Proper Radar simulation

[stormrider]

DCS needs a proper radar simulation.

[Worrazen]

Ah, I hoped for more explanation, are you familiar with how things work or just from the outside ?

 

Either way we need to remind ourselfs this is quite expensive on physics calculations.

[stormrider]

Ah, I hoped for more explanation, are you familiar with how things work or just from the outside ?

 

Either way we need to remind ourselfs this is quite expensive on physics calculations.

 

 

Ok, I see I'm talking to someone who understands it. What imo is a doable solution and which I thought it was obviously understood by this image, is to perform a all-aspect modeling like the picture above, and use the data in a a lut table. Although that's just a very small piece of the puzzle that I would like to see in DCS. There needs to be more complexity in its radar modeling.

[Pikey]

Don't think you can do it. Let me explain.

 

DCS has no concept of the unit in physical shape. None. It reads it from the lua description of the unit. I've explained this before, but the proof of this is by creating your own 3D model, for example, I created a periscope, and then providing the lua description for an aircraft carrier. You understand what happens on your radar screen, right? :) I also reported late activated units appearing on radar once, it's as clear as day, nothing comes out of the nose, its an illusion, but that doesn tmean its bad, its just limited.

 

So establishing that the model is not used for any calculations, you cannot do exactly what you see above, but what you can do is provide a longer list of RCS values (which ARE in the lua decriptions) provide a lot of different aspects, provide some for weapons loadouts (missing currently and is important) and extend the current model with some tables and simple calcs rather than flat values.

 

Make sense?

 

To some people. There are still many flat earthers who believe there is magic raytracing for air to air radar coming out of their radars nose, but we can leave them behind because I dont want to ruin the magic for everyone. They wont read this anyway.

 

So the counter proposal is a better RCS table and to include loadouts in the RCS, which is the most desperately needed.

[TLTeo]

Yep, running a table model rather than doing the calculations makes infinitely more sense and it's far more performance-friendly.

[stormrider]

Don't think you can do it. Let me explain.

 

DCS has no concept of the unit in physical shape. None. It reads it from the lua description of the unit. I've explained this before, but the proof of this is by creating your own 3D model, for example, I created a periscope, and then providing the lua description for an aircraft carrier. You understand what happens on your radar screen, right? :) I also reported late activated units appearing on radar once, it's as clear as day, nothing comes out of the nose, its an illusion, but that doesn tmean its bad, its just limited.

 

 

I'm not sure if we're in the same page here. I'm not referring to actual real-time radar or electromagnetic scattering simulation here, using a 3d model or any "model" per se. On the other hand, in the second paragraph, you seem to be in the same line as me or Teo:

 

So establishing that the model is not used for any calculations, you cannot do exactly what you see above, but what you can do is provide a longer list of RCS values (which ARE in the lua decriptions) provide a lot of different aspects, provide some for weapons loadouts (missing currently and is important) and extend the current model with some tables and simple calcs rather than flat values.

 

 

This kind of data is usually placed in what's called a lut table. The data is precalculated, the same way as missiles CFD, then a 2D or even a 3D table is generated. In this table, you have for example, in one axis, the Db values, another axis, the frequency and lastly, in another axis the aspect angles. For the pitch, it could be simply an aproximation like a log multiplier or a simple custom multiplier curve to add precision.

 

By itself this simple solution would already add another depth (frequency) to simplistic current radar model in place without compromising cpu capacity.

 

To some people. There are still many flat earthers who believe there is magic raytracing for air to air radar coming out of their radars nose, but we can leave them behind because I dont want to ruin the magic for everyone. They wont read this anyway.

 

So the counter proposal is a better RCS table and to include loadouts in the RCS, which is the most desperately needed.

 

 

In fact, even if out context, this is already possible. Ray tracing was already tested for a sonar model on a prototype submarine simulation. The sonar model was open source. The results were amazing in terms of accuracy, but they required a lot of computing power. Perhaps today with modern multicore CPUs, this could be doable. Regardless of that, it's OT here.

[Worrazen]

Yeah this one is hard to discuss because of the technical nature of it, I'm relatively weak in this area other than I understand PC hardware and some basics to get an idea how costly physics simulation can be, etc.

 

I'm not actually an expert either, anything to do with radar will not be the real thing, not with ATX PC hardware at home, it'll be an approximation with significant shortcuts and tricks. I know you didn't think ray-tracting, but for comparison, we're not even fully raytracing visible graphics yet in the whole gaming industry, let alone other EM frequencies and sound.

