I would like to ask a question about Mikrotik radio cards on channel bandwidth receive, for example if i use frequency scan from AP1 which is using 5320 and 20Mhz bandwidth, it shows signal strength of -24 @5700, -28@5865, -34@5200 from my other AP’s on the tower will these signals effect the receive signal on AP1 at 5320, and is the bandwidth gain of the receive side of Mikrotik if set for 20mhz actually 20mhz or does it remain full spectrum bandwidth regardless of what channel has been selected?
The question is also, is the transmitter only radiating within a 20MHz bandwidth or is it radiating more than this? So it could be the card is seeing what is really out there, or it could be the receiver is being overloaded and the signals are only inside the receiver?
The way to tell is to measure the level you’re getting now, insert 10dB attenuation and see if the signal falls by 10dB or more. If more, then the signal being received is being generated within the receiver.
Reducing TX power on your other cards by say 3dB, does that cause the levels elsewhere on other frequencies also to fall by 3dB? If everything is ‘real’, then the others should also fall by 3dB. If they are spurii or intermods, they will change by a different amount or even disappear altogether.
My question was on the receive (listen) RX bandwidth, is it strictly @20Mhz -3db and could it be lets say -10 @40Mhz, etc
I am curious how steep the bandwidth roll off curve is and if this would have any effect on the performance of individual AP’s on a tower having several sectoral AP’s
If all you’re interested to know is the theoretical rx bandwidth rolloff response figures of a particular card, you will have to research the manufacturer’s datasheet for the particular card you are referring to. My point still stands. Further to that, what should happen in theory is often wholly different to reality. ETSI have a mask it should comply to, but each card will be better or worse at complying with that spectrum mask.
Therefore relying just on the receiver’s theoretical selectivity performance to try to determine if interference will occur is not the only parameter that should be considered.
If all you’re interested to know is the theoretical rx bandwidth rolloff response figures of a particular card, you will have to research the manufacturer’s datasheet for the particular card you are referring to…
thanks for your reply,but so far i have not seen any data about the receive bandwidth spec on the manufacturer’s datasheets?
Therefore relying just on the receiver’s theoretical selectivity performance to try to determine if interference will occur is not the only parameter that should be considered.
Maybe i should have expanded my question to include the antenna as well, for example when using sectoral AP antenna with full spectrum 5.8 bandwidth.
My thoughts are to move each AP in/out on the support arm on the mast while still being directed to the target coverage area for any reduction in signal from the other AP’s on the mast (side lobes signal reduction)
What is the specific card you are looking for data on?
Maybe i should have expanded my question to include the antenna as well, for example when using sectoral AP antenna with full spectrum 5.8 bandwidth.
Please expand what you mean by that statement.
My thoughts are to move each AP in/out on the support arm on the mast while still being directed to the target coverage area for any reduction in signal from the other AP’s on the mast (side lobes signal reduction)
Always a wise decision to move antennae away from each other, but it depends on the antenna’s characteristics as to whether a vertical or horizontal separation will give the greatest benefit.
Then of course, there are the coax feed cables, pigtails and separation of the whole RF path between different transmitters, including the cards themselves, earthing points.
BUT if a transmitter is generating RF outside of it’s intended channel, or radiating on multiple channels due to some non-linearity, you can separate the bits as much as you like, you’ll still have co-sited interference!
What is the specific card you are looking for data on?
XR5 for starters?
Quote:
Maybe i should have expanded my question to include the antenna as well, for example when using sectoral AP antenna with full spectrum 5.8 bandwidth.Please expand what you mean by that statement.
For example if as mentioned if the AP in question is using 5320, using a XR5 with a 16dbi full spectrum antenna, this AP in RX listen mode what will the received signal be at 5200, and 5260, from the other AP’s all on the same polarity, there are other AP’s also used on this mast but the frequencies are much further apart and some are horizontal.
BUT if a transmitter is generating RF outside of it’s intended channel, or radiating on multiple channels due to some non-linearity, you can separate the bits as much as you like, you’ll still have co-sited interference!
