Last Post: June 28, 2017:
I hope someone might be able to put me straight. I have a Wifi design question that I would appreciate your thoughts on. I can’t go to any vendor engineers because I am doing an evaluation of tenders for a large LPV rollout to racecourses.
One of the vendors came in to present and their engineer put forward an argument that for high density you should design on the following basis
- Design for the worse client – ie 1SS, 11g device
- This implies a max rate of 54M
- Wireless is half duplex so half this rate
- Effective rate is therefore 27M
- So if you want a 1M per client design, you get ~25 per radio
- To keep clients in the sweet spot where they get the 54M rate, no more than 15m from AP to client
- Match the AP power with the client power because APs can transmit at higher power and have better antenna gain and therefore the danger is that the client could hear the AP but not be able to respond.
My contention would be as follows
- Design for 1SS devices, but allow an uplift for 11n/11ac and some 2SS devices (probably not very many) but it will increase the overall averages
- 11n and 11ac have the benefit of MRC and beamforming so should get better RSSI at client
- With a 10dB tx level and 15m, the FSPL is (as per https://wifinigel.blogspot.ie/2014/05/wifi-free-space-loss-calculator.html), 70dB to give a rx signal RSSI of -60dB.
- A Cisco AP at -77dB (https://www.cisco.com/c/en/us/products/collateral/wireless/3700-series-access-point/data_sheet_c78-729421.html) can get an MCS8 connection on 11ac with gives 78M.
- A Xirrus AP at -60dB (https://www.xirrus.com/wp-content/uploads/2015/02/tech-brief-access-points.pdf) can get an MCS 8 connection on 11ac which give 78M (86M with SGI=400ns) data rate (50% above his rate) and 65-72 on 11n.
- The ack is sent after a SIFS and is small in comparison to the data transmission so saying the bandwidth is halved because it is half duplex is incorrect.
- antennas as bidirectional in that they provide the gain for both the transmission and reception.
- my understanding is that the tx rate in each direction is separately decided by the transmitting STA (whether AP or client device). An AP has higher power, so it can transmit using high power and obtain higher rates and they should be received clearly enough to allow decoding by the device. The device, with lower power will have to transmit at a lower rate to get its messages through to the AP.
I believe that the upstream (device to AP) will be slow but the downstream will be faster – which is perfectly matched to a client/server environment – small requests and large responses.
Obviously, each vendor will have their own rules-of-thumb,for such things as "worst case(s)", rate/client, assumed noise levels etc. You should make sure you understand the differences between what each vendor assumes.
e.g. each different vendor's AP will support only so many clients - hopefully you get what you pay for..
I am assuming that when they say /g being the worst, it is because it is 2.4 GHz. A better assumption for 2.4 GHz is /b rates, and the best way to handle them is probably to just disallow them, and make you lowest allowed rate to be 11, 12, or 24 Mbps.
As some would say, design for the worst of the worst - not all clients of the same type perform the same - even from the same lot.
Others would say test with the "least capable, most important" client.
Can't tell from your post, but remember that rates, e.g. 54M is in bits, not bytes.
I would say that 27M out of 54M is being a bit optimistic, 24-26M might be a better estimate (in my experience).
In my office environment, you'd be lucky to get 54Mbps at 5 meters.
Currently VERY few mobile devices are more than single stream - even /ac ones. They don't have the room for multiple antennas needed for MIMO.
Beam-forming only gets you a couple dB at best - it does not make for range miracles.
Yes, (almost all) antennas are bidirectional, but the paths that signals take are not necessarily symmetric in a highly multi-path (MP) environment. Normal reciprocity considerations do not take MP into account.
It is critical that your network have balanced power levels between client and AP, although I like +1, +2 dB more at the AP.
It is critical that any installation have a Post site-survey performed to prove the mfg's claims. It should be included in the price.
Sorry I didn't have more time to answer your post.
Best of luck.
Thanks for taking the time to reply.
The assumption of 11g is not because of 2.4GHz, they make that rate assumption for 5GHz also i.e. 11a. Therefore they propose to ignore any benefit brought by 11n/11ac.
I agree with designing for the worst in general but in an LPV one should assume a mix of device capabilities and therefore allow for some newer device with 11n/11ac performance that will bring up the average throughput. Even without channel bonding and multiple streams, the newer specifications will see benefit over and above the 11a/11g devices?
Anytime you can increase overall throughput by increasing airtime fairness it's probably a good thing.
I am really not the best person to answer some of these issues as the devices I work with, although they support 40 and 80 MHz wide /ac signals, do not get any performance benefits from the higher rates - they work just as well at /b/g rates.
Under most circumstances, when you have a device whose performance is limited by physical limitations, you can't expect much performance increase from higher PHY rates.