 

Well, if we get a PSC one day, a Personal Super Computer, with the form factor the size of a refrigerator, external cooling in the basement hooked up with big pipes to cool the fins over an enclosure filled with non-conductive cooling liquid (oil) and the GPU/CPU submerged in it ...etc inside ... then we might be able to do a lot more in DCS.

 

Believe it or not, this is actually a serious idea I'm writing some notes from time to time, to make a draft. It's not that far fetched, many successful people who are retired would be able to afford 10-20K for something like this as a serious hobby, if they really want it, besides what about classic cars, they sell for 100K's and millions.

Edited September 17, 2020 by Worrazen

[stormrider]

Interesting paper: GPU Accelerated Ray-tracing for SimulatingSound Propagation in Water

 

 

http://liu.diva-portal.org/smash/get/diva2:1352170/FULLTEXT01.pdf

Edited September 19, 2020 by stormridersp

[deadpool]

You don't need a supercomputer. Having an RCS lookup depend on angle, that's absolutely easy with a look up table (LUT).

 

The problem is that the entire radar engine seems to be not working on a reasonable approach but on wtfs ...

 

If a target turns cold on you in the f-16 .. it will disappear at hardcoded 20nm .. the whole missile guidance radar stuff is completely borked ..

 

then if you see defects like:

missile is desynced and goes active, and your RWR then gives you a launch warning of the plane that launched it, even though it's turned cold already .. you just know that under the hood it's implemented in a VERY VERY rudimentary way that's not even taking the physical or world model into account.

[stormrider]

In other words, the current model is very dodgy at best.

 

We need a better radar modeling.

[Harker]

Yes, a more extensive LUT would be a good thing, with more angles, external stores modeling etc. I fact, given the current hardware state, RT could be used to generate a very accurate LUT for pretty much any shape and angle.

 

As for the underlying system, I agree, it needs a rework as well. Although I don't know how much of it is module-specific and how much is just the DCS core doing its thing.

[Northstar98]

As for the underlying system, I agree, it needs a rework as well. Although I don't know how much of it is module-specific and how much is just the DCS core doing its thing.

 

Does anyone know the RADAR technology employed in the F-14 and Viggen modules? I swear they did some form of ray-casting, and I swear this approach is different to aircraft like for instance the F-16C, F/A-18 and F-15.

 

I also swear the MiG-21bis' new RADAR system also does something along the lines of some form of limited ray-tracing/ray-casting if I'm not mistaken.

 

I'm pretty sure that the F-5E-3s RADAR is simplified, at least according to this.

 

I don't know about the RADARs of AI assets, such as our various ground units, ships and AI aircraft, presumably they're also pretty simplified.

 

There's also the issue of DCS terrains that model the Earth as if it was flat, which presents problems in and of itself. I'm pretty sure a workaround(?) is used for the RADARs but not sure what their limitations are. I guess some testing has to be done, but the flat Earth thing is really a discussion in and of itself.

 

EDIT: Quick test in the F-16C, my RWR picked up the RADAR of the Slava class CG while flying at 595 feet above the water at a distance of ~40.6nmi which seems about right according to this ( I should have LOS to the top of the ship at around 40nmi

Edited September 21, 2020 by Northstar98

[Harker]

Does anyone know the RADAR technology employed in the F-14 and Viggen modules? I swear they did some form of ray-casting, and I swear this approach is different to aircraft like for instance the F-16C, F/A-18 and F-15.

 

 

 

I also swear the MiG-21bis' new RADAR system also does something along the lines of some form of limited ray-tracing/ray-casting if I'm not mistaken.

That's what I was thinking too. Out of these three, I only own the Tomcat and you can clearly see ground returns and even ships, under the correct conditions and with the right radar settings.

 

The Viggen's AG radar rework was also impressive and it clearly didn't use ED's API, since it came out much before and its picture was better modeled than the APG-73 P2 of the Hornet, which wouldn't be the case IRL. It could be over-modeled, but still, it seemed like it used something different than a terrain layer, that ED is using.

 

Finally, the MiG's radar clearly takes ground clutter into account as well.

 

For all the above though, I have one question. Is target detection based on the same mechanism that generates ground returns etc? Or is it still the same mechanism as for other modules, added on top? It could be that two different types of returns are shown.

