Dual Antennae on Wireless Routers

Discussion in 'Wireless Internet' started by Nick, Apr 22, 2005.

  1. Nick

    Nick Guest

    I want to run a wireless router with only one of its antennae.

    The other antenna port on the router will be connected to an external
    yagi for a link to a neighbouring property.

    The other property will have a WAP configured in AP Client mode to act
    as a receiver.

    Will my wireless router still work with only one antenna?

    Why are they fitted with two antennae anyway?

    Any help/advice/comments would be welcome.

    TIA

    Nick
     
    Nick, Apr 22, 2005
    #1
    1. Advertisements

  2. Nick

    Bert Hyman Guest

    (Nick) wrote in
    Maybe. Mine only has one. What will happen if you connect another
    distant and directional antenna in place of one of yours, I
    don't know.
    Space diversity. In theory, having two antennas gives the receiver
    some immunity to fading due to multi-path signal propagation.

    This is why I wonder what will happen when the two antennas are
    really far apart, and one is directional.

    <pedantic>

    When you're talking about radios, the plural is "antennas"; when
    you're talking about bugs, the plural is "antennae".

    </pedantic>
     
    Bert Hyman, Apr 22, 2005
    #2
    1. Advertisements

  3. Obviously, the make and model of the wireless router is of no
    importancd. All wireless routers are identical, and interchangeable.
    So far so good. I suggest you look into dish and panel (patch)
    antennas instead of a yagi. The cost of a yagi for the amound of gain
    delivered is in my opinion excessive. If the range is fairly short, a
    simple patch, biquad, or reflector antenna will work.

    Also, if you use an ordinary CPE radio (client mode) at the neighbors,
    they will only be able to connect ONE computah to the CPE and have it
    work. If you want more computers at the neighbors, they will need
    either an additional router runing NAT or setup a wireless transparent
    bridge system, which involved additional radios. Another way is use
    WDS (wireless distribution system) which requires compatible (and
    specific) wireless boxes.
    Yes, with a very minor catch. If you're using wireless in your house
    from perhaps a laptop, the thruput between the laptop and the
    neighbors will be horrible. That's because the access point has to
    switch between the side and outside antennas. Some, not all, wireless
    routers and access point take excessive time to do this.
    Diversity receiption. 802.11a/b/g is suceptible to errors caused by
    reflections and multipath. The path from a client radio and the
    access point may not be perfect in all locations. However, by adding
    a 2nd antenna, and switching (scanning) between the two antennas, at
    least one antenna has a better chance of hearing the client. It's a
    cheap improvment in reliability.
     
    Jeff Liebermann, Apr 22, 2005
    #3
  4. Nick

    Nick Guest

    Thank you for your thoughts Bert.

    I guess I'll just try it. I have a fallback which is to connect a WAP
    via a crossover cable into the back of the router and use that for the
    external link.

    Regards,

    Nick
     
    Nick, Apr 23, 2005
    #4
  5. Nick

    Nick Guest

    I had more or less come to that conclusion. Most of these ADSL modem
    wireless routers only have one antenna. The Belkin seems to attract the
    least number of negative comments on the shopping sites and it has two
    antennas, so I guess it is try it and see. The fallback would be to
    connect a WAP to the router via a crossover cable and use that to feed
    the external antenna. I imagine this would overcome the "throughput
    problem" you mention below.
    Well the yagis are already fitted both ends from a soon-to-be-defunct
    community wireless project. The LOS is only about 30m. Elsewhere on
    the network we have used small patch antennas to good effect. They seem
    to work better than the 7-ele yagis, particularly in a multipath
    environment. I don't really understand this because I would have
    thought a narrower beamwidth would work better in the latter case and
    the patch antennas are much smaller than the yagis by volume!
    There is already an additional router in the neighbouring property
    connected back to back with the AP client box via a CAT5 crossover cable.
    That could be a benefit from a security POV. There is no requirement
    for client machines in the two properties to exchange data. I didn't
    realise that the WAP actually switches between the two antennas. I
    would have thought they would just be connectected to the electronics
    via a hybrid splitter.
    They must scan between the antennas synchronously with the tx/rx packet
    stream. Clever stuff.

    BTW I just looked at the pictures on your web site. It bears an uncanny
    resemblance to mine i.e. radio and computer junk everywhere. Guess you
    must be a ham too?

