What is the best frequency for SIP

VoIP with WiFi

Does VoIP and Lync work over "wireless network connections"? The answer to that is a resounding "yes, but".

  • Yes...
    because the VoIP stack and RTP do not really care which physical medium the data is transmitted over. This can even be UMTS or LTE and even the DSL connection at home is configured by Telekom as an "IP connection" from 3 Mbit downstream.
  • But...
    because the boundary conditions have to be right. The human ear and thus VoIP is quite sensitive when it comes to runtimes. In ISDN we are used to the fact that the voice arrives without interruption. ISDN uses the G711 codec with 64 bit for this. The amount of data is therefore predictable and the lines are dedicated. This is not always guaranteed with IP.

The question of the runtime and its variance (also called jitter) is the linchpin of every VoIP connection. This is not even limited to WiFi or other radio technologies, but the general nature of a LAN connection. This has a lot to do with the access method and the bandwidth. See also QoS - Quality of Service

Shared medium

In a WiFi network, on the other hand, the air is basically a "shared medium", even if there are different channels. If each end device were to occupy only one channel (frequency) and the access point were to work on all channels, then this would be a collision-free and exclusive, but also very slow connection. Only when end devices use several channels in parallel are gross data rates of 150, 300, 450 Mbit even possible. The very high rates also require that all end devices can handle the new coding method. An older or "simpler" terminal device may force the entire system to use a throttled "compatible" mode of transmission. Due to different ranges and locations, end devices cannot even reliably see each other and thus recognize collisions and free channels. All of this has to be "regulated". WiFi is Ethernet and uses the CSMA-CA procedure, i.e. everyone is allowed to send and a collision is recognized and the packet is sent again.

Do the experiment on yourself
Use Lync with a "wired" network and once with a WIFI adapter. You will notice limitations much more often with WiFi. Then if you move your notebook, things get worse. And now let's talk about mobile devices with WIFI.

From Lync's point of view, WiFi is simply a "network" and, similar to wired Ethernet, transmission performance and speed are not predictable and usually not guaranteed. So what should be so different in WiFi?

The simple cell

Let's assume a simple radio cell, which everyone probably has at home on their DSL router. A Fritz! Box or another device spans a radio cell, which varies in strength depending on the range. Station 1 and 2 are very close and can transmit at a very high speed. In simplified terms, we calculate even numbers and only one channel, i.e. no 2.4 and 5 GHz bands and no parallel channels.

  • PC1 sends 10MByte to a server via WiFi.
    If it were allowed to send all by itself, then with a 100MBit WLAN it would have taken 1 second.
  • PC1 and PC2 each send 10MByte
    Now "something" has to control who is allowed to send and when. The AP takes care of that. Ideally, it allocates the same bandwidth to both stations, i.e. each now only has half and the transmission takes 2 seconds. In reality, it takes even longer because there must be a gap between the packages

  • PC3 and PC4 are in the picture
    The fact that the AP has to take on a controlling function can best be described when two devices are so far apart that they cannot see or reach each other. If PC3 were to send and PC4 did not "hear" it, then PC4 would also send and the signals would therefore overlap. For a similar reason, the 10 Mbit BNC Ethernet networks were not allowed to be more than 185 m long and not to have more than 4 repeaters. The "Yellow Cable" ended after 500m

So the AccressPoint has to be a controlling component here. This superimposition with collisions can still be disruptive if several AccessPoints transmit on the same frequency. But more on that later

Distance, throughput and airtime

First I come to a much trickier problem, which most operators are not aware of, even with a single AccessPoint without external signals. I have now colored the rings in the picture because, depending on the distance, the signal strength becomes weaker and thus the ratio of useful signal and noise deteriorates.

With WiFi, the clients and access points react by simply sending more slowly, i.e. allowing more "time" to pass per bit so that the signal remains recognizable. We humans also do this by speaking more slowly when the other side doesn't understand us otherwise. Now look at this picture and remember the numbers from above.

  • PC1 and PC2 could send with 100MBit
    If you would do that at the same time, each would have 50MBit left
  • PC3 only sends with 10 Mbit
    It is further away and since the signal is weaker, it has to work with fewer "bit / sec" in order to be understood by AP. This also applies in the opposite direction. Because he still wants to send the same amount of data in bytes. the transmission takes longer. If it sends 10 megabytes with 10 Mbit, then it occupies the medium for 10 seconds.

The problem here is that the two PC1 and PC2 cannot use the 100MBit at all, despite their proximity to the access point. They send so fast, but their turn is much less frequent because the PC3 simply occupies the medium for such a long time due to its slow speed. If someone at the supermarket checkout puts the goods "slowly" on the belt, then you wait behind it too.

This is particularly bitter for Lync, since "waiting time" means a delay in the audio packets, i.e. the jitter is very different and packets could even come too late, so that they are displayed as "lost".