[Northstar98]

The Viggen's AG radar rework was also impressive and it clearly didn't use ED's API, since it came out much before and its picture was better modeled than the APG-73 P2 of the Hornet, which wouldn't be the case IRL. It could be over-modeled, but still, it seemed like it used something different than a terrain layer, that ED is using.

 

This is what got me, (mind you I'm on stable so things might've changed, I don't have DBS/SAR modes yet), in the Hornet, I can't see bugger all, even at close ranges; all I can make out are areas where there are a buildings, such as villages, towns and cities; I also noticed that roads also show up very well, not as clear as the Viggen, but way more obvious.

 

However, at close-range it is basically impossible to make out anything else, and no, I'm not talking about trucks and tanks and what have you, I'm talking about terrain features - I'm not doing any targeting here and trying to make out individual vehicles, I'm only focusing on terrain features.

 

In the Viggen, at the closest ranges I can distinctly make out rivers, runways, aprons/taxiways, pylons, bridges, roads and forests. Further away (beyond ~15km) everything is blurry and skewed as you might expect, but in general I can still make out fairly large bodies of water and major rivers. In the Hornet, on the Caucasus, I can't see anything apart from towns and cities and roads.

 

Obviously I can see where the coast is and in some cases the topography ahead. But I was pretty shocked at just how little I could see in the Hornet, compared to the Viggen, just for terrain features.

Edited September 23, 2020 by Northstar98

[Kev2go]

yea I too thought this to be unusual.

 

 

It would appear Heatblurs Radar tech ( think its Recasting?) technology is simply a more sophisticated and realistic method of emulating A/G terrain scanning.

 

Yea compared to stable build within Open beta the Hornet can get closer zoom of a terrain patch via 3 EXP modes and get much sharper image but in basic map mode even in OB is still seems inferior to the Viggens mapping radar. I dont know if it is overmodeled.At one point HB showed an excerpt of a Viggen A/G map.

 

 

 

 

 

That being said basic real beam mapping is supposed to be overall fairly blurry. havent seen any imags of actual APG73 Real beam mapping, but in general from images ive seen its supposed to be pretty blurry ( and is especially even more blurry on the out edges) to the point its really just really useful for making basic geographic features and large clusters of buildings to discern a settlement. Hell you can't even clearly discern an airfield until you are with say 10NM given in RBGM it lacks the angular resolution that you get with Doppler beam sharpening or let alone with SAR, which do of course need angular offset to see stuff.

Edited September 23, 2020 by Kev2go

[deadpool]

Does anyone know the RADAR technology employed in the F-14 and Viggen modules? I swear they did some form of ray-casting, and I swear this approach is different to aircraft like for instance the F-16C, F/A-18 and F-15.

 

I also swear the MiG-21bis' new RADAR system also does something along the lines of some form of limited ray-tracing/ray-casting if I'm not mistaken.

 

I'm pretty sure that the F-5E-3s RADAR is simplified, at least according to this.

 

I don't know about the RADARs of AI assets, such as our various ground units, ships and AI aircraft, presumably they're also pretty simplified.

 

There's also the issue of DCS terrains that model the Earth as if it was flat, which presents problems in and of itself. I'm pretty sure a workaround(?) is used for the RADARs but not sure what their limitations are. I guess some testing has to be done, but the flat Earth thing is really a discussion in and of itself.

 

EDIT: Quick test in the F-16C, my RWR picked up the RADAR of the Slava class CG while flying at 595 feet above the water at a distance of ~40.6nmi which seems about right according to this ( I should have LOS to the top of the ship at around 40nmi

 

The radar and the eyeballs of the AI are simplified to stupidity.

 

You fly a Viggen between hills at treetop level .. and you can see on tacview exactly when you are in line of sight of an enemy MiG21 .. because it will turn in on you.

 

Can't have been the radar .. so they must have perfect eyesight .. 30nm seeing a viggen at treetop level .. it's really boring.

[Northstar98]

That being said basic real beam mapping is supposed to be overall fairly blurry. havent seen any imags of actual APG73 Real beam mapping, but in general from images ive seen its supposed to be pretty blurry (and is especially even more blurry on the out edges) to the point its really just really useful for making basic geographic features and large clusters of buildings to discern a settlement. Hell you can't even clearly discern an airfield until you are with say 10NM given in RBGM it lacks the angular resolution that you get with Doppler beam sharpening or let alone with SAR, which do of course need angular offset to see stuff.

