REAL WORLD GPS:
The people that sell GPS gear are in the same league as those who sell computer gear. They use all kinds of acronyms and make up standards to suit themselves when quoting the performance of their gear. The harsh realities are often very different. Some things to watch out for are:
CEP. This is a common way of quoting accuracy. It stands for Circle Error Probable. It is the radius of a circle in which 50% of measurements will fall. Note that this also means that 50% of all measurements will fall OUTSIDE this circle and it takes no account of how far out they will be. In the real world most people want to know what the MAXIMUM error will be. In my experience it will be at least 2x the CEP. Note also that CEP is a RADIUS so if you take a large number of measurements in one place, the scatter will form a patch with a DIAMETER about 4x the CEP. i.e. measurements on either side of the patch can be up to about 4CEP apart.
Note also that CEP is always quoted under ideal conditions (i.e. a perfectly flat field with no obstructions and a geometrically excellent array of satellites). Accuracies drop off progressively as you include obstructions like trees, steep ravines, buildings and poor satellite geometry. For the average case in North Queensland bush I have found that CEP should be multiplied by about 1.5. Another nasty trick is to quote CEP after a period of averaging (up to 1 hour) which can improve the accuracy by 2 or 3 x but just is not realistic for real world data collection.
SEP. This stands for Sphere Error Probable and is the radius of a sphere that will include 50% of all measurements. i.e. it is similar to CEP except that it takes vertical error into account. It is a fundamental rule of geometry that vertical errors in GPS work will always be about 2x horizontal errors. Consequently SEP is usually about 2x CEP (and for this reason is very rarely quoted by manufacturers).
There are all kinds of GPS modes (and acronyms to describe them) but the most common are:
CA CODE. (or just plain GPS) This stands for Coarse Acquisition code. It is used by all hand held consumer type GPS units. The satellites send pseudorandom code sequences and the receiver uses a kind of "slide rule" technique to match the sequences and calculate the distance to each
satellite. The position is then calculated in the receiver by triangulation. Typical handheld units quote CEP of around 5-10m and I have found that this is reasonably true to life (given the caveats for CEP mentioned above). I use a Garmin GPS76 (which uses CA Code) with a handheld data recorder for all of my regional mapping and I have found it is substantially more accurate than even the best air photos. I only use high precision gear for mine scale work. There is a picture of the gear I use (with an older model GPS) on my website at http://www.geomap.net.au You can buy receivers that use CA code from about $200 to $1000. The more expensive ones tend to be those with data ports to connect to computers or handheld recorders.
DGPS. This stands for Differential GPS and strictly refers to any system that uses a base station at a known point to send corrections to a roving receiver. However, it is usually used to describe systems that use differential corrections for CA code receivers. The system requires two receivers. One is at a known point (base station) and the other is a rover that you carry with you. The "base station" receiver calculates it's position and compares it to the co-ordinates for the known point. The error in the distance to each satellite is then transmitted via a radio modem to the rover. The rover then calculates it's position incorporating the error in distance to each satellite. Systems of this type are often quoted with CEP of 1m, but I have found that 2m is more realistic. You also suffer the hassle of needing to carry all the radio gear and since it is usually UHF type, you need pretty much line of sight to the receiver to get the correction signals. if you lose the radio link, you are back to CA code positioning. Systems of this type are usually in the $10,000 range and for the limited improvement in accuracy I think it is probably not worth the cost.
RTK. This stands for Real Time Kinematic and is generally used to describe systems that use carrier phase calculations rather than CA code. All systems of this type use a base station and a rover + radio modems to transmit the error corrections. These systems try to calculate the number of wavelengths between the rover and each satellite on the signal that carries the CA code. This is actually quite difficult because each wave looks pretty much like the next one. Consequently there are numerous possible solutions for a position calculation. The systems use a kind of iteration approach incorporating error corrections from the base station until the best solution is found. This is an interger or "floating point" solution. Once you have an integer solution, it is relatively easy to detect very small movements by the change in phase of the signals as you move. This is then added to the original fix to give your exact location. This is called a "fixed point" solution. You can get either single frequency (L1) or dual frequency (L2) RTK systems. The dual frequency systems compare a second frequency transmitted by the GPS satellites with the L1 frequency to calculate the ionospheric delay in the signal from each satellite. The single frequency systems range from about $15,000 to $30,000. They quote CEP values of about 20cm, but I have found 50cm is the real world (and of course double that for Vertical). The dual frequency systems range from about $30,000 to $50,000 and quote CEP values of about 2cm, but I have found that 10cm is real world. These systems are very sensitive to obstructions. Heavy tree cover can easily knock you back to a CEP of 1m or worse. Another disadvantage is that if you lose that integer solution (from satellite or base station obstructions), it can take 1-5minutes to get it back again.
WAAS. This stands for Wide Area Augmentation System. It uses a series of permanent public ground stations to calculate a network of corrections and then transmit those corrections via an extra satellite to any GPS receiver. Many of the current handheld GPS units are WAAS enabled. This system claims 1-2m CEP. Unfortunately, all the ground stations are in the US so even though it is possible to pick up the WAAS correction in Australia, it is effectively useless because we are so far outside the control network. The Australian government has been particularly slack in regard to setting up a network here.
EGNOS. This is the European equivalent of WAAS and is just as useless in Australia.
SBAS. (Satellite Based Augmentation System) or Omnistar or L Band. This is similar to WAAS except that the network is set up and maintained by a commercial outfit (Omnistar). And there are ground stations all over the world including Australia. There are two variants: VBS (Virtual Base Station. CEP quoted 1m) and HP (High precision. CEP quoted 10cm). I have tested a VBS system and found that 1m is probably fair CEP on a good day (although it may be offset a metre or two from the free GPS signal). I have not tested a HP system, but I gave it the swerve after discovering that it could take up to 40minutes to re-aquire a lock if you lose the signal from the correction satellite. The big advantage of these systems is that they are completely autonomous so you don't need a base station, line of sight or radios. The hardware for both systems is about $10,000. The snag is there is an annual fee for the correction signal. for VBS it is about $4000/yr. For HP it is about $8000/yr. These are pretty rough figures, but in the ball park.
POST PROCESSING. An alternative to differential systems which transmit error corrections directly is to record all the satellite data at the base station and the rover and then match them up in the computer at the end of the day. This has the advantage that you don't need radio transmitters and you don't need line of sight from the base station, but you do need an extra computer at the base station to record the data and you also need some rather expensive software to do the data matching ($5,000+). You also get some pretty odd looking maps during the day before the corrections are applied and some of the data my not be matchable because the base did not see enough of the same satellites so parts of your map may remain wobbly.