reported F16 Look Down

[skywalker22]

1st 5 cases are indeed interesting, in a Look-Up situation there is surely something wrong. There is no ground clutter or anything (maybe only in case no.1).

As @BIGNEWY said, @Mr. Wilson paste the track files, so they can evaluatr them.

Didn't try that my self, but as said, 1st 5 cases are indeed strange.

Detection range is way too short, should be twice as that at least.

[BIGNEWY]

12 minutes ago, skywalker22 said:

Detection range is way too short, should be twice as that at least.

DM me your evidence we can show it to the team

[skywalker22]

1 minute ago, BIGNEWY said:

DM me your evidence we can show it to the team

Sorry, I would, but can't, I'm far from home atm.

OP (@Mr. Wilson) has to do it instead.

Anyway, I haven't tested for my self his 1 to 5 cases yet, but indeed they seem odd. Can you test that out @BIGNEWY?

[Mr. Wilson]

11 часов назад, BIGNEWY сказал:

Sorry is this a bug report? or just an observation?

please include track replays

 

Yes, this is a detailed bug report that shows what is between Detection range and Lock range. Also there is a strong drop in the detection range at low altitude of the attacking aircraft

The track file was very large and because of this I did not publish it, because earlier you made a remark to me that large files do not need to be published because the team does not have time to devote so much to analysis...

5 часов назад, skywalker22 сказал:

1st 5 cases are indeed interesting, in a Look-Up situation there is surely something wrong. There is no ground clutter or anything (maybe only in case no.1).

As @BIGNEWY said, @Mr. Wilson paste the track files, so they can evaluatr them.

Didn't try that my self, but as said, 1st 5 cases are indeed strange.

Detection range is way too short, should be twice as that at least.

To do this, I will most likely need to create a separate track file that will be much smaller...

[BIGNEWY]

Hi you want us to assume your data is correct without checking it and then you want me to ask the team to adjust it based on your findings. 

Sorry but that is not how it works. 

I can only go to the team with a bug I can reproduce.

When I test the results are as the team intended them. 

[skywalker22]

2 hours ago, Mr. Wilson said:

Yes, this is a detailed bug report that shows what is between Detection range and Lock range. Also there is a strong drop in the detection range at low altitude of the attacking aircraft

The track file was very large and because of this I did not publish it, because earlier you made a remark to me that large files do not need to be published because the team does not have time to devote so much to analysis...

To do this, I will most likely need to create a separate track file that will be much smaller...

Just do a short missions, you and AI flying at different altitudes. 5 different missions (your 1-5 case).

I would also help ED out here, but Im on vacations, so can`t help it out.

[Mr. Wilson]

3 часа назад, BIGNEWY сказал:

Hi you want us to assume your data is correct without checking it and then you want me to ask the team to adjust it based on your findings. 

Sorry but that is not how it works. 

I can only go to the team with a bug I can reproduce.

When I test the results are as the team intended them. 

Ok I will make a track with a few test cases to make it as short as possible, your data is very different from mine ?

[DCS FIGHTER PILOT]

Similar to,

 

the radar look down penalty for the Viper seems too severe in some cases and appears to take on a static value, (if you will) past a certain differential altitude. As one can see in the tracks, given look up, co-alt, or very mild look down conditions , a Viper can detect and track another Viper at around 40nm. From here, if the differential altitude between the two planes is increased very slightly from the very mild look down condition, detection and tracking range plummet from around 40nm to 27nm. This number more or less holds as the differential altitude continues to increase. I'm not 100% sure how look down penalty is modeled in DCS, but I would imagine the real world mathematical relationship between detection range and differential altitude (and likely closer speed) is transcendental in nature (i.e. no sharp drop off's/discontinuities). In game right now, it appears as if (whether it is intended or not), there is a simple "step function" like relationship between all the parameters. If this is the most complex thing the team can model now I understand but if it is not, it would be nice to have this addressed in the near future as it very much impacts gameplay. Thanks. 

Co-Alt.trk Look_Up.trk Mild_Look_Down1.trk Mild_Look_Down2.trk Mild_Look_Down3.trk Extreme_Look_Down.trk

[Mr. Wilson]

@BIGNEWY I created a new track file much more abbreviated to make it easier for you to research...