    Regards,

    Nick
     
    Nick, Apr 23, 2005
    #5
  6. It was not obvious that you had not purchased anything yet. On the
    assumption that you might have some existing equipment, I asked for
    the manufactory and model numbers. (Hint: Tell us what you are
    trying to accomplish and what you have to work with).
    If you look carefully at the various boards, most of them come with a
    diversity switch and two antenna connectors. That's because the
    MiniPCI boards found in many wireless routers are also use in laptops
    and client radios, where the packaging determines whether one or two
    antennas are used. I'm looking at some wireless PCI cards and note
    that the 2nd antenna is terminated with a 50 ohm chip resistor.
    There's nothing in the design of an "ADSL router" that requires either
    one or two antennas.

    Also, permit me to disuade you from purchasing an "all in one" unit
    with ADSL modem, router, and wireless in one box (even if they are
    somewhat cheaper). If you move, and your new location has a cable
    modem instead of DSL, you get to throw the whole box out. The
    wireless part tends to add a few acronyms every few months making
    obsolescence a real problem. However, the worst issue is over the
    location of the box. The modem and router parts want to live near
    where all the wires come together. AC power, phone, and CAT5 LAN
    cables all want to live near the floor, under a desk, in a closet, or
    in the basement. However, the wireless part wants to live in the
    open, up high, with a maximum view of the coverage area. These are
    incompatible. You can place the "all in one" unit on the floor and
    use a coax cable to relocate the antenna(s), but the coax losses are
    horrendous. Anyway, I suggest you purchase seperate boxes for the DSL
    modem, the router, and the wireless access point.
    I now have 3ea older unreliable Belkin 802.11b access points. I don't
    have the model number handy. The problem was that Belkin does not
    seem to be interested in updating the firmware in older products. If
    you look at the previous generation of Linksys, Netgear, and DLink
    products, you will see continuous product development well after the
    product is considered to be obsolete. Belkin and some others do not
    bother to do this. The reliability issues I had with Belkin are
    apparently known, but because development stops immediately after the
    product release, the issue becomes permanent. I suggest you keep
    looking.
    Methinks it would be nice if you describe exactly what you are trying
    to accomplish and what you have to work with. There are far too many
    topology options in wireless to provide universal solutions.
    At 30 meters, almost any external antenna will work. You should have
    a good strong signal with your unspecified gain yagi antennas.
    However, there's always a way to screw things up. One of my
    installations involved a pair of 19dBi dish antennas at about 100
    meters. Signal was very strong and I was getting about 25Mbits/sec
    thruput. No problems until someone installed an access point that was
    directly in line with one yagi, but somewhat furthur behind the yagi's
    intended target. Because it was in line, the gain of the antenna made
    the interference problem much worse. Oops. They refused to move to
    another channel (because the others were polluted) so I reduced the
    gain of one of the yagis and repositioned it somewhat until the
    interference was somewhat less of a problem. Final thruput was about
    15Mbits/sec with no interference and about 10Mbits/sec when the
    interfering access point is live. Watch where you point high gain
    antennas.
    Working better measured how? More signal? Better S/N ratio? Better
    thruput? What are you measuring?

    I much prefer patch (panel) antennas over yagis. It's not just the
    cost per dBi of gain issue, but also bandwidth, sidelobes, and
    sensitivity to environmental issues, that make the patch superior.
    The yagi's rotten sidelobes and f/b ratio will cause reflections and
    interference pickup. Also, the yagi is MUCH more sensitive to
    mounting issues than a panel antenna (with a solid ground back
    surface). Yagi's also have a practical limit on gain. For every 3dB
    of gain on a yagi, the antenna becomes twice as long. 15dBi is about
    the practical limit for yagi's, while panels will go to 19dBi and
    dishes to 24dBi. (Yes, I know there are higher gain dishes, panels,
    and yagi's, but if you read the fine print, they will be very narrow
    bandwidth and not cover the entire 2.4GHz band).

    It's been a while since I've played with the numbers so let me do some
    digging (mostly on the fab-corp.com web pile)

    Type specified -3dB bw Cost Cost per
    gain dBi degrees $US dB gain

    Patch
    Rootena RT24LP14 14dBi 35 $42 $3.00
    Maxrad Wisp 13dBi 35 $39 $3.00

    Dish
    PacWireless 15dBi 19 $35 $2.33
    PacWireless 19dBi 17 $41 $2.16

    Yagi
    Maxrad 15 15dBi 30 $59 $3.93
    Antennex 14.6dBi 30 $65 $4.45

    Yep, the yagi is still the most expensive (per dBi gain).
    Splitters and hybrid combiners have loss and loss is a bad thing. If
    the path length to the two antennas is different, there's a real
    chance that the signal phase at the two antennas will be different.
    If 180 degrees, they will cancel. The result will be a rather nasty
    series of nulls. There are systems that do use combiners and
    splitters, but only with isolated antennas, where there's no chance
    that a given signal will be heard by both antennas. On the other
    foot, the switch does not have any (major) interaction between
    antennas and can be run without worrying about creating nulls (i.e.
    dead spots).

