The NV2 manual page in the Wiki reads:
Features that Nv2 DOES NOT HAVE YET:
synchronization between Nv2 APs
Just curious - What about sync between the cards located in the same AP?
When will this feature be available for testing?
also not made yet, in planning stage.
Sync need a precise master clock. jitter needs to be kept minimal.
To achieve this :
-
the master clock needs to be stable and jitter free
-
the master clock need to be distributed by a star network, not a daisy chained network (in a daisy chain, jitter rise at each node).
If sync at a same tower site can be important to avoid interferences, sync through different sites can be interesting as well to allow for perfect roaming without sync loss during roaming. The receiver always keep the same base clock frequency constructed from the received channel with a gold PLL, and just change the channel frequency during roaming, it does not need to resync. GSM and GPS networks are working like this. They are multichannel networks, with intantaneous roaming thanks to absolute synchronisation between transmitters.
Wifi devices would certainly need to be modified to allow for fast roaming, it is only implemented inside proprietary hardware today.
The most usefull use of fast roaming is VoIP use where it is not tolerable to have drops.
There is no reason absolute synchronisation could not be done with low cost WIFI networks. NV2 is a good start, synchronisation could be added to allow for fast or instantaneous roaming, like we can have on GSM or DECT networks.
For absolute synchronisation between different towers, it’s possible to extract absolute time from a low cost GPS, and generate a very precise clock using a digital PLL, as soon as the local oscillator is precise enough (high end Quartz or rubidium oscillator). There are opensource design available for this.
Such design can keep sync to a few microsecondes even if GPS reception is lost during 24 hours. With GPS reception, precision is almost perfect, can be easily better than 10-12 with high end (low cost thanks to GSM) local quartz oscillator. A rubidium oscillator can keep precision to parts in 10-13 sourced from a GPS signal. Best hydrogen Masers can keep parts in 10-16 (used on Galileo sytem) !
I played with this a few years ago, it’s very impresive to see a frequency counter with 12 digits showing 10.000 000 000 Mhz comparing a GPS time reference to a local rubidium or cesium atomic clock.
Perhaps it is even possible to replace the GPS unit by a NTP server, as soon as the NTP server get his time from a precise atomic synced clock and the network jitter is low enough.
Like the precise oscillators made by Rakon
I believe this one would be suitable for NV2 timing purposes http://www.rakon.com/Products/Public%20Documents/Specifications/RTX7050A.pdf
You could possibly have a mini-pci (or USB if it had low enough latency) GPS/Timing card that can be put into RouterBoards around the network, these could then act as timing nodes that the other nodes synchronize off.
They do no give short term stability.
Long term stability is nice, ± 1 ppm / year, but for Time references synced to atomic time through a digital PLL, short term stability (24 hours period), is the most important parameter.
Most GPS timing reference manufacturers are using a precision quartz in a small temperature tightly controlled enclosure, eventually double enclosure, or a rubidium oscillator for high end devices.
Nevertheless, it is possible that those new low cost devices are good enough for Wifi use, as they are used inside GSM and Wimax access points.
The local oscillator give short term stability (hours) and the atomic time source (GPS or NTP), gives long term stability. The digital PLL circuit smoothly adapt the output frequency of the local oscillator so that it keep in sync with the long term time reference.
The long term time source cannot directly give a precise sync, because of important jitter of the transmission channel. That’s why a precise local oscillator is needed.
Imho this all sounds very nice. But what about AP’s and stations of 3rd party network you have no control off?
The biggest issue in Wifi is interference. If you are the only and sole user of the spectrum you have no problem. But in real world you see more and more parties using same freq’s (even in 5ghz) and to fight interference from these time synch in your own network is not helping…
I think manufactures better invest their time and energy in making radio’s such that links between designed units are more robuust.
The example of network wide synch on mobile phone networks counts, but only for these spectrums. Telcon provider owns the spectrum so he has no issues with competition. Free spectrum Wifi is just a jungle where setting your watch is not enough to keep in contact with the group.
