In article , Filthy Pierre wrote: :I'm in the process of making some indoor antennas for my home network. I :have a few different designs etc, but have so far been unable to locate :an impedance for the cable I should use...
:is it just 75ohm co-ax like for TV? or something different?
No, don't use that co-ax, it won't work!
A google search on wifi antenna impedance shows that most models are 50 ohm.
Well, here I get to challenge the traditional orthodoxy. Any book on system design will declare that it's a really good idea to match all the impedances. If the antenna is 50 ohms, and the access point radio is 50 ohms, it would seem to be proper to use 50 ohm coaxial cable.
However, if one uses 75 ohm cable, it's not as detrimental as it would seem at first glance. The 50 to 75 ohm mismatch is a VSWR of 1.5:1 which generates a loss of 4% of the power or 0.18dB.
that's tolerable, then 75 ohm coax should work. There will also be some detrimental effects to the the antenna pattern and frequency response if the antenna is fairly high Q (narrow band). However, this is normally not a problem with most (not all) 2.4GHz antennas. Similarly, the ceramic input filters in the access point might not like the mismatch. I haven't seen any problems, but I also haven't done any proper testing.
Incidentally, most 2.4GHz antennas are nowhere near 50 ohms across the entire 83.5Mhz band. For example, see:
note that the VSWR of the common coffee can antenna is 2:1 (either
25 or 100 ohms) near the band edges (2400/2450). Not exactly a perfect match but good enough as a 2:1 mismatch is "only" a 0.5dB loss. My guess(tm) is that either 50 or 75 ohm coax would work equally well with such an antenna.
I have a fairly good supply of 75 ohm surplus CATV coax and have used it in a few installations where price was more important than performance. I've also designed and tested a few 75 ohm antennas. My
50 ohm test equipment has input/output pads to deal with 75 ohm systems.
A big problem is proper connectors. The common N, BNC, TNC, SMA connectors are available in both 50 and 75 ohm diameters. However, the more exotic R-TNC and R-SMA found on access points are not available. I have to use coax adapters (or butchery) at the access point and switch to standard connectors.
As for the selection of coax cables, don't just grab a chunk of RG-6/u or RG-11/u and start using it at 2.4GHz. These have turned into a family of cables with radically different characteristics. Even the "satellite grade" of TV coax is only rated to 2GHz and is horribly lossy at 2.4GHz. For comparison, LMR-400 is 6.7dB/100ft, while the best RG-6/u is approx 11dB/100ft. For short runs, such loss is tolerable, but don't try it for long runs.
Anyway, enough heresy. Use 50 ohm coax if possible, and only use 75 ohms if you have a good reason to do so.
The previous post is true for a 75 ohm source feeding into a 50 ohm load, but the situation is a little more complicated when a 50 ohm source feeds a
75 ohm transmission line which terminates with a 50 ohm load.
In this situation the apparent impedance that the source sees is the load impedance phase shifted by 2 x pi x coax_length / wavelength_in_coax
The wavelength in coax will be roughly 2/3 of the wavelength in vacuum, (ie
8cm and 12.5cm respectively).
If the coax happens to be an integer number of half wavelengths long then the impedance as seen by the source will be 50 ohms. This will give a VSWR of 1:1.
If the coax happens to be an odd number of quarter wavelengths long then the impedance as seen by the source will be 75*75/50 = 112.5 ohms. This will give a reflection coefficient of 0.384, and a VSWR of 2.25. 14.8% of the power will be reflected, ie a transmission loss of 0.7dB. This ignores any losses due to the coax cable itself which will rise linearly with cable length.
Other coax lengths (ie not an integer number of quarter wavelengths) will make the load appear reactive to some degree, with a transmission loss of between 0dB and 0.7dB.
The reasons for using the right impedance cable are:-
There is a no transmission loss due to reflections of power back to the source at changes of impedance.