    I'm still trying to decode what you are trying to accomplish. If you
    are worried about security between wireless clients, then please
    consider the WRT54G router. It has a feature misnamed "AP protection"
    which is really "client isolation". No wireless to wireless bridging
    is allowed.
    No, it's much simpler than that. The access point stores which
    antenna heard a specific MAC address successfully last. If it fails
    to hear another packet within a specified time (varies from 100msec to
    several seconds), then it switches to the other antenna and waits for
    the resends. Note that it does NOT work like a radio scanner and does
    NOT make any decision as to which antenna has the best signal strength
    of S/N ratio. There also a mess of other algorithms and variations.
    Some AP's scan between antennas when there's no traffic. Also,
    versions of 802.11n (and Pre-N) are completely different and have
    seperate receivers on each antenna to try to reconstruct the data from
    multiple antennas.
    See the call sign in the signature. Yeah, I have the ham radio
    disease and suffer accordingly. Most of those I know that are
    involved in wireless design and development have ham licenses. Most
    of the sane ones are inactive. The big difference between ham radio
    and product design is that the typical ham can make *ONE* of anything
    work. I have to make my stuff produceable. (Now back to doing battle
    programming a Motorola GTX-900...).
     
    Jeff Liebermann, Apr 23, 2005
    #6
  7. (blah-blah-blah...)

    I forgot to mention some detail on splitters. In receive, the
    splitter with take signal from either antenna and deliver it to the
    receiver with about 1dB of loss. That's not too bad. However, in
    transmit, the splitter splits the signal equally between the two
    antennas for a -3dB loss per antenna plus a -1dB loss per port.
    Therefore, if you have two independently located antennas (outside
    yagi and inside omni), you will have a -4dB tx power loss at each
    antenna. Yech.
     
    Jeff Liebermann, Apr 23, 2005
    #7
  8. Jeff was being sarcastic. His point was that its impossible to say
    how a router will behave if you don't say /which one/ it is.
     
    Mark McIntyre, Apr 24, 2005
    #8
  9. Nick

    Jim Beam Guest

    Woops - Once in a million posts even the group expert misstates after a
    too long of a day. You are still world class though since we allow you
    a few every week!
    Above splitter loss is identical in both directions (about 4db).
    In the receive path 1/2 (-3db) of the signal is dissipated in the
    internal matching termination.
     
    Jim Beam, Apr 24, 2005
    #9
  10. Hee hee, I wasn't going to pick on him...

    All of the above is why space diversity only works well with a
    baseband combiner. That requires two complete receivers, not
    just a pair of antennas.

    If the demodulated signals from each of two receivers sent to an
    analog combiner equal signals will contribute equally to the
    output, rather than one better signal being switched in, and
    there is effectively a 3 dB *increase* in signal to noise ratio
    (the signals are added but the noise isn't). The control
    voltage would normally be an out of band noise slot, which will
    have less noise as the input signal increases.

    Such an arrangement is required when there is no error
    correction or flow control of digital data (for example, if the
    data is analog!), but is also significantly more expensive to
    implement than the hot switched antenna algorithm being used by
    802.11 radios.
     
    Floyd L. Davidson, Apr 24, 2005
    #10
  11. Argh. Y'er right. I screwed up. It wasn't a long day. It was a
    long night doing battle with one of my pet projects, preceded by me
    destroying my diesel engine last weekend, and buying a new (used) car.
    The loss in the splitter is the same in both directions and not
    different as I proclaimed. What's scary for me is that I made the
    same exact mistake in the NEC-LIST mailing list back in 7/2001. I
    just found the posting. Maybe vacation...
    Per week? I make that many mistakes? I thought I was (almost)
    perfect.
    Yep. I agree. The jury will kindly disregard my stupidity.
     
    Jeff Liebermann, Apr 24, 2005
    #11
  12. Nick

    Jim Beam Guest

    Sorry but I completely missed your point. I assume *space diversity*
    means two isolated (or at least different amplitude/phase) antenna
    patterns relative to the linked client/clients.

    I can not picture the configuration of your system with dual receivers
    and an analog baseband combiner. Of course I don't know exactly what
    you mean by "analog" or "baseband" in a system like this.