Most administrators are not aware of the fact that more distant WLAN clients can not only use less bandwidth, but also damage the other clients in the cell. Incidentally, this also applies to clients that simply cannot send faster yet. An ancient 11MBit client can be very annoying. In addition, there is a higher risk that during a long transmission a fault will render the entire package unusable and the slow client will have to send the package again.

When using WiFi, it can still happen that two clients cannot see each other directly and therefore the CSMA / CA collision avoidance cannot take effect. Then the clients switch to your RTS / CTS procedure, i.e. the client requests "transmission authorization" and the access point controls. Not only can collisions be reduced in this way, but the AP can even "control" and prioritize.

I did it once with the following setup. A gigabit server is connected to a Fritzbox 3940, which spans a 5 GHz WLAN in which there are two clients

A server process was started on the two clients, which sends all the packets from the two clients back to the central server. You can't "see" anything on the server, but you can also see the two clients next to each other.

Here the measurement when both clients are next to each other approx. 1m away from the access point:

Unfortunately, both clients are not "equally good" with regard to their WLAN network card, but you can clearly see that one client on the left is already transmitting 70-80 Mbit and the client on the right is still 10/20 Mbit. Now I've removed the right client so far that it had to be slower. from the 300Mbit "connection speed" it has dropped to 50 Mbit.

The right "moving" client breaks down to 9/12 Mbit and the round trip time goes up to 170ms. But the left client now also has a round trip time of 9.7 ms and the data rate is only 26/30 Mbit. Note that the scaling of the Y-axis has changed compared to the first image.

This test is only a rough approximation to show the effect. Strictly speaking, you would have to take two identical systems, which then get the same throughput through longer measurements with mean values ​​and then have to go again with a certain distance. Unfortunately, the tools must always be "accurate to the second", which leads to a poorer representation here. I hope to be able to update the data with professional equipment in due course.

The solution to this question is a finer segmentation, i.e. a reduction in the size of the cells by deliberately weakening the signal at the access point and installing several APs.

overlap

But if I now install several APs, then of course I am not allowed to operate neighboring APs on the same frequency band. In the 2.4 GHz band, there are channels 1-13, which have a width of 5 MHz. These can, for example: be summarized in three groups

Unfortunately, not all operators stick to switching to one of the three bands, and so a WiFi frequency band can sometimes look messy


Source: Snapshot with inSSIDer in my office

In public places this can still be quite tight. This then inevitably leads to "collisions" and a significant reduction in the actually available bandwidth. Therefore, it is also interesting for companies with suitable end devices to switch to the 5 GHz band. There are significantly more channels here, so that you have up to 19 areas without overlapping. However, higher frequencies not only have the advantage of higher data rates but are also more attenuated. So you have to install more APs.

Smaller cells are more interesting for companies anyway. There are supposed to be companies that do not plan any network cabling to the table at all, but instead give each room a small AP that supplies this room but is shielded from the neighbors by the walls

With three bands you can try to create areas without overlapping at least on one level. This is of course an idealized picture, because walls, furniture, cables etc. have a strong influence on the lighting.

With such an area, however, you should hope that the ceilings and floors of this transmitter shield sufficiently so that you do not have to plan in the third dimension. Then 5 GHz is really more important.

Change of AP

The shielding and clear delimitation of radio cells makes sense for another reason. Most clients with WiFi are no longer stationary, but can move. Notebooks and smartphones are classic examples of this. Another peculiarity of WLAN clients comes into play here, which only search for a new AP when the old AP is no longer visible.

The client is first connected to the green AP1 at position A and moves to position B. However, since he can just barely see the AP1, the association is retained. The client then brakes all other clients in the area of ​​AP1. However, it is closer to AP2 and could not only transmit faster and more securely there, but also would not worsen the bandwidth of AP1.

Unfortunately, currently no client supports an intelligent change of the radio cell. There are already APs (e.g. Aruba) that determine the location of the client and determine the best AP for this.Then all other APs are instructed to disconnect from the client so that the client has to find a new contact. However, this is not entirely trivial either, since such an AP change means a short interruption in the connection and, especially in an environment with many SSIDs, the client first has to see which SSID can be used further. Possibly. authentication procedures are still required. With such a handover, the client should definitely stay with the same SSID and also keep the IP address. Company systems can do this without any problems.

Lync in the company

So if you want to work with Lync and WiFi in a company, you first have to take note of the fact that more and more clients are romping about in the WiFi network and thus the bandwidth is divided. In office communities or downtown locations, you also have to deal with the fact that other companies also work with WiFi and that there is therefore overlap and radiation. in the 2.4 GHz band it can get tight. So make sure that the new end devices can not only use 5 GHz, but also use it.

In addition, it can be interesting to "measure" an environment once, i.e. you go through the rooms with a PC and note the irradiation of the various APs in order to then create a small map. It is definitely worthwhile for both of you to speak to your neighbors on their WiFi channel. I have already seen some WiFi APs that did not adhere to the channel assignment but were "optimized" manually and thus partially spanned two areas.

Otherwise, setting up a WiFi network in a company is a very interesting thing, for which you should definitely consult the appropriate experts. With the simple wild installation of several APs, with what you now know, you often make it even worse.