No - not generating rf outside it’s assigned channel but if i could measure direct from the AP using a spectrum analyzer what is actually being picked up and at what level.
The XR5 is based on the Atheros ROC. Therefore specific performance data is only available by directly contacting them. They are renown for not publishing any technical data beyond that which will sell their chips except to OEMs.
If you want to learn a little more on co-sited interference perhaps have a read of “Adjacent Channel Interference in Dual-radio 802.11a Nodes and Its Impact on Multi-hop Networking” by Harvard University.
Still not sure what a “full spectrum antenna” is exactly. Full spectrum would suggest DC to infinity. Do you actually mean an antenna designed to operate over a band of frequencies with a performance that is not necessarily consistent across that band, but is able to operate without too much return loss or too high a vswr so as to cause any serious issues for the transmitter? One that will have a widely varying gain figure, but the manufacturer always picks which ever one is the greatest gain figure, without telling you for which spot frequency that was for? One that is therefore listed by the manufacturer as being “5GHz band” or maybe even “5200-5800MHz”? 
Your example given is impossible to quantify as you have not provided enough data to calculate the interference figures. The amount of signal from one AP, being picked up by another on the same site, depends on many more factors than just the aerial type, polar diagram and gain.
no - not generating rf outside it’s assigned channel
All 802.11 transmissions radiate RF outside their assigned channel. Especially so on 2.4GHz! (Because the radiated channel bandwidth is greater than the channel spacing). The amount radiated also depends on if the TX device is being operated outside it’s published test parameters and if there any other non-linear devices in the vicinity, or connected to it (amplifiers for example).
Still not sure what a “full spectrum antenna” is exactly.
- 16dBi 4.9-6.0Ghz Sectoral Gain, http://www.itelite.net/offer.php?n=36&lang=en
This antenna with a XR5, which if i assume correctly will receive equally any signal in the 4.9 to 6Ghz spectrum,
while in transmit cycle the bandwidth maybe 20Mhz but as soon as the radio card goes into receive mode, will the bandwidth be much wider,
Ok it follows that if used a limited bandwidth antenna’s like say 5.4 to 5.7 would have a higher rejection to signals outside these frequencies but losing the flexibility to change frequencies outside the bandwidth.
Firstly the antenna given as an example of a “full spectrum bandwidth aerial” operating between 4.9 and 6.0 isn’t. It is ‘only’ 5.1 to 5.9GHz.
It is also clearly only really resonant at the highest point of it’s stated band coverage. Somewhere around 5.8GHz. Which is probably where the gain of 16dBi has been measured. At all other frequencies it is operating at less than peak performance. Not surprising given it’s construction design. The sidelobes at +/-80 degs are quite high (around -20dB), meaning that if there were more than one on a mast, mounted side by side, they would cause RF to get into each other unless you could get them to be very far apart. But as this is clearly designed to be a CPE, this is unlikely at a Base Station as this would not be the right aerial to use for that. As a CPE unit it’s fine and ‘fit for purpose’.
Also the phrase “full spectrum” is meaningless. Like I said, “full spectrum” means exactly that, the full spectrum. I.e. DC to iffinity. a 800MHz bandwidth from approx 5.1-5.9GHz is clearly not “full spectrum”. Probably better to describe it as “full 5GHz band” if you really wanted to call it anything at all.
The receiver will NOT immediately open up and thus “see” the whole band. The filtering is not done by the antenna bandwidth. It is done with a BAW resonator or in some cases with a single LPF in front of the LNA within the Receiver and software filtering using a FIR Digital filter. However, the better the filter characteristic, the longer the filter length, the longer the time taken for a signal to pass through it, thus the slower it performs, the worse the latency. Therefore it is always going to be a compromise between getting a filter mask to make it work and not too complex to affect the throughput performance. But the short version of the story is that the receiver filtering will be very nearly the same as the spectral mask used for the TX filtering.