    How are the two receivers and single transmitter configured relative to
    the two diversity antennas? Certainly not with a power
    splitter/combiner.

    Your explanation seems a little over simplified or "twisted" because the
    two *diversity* analog demodulated signals from each of two receivers
    probably will not be equal amplitude or in phase so there probably will
    be no 3db S/N improvement when combined.

    I realize of course that the outputs of two receiver channels with
    identical antenna patterns will combine with a 3db S/N gain. And I
    accept that two outputs of two complete (receiver, transmitter, antenna)
    systems can be combined digitally using flow control but that doubles
    the expense. That is not the system being discussed though.

    BTW, I realize that you are also a "group expert" but Jeff has seniority
    and a higher post count so hence the title ;-))
    Both of you guys really impress an old senile retired RF design
    engineer!
     
    Jim Beam, Apr 24, 2005
    #12
  13. Two separate antennas. Amplitude and phase are not significant, and
    will vary at different times on the two signals.
    Baseband is whatever signal is recovered from the initial
    demodulation of the RF carrier. Typically that is a signal
    which multiplexes several other separate signals together (and
    might be an AM or FM or whatever carrier itself).

    An "analog" device is continuously variable, while a "digital"
    device has a discrete set of values. Hence a "hot switch"
    system such as that being used to switch a single receiver
    between either of two antennas might be called a "digital
    combiner" (a binary, or two level switch).

    But something which continuously adjusts the output to different
    proportions of signal from either of two devices would be an
    "analog combiner".

    A hybrid that connects two antennas to one receiver would also
    be an analog combiner, but since the phase differences would
    result in unwanted signal variations, it doesn't work well (as
    Jeff correctly described) and merely amounts to guaranteed
    multipath interference.
    (Note that this can also be done with spectrum diversity, using
    two transmit signals and just one antenna.)

    One transmit signal and two paths to two different antennas
    separated by physical space, is called "space diversity". The
    two separate signals will each suffer from various degradations,
    but almost always at different times and/or to different
    degrees. A multipath fade on one antenna will not likely happen
    at exactly the same time as a multipath fade on a second antenna
    located a significant distance away (significant in terms of
    wavelength, so 4 inches is significant for 2.4GHz equipment).

    The problem with using a single receiver with two antennas is
    that those multipath fades are associated with a phase shift,
    and if the two signals from the two antennas are combined at RF
    (before demodulation), they will varyingly add or cancel, and
    thus defeat the entire purpose of using diversity.
    Well, yes it is certainly simplified! I suppose this topic
    could fill a chapter, at least, in a fairly difficult
    text... :)
    They *should* have a relatively constant amplitude and phase
    difference. The phase shifting of the RF signal would not cause
    a linear change to the phase and amplitude of the demodulated
    signal. The better the signal, the less variation... and whatever
    there is is *noise*, not signal. I.e., with a good signal the
    baseband output will have a high signal to noise ratio.

    The combiner takes advantage of that by combining two signals
    that are *coherent* (or, they were to start with, when they came
    from one transmitter), and thus the signals will add. The noise
    is not coherent (the noise on each signal is from a different
    source) and thus does not add. That results in an effective
    increase in the signal to noise ratio.
    Only at an instantaneous moment when multipath signals are
    minimal. Which is to say, if the antennas are close enough
    together, yes. But of course then any instantaneous fade that
    affects one will also affect the other, and there is no
    diversity because it is effectively just one antenna.
    That is one possible implementation of what is being discussed,
    and would be an extension of the existing system used by most
    802.11 radios today.
    Jeff has worked for years directly with "wireless" systems. He
    is far and away the most expert person posting in this
    newsgroup.
    You are probably senior to both of us! You will no doubt get a
    good chuckle from what I am basing the above discussion of space
    diversity on! I came to Alaska in the 1960's to work on the
    White Alice Communications System, which used L-Band
    (700-1200Mhz) forward troposcatter radio links that implemented
    space diversity exactly as described. A typical link used a
    single transmitter with an output of perhaps 3Kw and the receive
    end employed two antennas and two receivers with exactly the
    type of combiner described. The antennas of course were 75 foot
    high billboards with 60 foot parabolic reflectors (something
    like 45 dBi gain if I remember right).

