplined Oscillator (GPSDO) can be useful. They're about US100. 'BG7TBL' is one designer. These gadgets contain an Oven Controlled Xtal Oscillator (OCXO), brought into alignment to GPS by a uC. They typically produce plain-Jane digital 1PPS and 10MHz signals on their front panel BNC connectors.
Cool video, never knew time could be managed at home with my own server. When CM4 prices comes down, maybe I'll install it next to my Pi-hole! Thanks again for the video, I only wish you would post more often!
PTP works by using profiles and the size of the network, routers, switches, etc are taken into account with PTP. PTP measures the end to end latency form the client to the PTP server to calculate the loss to maintain time sub microsecond.
I did some work with small 4g pico cells, the ones outside used GPS for time, the ones inside we fed with PTP from our backhaul radio equipment. That had a GPS and generated the PTP pretty much as the PI is doing, we had a load of interest from companies that wanted the GM functionality and didn't need the radio link. I left soon after, always wondered if it went anywhere was cheap compared to a GM IRO of $1000 IIRC but maximum of 6 slaves, would probably do more but that was all we tested. Somehow the solutions now are going to be way cheaper and this was only 6 years ago.
Excellent video Jeff. I would love to see a video about the PI’s ability to generate a reference frequency. Not a reference time via a pulse interval. The frequency domain is very important for synchronization of communications. You know as in frequency hopping spread spectrum. If the PI could give us a cheap way to generate an accurate frequency it would enable all kinds of elaborate data exchange at a very low cost. It would be a game changer.
There is a good reason why the "traditional" devices are so expensive and it has nothing to do with the actual hardware. Those devices come with official certification and operating guarantees. Something this open source piece of hardware does not have. For hobbyists this might not matter, but in the company's this is mostly relevant for and in science that needs to be validated this is absolutely very important. This new card is nice and cool and absolutely no threat to the expensive models. The same goes for almost all expensive measuring equipment. Be it from fluke or agilent or Tektronix or whatever.
Well, the hardware has something to do with the price. In order to get a new "traceable to NIST" sticker, the equipment needs to be sent in for calibration. And equipment designed to be opened up and tweaked at least once a year for decades needs to have components inside that can stand up to that much use.
@@StringerNews1 compared to the cost of calibration, certification, tracing, support and what not, the cost of hardware is minor. Yes it has some influence, but it might as well be nothing.
@@bzuidgeest you're kidding, right? It's only a couple hundred to recertify most equipment, and the purchase cost ranges from thousands to tens of thousands. Which of those numbers is bigger?
@@StringerNews1 recertify! That is not the same as the original certifications of the design. Certification and calibration are also not one thing. Also with that kind of equipment economies of scale are not working. Long term support instead of the consumer 1 or 2 year is expensive. Yes they are no doubt milking their customers but it's not all nonsense. Look at how long fluke keeps certain models on the market because governments and companies have standardized on them. Hardly mass market. Therefore more expensive than the common trash.
Some guy made this idea work with starlink aswell, so eventually getting the starlink dish and router might throughput PTP through existing LAN cables.
Nice. With this i will be never again to late to bid in the last second of an ebay offer 😁 Some of us are happy if their windows clock isn't out of sync by hours. "Sorry, boss. I'm not a little late at work this morning. I'm just out of sync"
I've been playing with GPS PPS on a Pi4 and Sparkfun GPS module and getting nanosecond level accuracy. Your statement about GPS is accurate though. The better the access to open sky the better the accuracy. Nobody needs any of this for home network stuff but it sure is fun to set up your own PPS server for a few bucks.
Great topic. I have several Casio watches that have the shortwave radio update feature. They are affordable and amazing. This is a very underrated technology. I recall there was some conversation in recent past about Govt reducing their funding in the NIST program. Hope this never happens.
Yeah, in the WWV tour video I linked, the engineer mentioned that from time to time there are discussions about cutting off WWV since GPS has all but taken over. But I think it would be a bit disruptive since there are still so many radio-based clocks! Also, for a Time Card upgrade, they might include a WWV receiver since, once you know your location with GPS, WWV can be a very accurate backup time source if GPS is jammed or the signal is lost.
@@JeffGeerling Seems worth pointing out here that the WWV(/B/H/L) clusters serve not only as time standards, but -- as I think their top-of-hour announcement says -- also *frequency* standards. That's *much* harder to ship over the internet, and still important. And you can't get it from network TV anymore...
watching this in the car over Bluetooth. quite ironic I'm watching a video about precise time keeping & syncing. and having the audio 2 seconds later than the video!
