Jump to content

Multipath and the shortest path


TOPONA

Recommended Posts

Disclaimer: I am not GPS-savvy, I only know the very basics. I tried to browse through all the documentation made for normal users as me though; that lead me to understand something, but I still miss a big point.

It is about multipath.

As far as I've understood, the problem is that it under some circumstances (most noticeably in canyons, urban or not) GPS signals are deflected by obstacles, so they get to the unit across multiple paths, as pictured in Wikipedia:

Gps-multipath-efect.png

And on kowoma.de:

multipath.gif.

So the unit gets more than one signal from the very same satellite.

So far, so good; assuming that I got it right.

 

Now, the question is.

Being the correct path always the shorter (any deflected path is obviously longer than the straightforward one), couldn't the unit filter out all of them but the shortest, so to solve the issue?

 

The answer is of course not, elsewhere they would and multipath would not be an issue. What I'd love is to understand the reason why, and this probably is related with the fact that I miss some important piece of information.

 

Would please some GPS-saviour enlighten me and the other curious people around?

 

Any information would be truly appreciated.

 

Thank you all very much!

TOPONI

Link to comment
Now, the question is.

Being the correct path always the shorter (any deflected path is obviously longer than the straightforward one), couldn't the unit filter out all of them but the shortest, so to solve the issue?

Imagine the situation where the GPS is near the westerly canyon wall, and satellite A is not in direct line of sight (i.e. it's even farther west). However, the GPS does receive a reflected signal from satellite A off of the easterly canyon wall. It doesn't have the "shorter" signal to compare to, so it just uses the reflected signal. But this signal is now, say, 500 feet too long. So it's calculations will place it in the wrong spot.

 

Add 2 or 3 other satellites that are also out of sight over the rim of the canyon, but still close enough that you receive reflected signal, and your GPS could think you are miles away.

Link to comment

To answer your question a bit more directly - if the unit sees multiple signals from the same sat - yes it will reject the longest ones. However, most consumer grade units are unable to do a further signal analysis of a single signal to see if it has been reflected.

 

Thanks mate.

 

For what matters the first answer, "that longer path could be coming from a satellite further away" I think it's not the case as far as I have understood, since every sat has an unique identifier.

 

So in short: when multiple paths do arrive, the unit does filter all but the shortest path.

The problem is, that the shorter path is not necessarily the direct one, being possible that it is out of "sight".

 

I would call this issue the "deflected shorter signal" issue, rather than the "multipath", since it's not the multiplicity of the signals to cause the problem itself. Does this make sense?

Thanks again, I am finally clearing my mind out. As far as I'm getting this, SBAS are not going to help in this case.

Link to comment

I saw multipath in action for the first time ever recently while in a bus which was driving through a canyon. Reported location was 500-600 meters off the true location on an eTrex 20, placing us in the middle of a lake instead of on the road.

 

It would follow that those new, sensitive receivers have greater problems with multipath than the older ones - an old, not sensitive receiver will not pick up the rock wall reflected signal at all and be more prone to use only LOS signals from the satellites. But then, maybe better with a crude position than no position at all but that really depends on the application.

Link to comment

I saw multipath in action for the first time ever recently while in a bus which was driving through a canyon. Reported location was 500-600 meters off the true location on an eTrex 20, placing us in the middle of a lake instead of on the road.

 

It would follow that those new, sensitive receivers have greater problems with multipath than the older ones - an old, not sensitive receiver will not pick up the rock wall reflected signal at all and be more prone to use only LOS signals from the satellites. But then, maybe better with a crude position than no position at all but that really depends on the application.

 

Whoa. 500 meters is MUCH.

How deep (and narrow) the canyon was?

 

Thanks

Link to comment

I saw multipath in action for the first time ever recently while in a bus which was driving through a canyon. Reported location was 500-600 meters off the true location on an eTrex 20, placing us in the middle of a lake instead of on the road.

 

It would follow that those new, sensitive receivers have greater problems with multipath than the older ones - an old, not sensitive receiver will not pick up the rock wall reflected signal at all and be more prone to use only LOS signals from the satellites. But then, maybe better with a crude position than no position at all but that really depends on the application.