    I've been twisting knobs on microwave systems since the early
    60's and those troposcatter systems were the only combiners like
    that (and the most fun radio systems too) that I've ever worked
    on. The manuals had a set of routine maintenance tests that one
    could do, which might take a couple days to accomplish on each
    set of receivers, that would indicate if the receivers were
    operating correctly or not... But what we did was hang a VTVM
    on the output of the combiner control voltage amplifier, and
    terminate the receiver antenna input. If the resulting maximum
    possible combiner control voltage was up to snuff, that *had* to
    be a good receiver, and we could sign off the routine and go
    play pool, catch fish, or chase girls, for another day or
    two... ;-)
     
    Floyd L. Davidson, Apr 24, 2005
    #13
  14. Now just wait a minute... We've got a rifle team showing up at
    the wall for a bit of shooting practice at sunrise.

    If we let you cop a plea, who they gonna execute?
     
    Floyd L. Davidson, Apr 24, 2005
    #14
  15. Sorta. The two receivers might have a common local oscillator if it's
    a heterodyne system, so the demodulated data will fairly close to in
    phase. However, most modernish 2.4Ghz radios are direct conversion,
    so that's not an issue. Let's play with the numbers.

    If we measure the phase difference or time difference of arrival of a
    signal between two antennas, the difference could easily be a full
    wavelength or more. If there is a reflection involved, or the path to
    each antenna is different (one direct, the other reflected), then it
    can be even more. If I separated two antennas by exactly 1 wavelength
    (12.5cm), and combined the signals with a resistive combiner or 0
    degree power divider, then as I move my transmitter around the
    antennas in a circle, there will be 2 very deep RF nulls when the
    transmitter is inline with the antennas. Obviously, this is not the
    way to improve reliability.

    With 802.11b/g, the base phase data rate is about 2Mbits/sec or 500
    nsec per bit. The rest of the speed comes from moving the phase
    around and amplitude modulation to form 8 or more points on the
    constellation diagram, plus multiple data carriers in the case of
    OFDM. If I space the antennas again exactly 1 wavelength, but this
    time use two receivers to demodulate the data, the antenna spacing
    necessary to created a null at baseband when the transmitter is
    in-line with the antennas is:
    1/2 * 500 nsec * 3*10^10cm/sec = 75 cm
    That's now 6 wavelengths instead of 1 wavelength. A data de-skewing
    register will resynchronize the decoded data and prevent a data null.
    This is very very very roughly what a "RAKE receiver" does when
    dealing with multipath. I vaguely recall a nifty Intersil application
    note that described exactly how it worked but I can't find it.

    If we demodulate the data to a lower frequency (as above), the delays
    required to produce nulls are much larger. Given the fairly narrow
    antenna separation on the common access point, the separation at the
    modulated data rate is a sufficiently small percentage of data bit
    time, that simple summing of the demodulated data would yield an
    improvement in reliability. With de-skewing (bit alignment) the phase
    error can be mostly eliminated.

    Note that I haven't mentioned anything about inter-symbol interference
    caused by excessive delays and reflections. There's a limit to how
    long a delay can be accomidated at baseband.

    There's no 3dB improvement in signal to noise ratio with two
    receivers. That's because each antenna picks up the signal at a fixed
    S/N ratio. Combining these two signals in any manner will result in
    the noise increasing along with the signal. What it can do is select
    the antenna with the *BEST* S/N ratio and thus prevent dropouts,
    nulls, fades, and multipath effects.

    A simple analogy would be if I placed 100 identical antennas and data
    receivers on my kitchen table. As I move the transmitter around, the
    BER (bit error rate) from each receiver will vary from whatever works
    out as a maximum at the distance and connection rate, down to totally
    gone in the case of reflective cancellation. Each of the 100
    receivers will vary exactly the same way, but not exactly at the same
    time or when I am at the same exact location. If I had a way to
    combine the data output from each of the 100 receivers (and discard
    corrupted packets), the BER would whatever is coming from the best
    receiver at that instant. There will always be at least one of the
    100 receivers with a good signal. However, the BER will never be
    better than what is offered by the worst case path loss at the
    connected data rate. If my transmitter is very low power or the
    distance is large, then all the 100 receivers will hear a weak and
    rotten signal and respond accordingly.

    Incidentally, OFDM is frequency diversity. With multiple carriers
    (802.11a/g are 48 carriers), on slightly different frequencies, the
    chances of at least one of the carriers not ending up in a situation
    that creates a null is dramatically reduced. (Null's are very RF
    frequency dependent). Add two antennas and a diversity switch, and
    the typical 802.11g access point has both spacial and frequency
    diversity features.
    I beg to differ. Because noise is not coherent, it cannot be removed
    or cancelled by summing. Noise plus noise equals twice the noise no
    matter what form of combination or summation is used. If I take two
    identical signals at any frequency with perhaps a 10dB S/N ratio and
    simply combine them with a resistive combiner, I'll end up with a 10dB
    S/N ratio output. If what you say were true (noise does not add),
    then it should be possible to take one noisy signal, split it between
    two ports, cutting both the signal and the noise in half, and then
    combine it back to yield twice the signal and half the noise. This is
    obviously impossible.

