Tired of Bugs in Navico Equipment

[Dockhead]

Originally Posted by tanglewood

The answer to the headline question "Does Navico 4G radar (branded as Simrad, Lowrance, or B&G) have a "ridiculously broken MARPA" function?", is Yes after you read through the comments in the article.

You will find that those "obvious reasons" are completely invalid and that being "small" does not in any way handicap a radar's ability to perform MARPA. Furuno's small radars prove that, delivering excellent MARPA performance, and performance comparable to their larger radars.

Well, I don't agree that being small has nothing to do with a radar's MARPA performance. For good MARPA performance you need a good compass, and you need a radar which gives accurate target bearing. A larger antenna with narrower beam is inherently better at that.

However, software becomes more and more important, as it gets cleverer and more able to enhance the basic capabilities of the antenna. Considering the multitude of other software problems I've encountered with my Navico gear (the original subject of this thread), I completely buy the idea that my radar's poor MARPA performance is a software problem. I haven't tried MARPA (or ARPA, technically) on Furuno radars, but I find it plausible that it is far superior.

I think that MARPA is an important function, not at all obsoleted by AIS. Radar in general is crucially important for the kind of sailing I do, with all the night sailing, sailing in bad weather, and sailing in poorly charted areas. I spent time on a big Swan with two radars, and might consider something like that for my next boat. A small domed one up on the mast, and a large open array one on a radar pole. Unless Navico has some big turnaround by then, they will not be Navico radars, I think. Probably Furuno or Raymarine.


I loved using the big (6') open array on the Swan I sailed down the coast of Baja California, and up around in the Sea of Cortez, even though that was a decade ago and a generation ago of technology. It had very high resolution, very long range, very great target discrimination, and was very powerful. The charts down there were off by up to a mile, so we used it a lot for navigation. But you could also watch weather systems go by at 50 miles and more away, and detect vessels of absolutely any size down to dinghies. You could watch not only flocks of birds, but often individual birds. It was an amazing radar. I can't remember who made it; it was some commercial set with a dedicated, CRT monitor. Northstar or something like that. IIRC, it put out 25kW (!).

[colemj]

Originally Posted by Dockhead

For good MARPA performance you need a good compass, and you need a radar which gives accurate target bearing. A larger antenna with narrower beam is inherently better at that.

Somebody did not read that article all the way through the comment section…

Mark

[Dockhead]

Originally Posted by colemj

Somebody did not read that article all the way through the comment section…

Mark

Someone certainly did read it

Nowhere did anyone argue that bearing accuracy is irrelevant to MARPA performance. What was argued -- and I accept it -- is that the poor performance of Navico MARPA cannot be attributed to beam width etc.

As I said, clever software has done a lot to make small radars before like big ones of old. The vaunted close-range resolution of the Navico CW is not remarkable except in comparison with other small (18") scanners.

Navico has carp software, I think, which is the root problem.

[colemj]

Maybe I'm remembering it wrong, but I remember Tanglewood schooling Ben on radar beam width having very little to do with MARPA accuracy. He gave very good explanations as to why that is. Essentially, I remember that the leading edge of the return is what is important, not the width resolution. Maybe he will jump in and correct this.

Mark

[Dockhead]

Originally Posted by colemj

Maybe I'm remembering it wrong, but I remember Tanglewood schooling Ben on radar beam width having very little to do with MARPA accuracy. He gave very good explanations as to why that is. Essentially, I remember that the leading edge of the return is what is important, not the width resolution. Maybe he will jump in and correct this.

Mark

Peter argues that beam width is not as important because MARPA doesn't care about the width of the target -- it only needs to know where is the center of the radar return.

This is probably true to some extent, but a lower resolution return provided by a wider beam will not magically give you the same accuracy. The "edges" will not be as sharp, and the deviation between calculated center of the target, and the actual center of the target will grow, as resolution goes down.

So Peter's point does mean, and I agree with it, that the quality of MARPA function may not be a linear function of resolution. But resolution is certainly not irrelevant. The wider the beam is, the less information you get coming back. You also have less target discrimination, which is a big factor in MARPA performance, for obvious reasons -- the tracking is aborted when MARPA can't distinguish your target from other returns. Clever software can also reduce this effect -- by projecting the target's course and speed and looking for it where it expects it should be despite the target's having dropped off for a moment. But you can only do so much, with so much information.