 

That could have also been an issue with the map. Many times a map might be drawn from a source that is incorrect. Also there are places that have moved since the original source maps were drawn. For instance in Sitka Alaska my GPS shows everything to be 100 ft NW of where it really is. I have been told this is because of plates shifting since the maps were originally created.

Link to comment

Consider that if one is in a canyon (natural or man made), the chances are that the best signals will be from nearly overhead, and those will be the ones 'chosen' by the GPSr. Since the best results will come from satellites widely spaced and lower to the horizon compared to almost straight up, you won't ever get good results in places where the sky view is limited by solid surfaces. That is without consideration of reflected signals.

 

In the OPs original diagram, he uses the left and right walls for blocking (good) and for reflecting (ok) but since several sats would be involved, that far peak could be the relflector for at least one signal, and one behind the camera doing the block. GPS depends on the difference in time between transmission (by sat) and reception (by GPSr) to establish the distance from the sat. The sat's location is determined by the information in the almanac (I always guess it's Keplarian data) which determine it's position over the Earth. So you get a point in space and a distance from that and when you look at the intersections of several you determine your terrestrial position by trilateration. Since radio travels at lightspeed (300,000 m/sec), very small intervals caused by reflection can affect the calculation if that signal is selected. Bouncing of position readings can be caused by switching signals between different sats or between multiple changing paths. Satellites always move anyway, so the calculations do as well. So the whole issue is extremely complex and has been studied if not compensated for. One could also consider that two paths could result in the same signal arriving at the antenna 180 degrees out of phase and cancelling out completely or partially. That is the signal peaks are opposite and the result is no signal. This happens in voice radio communications at lower frequencies like 2m HAM. Simply move the antenna a bit and you can be out of the null. At GPS frequencies it would be harder to experience, because the wavelength is shorter, but it could happen since the antennas are smaller as well.

 

Basically I'm saying that calling something multipath or reflection or anything else is going to have lots of additional factors thrown in to confuse things, but may or may not be what you think it is.

 

Doug 7rxc

Link to comment

 

That could have also been an issue with the map. Many times a map might be drawn from a source that is incorrect. Also there are places that have moved since the original source maps were drawn. For instance in Sitka Alaska my GPS shows everything to be 100 ft NW of where it really is. I have been told this is because of plates shifting since the maps were originally created.

 

Could have been, in my case it definitely was not. The receiver was reporting the wrong position or displaying it incorrectly on the map. Whether the actual reason was multipath or not I don't know, but conditions were pretty optimal for such a thing to happen so I assumed multipath was the reason.

 

100ft. error is quite normal for some map detail, 1000ft+. is not!

Edited by tr_s
Link to comment

For instance in Sitka Alaska my GPS shows everything to be 100 ft NW of where it really is. I have been told this is because of plates shifting since the maps were originally created.

A 100 foot displacement would take a LOOOONG time, 100,000 years or more. The San Andreas fault displacement is about an inch per year, or will be after the next big earthquake, and that's extreme. If by "everything" you mean manmade features like roads then the map's probably just drawn wrong.

 

FWIW, this is what multipath can do. A the bottom of the canyon some points are a mile off. At the top of the canyon the track is right on.

1272832453.jpg

And the rocks are where they're supposed to be.

Link to comment

For instance in Sitka Alaska my GPS shows everything to be 100 ft NW of where it really is. I have been told this is because of plates shifting since the maps were originally created.

A 100 foot displacement would take a LOOOONG time, 100,000 years or more. The San Andreas fault displacement is about an inch per year, or will be after the next big earthquake, and that's extreme. If by "everything" you mean manmade features like roads then the map's probably just drawn wrong.

 

FWIW, this is what multipath can do. A the bottom of the canyon some points are a mile off. At the top of the canyon the track is right on.

1272832453.jpg

And the rocks are where they're supposed to be.

Some recent large earthquakes are known to have shifted fixed established points as much as 20 feet in a single event.

 

There is some evidence that continental drift is somewhat variable and may move entire continents as much as 5 inches some years and as little as .5 inch others years.

 

Further some GPS data has shown that within the larger plates some smaller areas are moving off in various directions by as much as 8 inches per year - not to mention movement of instantaneous quakes.