I know a project with this in mind. Central clock replacement. I think there still is a lot of old mechanical clocks needed to be replaced by a raspberry pi.
I have been replacing clocks in my house for a few years. I have four pi zero (original, w, 2w) based clocks scattered about. Two are clocks ONLY. Two have clocks in the UI; but, also play audio from my home server. I have a pi 4 based GPS enabled NTP server supporting all my networked CPUs. My wife puts up with me. 😁
The TimeCard uses a Xilinx FPGA, the software to program this is not open source, whilst you can get an open source bit stream to program it, if you want to change the programming you still need commercial software tools from Xilinx these are not open source although they are currently free to use
It's worth mentioning that 1PPS signals (done correctly) are aligned with UTC seconds, so that all the 1PPS signals on Earth should transition (upwards) at the very same instant.
A News Group called 'Time Nuts' has been going for decades. One hobbyist even brought a suite of atomic standards up a mountain to confirm gravitational time dilation, look up Project GREAT circa 2005.
GPS also has the roll over issue which can be fun to deal with. Do the cards have an auto update feature so they don’t fall victim to the GPS epoch issue?
I think NTP and PTP are different usage, PTP and GM clock are for digital transmission usage … this is prevent digital packet collisions and crash. It can apply on device on the LAN, router, switch, ap … etc. and also apply on multicast streaming … like SMPTe 2110, Dante AV and NDI. PTP is like Metronome
Most radio controlled clocks use the signal from WWVB which runs at 60khz and there is no voice. While you can use WWV and WWVH too that would be more for a diy solution since there are multiple transmitters at 2.5, 5, 10, 15, and sometimes 20 and 25Mhz and depending on the time of day different ones will be receivable, whereas 60khz is available for most of the US, typically stronger at night.
Heh, I can imagine someone writing an AI engine just to translate timestamp information from the voice... when the data is literally encoded in the signal already :D
@@JeffGeerling Lol that would be crazy. The low frequency is used more for coverage and reliability on one set frequency than extracting the signal format. Knowing this is what happens when you're both a RHEL admin and a ham radio operator, you learn lots of mildly useful knowledge.
@@JeffGeerling HF signals are generally propagated via the ionosphere, so the path length is quite variable (on the scale of milliseconds), even moment to moment. The VLF signal, being groundwave, would tend to have a more consistent pathlength. Not that either source is intended to be used for precision time transfer anyway.
LOL, time stuff there has been a few vids i have been watching about master clocks, lol down the road from me is CHU it is the canadian time brodcast and the only one that still broadcasts Bell 103 compatible time data
There are network switches that can fan out PTP. So if you connect the Pi to a few of those switches, you don't need those expensive grandmasters. In this case those switches handles downstream devices however many of them, and aggregates all of those into one PTP stream from the Pi.
That's my hope! AES67 is getting more widespread, it would be nice to have a more off-the-shelf and inexpensive option compared to the master clocks out now.
@@JeffGeerling Absolutely agree. It feels waaay nicer to work with than Dante as per my experience. Likewise I'm hoping that ST2110 is getting way more traction in the future, as it feels like a lot more suitable to complex networks than NDI that just tries to do magic and failing... as it's in complex corporate networks, not in flat one. (Some video ideas there? :fingers_crossed:)
The official abbreviation for Coordinated Universal Time is UTC. This abbreviation comes as a result of the International Telecommunication Union and the International Astronomical Union wanting to use the same abbreviation in all languages. English speakers originally proposed CUT (for "coordinated universal time"), while French speakers proposed TUC (for "temps universel coordonné"). The compromise that emerged was UTC,[6] which conforms to the pattern for the abbreviations of the variants of Universal Time (UT0, UT1, UT2, UT1R, etc.).[
Like how iso stands for international organization for standardization, sometimes the acronym doesn't match in all languages
@@JaredConnell roughly the same reason. Additionally ISO is derived from the Greek word isos, which means "equal".
The more you know... 😊
I just think of it as "Universal Time, Coordinated" and move on.
UTC stands for Universal Time Coordinate and refers to the baseline that coordinated time is set from (which is why it is represented as UTC+x).
So yes.. the acronym is correct.
Having worked on PTP professionally, I know that a few years ago big players used to charge $1,000,000+ to license the software for the algorithm that does the synchronization. Then it would still take many months of work to integrate it into a working product. The fact that this is open source is awesome!