[tanglewood]

I actually am arguing that the bearing accuracy is NOT related to beam width, and that the center of the echo is the center of the target within the aiming accuracy of the scanner. That aiming accuracy is 1% in the radar's who's specs I've looked at, and is the same for a 24" Furuno dome as well as my 6' 12kw FAR radar. The only difference the beam width makes is how much extraneous schmeer you get on either side of the target, thereby distorting the apparent width of the target. But the center is balls-on within the 1 deg rotational precision.

[tanglewood]

Oh, and I'm just talking about nominal conditions, single target, etc. As you start to get into challenging conditions and varuous corner cases, then I agree beam width comes into play in terms of the radars ability to identify a target. If you can't identify it, then you can't track it, so those issues clearly impact ARPA

The classic pathological example of a wide beam width is the inability to distinguish two side by side target. The first will still be giving an echos return when the beam starts to pass over the second target, so the echo return is continuous for both targets and they appear as one larger target. ARPA will track the center of the echo, which of course neither of the two boats. If you get close enough or the targets spread far enough apart such that the radar can distinguish them (they need to be separated by more than the beam width), then ARPA will jump and follow one of the two targets, or perhaps drop the target all together. Something similar often happens when a tracked target passes close to another boat or a buoy and the ARPA lock jumps to the other boat or jumps to the buoy.

[Dockhead]

Originally Posted by tanglewood

I actually am arguing that the bearing accuracy is NOT related to beam width, and that the center of the echo is the center of the target within the aiming accuracy of the scanner. That aiming accuracy is 1% in the radar's who's specs I've looked at, and is the same for a 24" Furuno dome as well as my 6' 12kw FAR radar. The only difference the beam width makes is how much extraneous schmeer you get on either side of the target, thereby distorting the apparent width of the target. But the center is balls-on within the 1 deg rotational precision.

You are assuming that different width beams have the same resolution at their edges, and differ only in width. But this is false.

In fact the edges of the beam are not "edges" at all; they are gradients of reflected energy, the steepness of which depends on the beam angle. So the accuracy of your calculation of the center of the beam, depends on its width. The wider the beam, the shallower the gradient, and the softer the edges, and the less certainty you have exactly where a target has entered the beam. Thus the less accurate will be your calculation of where the middle of the beam is.

You can check it with a thought experiment:

Imagine a beam which is 90 degrees wide. Do you imagine that it can give you a bearing with the same precision as a 1 degree beam?

The radar "beam" is not a beam actually, but a lobe:



https://www.ibiblio.org/hyperwar/USN...H-P-08-05.html



So as I said -- you can do some fantastic things with signal processing, and with software. Furuno software clearly blows Navico away. But there is only so much you can do, with so much information. With the same quality of signal processing, a larger, higher energy radar, which generates more information, is going to give you better bearing accuracy, and better MARPA performance, every time.

[Dockhead]

Some more information on this question:

"Bearing measurements can be made more accurately with narrower horizontal beam widths. The narrower beam widths afford better definition of the target and thus, more accurate identification of the center of the target." [emphasis added]

Radar Navigation and Maneuvering Board Manual

National Geospatial Intelligence Agency

Maritime Safety Information

[tanglewood]

Yes, I agree that the beam does not have sharp edges, so actual detection begins gradually and ramps up as the beam progresses over the target. In this respect target size and reflectivity matter as well. But at some point, let's call it C+n to represent the beams center bearing (C) plus how every many degrees before the beams center (n), the target begins to produce an echo. At all times C is known to within 1 deg. What's not known is exactly what n is. It might be the specified beam width/2, or it might be less for a smaller target, or it might be more for a very large target.

The target then paints for n degrees, at which point the C is at the left edge of the target. The target then continues to paint for the width of the target at which point C is at the right edge of the target.

So far we have a target echo that is n degrees plus the target's actual width wide. Note that I am mixing angular and linear measures here, but the conversion between the two is straight forward.

The beam center continues past the right edge of the target and continues to paint for another n degrees, at which point the return is lost. The right edge of the target is at C-n when this happens.

So the painted target is 2n degrees wide plus the actual width of the target. The 2n degrees of "false" echo are of equal size on both sides, so you know the target is centered between them, so the center of the target is the same as the center of the return, and it doesn't matter how large or small n is. All that matters is how accurately the radar knows the position of the beam center.