 

I grant you 100 feet in less than 100 years is a large distance and not likely the sole result of plate movement - but it is indeed at least theoretically possible.

Link to comment

To answer your question a bit more directly - if the unit sees multiple signals from the same sat - yes it will reject the longest ones. However, most consumer grade units are unable to do a further signal analysis of a single signal to see if it has been reflected.

 

Thanks mate.

 

For what matters the first answer, "that longer path could be coming from a satellite further away" I think it's not the case as far as I have understood, since every sat has an unique identifier.

 

So in short: when multiple paths do arrive, the unit does filter all but the shortest path.

The problem is, that the shorter path is not necessarily the direct one, being possible that it is out of "sight".

 

I would call this issue the "deflected shorter signal" issue, rather than the "multipath", since it's not the multiplicity of the signals to cause the problem itself. Does this make sense?

Thanks again, I am finally clearing my mind out. As far as I'm getting this, SBAS are not going to help in this case.

 

Huh? Ummm. No?

 

If you've listened to FM radio in your car you've most likely experienced multipath fading. It usually happens when you are moving slowly. Like when you are creeping up on a red light. The strong signal from the radio station will begin to disappear until you are listening to nothing but static. It will then return to full strength as you continue to creep forward. If you creep forward some more the signal will start to fade out again.

 

In digital communication you have a type of multipath interference called intersymbol interference. This is where one symbol interferes with subsequent symbols. Imagine that someone is yelling out hi and low. It is your job to write down what the person says. Easy right? Now stick that person in a cave that echos a lot. I guarantee that you mistake some of the echos for the actual person.

Link to comment

FWIW, this is what multipath can do. A the bottom of the canyon some points are a mile off. At the top of the canyon the track is right on.

1272832453.jpg

And the rocks are where they're supposed to be.

@SeldomSN,

....been there,done that....(that situation, not physical location), and that's the reason why, when mapping an important trail, I carry 3 GPS's simultaneously. That "usually" (not always) results in a data set that can be edited into an accurate location for the trail. Sometimes even that is not enough and a repeat trip is required.

 

Strategically carried/placed units, logging type and interval, type of unit, current software version sensitivity, sometimes external antenna, etc, all have an effect on whether you are successful in minimizing multipath errors. Operator actions and habits can definitely influence the outcome.

 

Early results from an Oregon 550 were LOUSY, but after several communications with Garmin and resulting software/firmware changes, they improved greatly. Sadly, sometimes new S/W versions for different models "fix" one thing and "screw up" other capabilities that weren't broken. Every trail mapping session is a new day!

Link to comment

FWIW, this is what multipath can do. A the bottom of the canyon some points are a mile off. At the top of the canyon the track is right on.

1272832453.jpg

And the rocks are where they're supposed to be.

Grasscatcher, when I made that track I was also carrying an OR300. Results were as bad, but different. I haven't tried averaging combined tracks though. I got similar, but not as bad results on North Kaibab trail in the Grand Canyon using a 62s.

Link to comment

As others have said map displacement is generally not due to geological movements but just plain erratic data from geodesic institute, alternatively poorly converted to the Garmin format.

 

Garmin's basemap is a prime example of this. It's a shame that the few data it has isn't even accurate, but off by hundreds and hundreds of meters on average, it seems!

Link to comment

As others have said map displacement is generally not due to geological movements but just plain erratic data from geodesic institute, alternatively poorly converted to the Garmin format.

 

Garmin's basemap is a prime example of this. It's a shame that the few data it has isn't even accurate, but off by hundreds and hundreds of meters on average, it seems!

Sorry, but Garmin's Basemap is not an example of this. The issue with the Basemap is scale, not accuracy. It is accurate to it's scale of 1M.
Link to comment

Sorry, but Garmin's Basemap is not an example of this. The issue with the Basemap is scale, not accuracy. It is accurate to it's scale of 1M.

 

1:1000000 ?

 

Are there general rules on how accurate one can expect the data to be, based on scale? More than I knew... It's a vector map...

Link to comment

 

1:1000000 ?

 

Are there general rules on how accurate one can expect the data to be, based on scale? More than I knew... It's a vector map...