    Bah. Quantity is not a very good replacement for quality. I make all
    too many mistakes but figure that if corrected quickly enough, that's
    acceptable. The problem is that I haven't done full time RF design
    for about 20 years. I'm catching up quickly, but discovering that I
    have some holes and misconceptions to deal with along with the usual
    age related memory problems. I'm also weak in the all important
    protocols area, probably because I hate reading RFC's and IEEE specs.

    Quantity test for alt.internet.wireless:
    | http://groups-beta.google.com/groups?&as_ugroup=alt.internet.wireless&as_uauthors=Jeff+Liebermann
    1400 postings. No wonder I need a vacation.
    Thanks. However, you've caught me at two giant goofs and will
    probably find more if you dig deeper. Experts should be infallible
    and I'm certainly not. Perhaps a lesser title might be more
    appropriate.
    I'm 57 and am nowhere near retired. The family tradition is to drop
    dead at work, and I'm doing my best.
    Without the antenna, you're just measuring the front end noise level.
    For a system test, that's a good go/no-go test, but can be easily
    screwed up by some common failure modes. For example, an oscillating
    RF stage will produce more noise, instead of less noise, and end up
    with a deaf receiver. Same with the then common carbon comp resistors
    becoming a noise source. It also won't test the antenna since it's
    disconnected. The good news is that a failure anywhere else, except
    the front end, will result in a dramatic drop in detected noise, so
    this is probably an acceptable test.

    VTVM is a "Vacuum Tube Volt Meter". I just noticed I still have an
    Eico VTVM in the pile. Why, I have it, I dunno.
     
    Jeff Liebermann, Apr 24, 2005
    #15
  16. I assure you that is not true. The received signals in the two
    channels are coherent, but the noise is not. There will in fact
    be an improvement in SNR *of the baseband*. That is a *very*
    *different* thing than the received SNR for RF signals at the
    receiver inputs. The two are related only to the degree that a
    better signal (both in terms of absolute level or in terms of
    SNR) at the antenna will (within the dynamic range of the
    receiver) result in a higher SNR at the baseband output.

    Not that phase distortion of the RF signal won't cause
    inter-symbol interference and reduce that SNR. But it will
    always be *at least* as good as the best signal and when both
    signals are equal the SNR will be 3 dB better than it would be
    with only one signal.
    That's a good description. However, that is a digital
    combiner, not an analog combiner. What you gain there is
    entropy, not SNR. Which is to say that you could have 50 of the
    100 receivers out of service and the BER of your data would not
    change! All 50 of them are just passing information that is
    redundant with the other 50.
    That's precisely true. I'm sure you've seen an HF RTTY signal
    using 2 tone Frequency Shift Keying, where the two frequencies
    less than 200 Hz apart fade in and out totally independent of
    each other. Watching RTTY on a scope is about the best
    demonstration of frequency diversity I've ever seen.
    Exactly. Summing the incoherent noise will *not* produce the
    same result as summing the two received signals, which are
    coherent.
    If the two noise sources are coherent, that will be true... and
    won't be if they are not.
    And that is obviously true because those two noise signals *are*
    coherent!
    I agree with you about "post count", but even so... you've
    still got at least twice as much hands on experience with
    wireless as probably any other 3 or 4 of us put together here.
    You ain't doing bad. Google says 2710 posts total, to all
    newsgroups, over the past 5 years.

    Looks like a simple case of Usenet addiction to me.
    Maybe you can get an appointment with my shrink...
    Nah, experts make goofs. The only reason your "two giant goofs"
    look giant is because you made them. But, lets face it... it
    literally takes someone who has been doing this stuff as long as
    you to see half the nuances of what you are saying most of the
    time, never mind to correct some part of it.
    Well, Jim is almost certainly older than both of us then. I'm a
    couple years older than you. But I bet I'm a *lot more retired*
    than even Jim is! My most stressful decisions these days
    involve which cat to pet first... or maybe what to eat for
    dinner. Or whether I want to take a nap now, or later, or both!
    It's an FM radio. The control voltage is developed from an "out
    of band" slot above the normal baseband bandpass. The noise in
    that frequency slot goes up as the amount of quieting decreases,
    and inversely proportional to the signal input. No
    signal... lots of noise. Just how much is a very good indicator
    of how functional the entire radio is.