 

Oh yeah, it's the same in the Viggen, your image looks fairly close to what we have right now, at the range indicated. I can only see things in detail at <15km, the rest is a blurry mess, and sometimes even skewed but I can sort of make out what it's supposed to be when comparing to the F10 map.

 

The Hornet however, I can only make out coastlines, clusters of buildings or roads (roads actually show up really well, just as well as towns which I thought was odd), sure they're all blurry, but I can see what it's supposed to be on the map. But in the Hornet it's impossible to make out anything else, even at the closest range setting (5nm).

 

In the Viggen, under 10km you can clearly see rivers, trees, roads, pylons, bridges and buildings, they don't stand out as much as the Hornet does, but you can make them out.

 

The Hornet however, is very good at spotting towns, and has more range than the Viggen, but it looks incredibly cluttered, and not as clear as the Viggen.

 

I'll get back to you with a test.

 

 

 

Testing

 

So far this is what it looks like in the current 2.5.6 stable, I don't if the Hornet's AG RADAR has changed in successive OB updates, apart from the obvious DBS modes. Note that I'm not trying to say which is more realistic, just highlighting the difference between the 2.

 

For this test I'll simply place down a Viggen about 5-6~ km north of Vaziani at 1000m MSL, I'll get a few pictures of what the picture looks like, and then swap it out for a Hornet and repeat. Both aircraft will stay in active pause at mission start, so they should be in identical positions at identical altitudes.

 

View from out the front, we can see Tbilisi-Lochini airport directly ahead, as well as much of the city. We can just make out Tbilisi-Soganlug, with Tbilisi-Vaziani just off our left. Note the trees, road and electrical pylons and rough landmarks.

 

 

 

Starting off at 15km

 

 

We can clearly see the the river, the buildings and the electrical pylons. The airbases/airports as well as the roads are faint, but visible. In the case of Tbilisi-Vaziani, we can just about make out the runway, taxiway, ramps and aprons if we look hard enough. Lochini is visible, but more skewed, we can just about make out the general outline-ish. Also note how easy it is to tell the difference between trees and buildings.

 

 

30km

 

 

Again, same as above, it's very clear what's what, even if details such as airbases and roads are quite faint. This is because we can see a very faint reflection from the ground itself, and almost nothing from rivers, roads and airbases, so there is some contrast even if it's faint. In the Hornet we don't get any reflections from the ground, so it's basically impossible to make out rivers, roads and airbases unless there's a load of buildings outlining them.

 

Now for the Hornet, here I'll try 5, 10 and 20nm, at the default and maxed gain.

 

 

Starting off at 5nm, at the default gain (5)

 

 

Can't see the terrain at all, can just about make out buildings (which I only know are buildings because I can see them with my eyes, otherwise it would be pretty difficult to distinguish). In general it's far less clear as the Viggen, only the closest pylons are visible and we can't see trees, or make out Vaziani, or see the river and trees.

 

 

Dialing the gain up to max (9)

 

 

Now we can see a bit more, though if I didn't know what I was looking at already, it would probably be quite difficult. I think the main problems is that, unlike the Viggen there's no faint reflection from the ground, and not much contrast between a weak return and strong return, it's also less well defined.

 

 

Now out to 10nm and gain back to default

 

 

Pretty much the same story, just reaching further out, with some faint returns it is pretty difficult to gauge what it is. Note that the Hornet seems to be able to pick up certain features such as buildings, roads/railways(?) further away than the Viggen, but there's a lot less contrast and generally more blurriness. I am far better able to positively identify and correlate features in the Viggen (even if it has a shorter range) than the Hornet, where I can maybe see some more stuff, it's just way more difficult telling what it is.

 

 

Gain to 9

 

 

Now we can resolve out some of those question marks and identify them as buildings, the river and the airbases are still basically invisible, there's just a few faint splodges. One thing I will say is that RADAR shadow seems to be modelled slightly better in the Hornet, than the Viggen, but it probably needs more testing and this area isn't exactly the most suitable.

 

 

Now out to 20, default gain

 

 

Much the same thing again, but note how the buildings are better defined now, they all appear bright, whereas at 10nm at gain of 5, some of them were much fainter.

 

 

And finally gain maxed

 

 

I think I can just about see Tbilisi-Lochini, but note how well the roads and railways to the left of the screen are defined - very obvious. Unfortunately, the rivers are still basically impossible to see, as are large clusters of trees. If we got a slight reflection from the ground itself they might be more visible.

 

 

Edited September 23, 2020 by Northstar98
Initial testing