There is something further to be taken into account and that is the AGC characteristic of the Receiver. Therefore when a receiver is picking up a strong signal, on channel, the gain of the LNA will be driven down so as to always keep the signal to around -40dBm (exact value depends on model) and therefore the co-channel response shape will improve and reduce co-channel interference.
So.. once again, here is another parameter that is required to be taken into account when trying to calculate if adjacent channel interference will occur or not!
The antenna will hardly offer any rejection at all of out-of-band signals (e.g. signals at 2-5Ghz or 6-8GHz). That role is nearly all down to the quality of the RF filtering and 3rd order IMD performance of the LNA of the wifi card. At out of band frequencies, the polar diagram changes significantly, for example at 4GHz, the main lobe may now be at 45degs from the centre line, admittedly with reduced gain, but it will still pick up the signals.
It is possible to obtain inter-digital filters or cavity filters to reduce co-channel interference, but these are large and if narrow enough to cover just one channel, are very expensive. It also means that if you need to operate on a different channel to that for which the filter is tuned for, you have to either retune the filter (a skilled process and time consuming, even WITH all the right equipment) or buy another one. I can recall working for a microwave company where my sole job was to tune 7GHz inter-digital cavity filters for a military contract, which had to have a flat response curve within +/-1dB over a 1 GHz band and it took 5 days to do just one!
Bottom line - what do you think the performance will be like when you only spend a few dozen bucks for a wifi card?! It’s a miracle it works at all!
rarely seen such an interesting topic! I think we all can learn from it. so I have some questions/remarks too:
Why is this? I would think it is the middle frequency. So a 5.4 to 5.7Ghz antenna would perform best at 5500Mhz?
Or lies the answer in the fact that the actuall 5Ghz band is devided in two parts. One lower and one upper part where the last is for outdoor use. Thus by designing antenna’s for some frequency in the upper part it makes more sense than just the centre of the 5Ghz band spectrum? After all, who wants a 28dBi raster in his living room! 
The receiver will NOT immediately open up and thus “see” the whole band. The filtering is not done by the antenna bandwidth. It is done with a BAW resonator or in some cases with a single LPF in front of the LNA within the Receiver and software filtering using a FIR Digital filter. However, the better the filter characteristic, the longer the filter length, the longer the time taken for a signal to pass through it, thus the slower it performs, the worse the latency. Therefore it is always going to be a compromise between getting a filter mask to make it work and not too complex to affect the throughput performance. But the short version of the story is that the receiver filtering will be very nearly the same as the spectral mask used for the TX filtering.
There is something further to be taken into account and that is the AGC characteristic of the Receiver. Therefore when a receiver is picking up a strong signal, on channel, the gain of the LNA will be driven down so as to always keep the signal to around -40dBm (exact value depends on model) and therefore the co-channel response shape will improve and reduce co-channel interference.
BAW resonator = ??
LPF = ??
LNA = ??
FIR Digital filter =??
AGC = ??
“co-channel response shape” = ??
I know we can all google for it, but it would make it much more readable, and thus learning, to have these terms explained here. Can you please add these to your further very interesting text?
Can you give us a better understanding of what the “length of the filter” is and why it is delaying the signal so much we would notice the increase in latency?
I am from the times when I as a sailor learned about radar technology, where the filtering was done with basically physical filtering of signals. I can remember something about electronic filtering in these days but must agree the theory behind all that in my head is pretty much gone, if not outdated…
3rd order IMD performance of the LNA of the wifi card.
IMD = ??
Bottom line - what do you think the performance will be like when you only spend a few dozen bucks for a wifi card?! It’s a miracle it works at all! >
Ha! That’s good one! Hence MT has so much problems getting nv2 properly to work in heavy crowded spectrum situations.
WirelessRudy - Thank you.
Why is this? I would think it is the middle frequency. So a 5.4 to 5.7Ghz antenna would perform best at 5500Mhz?
