Ask the Experts: Synchronization

Synchronization is the key ingredient to designing a network that gets efficient use of spectrum and scales to satisfy new customers with conistent thrroughput as the network grows. Forrest Christian from PacketFlux Technologies will join the Cambium Networks developers in this special session on June 03. 

Ask us Anything about Synchronization and find out about

  • Designing Synchronization into your network
  • Best use and re-use of channels
  • Monitoring and measuring network performance

 

Participating is easy. Just login and join this thread at any or all of the following times on Wednesday, June 3rd:

  • 1 – 3 PM CDT

Pass this information to your colleagues. If you are not yet a member of the Cambium Community, join at http://community.cambiumnetworks.com/ by using the link in the top right corner.

We look forward to the discussion and meeting with you. Add the session to your outlook calendar by clicking on the attachment below:

 

1 Like

Hello.

I’m Forrest Christian from Packetflux Technologies and I’d like to give you a little background on myself and my company.

I help start a Dial-up ISP back in 1994 (my idea but that’s a different story). In 1997 the ISP started putting customers on our network via wireless. After using several other technologies, the now-WISP started using Cambium gear (Motorola back then) and they have been using it very successfully ever since.

As the WISP network grew, it quickly reached the point where synchronization was important to avoid tower to tower interference. Because almost all of the sites were very small and we had quite a few of them, it became apparent quickly that a lower cost synchronization solution was needed. I designed a small GPS synchronization receiver suitable for a couple of radios at a site, and fairly quickly thereafter a sync over power solution as well. I started to sell this solution to other WISP operators and it became very popular very quickly.

Over the past ten years or so, those original two products have grown up to include a whole line of products suitable for wireless operators. Although I now have staff which handle things like bookkeeping and actually assembling the products, I still do the product designs myself and still help troubleshoot synchronization problems my customers encounter. As a result I’m very aware of how synchronization works (or doesn’t work) and common misconceptions about it - both positive and negative.

I’d be happy to answer any questions you might have.

-Forrest

1 Like

Hi Forrest:

That being said, can you talk to the two points here?

What are the network benefits to using synchronization?

 

Are there any benefits of NOT using synchronization?

 

 

 

Thanks!

Some additional questiosn:

If you use Synchronization in PMP 450 are you required to have any guard band between channels on Aps within the same tower?

If you use antennas less the 90 degrees, say 40 could you run frequency reuse without the antenna being fully front to back?

Do you need to configure specific parameters at all sites to match in order to have a truly synchronized network? If yes, what are those parameters?

Can you speak to the density per tower you’ve seen, or personally achieved with a synchronized network?

Will synchronization help you mitigate interference from an outside source such as another ISP?

@Matkix - Just hitting some of your 'additional questions':

Frequency reuse on a tower (with reliable sync) is really more a question of whether client devices can see both APs, or an AP can see clients NOT its own, at high enough signal levels to interfere.  With sync, both APs will be transmitting at the same time, and both listening for SM uplink at the same time.  If there is crossover (either sync issues leading to uneven timing and APs seeing each other, or AP/SM seeing the other sector) then there is a problem.  If two same-channel sectors are only 40 degrees apart, there's a much higher likelihood of them seeing each other's clients, and their clients seeing both of them.

WRT PMP450 (at least 3.65GHz) Up/Down ratio, max range, control/contention slots, Frame Period, and cyclic prefix will all (AFAIK) impact frame timing.  Simplest approach is to maintain the same standardized settings across your whole network, but at least anywhere towers' footprints overlap.

Sync will absolutely help with outside interference, IF that source is ALSO using sync, with parameters affecting frame timing set similarly.  This is really the exact same scenario as self-interference between your own adjacent towers.  If everything is set the same, and sync is stable, you should be able to make it work.  If your sync is off, or settings differ enough to push the frame timing off, you'll interfere.