I think manufactures better invest their time and energy in making radio’s such that links between designed units are more robuust. The example of network wide synch on mobile phone networks counts, but only for these spectrums. Telcon provider owns the spectrum so he has no issues with competition. Free spectrum Wifi is just a jungle where setting your watch is not enough to keep in contact with the group.
That’s perfectly true, but :
-
carrier sense multiple acces is not the ideal solution because of the heavy collisions it can produce. it would be better if all carriers would have synced networks, so that they can share efficiently their bandwith with TDM (NV2 like) protocols, even if they are on the same channels.
-
synched networks, even if privates, can certainly help giving a more stable wave at receive sites when there is more than one tower transmitting on the same channel with absolute sync and true TDM.
-
i’m not sure that it is possible today to greatly improve the wifi technology we ever have with NV2 like protocols. The modulation used for Wifi is ever optimized. Wimax is better because it is using frequency hopping, but this is a totaly different technology.
Wimax products are expensives, i think there is some room for Enanced Wifi, even if proprietary, as soon as the price is kept low, as it is the case with Mikrotik products.
Syncing could be something doable on Wifi, without changing everything in the protocol. GPS are low cost devices today, and precision low cost miniature local oscillators are available as well.
try RTS/CTS. That is already a great improvement on CSMA.
it would be better if all carriers would have synced networks, so that they can share efficiently their bandwidth with TDM (NV2 like) protocols, even if they are on the same channels.
hmm, dreaming of the perfect world… in a free spectrum environment? Won’t happen… its ´free´spectrum. the word “free” means ´cowboy land´. Specially now big players are start to sell powerful domestic routers in 5Ghz band… how do you ever see household synch with your AP’s in a tower voluntary? 5Ghz is becoming what 2,4Ghz already saw; overcrowded. The only forward is to use smart technology and improving the standard so we can stay ahead of the crowd. Or look at Ruckus. They have a smart, patented, wifi system developed overcoming lots of present wifi issues. But now we are talking bucks here and it is not clear to me if it will work in wider outdoor environment.
- synched networks, even if privates, can certainly help giving a more stable wave at receive sites when there is more than one tower transmitting on the same channel with absolute sync and true TDM.
Not true IMHO. Towers at distance create a time delay on the other side due the travel time it takes for the radio wave to travel to this other tower. It’s hard to get the waves at one tower synchronized, let alone two, let alone if there are more in reach..
Wimax products are expensives, i think there is some room for Enanced Wifi, even if proprietary, as soon as the price is kept low, as it is the case with Mikrotik products.
True, but putting options like GPS synch etc. in these cheap products won’t keep them cheap… At the same time Wimax stuff is dropping in price.
I have nothing against time synch embedded in units and a big fan of 802.11a/b/g/n products and their free spectrum usage. But IMHO it think the chosen route in nv2 together with MT’s proprietary management protection protocol is a path I would like to see deployed in full. Time synch is more for licensed spectrum where it is known who is using what frequency and where.
If you can make units talk to each other, and to each other ONLY while even other radio’s in same frequency are around then 802.11a/n still stands a change in the future. Time synch can help here and there, but more is needed.
Ok… Allow me to un-hijack this thread…
Normis,
Thanks for the quick reply! While the idea is being planned, I wanted to modify the request slightly, in a way that may be easier to implement within a single radio.
I would like to configure two wireless cards with Nstreme polling enabled, in whatever configuration. Then have an option to “bind” the nstreme polling of the two cards. So instead of just running an individual polling loop on both cards, both cards run in the same polling loop. Let me provide an example.
wlan1 has two stations, st1 and st2. wlan2 has two more stations, st3 and st4. Same or different channels, doesn’t matter. Same or different bands, doesn’t matter. As it is now, with nstreme polling enabled:
wlan1 poll → st1 (simultaniously) wlan2 poll → st3
wlan1 st1 reply (simultaniously) wlan2 st3 reply
wlan1 poll → st2 (simultaniously) wlan2 poll → st4
wlan1 st2 reply (simultaniously) wlan2 st4 reply
Over and over again. What I propose is an option to make this happen instead:
wlan1 poll → st1
wlan1 st1 reply
wlan1 poll → st2
wlan1 st2 reply
wlan2 poll → st3
wlan2 st3 reply
wlan2 poll → st4
wlan2 st4 reply
Over and over again.