I can't fathom why people care so much about this.
@@tim-w Do you use a cell phone? Tight synchronization is vital to making the radio network work. There are many other applications, but this is probably the one normal people interact with most often. The electrical grid also relies on tight synchronization, but the transition to digital is happening more slowly as they are ultra-conservative in that industry.
@@tim-w Making cell phones networks possible. Making GPS possible. Making banking transactions secure. Proving relativity. Calibrating equipment not to mention more serious applied and pure scientific research.
@@tim-w I currently have better than 1ns timing jitter between my Grandmaster and switch using PTP and Synchronous Ethernet in my home lab and the GM is accurate to about 40ns of UTC. As has been mentioned, timing is critical in communications systems. It is looking likely that 5G will transition in later specifications to a requirement of 5ns timing accuracy. It is fun to play with but there are absolutely real-world applications for sub-ns timing. In the industrial networking space accurate time is needed for synchronising different parts of a production facility and is provided by industrial field bus systems such as EtherCAT. With standard crystals I’ve had better than 20ns of timing jitter between devices on a network.
One of the problems is that the GNSS constellations such as GPS, Galileo, GLONASS etc are prone to blocking so it is common to have multiple Grandmasters which may be 100’s of miles apart for resilience. In the UK there are exercises where GNSS is deliberately blocked and these are listed on Ofcom’s website. IT is also possible to have a GNSS firewall where if the GNSS signal is blocked the timing can be sourced from either another remote GNSS Grandmaster or a local Caesium / Rubidium clock.
Accurate time is also critical for trading to track when exactly a transaction is placed and also in cars. Many new automotive applications rely on time for position doing and synchronising RADAR systems for spatial awareness.
Oof, this makes me think of how big companies will use open source work, not contribute, and then try to draw away attention from the open source work by rebranding and popularizing their own version that builds on it. I’m thinking about video capture software, that most streamers use. I hope whoever decides to make use of this put in a few of the millions into the work.
Glad to see you and Dr. Byagowi debunking the radioactive hype. Looks like the average Rubidium oscillator consists of 160 picoCuries of radioactivity. For comparison, your average banana measures in at 520 picoCuries. And you *really shouldn't be eating* any Rubidium oscillators you find lying around.
During undergrad physics we did a ton of experiments with radioactive button sources and a few using our college's very own neutron source. At the beginning of every lab we were given a safety lecture about the proper way to handle... the bricks of lead we were using for shielding, because those were the real health hazard.
I would also recommend against eating bricks of lead :D
Thank you for the explanation Gilad. It makes me feel more safe playing with these toys :)
@Astrocat 3D I'm pretty sure he will let us recompile the Linux kernel
Must be why I glow in the dark, I eat a lot of bananas.
@@anon_y_mousse This issue came up. Person in the class asked if they should be worried about their radiation dose because their significant other's diet was so rich in potassium. Physics professor replied, with a completely straight face, "Well, dosage vs. proximity follows the inverse square law."
They should have more of those WWV type of transmitters all over the US and force auto makers to tune into it for turn signal flashing. No more sitting at red lights watching a bunch of flashing lights sync and unsync.
Hahaha, I actually made a similar comment on a Technology Connections video about turn signal timing :D
Make the whole city blink in sync, would be a nice sight on a dark night, it would probably reflect in the sky :)
I love that
@@JeffGeerling I was just about to mention Alec's clip on this, and you beat me to it.
You can just use cameras already in cars for this. :-)
I have used PTPv2 in wind tunnels, where multiple computers record audio signals from a few hundred microphones. The data is all timestamped, which allows software post-processing to calculate where certain sounds came from. The same effect as a dish that your would physically move. For this to work well, you want nanosecond-scale accuracy.
And since wind tunnels are generally large metal structures, running a GPS antenna for each computer is not preferable.
large metal structure, like an antenna ? ;-)
@@jyvben1520 Like a faraday cage, about 20 meters in height. We try to keep the computers close to the microphones (near the ground) and they're generally standing in the wind (300kph). Running long wires in that environment doesn't seem wise.
What are you using microphones for in a wind tunnel and is it even possible to do acurate audio location when wind is a factor?
@@hyperfluff_folf about 50% of aircraft noise comes from the engines. The rest comes from the aircraft itself. Landing gear, flaps, but also small gaps between components. Since noise of landing aircraft is a big factor for airports, manufacturers want to find the sources of the noise. This is generally too complex to simulate with a computer I'm told.