If the outside interference source is NOT using sync and compatible frame timing, then your own sync won't fix it.

j

2 Likes

@newkirk wrote:

@matkix - Just hitting some of your 'additional questions':

 

Sync will absolutely help with outside interference, IF that source is ALSO using sync, with parameters affecting frame timing set similarly.  This is really the exact same scenario as self-interference between your own adjacent towers.  If everything is set the same, and sync is stable, you should be able to make it work.  If your sync is off, or settings differ enough to push the frame timing off, you'll interfere.

If the outside interference source is NOT using sync and compatible frame timing, then your own sync won't fix it.

j


Many WISPs who operate adjacent networks work together on synchronizing, as it does benefit both networks in the long run. I would like to hear from the field your experience with working with adjacent WISPs and synchronization.

Here is my list of questions.


Regarding 3.65GHz in the PMP450 only:

Do you recommend 10MHz or 20MHz channels with 3.65?

Is a guard band needed when using 10MHz channels in 3.65 with the PMP450?

  •  The manual mentions 3MHz or 5MHz guard bands.
  •  Likewise using 10MHz guard bands results in 3 channels 3655,3675,3695
  •  Using no guard bands gives 5 channels 3655,3665,3675,3685,3695

What is the expected performance hit when using Adjacent Channel Support?

  •  The manual mentions that SM’s will be limited to 23db
  •  If 23db is true, how would this affect NLOS customers.

What is the expected bandwidth when using 7MHz channels in 3.65?

What can be done to limit interference from other 3.65 APs and towers in RF range?

  • Site planning becomes increasingly difficult when dealing with only 3 x 10MHz channels and one polarity (MIMO) when deploying multiple tower sites within RF range of each other.

Any comments on 900MHz PMP450i?

Any idea when FW v13.4 will be released?

Hi there... I was asked to introduce myself in this thread also...

My name is Matt Mangriotis and I am the product manager for the PMP 450 product line. As our current customers know, synchronization is of the utmost importance to maximize spectral efficiency, limit spectral pollution, and minimize self-interference.  This is especially true for those that want to scale their business in both density and area.


My favorite color is RAL 5015.

1 Like

@CambiumMatt wrote:

My favorite color is RAL 5015.


Ha! That makes sense... otherwise known as 'Cambium Blue'? Pretty close... my favorite blue is Leguna Seca Blue:

bmw-750-laguna-seca-blue-21-655x436.jpg

1 Like

@matkix wrote:

What are the network benefits to using synchronization?

 

Are there any benefits of NOT using synchronization?

 


Synchronization generally only fixes self-interference problems.   In a PtMP system you can generally have five types of self-interference:

1) An AP not being able to effectively hear it's own SM's due to a different AP transmitting.

2) An AP not being able to effectively hear it's own SM's due to a SM associated with a different AP transmitting.

3) A SM not being able to effecively hear the AP it's associated with due to a different AP transmitting.

4) A SM not being able to effectively hear the AP it's associated with due to a SM on a different AP transmitting.

5) AP's and SM's on the same AP not being able to hear each other effectively because multiple radios associated with the same AP are talking at the same time.   This could be two SM's or an AP and a SM or some other combination of one or more radios.  The difference between this case and cases 1-4 is that in cases 1-4 the device causing the interference is associated with a different ap, where as in this case (5) the device causing the interference is associated with the same ap.

So let's take the easy case - #5.   This is a common problem with CSMA (Carrier Sense Multiple Access) or "listen before talk" implemenations, and is most often caused by one SM not realizing another SM is talking to the AP since it can't hear it due to antenna patterns and/or other RF blockage.  (Do a search for "hidden node").  The solution that is usually implemented to fix this is some sort of scheduled/polled MAC - these are almost all proprietary.   Once you do this, almost all of these problems go away (at the cost of interoperability between vendors) since the AP can control which device is permitted to transmit at any given time - there still has to be a bit of time for SM's which aren't scheduled to be able to tell the AP that they want to talk, but generally this is confined to a very small amount of airtime, and won't affect data transfers, only the ability of a SM to register or request a uplink slot.  

