[HamWAN PSDR] [Netops] Proposed Revision to Spectrum Allocation Standards

Bart Kus me at bartk.us
Fri Dec 25 10:06:32 PST 2015

When you propose elimination of the 5MHz and 20MHz channel widths, are 
you referring to their central frequency spacings (staggering) or to the 
bandwidths themselves?  There is a 3dB gain of 5MHz bandwidth over 
10MHz, which really helps some of our weaker clients:

[eo at Baldi-S1] /interface wireless> registration-table print
  # INTERFACE                              RADIO-NAME MAC-ADDRESS       
  0 wlan1-gateway                          WE7X/Baldi D4:CA:6D:7A:B8:81 
no  -83dBm          4.8M... 19h38m30s

[eo at CapitolPark-S2] > /interface wireless registration-table print
  # INTERFACE                              RADIO-NAME MAC-ADDRESS       
  0 wlan1                                  KU7M/Newcastl... 
D4:CA:6D:9D:5C:07 no  -86dBm          3.2M... 9h21m7s

[eo at SnoDEM-S2] > /interface wireless registration-table print
  # INTERFACE                              RADIO-NAME MAC-ADDRESS       
  0 wlan1                                  AE7Q/MillCree... 
D4:CA:6D:54:B4:F5 no  -87dBm          2.2Mbps 18h25m26s

I can't foresee us taking 3dB away from these folks.  So as far as the 
bandwidth is concerned, I think 5MHz bandwidth is here to stay because:

     1) it's the highest power spectral density we can achieve with the 
hardware, and
     2) we don't have throughput saturation on our sectors,
     3) which means we also use 5MHz to increase reliability (reduce 
retransmits) of the service.

Now, about the interference between 5MHz and 10MHz plans.  The PSDR 
group here recognized this would be an issue if we wanted to run some 
sectors @ 10MHz bandwidth, so we migrated the network onto a 5MHz 
bandwidth / 10MHz staggering plan.  This gives us the ability to run an 
arbitrary mix of 5MHz and 10MHz bandwidths without any 
self-interference, since the radiation angle for each sector is the same 
in both bandwidths.  This provides an easy next-gen upgrade path without 
further spectral changes.  This upgrade would only take place once we 
hit saturation on a given sector and we'd have to weigh the impact to 
the marginal users vs. benefits to the high speed users.

To support this new 5MHz bandwidth / 10MHz stagger combination we simply 
pushed an update to all the client modems with an additional channel 
plan, and updated our client setup docs to include the new channel 
plan.  This doesn't mean we had to retire the original 5MHz plan, 
though.  That can stay programmed into client modems for any areas that 
wish to use it.  It's still the tightest packing of the channels for 
areas that have very limited access to Part 15 frequencies and therefore 
need to use the rest of the Part 97 5.9GHz segment to run their PtPs.

Any new useful channel plans can simply be added to the official list 
(BTW: we need to publish an official list) without the need to kill 
existing channel plans.  For the sake of reducing search time in the 
clients, we can prune the official list down to a "known active" list.  
Maybe still uploading all the channels, but in a disabled state for the 
unused ones.  This optimization would reduce client modem compatibility 
when new networks come up, so there's pros and cons here.  I'm fine with 
recommending the widest list by default, and if the user so chooses, 
they can self-optimize their local channel list.

Now, given this client flexibility, the new b-channels you propose look 
like a fine idea, and I can appreciate your argument with very nearby 
sites not having perfect directional shielding.  We'd likely have the 
exact same situation if our Council House site came on the air, since it 
would be very close to the Capitol Park site.  Capitol Park also uses 
gen 1 antennas, so their directional shielding is poor.

There will be added complexity if 2+ HamWAN networks using different 
channelization start to share a border or a region.  In those cases, 
they'd have to work together to only use compatible channels in the 
overlapping regions.  I don't even wanna touch on what'd be compatible 
among all the possible channel plans.  Sounds like a headache.  The 
clients should remain unaffected in such cases though, provided they 
have the full channel list uploaded.

The use of omni channels is a very complicated issue.  Can you tell me 
more about how you're using them today without causing 
self-interference?  I'm facing the same design challenge on the 900MHz 

The use of MIMO is a fully compatible upgrade, so there's no problems 
there.  I don't think this has any impact on the standards other than 
maybe updating the "recommended hardware" section.


On 12/25/2015 12:43 AM, Ryan Turner, K0RET wrote:
> Netops, members, and interested parties of PSDR and Memphis,
> I propose the following changes to our official band plan for point to 
> multipoint:
>   * Elimination of the 5 MHz and 20 MHz channel width standards
>   * Introduction of a second set of sector channels
>   * Introduction of channels specifically to be used for omni use in
>     areas where sectors are not feasible (user density, space, budget)
>   * Adoption of dual chains, horizontal and vertical, for the last mile
> See a chart of the proposed band plan here: 
> https://jsfiddle.net/ryan_turner/n13cy7ot/4/embedded/result/
> I propose these changes in order to address a number of issues:
>  1. 5 MHz channel width is not particularly beneficial; its
>     performance improvements are only marginal over 10 MHz, and it is
>     a rare scenario for this to be useful in a last mile situation.
>     Frequently signals that benefit from 5 vs 10 MHz would be much
>     further improved by instead better client antenna positioning. 5
>     MHz channels may be reasonable temporarily should we place them
>     somewhere within an unoccupied 10 MHz channel, but today as the
>     two 5 mhz and 10 mhz plans coexist, they overlap and interfere.
>  2. 20 MHz channel width is not likely to be adopted locally due to
>     marginal nature of all of our links and the limited benefit from
>     the afforded bandwidth. I do anticipate hardware limitations may
>     drive us to eventually use 20 mhz channels, however.
>  3. Since our antenna systems are not ideal radiators, inter-cell
>     interference remains a problem for sites that have LOS between
>     eachother; two of our sites today are able to see eachother
>     entirely -- that is to say, each sector on one can see all five
>     other sectors; introducing more channels for this scenario would
>     greatly help
>  4. In dense urban areas and areas where towers and tall buildings are
>     specifically banned, in order to reach our users we need more
>     frequent, smaller points of presence
>  5. Dual chain radios have already been deployed at multiple sites and
>     found to be better performing than the slightly higher output
>     power mikrotik metals; cell site sector antennas already support
>     dual polarity, and with Mikrotik's mANT30 dish the end-user cost
>     difference is about $40 over our current day recommended solution.
> Please share your thoughts and opinions.
> Thanks,
> -- 
> Ryan Turner, K0RET
> +1.901.300.0039
> HamWAN Memphis Metro <http://www.memhamwan.org/> - Memphis' Amateur 
> Radio Multi-Megabit IP Network
> _______________________________________________
> Netops mailing list
> Netops at hamwan.org
> http://mail.hamwan.net/mailman/listinfo/netops

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.hamwan.net/pipermail/psdr/attachments/20151225/fb7f83f1/attachment.html>

More information about the PSDR mailing list