Reflections bouncing up and down the mismatched coax will give rise to a time delayed signal interfering with the desired signal. In the above scenario (75 ohm cable and 50 ohm load) there will be 14.8% x 14.8% = 2.2% of the signal present as time delayed interference, ie about 17dB down on the desired signal. This is probably low enough not to be a problem with digital signals.
The reflected signal will add to the peak voltage and peak current strain on the source. Hopefully the wireless link designers will have allowed for an awful mismatch in case of bad siting of the antenna, so this should not be a problem.
Agreed on all points. However, you're ignoring the effects of coax cable loss. Let's say we're playing with a 50ft chunk of CATV coax with a loss of about 3dB. That means that only half of the tx power makes it from the xmitter to the load, and only half of the reflected power makes it back from the load to the source. I just burned an hour trying to work out the exact VSWR at the source for a 3dB cable loss, and keep getting rediculous answers. I can pull it off a nomagram in the ARRL Handbook, but can't seem to get the numerical answer to agree. Maybe 8 hours of snooze and some fast reading will provide an answer tomorrow.
Yeah, but when the mismatch loss is 0.7dB maximum, which is about the same as contributed by a typical connector pair and a short pigtail, I wouldn't consider that a serious problem. Again, let me point out that the antenna VSWR is often 2:1 or worse at band edges, which is a far worse loss than the mismatch loss.
Of course, I could "build" a 61 ohm 1/4 wave coaxial matching section to deal with the impedance mismatch. It's broadband enough to work over the entire 83.5Mhz. However, if I include connectors, the loss of the matching section is about equal to the worst case mismatch loss, and is therefore a waste of effort.
Again, you neglected the cable loss. If we use my convenient 3dB coax cable loss figure, the reflected power will be 20dB down at the load and 23dB down back at the source (where it can do the most damage). It's actually a fairly large signal. Starting with +15dBm tx power,
-23dB yields a reflected signal of -8dBm which is strong enough to clobber any received signal if suffiently delayed. However, the reflected signal in my 50ft cable will be delayed approximately: 2 * 50ft / 1ft/nsec / 0.67 = 44 nanoseconds. The "2" is because the signal must go up the coax and back. The 0.67 is the coax cable velocity factor.
802.11g OFDM uses an 800 nsec guard interval to allow for reflections, bounces, and multipath delayed transmissions that cause intersymbol interference or delay spread. 44 nsec out of 800nsec is not a problem (at any amplitude).
Strain? At 50 milliwatts, there's no strain on the transmitter devices. 802.11b/g require that the transmitting device be linear. That give the power device considerable tolerance to reflected power. The devices are also made to survive transmitting into open circuits and shorts without VSWR senseing and protection. Actually, I think the original Prism 1 and 2 cards would detect VSWR:
However, I don't think they did anything with the numbers to protect the xmitter. Incidentally, I've used the above program to test and "sweep" antennas. You haven't seen VSWR until you've build a coffee can antenna.
Anyway, I don't recommend using 75ohm CATV coax if 50 ohm coax is available. However, it does work if needed.
However, since my convenient 3dB coax cable loss cuts the forward power in half, and the also the reflected power in half, we get: (reflected_power/forward_power) = 0.040 / 4 = 0.010 Working backwards, the reflection coefficient is: sqrt(0.010) = 0.100 which is an input VSWR of about 1.2:1
Using the same proceedure for the worst case VSWR (odd multiples of
1/4 wavelength electrical), where the reflection coeficient is 0.384, I get a lossless coax VSWR of 2.25:1. However, when I introduce the
3dB cable loss, the reflection coefficient becomes 0.192 or a VSWR of about 1.4:1.
More crudely, a high loss coax cable improves the input VSWR picture by attenuating any reflections. I also use the same idea for running
10base2 (cheapernet) coax cables using 75 ohm RG-6/u instead of the usual RG-58a/u. I works quite well if you only have stations at the ends of the cable run (no taps). I've done about 900ft runs without the slightest problem. Same idea. The high cable attenuation eats the reflections.