Now lets take the other 4 cases:

In case #1, you have a situation where you have an AP trying to listen to the SM, and it is instead hearing a lot of interference from a different AP.   You can solve this either with RF planning and design - which can be difficult since  AP's on towers are all up high and tend to be visible to each other, even over several miles - OR you can synchronize the AP's transmissions.   By synchronizing the AP transmissions so they all start and end transmitting at the same time, you can ensure that one AP under you control is never transmitting when another AP is attempting to listen.

One other note about case #1 is that it tends to be the only case where you also have to consider adjacent channel interference.  When you have multiple AP's on the same tower there is enough bleed outside the channel an AP is set to that it can actually interfere with AP's on other frequencies.  How far away in frequency this will occur depends on a lot of variables and it really isn't relevant to this discussion other than to say:  If you turn on synchronization this fixes this problem.

Case #2 is one that synchronization doesn't fix.   In fact, it can make it worse, because with synchronization, SM's will only transmit when AP's are listening - without it, the interference-source SM will sometimes transmit while the second AP is transmitting, causing a different ype of interference, but not this one.  Fortunately this can be solved relatively easily with RF engineering - I'll come back to this case in a minute.

Case #3 is the opposite problem of case #2.   Much of the discussion on case #2 applies here, just the roles between the SM and AP are reversed.   The solution to case #3 is also the same, which is why I'll come back to it in a minute.

Case #4 is fairly rare, but is also solved by synchronization.   In this case, you have two SM's which can 'hear' each other loud enough that they can't hear their AP.   In the case I've seen this, we had a SM on two different houses, pointed at two different AP sites.   The SM's also happened to be pointed at each other.    Whenever we put the two APs that these two SMs were assocated with on the same frequency, they quit working.   The fix was to actually add synchronization to the one AP site which for whatever reason didn't have it (or it wasn't working, or whatever).   By synchronizing these, each SM would never be listening when the other SM is transmitting.

So lets go back to cases 2 and 3.   These are both solved by simple RF engineering.   The RF engineering is actually made a bit simpler with synchronization as well since you only have to consider possible AP<->SM paths.   It's hard to describe in text but I'll try by using an example.  Let's assume you have a SM, and you have it pointed at a given AP site.   Let's also assume that 20 degrees to the right of that AP site from the SM's perspective you have another AP site which is almost the same distance away.   Also assume the APs pointing roughly in the SM's direction on both AP sites are on the same frequency.    Let's also assume that the SM has a 60 degree antenna.  

Thinking about this logically you'll see that the SM will hear both AP's at roughly the same signal strength.   Because the 60 degree antenna in the SM is going to be at most 3db down at 30 degrees off axis (half of the beamwidth), then you're going to have somethng less than 3db of reduction 20 degress off access (aka the second AP).   In addition, both AP's will be hearing a similar signal from this SM, which will cause interference with the second AP.    The solution is simply to reduce the beamwidth of the SM's antenna such that it both picks up a strong singal only from the AP it's interested in, and only transmits toward the AP it's interested in.   Or put very shortly:  Put a dish on the SM to narrow the antenna pattern and reduce the second AP as an interference source.

You could actually figure out how much interference the SM is causing to the AP and vice versa by doing a RF path analysis by looking at the effective gain on the chosen antenna in the direction of the other end of the link (see the antenna directional gain plot), and using that as the antenna gain.   By computing worse-case signals received, you can come up with a frequency plan and "antenna pattern" choice which works for your situation.  If that didn't make sense, and you really care (or someone else cares), ask and I'll try to dive into that a bit deeper.

You also asked about benefits of NOT using synchronization.  I briefly touched on one which I don't think really is a benefit - that is by not using sync your interference is much more random, which might be a good thing in a very lightly used network.  In my experience, random interference is not a good thing, even in a lightly used network.  Beause sync makes the interference more consistent, networks tend to operate more consistently which is good for both operation when they work, and troubleshooting when they don't.