Re the 90 degree beam width example, yes, I do think it will give the same result. The target will paint for a full 90 deg of the sweep, 45 deg before the actual target and 45 deg after. But midway between the start of the echo and the end of the echo is the target's center. This is why in all commercial radar training you are told to take bearing measurements to the center of the target.

[tanglewood]

Originally Posted by Dockhead

Some more information on this question:

"Bearing measurements can be made more accurately with narrower horizontal beam widths. The narrower beam widths afford better definition of the target and thus, more accurate identification of the center of the target." [emphasis added]

Radar Navigation and Maneuvering Board Manual

National Geospatial Intelligence Agency

Maritime Safety Information

OK, I'll check it out. But do you see my point about half of the extraneous echo being before the target, and half being behind, so the center of one is the center of the other?

[tanglewood]

Originally Posted by Dockhead

Attachment 114905

https://www.ibiblio.org/hyperwar/USN...H-P-08-05.html

This article seems to be focused on manually training a radar on a particular target, presumably to shoot it. And it correctly shows that looking at the strength of the return at any given bearing will not help you locate the target's center. This is due to beam width.

But it then goes on to describe exactly what I described as a way to accurately find a targets center, namely to sweep the target then bisect the echo return. A marine radar does such a sweep constantly.

The context is still a bit different from a recreational or commercial marine radar, but I think the approach is the same.

The article also brings up another interesting pathological case, namely a fast moving target. In that case the radar can be "chasing" the target and create an echo larger (I think) than if the target were moving slowly or standing still. In that case I'm pretty sure the target's center will be skewed off the center of the echo return, but I'd have to think about it some more to be sure.

[Dockhead]

Originally Posted by tanglewood

Yes, I agree that the beam does not have sharp edges, so actual detection begins gradually and ramps up as the beam progresses over the target. In this respect target size and reflectivity matter as well. But at some point, let's call it C+n to represent the beams center bearing (C) plus how every many degrees before the beams center (n), the target begins to produce an echo. At all times C is known to within 1 deg. What's not known is exactly what n is. It might be the specified beam width/2, or it might be less for a smaller target, or it might be more for a very large target.

The target then paints for n degrees, at which point the C is at the left edge of the target. The target then continues to paint for the width of the target at which point C is at the right edge of the target.

So far we have a target echo that is n degrees plus the target's actual width wide. Note that I am mixing angular and linear measures here, but the conversion between the two is straight forward.

The beam center continues past the right edge of the target and continues to paint for another n degrees, at which point the return is lost. The right edge of the target is at C-n when this happens.

So the painted target is 2n degrees wide plus the actual width of the target. The 2n degrees of "false" echo are of equal size on both sides, so you know the target is centered between them, so the center of the target is the same as the center of the return, and it doesn't matter how large or small n is. All that matters is how accurately the radar knows the position of the beam center.

Re the 90 degree beam width example, yes, I do think it will give the same result. The target will paint for a full 90 deg of the sweep, 45 deg before the actual target and 45 deg after. But midway between the start of the echo and the end of the echo is the target's center. This is why in all commercial radar training you are told to take bearing measurements to the center of the target.

You're still assuming a hard edge -- and you're assuming the same accuracy of determination of the edge of the beam ("start of the echo" and "end of the echo"), with different beam widths. That's the basic problem here, because that's not true.

The "edge" of the beam ("start" or "end" of the echo) is where the strength of the reflected signal reaches a certain percentage of max (71% is one convention; see below). The strength of the reflection rises and falls at a rate which is a function of the width of the beam. But your equipment has a certain sensitivity -- it can resolve the difference between x mW and y mW, say, but cannot resolve a difference of less than that. So the shallower the rate of change (which is a direct function of beam width), the less accurately will it be able to determine the exact time the threshold has been reached. Thus the accuracy of detection of the "edge" of the beam is a function of the beam width, assuming that in all cases the electronics are equally sensitive.

The accuracy of the calculated center of the beam, is a direct function of the accuracy of the determination of the "edges".

Thus it is not true, that you can calculate the center of the target with equal accuracy, with radar beams of any width.

To illustrate:

Say you have a radar with a 10 degree beam. An ideal target (zero size and 100% reflectivity) located due N from your ship, detected by an ideal radar set (infinite sensitivity and processing power) will go from "no reflection" to "reflection" in 0 time at 355 degrees and from "reflection" to "no reflection" at 005 degrees. Thus it is trivial to determine with total accuracy that the bearing to the center of the target is 000 degrees. That's the way you look at it, Tanglewood.