Because, looking at the VSWR response curve for the aerial given as an example, it definitely dips at around 5.8GHz. If you notice along the whole of the rest of the band it is waving up and down due to the multiple resonances of each of the component parts that make up the antenna design. I.e. each bit of the copper clad etching on the circuit board is slightly tuned (by design) to different segments of the band to broaden the original designed frequency it is really most resonant on. In this case, it would appear that it was originally for Band C (5.8GHz+/-) and was stretched downwards by adding more copper or slightly increasing some of the line lengths between them. We are of course, only talking about very tiny amounts of adjustments. But as I have no intimate knowledge of how this one was made, besides having taken one apart last year, it is difficult to be sure.
I would also expect to see the gain figure vary by around -2dBi to -3dBi across the band from the figure obtained at 5.8GHz. So if perhaps you were operating at 5.3GHz, the same aerial may be providing -14dBi gain and with a slightly different polar diagram even? I have seen aerials that have shifted their angle of fire by a few degrees depending on the frequency being used!
(I am also always deeply suspicious of any manufacturer’s claims of an antenna gain unless they can guarantee that every aerial is tested rather than the first one they built in the R&D lab and sent away for testing. I have seen aerials claiming on paper one thing, but reality is a lot lower. I have see some Omni aerials that are physically much longer in length, making one think they have a higher gain, yet when you open them up, inside they contain the same identical copper wiring or PCB boards inside as their lower gain models! Once again - “you get’s what you pays for”)
Sorry about the acronyms, but this is my stock and trade.. OK
BAW : http://en.wikipedia.org/wiki/Electronic_filter
LPF : http://en.wikipedia.org/wiki/Low-pass_filter
LNA : http://en.wikipedia.org/wiki/Low_noise_amplifier
FIR Digital filter : http://en.wikipedia.org/wiki/Finite_impulse_response
AGC : http://en.wikipedia.org/wiki/Automatic_gain_control
co-channel response shape - that was a error. It should have read ‘adjacent channel response shape’. Sorry for any confusion.
Anyway, “adjacent channel response shape” is a phrase made from “adjacent channel” i.e. I am referring to the response of the receiver on the adjacent channels. “response shape” refers to the receiver’s selectivity curve. Or how fast the roll-off is in the receiver’s performance at off-channel signals. Or in other words the shape of the selectivity filter. The faster the roll off, the greater the slope angle, the greater the attenuation of signals that are away from the centre frequency of the channel it is actually on. Filters also need to be wide enough to take the smaller and lower in level side lobes of the transmitted phase modulated signal. A Wifi signal does not generate all it’s energy in a neat square shaped channel. There is a main peak in the middle, it may even be quite flat across the top (especially with 802.11n), but there is always energy contained further down on each side of that main lobe of energy which must not get filtered out as it is still an important part of the signal that is required to be processed.
“length of the filter”. How many iterations the software program has to loop around. Obviously, the more times the software has to loop around the same piece of code, performing the computations, the longer the amount of time it will take for the answer to “pop” out the other end. Therefore the longer the period of time it takes from the time the signal arrived at the start of the filter to the time the signal is finished being processed at the ‘other end’. Therefore the greater the latency. Also, there is only so much CPU power in these little chips, so any software needs to be optimised as well and compromises made!
It could be argued that hardware filtering is always better, as it requires no processing at all, being a natively inert process of ‘just’ using capacitance and inductance or reactance. However with the requirement for these boards to cover ever increasing bandwidths, those hardware filters will require tuning to enable them to cover the full band. To engineer into a design this auto-tuning under computer control is very difficult when dealing with microwaves and in turn, their millimetric wavelengths as everything interacts. Or in simple terms, it simply becomes a nightmare to design. Therefore by using software filtering, the costs are much reduced.
IMD : http://en.wikipedia.org/wiki/Intermodulation
My original training was as a Radio and Electronics Officer in the Navy, with Radar Maintenance, SOLAS and of course, Morse Code at a minimum of 25 wpm! So it appears we may have some similar backgrounds!
The sidelobes at +/-80 degs are quite high (around -20dB), meaning that if there were more than one on a mast, mounted side by side, they would cause RF to get into each other unless you could get them to be very far apart.