 Probably the biggest reason one would not choose to synchronize is if they have a very small network (a couple of AP's), and they don't have any interference sources they can reduce or eliminate via synchronization.   Having 'a lot' of interference from third parties is not in itself a reason not to choose to synchronize your network - if anything it's a reason to synchronize so you can better utilize the few channels you might have access to.

The other downside of radios using synchronization is that it generally requires a fixed uplink/downlink schedule and ratio.   This, as a side effect increases apparent latency across the network.   Ratio first:  75% down/25% up is a common ratio.    This means that if you had a 100Mb/s radio, the most you could do is 75Mb/s down, and 25Mb/s up - and it won't change dynamically.   In fact, it must be effectively the same across your network.   You can't set one AP to 50/50 and the next to 60/40 and the next to 90/10 and have sync work correctly.   This means that if for example the 25Mb/s up fills, and the 75Mb/s down isn't yet full, you can't use the 75Mb/s down capacity temporarily.

In relation to the latency:   Because the MAC uses a fixed schedule, the AP transmits for a period, then waits for a bit for the air to clear, and then receives, then waits for a bit for the air to clear again, and then the cycle repeats.  This cycle causes a bit of built-in latency since a transmit/receive switch has to wait for a complete cycle.  Without a synchronized frame, a SM can respond immedately with the ping response, showing an improved ping rate.   I'm not actually convinced that the few ms of latency reduction you get without a fixed-scheduled MAC is worth it.   For instance, each transmit/receive switch takes time and sucks up radio capacity.  By only doing ~800 of them (PMP450) per second you gain a lot of efficiency vs a traditional polled or CSMA MAC which will do thousands per second.

I think that's a good start on this topic.   Hopefully that touched on what you were curious about.

-forrest

3 Likes

@matkix wrote:

Some additional questiosn:

If you use Synchronization in PMP 450 are you required to have any guard band between channels on Aps within the same tower?

If you use antennas less the 90 degrees, say 40 could you run frequency reuse without the antenna being fully front to back?

Do you need to configure specific parameters at all sites to match in order to have a truly synchronized network? If yes, what are those parameters?

Can you speak to the density per tower you’ve seen, or personally achieved with a synchronized network?

Will synchronization help you mitigate interference from an outside source such as another ISP?


I'll let cambium handle the guard band question. 

In relation to the antenna pattern/frequency reuse question:   It all comes down to the patterns for the antennas you choose.   The engineering is actually rather simple.   Take the antennas you are looking at.  Look at the difference in signal gain between the two antennas based on the angle apart you are mounting them and where the SM's are going to be.  For instance if you mount one antenna at a bearing of 0 degrees, and a second at 90, and a single SM at 0 degrees, you'll want to compare the signal gain of the first antenna at 0 degrees and the second at 90.    If you had SM's out to say 15degrees, you'd want to compare the first antenna at 15 degrees and the second at 75 degrees. 

Once you understand the difference in signal gain, you then have to determine whether that is sufficient or not.   Of course you have to have at least as much difference as is needed to maintain the modulation you desire.   Plus some for things like uplink signal strength differences (this is where power leveling helps a lot), and so on.

In an extreme case,  you may be able to do several reuses of a given frequency with an extremely tight dish antenna - such as with the case with backhauls.    Panels are much more difficult.   But you could also mix and match - say a BH with an AP - assuming the BH is set to master - remember a slave is the same as a SM.

In relation to the "matching parameters' question:  I'll let cambium be specific about what needs to be set to be truly synchronized, but yes, they need to be all set the same.  Or more accuately, whatever parameters you set needs to result in a similar transmit/receive timing.   There are frame calculator tools in the radios and white papers to help you get everything set correctly.  Things like uplink/downlink ratios, control slots and max distance settings along with specific versions and radio types change the framing, and I've sort of lost track of all of the details since I haven't had to do the math for some time now.