You are right that wind is a factor in the propagation of sound. These effects have been modeled, and we apply corrections to our measurements in an attempt to undo these effects.
In order to apply the corrections we need to know things like wind speed in the wind tunnel, so these are measured too.
@Richard Cranium Yep, where PTP is used you sync after the fact.
In general the strategy in measurement systems is to directly store your measurements on disk, unmodified. Wind tunnels are expensive to rent, so you want to have your data to pass through as little lines code as possible. Every line of code is a possible bug, and you cannot recover from bugs in your acquisition code.
It's about time!....to watch Jeff Geerling!
The T-shirt was well Co-ordinated also... ;-)
My first job as a sysadmin 30 years ago was setting up and managing xntp on our network, so this takes me way back. Back then you had to ask permission to have our servers "ping" off of a university's stratum 1 clock over in the next province. It's fascinating to learn how NTP is not accurate enough these days. But wow, a Stratum one clock for so cheap sounds amazing.
And no, it did not impress the ladies back then either. 😆
[First date] Would you like to know how accurate my clocks are?
My clock's more accurate than any other out there.
I did an internship at a telecom company, every piece of lab equipment (network switches, oscilloscope etc.) was modded to be able to use the same clock signal via bare (coax) cabling.
I've run Stratum 1 servers from RPi 3s for a few years. While testing code for NTPsec, I had a group of 4 fed from the same antenna with carefully crafted cables to ensure that signal path was the same for each device. This was used to experiment with convergence differences between different NTP solutions.
Also, my first professional presentation was on time synchronization using xntpd as one of the options. I also have code that is part of that project for adjusting the time in the kernel to reduce drift, basically adjusting the value of each 'tick' to account for the variance in the systems base oscillator.
We use ptp to coordinate audio and video over a network in the Audio Visual field. It allows us to get video walls synced up for making sure they are frame accurate. We actually don't care about the exact time just that all the device are synchronized. It also means we have to use somewhat specialized switches that support QoS and non-blocking.
Are they purple switches?
@@PsRohrbaugh honestly for video we end up using the Netgear av line switches. They come in a good range of sizes and configurations preset that's make manufacture specs for encoders and decoders. faster and easier deployment that way.
Jeff, On that scope you can use Measurements to measure time between the two pulses so you don't have to count ticks. You can also use Cursors to manually get a time delta by placing one cursor on one pulse and the other on the other pulse. Measurements and Cursors are very powerful tools in the scope and it's well worth the time to learn how to use them. Best wishes.
True, true.
Got many CM4's on order, in the meanwhile, the sun is out today so the master clock in the garden (sundial) is all we have! PS, black for signal and yellow for ground???
Haha I was wondering if someone would notice. I had plugged them in backwards from the pins I thought they were and just left 'em.
came here only for this comment 😅
If you think about it, on a sundial it is black for signal and yellow for ground.
I made a gps based ntp server with my old pi a few years back. it used the PPS on that gps chip. It was fun
About 20 years ago, I read the Linux NTP README and discovered that I could buy an obsolete GPS receiver for $25 and a serial cable for another $10 to turn pretty much any Linux box into an accurate NTP server. The hardest part was making a battery eliminator, so I didn't need to keep replacing the AA cells in the GPS receiver. Today you could probably use an old cellphone and a USB OTG cable for next to nothing.
@@StringerNews1 mini pcie and M.2 lte and gps cards exist
@@StringerNews1 afaik there are GPS/modem modules with serial interface for around 30$ to be used for Arduino, drones and other microcontroller-based stuff
@@bobbydazzler6990 why are you asking me this? I've never found a 1 Hz tone useful for anything I do, certainly not a replacement for the standard 10 MHz tone carried over a coax cable. I was talking about making an inexpensive NTP server, as was the OP. If you really need to count one second intervals, you could use the phone's clock or something. What exactly is your use case?
@@bobbydazzler6990 sorry, but I'm not your parents. You're not entitled to my time. You're acting like I owe you something, and that's simply not true. If you want strangers to respond to you, try asking nicely, and learn how to hold a conversation.
I get pumped when I see a Geerling notification in my feed. I know I'm gonna learn something cool.
2:40 Radio Clocks "...will sync-up after a few minutes..." Only locally, from a 'continental' perspective. More-distant locations will have to wait until the wee hours (e.g. 2:00 AM local), when the VLF signal can reach their location.