But real radar doesn't work like that. There will be a period of uncertainty between "reflection" and "no reflection", because real radar sets don't have infinite sensitivity. The same amount of radar set sensitivity, will produce a longer period of uncertainty, the wider the beam is. Would need to do the math, but I bet it's roughly a constant percentage of the beam width. So if your radar can get the threshold signal strength to an accuracy of +/- 25%, then it will detect "reflection" at 355 +/- 2.5 degrees, so might say 352.5 or 357.5 or anything in between. Likewise with "no reflection" -- 002.5 to 007.5. The accuracy of calculation of the center between these two points will be the same.

But if the beam is 5 degrees wide, this uncertainty will be half. Thus the accuracy of calculation of the center will be twice as accurate.

These are somewhat simplified guesses, for the purpose of illustrating the principle. If anyone has the real math, that would advance the discussion.



"Beam width is the angular width of a radar beam between points within which the field strength or power is greater than arbitrarily selected lower limits of field strength or power. There are two limiting values, expressed either in terms of field intensity or power ratios, used conventionally to define beam width. One convention defines beam width as the angular width between points at which the field strength is 71 percent of its maximum value. Expressed in terms of power ratio, this convention defines beam width as the angular width between HALF-POWER POINTS. The other convention defines beam width as the angular width between points at which the field strength is 50 percent of its maximum value. Expressed in terms of power ratio, the latter convention defines beam width as the angular width between QUARTER-POWER POINTS."

http://msi.nga.mil/MSISiteContent/St...RNM/310ch1.pdf

[tanglewood]

Originally Posted by Dockhead

But real radar doesn't work like that. There will be a period of uncertainty between "reflection" and "no reflection", because real radar sets don't have infinite sensitivity. The same amount of radar set sensitivity, will produce a longer period of uncertainty, the wider the beam is. Would need to do the math, but I bet it's roughly a constant percentage of the beam width. So if your radar can get the threshold signal strength to an accuracy of +/- 25%, then it will detect "reflection" at 355 +/- 2.5 degrees, so might say 352.5 or 357.5 or anything in between. Likewise with "no reflection" -- 002.5 to 007.5. The accuracy of calculation of the center between these two points will be the same.

OK, I see what you are saying now. I was making the simplification that the garbage on either side of the echo would always be the same. You are saying that it's +/- based on the echo return becoming strong enough for the detector to trip. So the garbage on either side is only +/- equal, not exactly equal.

Wouldn't the sharpness of the edge depend primarily how rapidly the beam power drops off at the edges rather than the beam width? I've seen the "edge" specified as 3db less than the center power or something like that, but that's just a paper spec. I know nothing about beam shaping and the degree to which one can obtain sharp edge drop-off independent of beam width. You are saying they are related, which I don't doubt, but how much?

It would be interesting to know how many degrees of rotation the no-detect/detect transition takes.

What I do know is that Furuno specs bearing accuracy to 1 deg on a number of their radars, large beam and small beam. I take that to mean they can determine the bearing to target with that precision, taking all these factors into account.

[poiu]

Originally Posted by tanglewood

OK, I see what you are saying now. I was making the simplification that the garbage on either side of the echo would always be the same. You are saying that it's +/- based on the echo return becoming strong enough for the detector to trip. So the garbage on either side is only +/- equal, not exactly equal.

Wouldn't the sharpness of the edge depend primarily how rapidly the beam power drops off at the edges rather than the beam width? I've seen the "edge" specified as 3db less than the center power or something like that, but that's just a paper spec. I know nothing about beam shaping and the degree to which one can obtain sharp edge drop-off independent of beam width. You are saying they are related, which I don't doubt, but how much?

It would be interesting to know how many degrees of rotation the no-detect/detect transition takes.

What I do know is that Furuno specs bearing accuracy to 1 deg on a number of their radars, large beam and small beam. I take that to mean they can determine the bearing to target with that precision, taking all these factors into account.

In the Navico promotional information that they say they use a technique to improve the resolution of the radar. The 3g doesn't have the 'beam sharpening' 'target separation control' technology, but the 4g does and so improves effective beam angle from 5.2 deg to 2.6 deg. I am sure it is edge detection. Looking for a rising and falling return over the leading and falling edge of the return. How else could it be done? Also, if it is done like this, then it should be possible to obtain an even higher target resolution shouldn't it?