In relation to density:   Because at the WISP we generally are up against a mountain and are pointing all our AP's in the same direction, we don't get a per-tower reuse - but we do get to use all of the non-overlapping channels pointed in the same direction, and mounted very close together.    We do get very good tower-to-tower reuse - never losing a channel to self interference (although we do have to pay attention to placement of AP's in relation to SM's when reusing frequencies - don't want two close AP's on the same channel pointing the same direction).

I also have several customers (of packetflux) who rely heavily on the frequency reuse inherent in the synchronization system.  One customer in particular, who is in a fairly populated part of the US, regularly puts up a full ring of both 2.4 and 5.7 - that is, two radios on each non-overlapping frequency (12 radios on 5.7 and 6 on 2.4).   He has now expanded into the rest of the 5.x band to get even more capacity.

In relation to outside ISP interference:   The answer is 'depends'.   If you both use the same gear and can coordinate, definitely.   If not, then the synchronization doesn't help you reduce interference from them, but instead helps you use the remaining channels more efficiently.

1 Like

@matkix wrote:

Some additional questiosn:

If you use Synchronization in PMP 450 are you required to have any guard band between channels on Aps within the same tower?


With PMP 450, there is no guard band required between channels on APs within the same tower. Only caveat is the 3 GHz PMP 450 where you will need a guard band unless you enable "Adjacent Channel Support". When "Adjacent Channel Support" is enabled, the SM limits its Max Tx Power to 23 dBm combined vs. 25 dBm its capable of. If "Adjacent Channel Support" is not enabled, you will need a guard band of 5 MHz between two 20 MHz channels, 3 MHz between two 10 MHz channels and 2 MHz between two 5 MHz channels. 

1 Like

@Logonix wrote:

Here is my list of questions.


Regarding 3.65GHz in the PMP450 only:

Do you recommend 10MHz or 20MHz channels with 3.65?


This is for you to decide depending on the bandwidth demand of the subs in your deployment. PMP 450 supports both the upper and lower 3.65 band. So you have 50 MHz of spectrum to play with. If you are deploying 4 APs on a tower, you can certainly use two 20 MHz channels with frequency reuse on the 4 APs. 


@Logonix wrote:

Here is my list of questions.


Regarding 3.65GHz in the PMP450 only:

Is a guard band needed when using 10MHz channels in 3.65 with the PMP450?

  •  The manual mentions 3MHz or 5MHz guard bands.
  •  Likewise using 10MHz guard bands results in 3 channels 3655,3675,3695
  •  Using no guard bands gives 5 channels 3655,3665,3675,3685,3695

What is the expected performance hit when using Adjacent Channel Support?

  •  The manual mentions that SM’s will be limited to 23db
  •  If 23db is true, how would this affect NLOS customers.

As I mentioned in my post above and something you have pointed out, guard bands are required if you do not enable "Adjacent Channel Support". When enabled, the SMs limit thier Max Tx power to 23 dBm (2 dB less than what they are capable of). In terms of NLOS performance, that is what it is - you lose 2 dB of power on the Uplink. In typical deployments 23 dBm is usually sufficient to at least get your SMs transmitting at 6X (64 -QAM) rate which is more than enough for a single sub. 

1 Like

"

4) A SM not being able to effectively hear the AP it's associated with due to a SM on a different AP transmitting.

<snip>

Case #4 is fairly rare, but is also solved by synchronization.

 "

I wanted to point out that the rarity of your fourth case may vary in some areas.  In suburbs you might have hundreds of rooftops that all have line of sight to each other.  This can be a very big deal, and as you say it's eliminated by sync.

1) I've seen a wimax product that requires 20+ minutes to acquire GPS lock on a cold start.  I'm told in the carrier Wimax/LTE/cellular space that's actually fast, and some products require up to an hour, and I heard of one older product that supposedly needed 24 hours.