The track file is based on 2 test cases: 

CASE 1 - Fighter h=1 000 ft / Target h=10 000 ft  

CASE 2 - Fighter h=10 000 ft / Target h=20 000 ft 

In the first test case, we see an extremely small target detection distance and the second test case was chosen to show the delay between target detection distance and target lock distance. 

The delay between target detection distance and target lock distance is present in all test cases and also in all situations when you are trying to lock the target..

For example, in the second test case, I managed to press the TMS forward button 4 times before the radar took it to the lock. It is worth noting that there is no such behavior of the radar for any aircraft in dcs

I want to summarize everything that I wrote earlier, I see two problems:

- extreme reduction in the radar detection distance at low flight altitudes of an attacking aircraft

- delay between target detection distance and target lock distance 

server-20220731-015954.trk

Edited July 30, 2022 by Mr. Wilson

[BIGNEWY]

posts merged

[Mr. Wilson]

31.07.2022 в 09:27, BIGNEWY сказал:

posts merged

Сould you write how much your data differs from my data from the tests? Are there any problems already identified in the course of the research?

[skywalker22]

On 8/2/2022 at 10:57 PM, Mr. Wilson said:

Сould you write how much your data differs from my data from the tests? Are there any problems already identified in the course of the research?

My are exactly the same as yours, even after last update in August 2022. So I would say there is practically no difference before and after an update.

From last update:

 

Edited August 6, 2022 by skywalker22

[Tiger-II]

On 7/24/2022 at 9:33 AM, GGTharos said:

@NineLine I think this is an issue of simplification.   There is geometry involved, where eg. look-down may not actually be look-down due to the curvature of the earth, as well as the type of background clutter present.  Similarly, there exist look-up that would give you 'look-down' results, ie. a SAM (or low flying aircraft) looking up at the side of a mountain.

DCS does not cover these scenarios because instead of attempting to compute SNR and other actionable attributes (all knowledge which is readily available to anyone and known by ED, including various equations to calculate these things), it attempts to simplify them into antenna slew down vs antenna slew up.

That is the impression that the current situation gives, I figure Maestro should be able to clear this up.

That pretty much sums up my understanding of the modelling, too. It's over-simplified. The RADAR literally looking below the local horizon is insufficient. The question is does the target have a background, what angle is the background at (flat ground vs. mountain), how far behind the target is it, and how far from the emitting RADAR is it?

Edited August 7, 2022 by Tiger-II

[FusRoPotato]

17 hours ago, Tiger-II said:

That pretty much sums up my understanding of the modelling, too. It's over-simplified. The RADAR literally looking below the local horizon is insufficient. The question is does the target have a background, what angle is the background at (flat ground vs. mountain), how far behind the target is it, and how far from the emitting RADAR is it?

 

Angle of surface and surface type might be too high-fidelity and complex to worry about because it concerns more of a finer detail than it does basic logical function. It matters only a little and you'd probably not enough difference for it to ruin the experience. What's far more important is whether or not the target is within the section of picture where the latter portion of the range bin dips below the surface. No matter what, if a target on the deck is approaching you, it should have some rare chance of showing through the noise just via doppler, then 100% chance of showing through when it rests within picture where the range bin is above ground. This of course happens more often the higher a target is, but there will not be any noise at all in some lookdown cases because it is filtered out by time. The range bin simply ignores anything that comes back too late. It may be impossible to lock them if they are far, slow, and low enough, but there should still be occasional hits with luck and it's what TWS prediction was created for. Not ever having anything show up at all makes a massive difference in combat tactics and this lookdown penalty just erases all possibility of that.

The same can be said for lookup against side lobe interference. It's just overtuned and always interfering without any consideration for all the filtering techniques and randomness associated with how a radar paints a picture and determines if something is there.

[Mr. Wilson]

06.08.2022 в 18:19, skywalker22 сказал:

My are exactly the same as yours, even after last update in August 2022. So I would say there is practically no difference before and after an update.

From last update:

 

 

It's good now to understand from the developers in their version of the game the same or not

[skywalker22]

3 hours ago, Mr. Wilson said:

It's good now to understand from the developers in their version of the game the same or not

Of course its the same, the same for all.

[skywalker22]

What surprises me on APG-68, that it can barely look up at low altitude. 