Lync in the home

A company can install a system with one controller and several APs, but if you are thinking of owning your own home then you have to rethink this. Lync is also used "at home", be it as a home office, substitute, standby, etc. But at home there are completely different challenges.

  • Internal co-users / new media
    The LAN and WiFi in the home can be so fast, but everything has to go through the common bottleneck "Internet connection". DSL with 16MBit or 50MBit sounds like a lot, but what is left over for Lync when the roommates are watching videos, playing games, Skype or downloading software and updates cannot be foreseen, but has a direct impact on "real-time communication"
  • External neighbor
    If you work via WiFi and are not in a remote homestead, there are neighbors who usually also use modern means of communication. Accordingly, there can be overlaps with WiFi.
  • shielding
    In a current new building, the concrete reinforcements in the ceilings and floors are so tight that they shield WiFi very well. It is not very likely that a Fritz! Box in the basement can supply a client in the children's room on the 1st floor, except through the "gap" in the stairwell. But that is not reliable.

For many reasons it can therefore be interesting to lay a LAN cable or to connect stationary devices via PowerLAN, i.e. to transfer the data via the power grid in order to use the "air space" for the really mobile devices.

With regard to the external connection, it can be interesting to do a performance test. CacheFly, for example, offers test files (1MB, 10MB, 100MB) at http://cachefly.cachefly.net/speedtest/, which you could regularly download via WGET, PowerShell or similar in order to have the effective bandwidth. The files are supposedly generated dynamically so that proxies and caches do not falsify the measurement. Combine this with an HTTP sensor with PRTG and you already get an insight into when the internet line from your provider is delivering even less than you paid for.

But this does not solve the WiFi in the house LAN. There are several solution strategies for this problem as well. Here is a solution:

I started with a Fritz! Box to quickly notice that the shielding of the ceilings leaves me no choice but to work with several APs.

So I set up an access point on each floor that can handle 2.4 / 5 GHz. In the first consideration I gave everyone the same SSID, but that made it more complex to recognize the AP that the device is working with and which one it can still see. Therefore, each AP currently has its own SSID for the floor and the frequency. Of course, this means that I have to enter all six APs for each device. But on the other hand I can see immediately which AP I am connected to and I can switch manually. This can be useful, e.g. when I'm in the garden and can thus reach the AP on the ground floor as well as on the upper floor.

In my case, the APs were simply connected using a pre-laid CAT6 cable. If you live in an old building, you can optionally set up the connection via Power-Lan. I advise against using WLAN repeaters, because most of them send and receive on the same band, so that each transmission practically occupies the channel twice.

In the example, of course, only the central AP also assigns the IP addresses. The other two APs work as pure repeaters. This has the advantage that you change the AP when changing floors, but the IP address remains the same based on the MAC address. The only unpleasant thing here is that you have to add all the APs that you want to use once for a new device. However, this can be prevented if all APs use the same SSID with the same password. Then, however, it becomes almost impossible for the average user to recognize which AP he is connected to and in the event of an error, the search for such a "home environment" is significantly more time-consuming.

I have recently had very good experiences with it. Maybe at some point I'll change the SSIDs to a common name again, so that the entry on new clients is a little easier. for Lync and other protocols it makes no difference

The separate names for 2.4 and 5GHz also make it easy to determine which AP is present at which point and how strong it is and whether a terminal can communicate with 5GHz at all.

WiFi WMM

You have already read above that the access point is like the boss in the ring, allocating talk time. There are several extensions to the standard, especially to make audio / video or telephony via WiFi more effective. Unfortunately, these extensions often only support the enterprise components again. However, these only make it possible to make useful and reliable phone calls via WiFi using handheld devices such as the Spectralink 8400. If you want to replace your DECT telephones with WiFi, then it has to be comparably stable.

A PC that is temporarily not connected during a movement is less critical, but those who use the phone as a matter of course will have no understanding of interrupting calls.

  • QoS Wi-Fi Multimedia (WMM)
  • WMM Power Save
    This standard allows the transmission power to be controlled. For example, an AP can tell the mobile device how strong it is reaching it and the mobile device can reduce the transmission power accordingly. Without this function, the handset would always transmit at high power, which would be bad for the battery life
  • Support WMM Admission Control
    The AP can use this standard to control who is still admitted to the radio cell. a certain bandwidth will usually be provided for VoIP and if too many incompatible devices were associated, this guarantee could no longer be upheld

Some standards must therefore be met by the client in connection with the AP. Others are more related to the AP. If the infrastructure for WiFi does not exist and can only be created at a high cost, "IP over DECT" could be an option. DECT works with time slots

more links

Design Principles for Voice Over WLAN
http://www.cisco.com/c/en/us/solutions/collateral/wireless/4400-series-wireless-lan-controllers/net_implementation_white_paper0900aecd804f1a46.html

Network Planning, Monitoring, and Troubleshooting with Lync Server
www.microsoft.com/en-us/download/details.aspx?id=39084