    Because the noise slot is at a higher frequency than the normal
    range of the baseband the noise will be reduced if the bandwidth
    of the receiver is too narrow, and will be reduced if the gain
    of any of the receiver's RF, IF, or noise channel is too low.

    In this case the control voltage was supposed to be adjusted to
    something like -42 volts when there was no signal input to the
    receiver front end. With a receiver that had just been totally
    overhauled it could be adjusted as high as maybe -50 volts. So
    we liked to see anything from about -46 to -50 as the max
    possible. If it was less than that, it was time to work for a
    living.

    (These radios were made to Western Electric specifications by
    REL, Inc., which if I remember right stood for Radio Engineering
    Laboratories. I have no idea who actually designed them or who
    was behind REL. It was all early 1950's technology. Loads of
    fun...)
    The whole system has to be optimum to get that much noise in the
    out of band slot. Tune the IF too narrow... no noise. Cut out
    the RF or 1st mixer... no noise. Basically when the antenna is
    removed it causes every AGC loop to operate at full gain with
    the single exception of the combiner junction point, which will
    be at cutoff or squelched.
    Resistors are in fact a noise source... but far less than
    anything these receivers were going to sense! We're talking -95
    to -98 dBm input for 20 dB of quieting. The front end was a
    416B triode in grounded grid (actually, a specially selected
    416B called a 6280WA which back then cost about $150 each).
    Testing 60 foot parabola's isn't practical.

    But in fact at one point we *did* have a special crew go around
    and test feedlines and antennas, because we had one that had
    something like 20 dB extra loss on the transmit side. Which
    near as we could tell, once we had equipment to tell us which
    part to dismantle, was caused by improper installation and had
    resulted in some burning in a 4" rigid coax section that
    probably took place the first time they applied RF power to it.

    The problem was that we could only figure out where about 10 dB
    of that loss was, and until the day it was turned down that one
    antenna had significantly less gain than the others. (We used it
    for the backup transmit, which was a 10 Kw klystron running at
    only 500 watts output. Low enough not to interfere too much
    with the 3000 watts on the other antenna, but enough that if the
    main side failed completely it would still hold the path. On
    the other antenna only 5 watts would have been enough.)
    Exactly. And in fact it would catch a flat front end too.
    Instead of -47 volts, the max output would be maybe -43. Not
    much, but it definitely gave an indication.
     
    Floyd L. Davidson, Apr 24, 2005
    #16
  17. Nick

    Jim Beam Guest

    - snip - Lots of good stuff but the discussions are not on exactly on
    the same wavelength nor discussing the same topology, conditions,
    assumptions, etc.
    Jeff -

    I think Floyd was referring to the difference between combining/summing
    non correlated noise *vs* correlated signals.
    The absolute gain/loss reference got lost in the discussion!
    Keep in mind that these combiners combine/sum instantaneous *voltages*
    rather than power.

    Re-visit the "splitter" used as a receive combiner in your post that
    started this discussion. (For discussion assume an in phase combiner
    with 0 dissipative losses) Then:

    Any single signal or two non correlated signals will combine/sum with 3
    dB individual signal losses. Thus total output power will be 3 dB less
    than total input power.
    For example two 0 dBm input noise signals will result in a 0 dBm output
    signal.

    BUT two *correlated* input signals' will sum depending upon voltage
    phase. Total output power will equal total input power for "in phase"
    correlated signals (voltages will add). For "out of phase" correlated
    signals the voltages will subtract and output will be zero (the null you
    referred to) for equal level input signals.
    For example two 0 dBm input "in phase" correlated signals will result on
    a +3 dBm output signal.

    Thus the 3 dB S/N improvement he referred to.

    This discussion is not completely "apples to apples" so it results in
    lots of different answers.

    Thanks to both You and Floyd for all the help you provide here. I have
    followed this group since it was started and the present level of
    technical expertise by far the best ever.
    And the ass hole flame level is the lowest too!

    Back to lurking!
     