 

Yes; kind of. For most USA printed maps the accuracy standard is that not more than 90% of the tested points will be within 1/50 of an inch on the published map. For a 1:24,000 scale map (1 inch on the map represents 2,000 feet on the ground) that is 40 feet. The accuracy standard(s) can vary by map scale, time and country.

 

For most mapsets created for Garmin GPSrs the situation is more complicated. Coordinates are defined as 360 degrees divided by 2 raised to the 24 power (maximum). We call this the bit-level. Most mapsets will use three bit levels (some may use 2 or 4 depending on the type of data and what the map author is trying to show). Usually these are 23, 21, and 19 (OSM uses? 24, 22 & 20, likely because of all the building outlines available and the closer spaced street grid in the older cities). Either by default or by choice of the map author, the different features are defined to display to different zoomed-out bit-levels (it would over clutter a map display to show small buildings, ponds, intermittent streams, etc when displaying a large area on the GPSr). In each .img file in the mapset are three (+-) sets of map data; one each for the features and coordinates valid at each bit-level. The build software adjusts the more detailed coordinates and eliminates duplicate coordinate pairs and features too small to be shown as more than a point as it builds the more generalized maps.

 

As far as I can tell, a 1:24,000 scale map would be between the 23 and 22 bit-level. On the GPSr this is about the 500 or 800 foot diplay level. There appears to be a small difference depending on which model GPSr the mapset is being displayed at. The GPSr and MapSource and BaseCamp will also allow you to change the amount of detail displayed; however, this affects all feature types and zoom-levels. Issues are created when users keep zooming-in beyond the most detailed scale of the mapset; there is no more data available, nor any more detail in the data or the locations; only empty magnification - the features are simply seperated/enlarged at the same rate. Think of enlarging a paper map on a copy machine.

 

As far as Garmin's included basemap (probably 1:1,000,000 or 1:2,000,000); it's use would be when panning some distance and then zooming-in to the more detailed data.

 

As far as your previous concern about accuracy; road next to steeply sloping glacier shapped valley side, moving bus (metal roof and sides) and unknown scale and production method of mapset on GPSr, I do not think 1,000 off is unreasonable, perhaps not even 500 ro 600 meters. Did you note the 'circle of error' on the GPSr and the configuration of the satellites? With most of the NE 1/2 half of the sky blocked by the valley side and likewise some of the SW, the pattern was more linear the triangular.

Link to comment

I think the the multi-path problem may be old news. When the SIRF chipsets came out, their claim to fame was that their DSP algorithms could filter out multi-path signals.Others must have followed, I'm guessing.

The nasty tracks I posted above were with a 60csx (Sirfstar) and a Gilsson antenna. I've tried every Garmin I own on that hike, and none of them have given better results.

Link to comment

I think the the multi-path problem may be old news. When the SIRF chipsets came out, their claim to fame was that their DSP algorithms could filter out multi-path signals.Others must have followed, I'm guessing.

 

I have had very similar results in the Grand Canyon recently with an Oregon 450, 550t, and Montana 650. These units all have SIRF chipsets in them, I believe.

Link to comment

I have had very similar results in the Grand Canyon recently with an Oregon 450, 550t, and Montana 650. These units all have SIRF chipsets in them, I believe.

Which trails? I had fairly jumpy results with a 62s on the steep parts North Kaibab/Bright Angel, but also around Ribbon Falls.

Edited by seldom_sn
Link to comment

I think the the multi-path problem may be old news. When the SIRF chipsets came out, their claim to fame was that their DSP algorithms could filter out multi-path signals.Others must have followed, I'm guessing.

 

I have had very similar results in the Grand Canyon recently with an Oregon 450, 550t, and Montana 650. These units all have SIRF chipsets in them, I believe.

OOPS, should have quoted. I was responding to the first post and the topic "Multipath and the shortest path".
Link to comment

Which trails? I had fairly jumpy results with a 62s on the steep parts North Kaibab/Bright Angel, but also around Ribbon Falls.

 

I was hiking in Havasupai, along the Havasu Creek, between the Supai Village and the Colorado River. In some areas, my Tracks show me all over the canyon, off by ~500 ft from my actual position.

Link to comment

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...