Seems like it would be a really easy, quick code change.
What do you think?
I will add this :
adding physical clock connection between wifi cards, through a small coaxial cable, should be doable without heavy code and hardware modifications.
This would allow synchronization for same tower antennas.
Next, for remote synchronization, oem GPS devices today are not so expensive.
Sync is extremely valuable even in unlicensed spectrum.
It is THE key reason that many WISPs put up with the price and other problems of Moto Canopy gear. It makes a huge difference. Even if it did NOTHING else it allows APs on a tower to put their channels closer together, which increases the number of APs (and backhauls) you can put on the tower. That alone would be worth the price.
But it does much more than just that.
Yes it’s certainly not for pleasure that Motorola put sync ports inside their products. They even have a GPS sync module for remote sync (SyncPipe).
But sync is seen in a lot of professional applications :
-
Digital Audio and Video Networks inside recording studios and broadcast facilities
-
GSM phone networks
-
Satellite GPS (and GLONASS, GALILEO)
If we have two adapters in the same board running NV2, are they synchronized?
You are talking about poolling synchronisation. I think that in a NV2 domain, all stations are automatically synchronized, so that they talk in a TDMA fashion.
But we are talking here about wave synchronization, so that radios on the same tower, or on different towers, have the phase of their transmision synchronized.
This is low level sync. NV2 pooling sync can be considered high level sync.
Low level sync needs specific hardware support : a hardware sync link between cards, or remote sync through very precise local oscillators helped from an atomic time source like a GPS or a NTP server.
This level of precision cannot occur through a software sync on the PCI bus with actual hardware.
Just to make sure I’m absolutely clear in understanding this, a 433 with two 5GHz cards cannot synchronize tx between the two cards?
I can see the absolutely accurate time source with two different AP’s. I’m talking about two cards in one board. This is too complicated?
Same problem with two cards in the same box. They need to be physicaly linked by a sync bus. Generaly this is done by a coaxial cable.
For GSM networks, the time source is generaly a GPS unit with a 19.6608 MHz sinus output.
In the scientific area, sync is generaly achieved by a 1 PPS output, a digital time code, or a 10 MHz or 100 MHz sinus output.
100 MHz clocks are used on high end equipements, like hydrogen maser atomic clocks, where the ultimate precision sync transfer is needed (about parts in 10EE-15)
Can anyone from Mikrotik comment about GPS sync support for wireless AP’s? We are 100% Mikrotik but the need for GPS Sync and the availability from UBNT is forcing our hand. We will be moving from MT to UBNT unless MT can show us some movement in that direction. Would like to have a comment asap as we are going to be make a very large purchase next week that will be going UBNT’s direction.
I still don’t see why many of you are so frantic about getting Wifi AP’s time synchronised.
Imho this is a selling tool which in real world is not that interesting as it looks… for the free wifi bands.
It works definitely very well on all licensed bands like cellular phone, satellite etc. Its even critical for some of these. But because we talk about licensed frequencies interference from other 2nd ´unknown´ radio owner is not around. So synch is basically only used within same operators radio network so they can use their limited radio channels more affective and it is fully controllable by operator itself.
The moment you expect same for operators working in a free band like 2,4 and 5Ghz Wifi things are different. Most situations where more than 1 operator uses un-licensed band in a region or even on the same tower, they are also each other’s competitors. Good change they also work with different make devices. How do we think we can manage to synch all? It’s already explained synch has to be done with hard coax cabling between units. So now we need to hook up with competitor’s devices? Good luck! 
And what about all these private 2,4G and 5Ghz AP/routers?
Don’t hang your hopes on time synch of wifi AP only. Theoretically it looks nice, at first sight. But it only needs one other AP sending in same channel that is not synced to put you back to where you were… interferences.. And lets be honest, this will happen and its not even illegal!
But yes, if you are the only one on the tower and region and you want many sending/receiving units in one tower it could be making a big advantage.