About 1meter apart but was considering as the side lobes may vary from antenna to antenna to physically move the antenna while monitoring the signal levels from the other AP’s for min signal,
But as this is clearly designed to be a CPE, this is unlikely at a Base Station as this would not be the right aerial to use for that. As a CPE unit it’s fine and ‘fit for purpose’.
What is the best AP antenna for price and performance for 5.8ghz?
Bottom line - what do you think the performance will be like when you only spend a few dozen bucks for a wifi card?! It’s a miracle it works at all! >
Same question - what is the best radio card for price performance,
You mean moving the antenna backwards and forwards horizontally? The other problem is the casing of that CPE is plastic. Therefore any RF from other APs will easily get into the sides and get directly into the wifi card and electronics. Electronics that have transistors, transistors that have semi-conductor junctions, which are non-linear when driven outside their operational spec and what does non-linear devices then in turn cause? Interference. IMD’s and harmonics. That is why it is best to leave plastic boxed CPE devices to what they were designed for. For the CPE, where the chances of being near to other high levels of RF (in-band or out-of-band) are remote.
How long is the string and is it stretchable? OK, I know I sound like I’m being sarcastic - but that is such a horrible question to ask as it depends on what you want it to do and how deep your pocket is! I would always recommend separating the antenna from the base station equipment at a base station. You choose the type of antenna on what you want to achieve and checking the data sheets given by the supplier are validated. That pushes up the price as that means the aerials all need testing before they leave the factory. Or you do it.
But from then onwards there are so many criteria to take into account that I would find it difficult to answer that fully here because a) my lively hood depends on providing best practise advice to businesses that I can charge for (so why give everything away free here) and b) after 30 years in the business I am loathed to start writing an entire book on this forum on how to do RF installations properly, taking into account every possible eventuality and situation. Even if I did, I suspect that someone will cut a corner somewhere (as they don’t understand the technical consequences of skimping on something that appears to them as being a tiny little detail of no importance) and as a result whatever they have built will fail to work as well as expected and who gets the blame for that? Me? No thanks.
bet you can guess the answer I want to give to that?! But the Ubiquiti and Mikrotik cards are fine for what you pay for them. If you want even higher quality start looking for words like “carrier class”.
To get best performance from a link or understanding the complexities of site sharing means understanding the risks of utilising unprotected spectrum, with very low cost equipment, yet expecting it to perform the same as links that cost tens of thousands. Understanding how to minimise antenna coupling, the usage of cavities where necessary, also when you’ll need RF circulators and earthing everything correctly. Even knowing what types of equipment enclosures to use depending on the known extremes of the environment, the type of waterproofing tapes to use… the right type of CAT5, power and coax cables. Doing things the right way cost money. I have heard far too many people claim “why spend a hundred grand on a microwave link, I can do it for 100 bucks and get the same results!” Yeah, right.
You mean moving the antenna backwards and forwards horizontally? The other problem is the casing of that CPE is plastic.
Yes horiz movement - from your reply i would assume the case for the AP must perform two functions apart from the usual weatherproofing aspects (1) provide screening to reduce RF pickup from rear or all sides (2) to augment the antenna performance by acting as a “reflector element”
How long is the string and is it stretchable? OK, I know I sound like I’m being sarcastic - but that is such a horrible question to ask as it depends on what you want it to do and how deep your pocket is!
Aah no harm in asking
I would always recommend separating the antenna from the base station equipment at a base station
I have to date used intergrated antenna+board+radio card as the example listed, less interconnect cables used, regardless of quality as they can change the effective the rf load presented to the radio card, just calculating insertion loss for connectors and cable length can be misleading, can only reduce the TX power from a AP.
Once again thank you for your replies.