CASE 1 -- Fighter h=1 000 ft / Target h=1 000 ft / Detection range=14.33 nm / Lock=13.5 nm

CASE 2 -- Fighter h=1 000 ft / Target h=5 000 ft / Detection range=14.35 nm / Lock=13.3 nm

CASE 3 -- Fighter h=1 000 ft / Target h=10 000 ft / Detection range=14.35 nm / Lock=13.3 nm

CASE 4 -- Fighter h=1 000 ft / Target h=20 000 ft / Detection range=14.63 nm / Lock=13.8 nm

CASE 5 -- Fighter h=1 000 ft / Target h=30 000 ft / Detection range=13.22 nm / Lock=12.7 nm

 

What's destracting it?

For case 1 I might say its the ground clutter, but for all the rest? There is no noise of anykind, its looking up into the sky.

[Hobel]

From:

 

Edited August 8, 2022 by Hobel

[okopanja]

45 minutes ago, Hobel said:

Thanks, could you also post a link to this dashboard?

[FusRoPotato]

On 8/7/2022 at 9:29 PM, skywalker22 said:

What's destracting it?

For case 1 I might say its the ground clutter, but for all the rest? There is no noise of anykind, its looking up into the sky.

They are trying to account for sidelobe interference but, similar to lookdown logic, their approach to it is also misprepresenting how a radar functions and how much information it can return. When a radar paints an area, it does so with a main beam, but because of the radar's shape, there are additional beams that come out the side in a sort of water ripple pattern called sidelobes. Those can bounce off the ground and return noise, though typically only the bottom of the first sidelobe has enough vector strength to produce noise.

Sidelobe interference can interfere with lookup detection while flying low, but this interference is very dependent on the time it takes for the first sidelobe to return compared to the range of the target. The other sidelobes almost never interfere because their returns will always be extremely weak, but sometimes the first one can be strong enough to mask a target far away at high elevation when the range of the target matches the range of the first sidelobe hitting the ground. This results in the sidelobe return squeeking past the filters and increasing the chance of masking. All other sidelobes are typically too weak or easily filtered out by time.

What this typically results in blind spot that is dependent on a geometric calculation involving the targets range, targets altitude, and radar altitude. Of course, such estimation is simple and assumes flat terrain, but otherwise could be determined by the target's range, elevation, and a ray trace towards the ground from the bottom of the first sidelobe's forward vector. What might be left is a target you can't see at 20nm and 25kft up while flying at 1000ft, but you will be able to see it when it's at 25nm. It can be calculated geometrically if a source can be found of the AGP-68's first sidelobe radius angle, but even though it's probably not classified, such accuracy is in simulation is probably not a good thing. Just a generic representation of how sidelobe interference works would probably be more than good enough, but simply reducing range based on altitude misses the fact that radars flying low can still see far away targets up high. The interference is not always interfering, I wouldn't even say it usually interferes, but sidelobe interference is a real thing that can blind you.

Edited August 9, 2022 by FusRoPotato

[NineLine]

Hey all, just a small update, we found some excellent data on look down affect and we are investigating, more news when I have it. Thanks for your patience. 

[skywalker22]

2 hours ago, NineLine said:

Hey all, just a small update, we found some excellent data on look down affect and we are investigating, more news when I have it. Thanks for your patience. 

Awesome. Btw, look up also included?

[skywalker22]

Here is another demonstration how radar AN/APG-68 v5 looses lock in a clean head-on lookdown situation (5k of difference). That should really not happen.

 

f16_radar_issue4_looses_lock.trk

[cmbaviator]

Hello,

 

I don't know if its me but i don't understand why when below 20/30Nm, i can't bug a target after at least 10+ seconds. Even if I'm a DCS noob, i often get killed in BVR because the ennemy is able to lock and fire me way earlier than i'm able to. In the video below at 1:02, it took 30s to get a soft lock on that target and i wasnt even looking down, it took 30s!!!! to soft lock an aircraft around 30Nm and below ? I talked on the 4YA server and some players told me that the F16 has lately that issue, ie taking to much time to soft lock a target ? is it the same in TWS mode ?

 

Granted that target was slightly in a notch position, but given the low range, shouldn't I be able to TWS him earlier ?

Edited September 8, 2022 by cmbaviator

[skywalker22]

That's the current look-up and look-down issue of APG-68(V)5 radar. 

We did some test recently, you can check it out here, and it's still under investigation by ED team.