Anyway, if I understand well synch is also something that probably is more down to the chipset developer like Atheros. Personally I think it would be cheaper and easier to develop better receiver filters to keep interferences out of the door.
No, sync can be achieved remotely, through GPS or even NTP if an IP link is available. Every provider GPS or NTP linked to atomic time can be synched. This works like this for GSM networks.
And yes it needs radio chipset support, or external sync devices, but in this last case a hardware sync cable to the radio chipset is needed, so a sync input is needed. Not sure this is available on actual standard chipsets.
If there is only two or three providers, perhaps it can be usefull for them to try to sync themself rather than competing on the band without smartness.
Anyway you are right, it’s not easy or even perhaps usefull in most normal cases for 2.4 Ghz.
But for licenced bands, and for uses where a provider is alone, for realtime roaming inside buildings, for VoiP, even at 2.4 Ghz, this can be usefull.
Professional VoIP access points (DECT) are all synched, so that you can roam from access point to access point without loosing receiver sync. Commutation is instantaneous, the receiver don’t need to loose time to resync.
Wifi is not only 2.4 Ghz. Some manufacturers are producing 180-250 Mhz, 700 Mhz, 900 Mhz, 3.65 Ghz, 4 Ghz transmitters for licenced, military or governement use.
Anyway, syncing is certainly a good thing to propose for a low cost manufacturer, because actually to get this you need to buy higher cost hardware.
If you don’t need it, don’t use it. If you need it, then connect the GPS or NTP external synching unit, and you are synched.
Using a modular sync device approach would not rise the price of hardware too much. Only a sinus analog sync input is needed on radio devices.
To lower the cost even more, the RJ45 input could be used to carry the sync signal from an external sync box. CAT5 or CAT6 cable is enough to carry a sync 10 MHz signal on a few meters lenght. 4-5 or 7-8 RJ45 pair could be used, (the ones carrying DC power), leaving pairs 1-2 / 3-6 for fast Ethernet or half Gigabit data path.
Well, units still need to synch with each other in the transmitting or receiving phase. Otherwise they could be in opposite phase running or there must be a protocol which is depicting all industry devices every 0,000 second of each minute units start with sending.
If there is only two or three providers, perhaps it can be usefull for them to try to sync themself rather than competing on the band without smartness.
Ok, but that need some slave-master config. Because it needs to happen each time the ´master´ has had a hickup.
Lets say provider one is first and than provider 2 decides to synch with 1. Ok. Now comes no-know provider 3. If he is not going to synch he might ruin everything. So the software must be made so intelligent it ´listens´ first if adjacent radio is running as slave (which must be the default) and in such case one of them has now to switch to ´master´ mode whereafter all others (also no. 3 provider) automatically should follow as ´slave´.
This master/slave config needs to be become industry wide standard, even for consumer products. Otherwise a domestic router close to our professional tower can still disrupt everything… 
And what about distant units anyway? I believe a 5Ghz radio wave travels roughly 30 mtrs per µs. (If I remember well). If the synced phase changes every 1 µs (I presume?) than a unit at 15 or multiply of this meters away is in full opposite synch! In situations where many radio’s use same frequency this might make things worse!
Anyway you are right, it’s not easy or even perhaps usefull in most normal cases for 2.4 Ghz.
and 5Ghz. Within one or two years 5Ghz is as common as 2.4Ghz, even for consumers…
But for licenced bands, and for uses where a provider is alone, for realtime roaming inside buildings, for VoiP, even at 2.4 Ghz, this can be usefull.
Professional VoIP access points (DECT) are all synched, so that you can roam from access point to access point without loosing receiver sync. Commutation is instantaneous, the receiver don’t need to loose time to resync.
We fully agree on this! 
Wifi is not only 2.4 Ghz. Some manufacturers are producing 180-250 Mhz, 700 Mhz, 900 Mhz, 3.65 Ghz, 4 Ghz transmitters for licenced, military or governement use.
Yes, but we are mainly talking 2,4 and 5Ghz here. The others are not yet supported by MT or hardly in use in a ´free-`licensed radio networks like we discuss mainly in this forum.