Indeed. I currently live in the next state over from where the WWVB transmitter is, and my "atomic clocks" that use WWVB typically need several tries to sync up. The best case is they start at midnight, and finish at 12:04, but it's not unusual to have them trying again at 0100, 0200 ... all the way to 4 AM. When I lived in Chicago, these types of clocks almost never synced up. Too much electrical noise in the city.
Tom Scott he's amazing his videos are super informative and easy to understand the bcz the he explains about things in a simple way.
He crams a good amount of information into a short video that illustrates what he's talking about very well.
@@JeffGeerling yeah 👍
As someone who went through the hassle of researching and waiting until I could snag a good deal on a used enterprise GPS clock, I'm stoked to see how accessible accurate NTP / PTP sources are becoming for the home lab community now. (Yeah, you're not the first, and neither was I, but hey, it's a cool club, and we're happy to welcome you!)
I want to warn anyone who's eyeing old enterprise gear, make note of how long it's been end of life, check whether you can get firmware updates without a[n expensive] support contract, and verify that it was updated recently enough to handle the GPS Week Number rollover on April 6, 2019. The oldest and most affordable stuff out there cannot, so it's basically useless, and the gear that's new enough to still work is not as affordable as it should (IMHO) be. Caveat emptor, and whether you go refurbed enterprise or DIY, happy timekeeping!
This is fantastic for me. I have equipment that uses GPS for timing. Last year some of the GPS units stopped working and the equipment defaulted to NTP, which was not accurate enough. I then discovered they could use PTP, but the cost of a server was too much for me. Now maybe I can implement PTP.
It sounds like it's _time_ for an upgrade!
@@JeffGeerling ba dum ts xD, good one
But only if the network switches and other gear actually support PTP. If they don't, then you won't have a way to get the PTP signal across LAN segments so that the clients can talk to the server(s).
This video is a godsend. I am currently working on a project where positioning is determined based on Time Difference of Arrival between two signals. Even though the project is in very early stage and we don't need perfectly synced devices right now, but eventually we'll need it. Else nothing will work.
Thanks for making a video on this topic. It'll definitely help me immensely.
Fascinating. I love all of the things you can do with a RPi. Great video!
Nice job on finding yet another Raspberry angle on something I didn't know anything about!
I loved the vibe and enthusiasm of Ahmad!
Jeff.. look at the timing features etc for the upcoming WiFi 7. 802.11be includes full QoS and time services so all your wireless devices can also sync.
WiFi 7 will be quite interesting. Borrows some features from 5G (which also uses this kind of timing accuracy to get the bandwidth it does).
"Maybe i'll be the first homelabber with a time server in my rack"
Sorry to break it to you, but that prize was handed out a long time ago, even on a Pi.
Running GPS based stratum 1 servers on RPIs has been common for long.
What I meant was an atomic-clock-based PTP time server. No other *Pi*-based setup can do that besides a CM4 (and only recently at that).
What a strange coincidence. I saw recently that VMware 7.0 supported PTP as a time sync source, which led me down the rabbit hole of all this stuff. Very cool video!
You tease us so much. Most of us will not get our hands on ANY sort of RPi this year (or next, the way things are going). So you may as well do projects with, or educate us on, the hadron collider.
Great video as always Jeff, but I have to admit that for me the star of the show was the shirt you're wearing - I want one of those!
Yes Jeff great video you always go beyond the content details I expect too see. Great work!
Thanks! Very interesting stuff.
Jeff showing off 2 RPi CM4 when it's hard to get your hands on a single one 😂😂
I love videos like this, though this issue in time syncing doesn't affect me personally. I am fascinated by it because the lengths people go to, to perfect solutions like this is impressive and greatly underrated by the majority. I can't help but and geek out about videos like this. Keep up the great content Jeff!
This is a great post.... I just did something exactly like this... My Pi Hat uses multiple GPS sats to give me +/- 400ns time /date and I use NTP to get it to my networked servers...
PTP is my favorite clock!
Thank you very much. It was a new experience for me.
One of the most unique and interesting videos on the channel so far!
Yet another reason to ditch "Daylight Savings Time", with its twice/year shenanigans!
I always get hyped up from these videos to have extra precise network time on my LAN for no apparent reason; your presentation is great. :D Is there such thing as a database of NICs to see which support PTP protocol? Eg. can I check somehow if an onboard LAN on a motherboard supports this or not? Because having a CM4 PTP server is just one part of the equation. I guess the big investment is going to be on the PC side.