Canopy seems to only need 30-90 seconds to have GPS sync up and running, and the PMP320 might have been 5 minutes at the most.  I've been wondering what the difference is.  I suppose it's a long shot that you (Forrest) would know, but I thought I'd ask.

2) Forrest, do you foresee Packetflux making GPS sync products for any platform besides Canopy? 

3) The biggest downside I see to GPS sync is that you have a problem when it stops working.  What options do we have to "harden" our GPS sync setups?  Would it be worth having surge protectors on sync cables for example?

From Forest Post

"Thinking about this logically you'll see that the SM will hear both AP's at roughly the same signal strength.   Because the 60 degree antenna in the SM is going to be at most 3db down at 30 degrees off axis (half of the beamwidth), then you're going to have somethng less than 3db of reduction 20 degress off access (aka the second AP).   In addition, both AP's will be hearing a similar signal from this SM, which will cause interference with the second AP.    The solution is simply to reduce the beamwidth of the SM's antenna such that it both picks up a strong singal only from the AP it's interested in, and only transmits toward the AP it's interested in.   Or put very shortly:  Put a dish on the SM to narrow the antenna pattern and reduce the second AP as an interference source."

Having read this and the rest of the thread..

How much negative impact would it be to use 120 degree antennas on the AP in a crowded RF area?

One thing missed from the reply from Cambium_Sri or Forrest Christian is the performance penalty for not having synchronisation on a base site. Since the transmission from one AP antenna may be seen by an adjacent AP antenna, the adjacent channel performance requirement becomes unrealistically large particularly if all antennas are mounted on one pole. In order to make such a system work it is necessary to apply extra screening between each antenna, or mount the 4 antennas on different sides of a building where one antenna cannot 'see' another antenna. It may also be necessary to have a large guard band.

1 Like

@Logonix wrote:

Here is my list of questions.

What is the expected bandwidth when using 7MHz channels in 3.65?


With 7MHz, you can get up to 28 Mbps (8X), 21 Mbps (6X), 14 Mbps (4X) and 7 Mbps (2X) real world sector throughput (Link Tests will show higher throuhgput since its a 1-to-1 measurement). However, depending on the rates at which each SM is at under the AP, the sector throughput may be impacted. I strongly recommend using the PMP 450 Capacity Planner R13.3 available here to get a better prediction of sector throughput for your specific deployment. Alternately, there is a very good tool called LinkPlanner available here to aid with this.

 


@Logonix wrote:

Here is my list of questions.


Regarding 3.65GHz in the PMP450 only:

What can be done to limit interference from other 3.65 APs and towers in RF range?

  • Site planning becomes increasingly difficult when dealing with only 3 x 10MHz channels and one polarity (MIMO) when deploying multiple tower sites within RF range of each other.

If the other towers are yours its imperative that you sync the APs across towers to reduce self interference. If the other tower is not yours, then you have two options:

1. Work with the other owner to sync APs between the two towers if that owner is also using similar Cambium gear. 

2. Do channel planning so you don't interfer with each other in an unsynchronized deployment. 

If the above two are not options, then you are pretty much dealing with external interference and is like any other deployment. We believe PMP 450 provides class leading performance when it comes external interference. However, proper channel planning goes a long way in dealing with it. 

I invite other WISPs who are reading this response to chime in on their experience in dealing with external interference and how they have solved it. 


@Logonix wrote:

Here is my list of questions.


Regarding 3.65GHz in the PMP450 only:

Any comments on 900MHz PMP450i?

Any idea when FW v13.4 will be released?


We would like to keep this thread focussed on Synchronization topics.

However to quickly answer your off topic questions:

1. 900 MHz: A general search of this forum will provide you good information of this upcoming platform. One such thread is this where CambiumMatt and others talk about the product. 

2. Release 13.4: Not yet determined. This is currently in Beta. We're working through some final bugs and we are close! Keep checking back on the Beta forum.