1 yes, 2 no. The antenna is placed in the front lid and the back of the antenna has a steel plate behind it as ground plane / reflector. It also reduces RF getting back into the electronics. But as it is only a plate, and not a box, it is not a faraday cage. Therefore it offers no screening effect. The reason why those boxes have no screening is because they are low cost and they expect you, as the purchaser, to understand it’s correct application! (In one situation where a client did not wish to utilise a different type of box, we advised spraying the inside of their Mikrotik Outdoor case with Nickel RF Screening spray paint and this was very effective at reducing interference).
Placing multiple CPE boxes of that size on one mast begins to also look unsightly. I hope your neighbours don’t care!
less interconnect cables used, regardless of quality as they can change the effective the rf load presented to the radio card, just calculating insertion loss for connectors and cable length can be misleading, can only reduce the TX power from a AP.
Putting the enclosures a metre or two away from the antennae reduces RF getting into them. Using metal boxes, further reduces the chances of you getting interference from other sources and you causing interference to them. 1m of very low loss coax will only drop the signal by about 2dB approx, but the benefits of getting a much lower noise floor and less interference more than make up for the small signal loss. Anyway, by using better antennae with gain figures which are the same in reality as on paper can make up for that loss too.
1m of very low loss coax will only drop the signal by about 2dB approx
Yes but with reference to a PTP link that loss is doubled to 4db, a quick calculation using XR5 card is @54 is 23dbm = 631mw, 4 db reduction = 19dbm = 80mw
If you have a badly engineered system where you are getting 20-30dB of interference, the link keeps failing or you have a low throughput, but then you install it correctly with a 4dB loss, but the removal of that 20-30 dB of interference, are you still trying to argue that gaining the 4dB is more important than the loss of that interference?
Also, if you can’t afford to lose 4dB on a link, without it falling over, you need to re-assess your link budget! Any link with less than 20dB of fade margin is asking for trouble. Yes, you can make them work with figures of between 10dB and 20dB, but they will not be as reliable and there will always be the risk of it failing more often. If your fade margin is only 4dB, then you already have an even more serious problem than worrying about losing 4dB in cable losses.
If you have a badly engineered system where you are getting 20-30dB of interference, the link keeps failing or you have a low throughput, but then you install it correctly with a 4dB loss, but the removal of that 20-30 dB of interference, are you still trying to argue that gaining the 4dB is more important than the loss of that interference?
No - i have for example a 20.5Km ptp link which has a fade margin of 29dB, my policy is simple - don’t insert any item in the broadcast/receive chain unless absolutely necessary, i take all of your points and thank you very much for taking the time and the effort in the detail of your replies,
..we advised spraying the inside of their Mikrotik Outdoor case with Nickel RF Screening spray paint and this was very effective at reducing interference).
Good suggestion but i would be concerned that paint onto plastic will crack and break when the AP is vibrating in high winds on a exposed location, maybe silver foil stuck onto the inside of the plastic case and once again because of a wavelength of 5.172cms i don’t want to solve one problem and generate another?
All of this effort and research is into why just one AP on this tower has regular disconnects/reconnects when using NV2 but OK on 802.11 and the other AP’s are OK using NV2, I have tried tweaking this AP in the advanced setting tab?
The only way I can make a 20.5km link on 5.8GHz give me a 29dB margin is if the EiRP is 50dBm! (i.e. 100 Watts) Am I assuming this is your method of removing all obstructions or losses - like birds - you fry them out of the way? 
(XR5 at 28dBm, 1dB pigtail&connector loss, 23dBi antenna gain = +50dBm EiRP then -134dB Path Loss, 23dB Antenna gain, 1dB pigtail&connector loss, assuming 12MBps connection rate, -91dBm Receiver sensitivity required = 29dB SNR)
Tried aluminium foil once a long time ago but had some many practical problems trying to make it fit the shape of the box without one single little break in the foil (which then breaks it’s effectiveness) - so that’s why I now only use the paint. It works VERY effectively. It covers every single tiny little corner or bend thus ensuring a complete unbroken surface and provides a hard covering once dry. I have never seen it crack with use. I have seen boxes I’ve done years ago and they are still intact. But overall, you are still better off using a metal box, a short length of coax and a decent aerial in the first place!