Unfortunately, it seems like IEEE1588 support is often buried in tech specs for NICs :(
And not all that support it really seem to have full driver support, either. It seems like Intel is the best bet, even many older NICs support it (e.g. i210 I think).
@@JeffGeerling so this will take some "time" (ba dum tss) to get mainstream unfortunately. :(
PPT also works with software time stamping. We have a Raspberry Pi 4 with a GPS Hat (with PPS) running as a PPT Server. We use this to time sync our web server which is located in the basement where we haven't any change to get a GPS signal.
Using software instead of hardware time stamping reduce the accuracy to the micro second range. But that's at least two magnitudes better than using NTP from a time server via internet
@@ChrisA4A4 True! And I didn't even touch on that ability. And from some further investigation, it looks like it's _possible_ the chip in the Pi 4 could have 'unofficial' support for hardware timestamping, too.
Pis are getting more and more useful! I'm never going to get one at this point..
At work we're looking at using our Cisco ASR routers as PTP GM clocks, with GPS reference. It looks like it's a pretty straightforward setup, so I'm sure there's more than a few CCIE types who would have this set up in their home labs. Granted a couch change raspberry pie is a lot cheaper than a $100K router, but businesses retire these things all the time, and it's not uncommon for equipment to be redirected from the e-waste bin to employees' home labs. (EDIT: Not that I'm speaking from experience.)
You say "Until next time I'm Jeff Geerling", but every time you come back you are Jeff Geerling again! 🤪
I can guarantee you are not the first with a PTP time server in their home lab. I don't remember who it was, but I remember seeing that someone had multiple atomic clocks in his homelab. This guy even did a test with him going up a mountain with the atomic clock and getting down and comparing two clocks.
PTP is also used in the entertainment industry; synchronization for light and video shows.
https en wikipedia org /wiki/SMPTE_2059
Also AES67 for broadcast! I'm hoping to explore that a little more with my Dad (radio engineer) on a Geerling Engineering video in the future.
You learn alot from these comments 10/10 video
Very interesting and very educational. Excellent job Jeff.
The Dante audio protocol actually uses PTP to sync all of the audio packets that are traveling over the network.
Loved this video. Beautiful , elegant explanations! And to boot on a PI! woooo hooo
A few years back I had a Raspberry with GPS as the only time source, because that RPi was used outdoors without any network connectivity. I remember extremely good how I struggled to get GPSd and NTPd talking to each other.
2:26 I just picture Jeff toiling away with a soldiering iron while WWV plays in the background. 😄 that's true geek cred.
For test bench use, a GPS Disciplined Oscillator (GPSDO) can be useful. They're about US100. 'BG7TBL' is one designer. These gadgets contain an Oven Controlled Xtal Oscillator (OCXO), brought into alignment to GPS by a uC. They typically produce plain-Jane digital 1PPS and 10MHz signals on their front panel BNC connectors.
Informative and got me thinking (This is where I usually get into trouble)
Cool video, never knew time could be managed at home with my own server. When CM4 prices comes down, maybe I'll install it next to my Pi-hole!
Thanks again for the video, I only wish you would post more often!
Always great content
PTP works by using profiles and the size of the network, routers, switches, etc are taken into account with PTP. PTP measures the end to end latency form the client to the PTP server to calculate the loss to maintain time sub microsecond.
I did some work with small 4g pico cells, the ones outside used GPS for time, the ones inside we fed with PTP from our backhaul radio equipment. That had a GPS and generated the PTP pretty much as the PI is doing, we had a load of interest from companies that wanted the GM functionality and didn't need the radio link. I left soon after, always wondered if it went anywhere was cheap compared to a GM IRO of $1000 IIRC but maximum of 6 slaves, would probably do more but that was all we tested. Somehow the solutions now are going to be way cheaper and this was only 6 years ago.
Sometimes the answer to a problem is NOT a raspberry Pi. Jeff Geerling 'Hold my beer!'
You're right, sometimes the answer is RPi-Pico, or 42. 😉
Excellent video Jeff.
I would love to see a video about the PI’s ability to generate a reference frequency. Not a reference time via a pulse interval.
The frequency domain is very important for synchronization of communications. You know as in frequency hopping spread spectrum. If the PI could give us a cheap way to generate an accurate frequency it would enable all kinds of elaborate data exchange at a very low cost. It would be a game changer.
Absolutely amazing - Thanks!
Seriously impressive progress!