    Jim Beam, Apr 24, 2005
    #17
  18. Nick

    Nick Guest

    The objective is to share an ADSL connection between two properties to
    replace a community wireless solution for internet access. Both
    properties have external yagis connected to WAPs configured in AP client
    mode and then via a crossover cable to wireless routers. The yagis
    currently point to a WAP about 50m away which is connected to a
    satellite internet feed via a firewall. Internally each property has
    its own wireless network with Channel, ESSID and WEP key. Wireless
    laptops are used in each property. The existing wireless kit uses
    Pheenet APs and Linksys 802.11b wireless routers.
    I never took the lid off a wireless router and have never noticed that
    on PCI cards.
    Agreed. But I imagine that two antennas might give an 'average' 3dB
    improvement in SNR instead of heating a chip resistor.
    Point taken. But in this case the PSTN line is in the loft and so are
    all the power and network connections. The price of an 'all-in-one'
    unit is almost the same as a separate ADSL/ethernet modem here in the UK.
    The Belkin seemed to attract the least negative comment from other
    purchasers on the etailer's site I looked at. I'm also looking at 3Com,
    Dlink, Netgear and Linksys.
    Thanks. I will look at the WRT54G.
    That sounds like a bridge to me.

    I have pretty much decided to connect the new router to a separate AP on
    a different channel i.e. a separate wireless network to link the two
    properties. That way any nasty interference effects between the
    external yagi (with its sidelobes) and the router antenna(s) in the loft
    would be avoided.

    Regards,

    Nick
     
    Nick, Apr 26, 2005
    #18
  19. Satellite is usually quite slow and has sky high latency. ADSL should
    be a substantial improvement. However, since there is apparently an
    existing wireless system nearby, take care to select your channels to
    avoid mutual interference.

    I don't really need these to answer the antenna questions but it would
    be nice to know:
    How far away are the properties?
    Any line of sight problems or obstructions?
    Since you have existing equipment, what are the model numbers?
    How much antenna gain on the yagis?

    802.11b might be a problem if you purchase an ADSL line of 3Mbits/sec
    or faster download speed. We have 6Mbit/sec service available
    locally. You'll probably need 802.11g to take advantage of the full
    speed.
    The load resistor rarely gets tx power. The way the diversity switch
    works is that the radio selects the antenna that received the last
    successful packet. Both the transmitter and receiver are switched.
    Since the real antenna is getting all the successful packets, the
    radio spends all its time connected to the real antenna. It may
    occasionally scan the terminated antenna port, and perhaps transmit a
    beacon or two into it, but the greatest majority of transmission go to
    the real antenna. Not a problem.
    I don't see the problem. It's much easier to run one pair of
    telephone wire from the loft to a tolerable radio location, than it is
    to run coax cable for an antenna, or CAT5 for ethernet. Just locate
    the ADSL modem in a convenient location but concentrate on putting the
    radio part (CPE) in an RF ideal location. That's much easier in a
    separate box. It's easy enough to enclose a wireless access point,
    powered with PoE, in an outdoor box, on a pole, and next to the
    existing yagi. As for price, it's usually more expensive to initially
    purchase separate boxes, but far less expensive when only one part of
    the puzzle needs to be replaced.
    It's impossible to generalize solely on the basis of manufacturer.
    Almost everything comes from assorted manufactories in China or
    Taiwan. Dlink, Linksys, Netgear, etc all just put their names on some
    obscure factories products. Many are almost identical. For example,
    look at how many are similar to the DWL-900AP+ wireless bridge:
    http://www.vallstedt-networks.de/?vendors/GST_Vendors
    which are basically a list of TI ACX-100 boards. Each manufacturer
    also has winners and losers all mixed together. What differentiates
    the vendors is not the quality of the hardware, packaging, or
    construction. It's support and firmware. The apparent inability of
    Belkin to support its *EXISTING* customers with continuous firmware
    updates is why I think don't like Belkin. The same products, with the
    same chipset, possibly from the same factory in China, but from a
    different vendor, does get updates, fixes, and support. Before buying
    something, I suggest you check into the availability of firmware or
    driver updates or the date of the last update. If there's only one
    update, find something else.
    Also look into alternative firmware for the WRT54G. Sveasoft,
    HyperWRT, etc.
    Close, but not exactly. The radio decides which antenna works best
    with specific client radio. Each radio is identified by its BSSID
    also known as its MAC address. In this case, the MAC address is
    simply a radio identifier and has nothing to do with traffic handling.
    Incidentally, *ALL* wireless is based on bridging.

    In a bridge, the system sniffs the traffic to extract a table of MAC
    addresses heard on each port. The bridge then decides if a packet
    needs to cross the bridge based upon the destination MAC address. One
    distinction is to select antennas, there does not need to be any
    traffic or sniffing. Both the antenna and MAC address table expire,
    but the antenna selection has a much shorter expiration time because
    the path might change much faster than perhaps a client radio roams to
    a different access point.
    Good plan and good luck. Do a "site survey" to make sure there aren't
    any other users around.
     
    Jeff Liebermann, Apr 26, 2005
    #19
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.