Thanks for adding actual captions for the Deaf
Stop buying all CM4 I can't get my hands on one :)
Very educational video. I look forward to see that applied to online gaming
Thanks!
There is a good reason why the "traditional" devices are so expensive and it has nothing to do with the actual hardware. Those devices come with official certification and operating guarantees. Something this open source piece of hardware does not have. For hobbyists this might not matter, but in the company's this is mostly relevant for and in science that needs to be validated this is absolutely very important. This new card is nice and cool and absolutely no threat to the expensive models.
The same goes for almost all expensive measuring equipment. Be it from fluke or agilent or Tektronix or whatever.
Well, the hardware has something to do with the price. In order to get a new "traceable to NIST" sticker, the equipment needs to be sent in for calibration. And equipment designed to be opened up and tweaked at least once a year for decades needs to have components inside that can stand up to that much use.
@@StringerNews1 compared to the cost of calibration, certification, tracing, support and what not, the cost of hardware is minor. Yes it has some influence, but it might as well be nothing.
@@bzuidgeest you're kidding, right? It's only a couple hundred to recertify most equipment, and the purchase cost ranges from thousands to tens of thousands. Which of those numbers is bigger?
@@StringerNews1 recertify! That is not the same as the original certifications of the design. Certification and calibration are also not one thing. Also with that kind of equipment economies of scale are not working. Long term support instead of the consumer 1 or 2 year is expensive. Yes they are no doubt milking their customers but it's not all nonsense. Look at how long fluke keeps certain models on the market because governments and companies have standardized on them. Hardly mass market. Therefore more expensive than the common trash.
@@bzuidgeest I see the problem, you don't know what you're talking about. 🤦
Welp. Definitely learned some new stuff here!
@+①②①④⑥③⑨⓪③④WhatsApp Oh wow definitely will "Jeff"
Jeff: Nerdy about time
Me: Doesn't even use a watch anymore
12:26 "I use this material to pick up girls."
"Hey girl, are those clocks atomic? Because you make my heart go tick-tock."
I'll see myself out.
hahahaha...only the geeks will buy this pick-up line. =)
This is really interesting, once more available I might get one
I have dealt with GPS time devices and can't wait see the time hat.
Some guy made this idea work with starlink aswell, so eventually getting the starlink dish and router might throughput PTP through existing LAN cables.
Nice. With this i will be never again to late to bid in the last second of an ebay offer 😁
Some of us are happy if their windows clock isn't out of sync by hours.
"Sorry, boss. I'm not a little late at work this morning. I'm just out of sync"
I've been playing with GPS PPS on a Pi4 and Sparkfun GPS module and getting nanosecond level accuracy. Your statement about GPS is accurate though. The better the access to open sky the better the accuracy. Nobody needs any of this for home network stuff but it sure is fun to set up your own PPS server for a few bucks.
Love the channel
CERN's "white rabbit" protocol and hardware can get even better (picosecond range). Completely open source, but probably not supported by CM4.
Amazing video, Jeff! 💪🏻
Really cool stuff!!
Well thanks for making those board more rare.
Great topic. I have several Casio watches that have the shortwave radio update feature. They are affordable and amazing. This is a very underrated technology. I recall there was some conversation in recent past about Govt reducing their funding in the NIST program. Hope this never happens.
Yeah, in the WWV tour video I linked, the engineer mentioned that from time to time there are discussions about cutting off WWV since GPS has all but taken over. But I think it would be a bit disruptive since there are still so many radio-based clocks!
Also, for a Time Card upgrade, they might include a WWV receiver since, once you know your location with GPS, WWV can be a very accurate backup time source if GPS is jammed or the signal is lost.
@@JeffGeerling Seems worth pointing out here that the WWV(/B/H/L) clusters serve not only as time standards, but -- as I think their top-of-hour announcement says -- also *frequency* standards. That's *much* harder to ship over the internet, and still important.
And you can't get it from network TV anymore...
@Astrocat 3D Ditto. This is a great example of how analog complements digital information. The simplicity of the radio wave's protocol is brilliant.
@Astrocat 3D Yeah, what's wrong with running wire telegraphs along railroads, just in case?
Wow, the day after Sam at THIS MUSEUM IS (NOT) OBSOLETE channel shows us how to build our own master clock, we get this!
Fancy scope !
watching this in the car over Bluetooth. quite ironic I'm watching a video about precise time keeping & syncing. and having the audio 2 seconds later than the video!
My dad would listen to the WWV every week to set the 50+ clocks in our home.
I was going down the rabbit hole of using a USB gps module on my pfsense box and then you make this video. Heh.
Jeff ur videos always motivate me to go play with PIs but then I remember that I need to buy some 😭
Reads title, scratches head, plays video, reads title again, tilts head confused… “ooooooohhhhhhh”
I know a project with this in mind. Central clock replacement.
I think there still is a lot of old mechanical clocks needed to be replaced by a raspberry pi.
I have been replacing clocks in my house for a few years. I have four pi zero (original, w, 2w) based clocks scattered about. Two are clocks ONLY. Two have clocks in the UI; but, also play audio from my home server. I have a pi 4 based GPS enabled NTP server supporting all my networked CPUs. My wife puts up with me. 😁
People use raspberry pis in amateur radio too, that'd be fun for you to explore if you haven't already.
The TimeCard uses a Xilinx FPGA, the software to program this is not open source, whilst you can get an open source bit stream to program it, if you want to change the programming you still need commercial software tools from Xilinx these are not open source although they are currently free to use
It's about Time.
It's about Accurate.
We stay Precise.
We Configure!
Put in the work, put in the hours and take what's ours (ahoo)
Linux installed in my veins
It's worth mentioning that 1PPS signals (done correctly) are aligned with UTC seconds, so that all the 1PPS signals on Earth should transition (upwards) at the very same instant.
A News Group called 'Time Nuts' has been going for decades. One hobbyist even brought a suite of atomic standards up a mountain to confirm gravitational time dilation, look up Project GREAT circa 2005.
GPS also has the roll over issue which can be fun to deal with. Do the cards have an auto update feature so they don’t fall victim to the GPS epoch issue?
Awesome video, thank you!
I think NTP and PTP are different usage, PTP and GM clock are for digital transmission usage … this is prevent digital packet collisions and crash. It can apply on device on the LAN, router, switch, ap … etc. and also apply on multicast streaming … like SMPTe 2110, Dante AV and NDI. PTP is like Metronome
A second is quite literally the second division of the hour, after the first, minute division.
Now create a MIDI convertor / Main Clock for musicians. Stop the studio drift - Thanks in advance!
In a “crazy world…” here comes Jeff!!! Lol
I was unsubscribed from you by CZcams. I am angry. Now I have to go back and see everything I missed. Dang you CZcams!
Most radio controlled clocks use the signal from WWVB which runs at 60khz and there is no voice. While you can use WWV and WWVH too that would be more for a diy solution since there are multiple transmitters at 2.5, 5, 10, 15, and sometimes 20 and 25Mhz and depending on the time of day different ones will be receivable, whereas 60khz is available for most of the US, typically stronger at night.
Heh, I can imagine someone writing an AI engine just to translate timestamp information from the voice... when the data is literally encoded in the signal already :D
@@JeffGeerling Lol that would be crazy. The low frequency is used more for coverage and reliability on one set frequency than extracting the signal format.
Knowing this is what happens when you're both a RHEL admin and a ham radio operator, you learn lots of mildly useful knowledge.
@@JeffGeerling HF signals are generally propagated via the ionosphere, so the path length is quite variable (on the scale of milliseconds), even moment to moment. The VLF signal, being groundwave, would tend to have a more consistent pathlength. Not that either source is intended to be used for precision time transfer anyway.
Oh brother, you are reminding me of my Calibration lab, I use to run. Nist we use to use the GPS satellite timing. How you recovering mate?
LOL, time stuff there has been a few vids i have been watching about master clocks, lol down the road from me is CHU it is the canadian time brodcast and the only one that still broadcasts Bell 103 compatible time data
There are network switches that can fan out PTP. So if you connect the Pi to a few of those switches, you don't need those expensive grandmasters. In this case those switches handles downstream devices however many of them, and aggregates all of those into one PTP stream from the Pi.
Thanks, an amazing lesson.
Media and entertainment is also using PTP! Maybe I can propose replacing the GM clocks on our network with few CM4s?
That's my hope! AES67 is getting more widespread, it would be nice to have a more off-the-shelf and inexpensive option compared to the master clocks out now.
@@JeffGeerling Absolutely agree. It feels waaay nicer to work with than Dante as per my experience. Likewise I'm hoping that ST2110 is getting way more traction in the future, as it feels like a lot more suitable to complex networks than NDI that just tries to do magic and failing... as it's in complex corporate networks, not in flat one.
(Some video ideas there? :fingers_crossed:)