Aim-9X way too sensitive to flares

[Mad_Shell]

The Aim-9X has an imaging infrared seeker, giving it way mor capabilities to discriminate flares from targets compared to previous IR seeker generations. You can find an example of real life trials here: 

 

 

In the 1st case, the missile flies for more than 10 seconds, and hits the target while it's against the ground clutter and releasing a large number of flares at very high rate. 

 

I did some similar(ish) test in DCS, with a similar flight time, against a flaring, non maneuvering, non afterburning F/A-18C flying straight level at co-altitude (so no ground clutter, which is not modelled in DCS anyways afaik, contrary to the irl test). Track file attached.

The results:

Terrible hit rate of about 10% percents at best, with the missile going for flares almost every time. That seems very unrealistic, especially compared to what we see in the video of the trials.

I found that instead of the current ccm_k0 = 0.2, a way lower ccm_k0 = 0.025 gives a hit rate of about 60-80%, which seems reasonnable given that in the real life trials the missile manages to hit despite a higher flares release rate and ground clutter compared to my test.

Moreover in the "Apache over Libya" book, there is an engagement in which the Apache has difficulties defeating a MANPADS with flares, barely surviving. In DCS, even the Aim-9x will very easily be defeated by an Apache launching a few flares. 2 or 3 flares are enough for any MANPADS in almost all cases. That also suggests that the resistance to flares of the Aim-9X, and possibly other missiles, is too low in DCS.

9x_IRCCM_effectiveness.trk

[Glide]

I have a lot of experience with the 9x in DCS, and I feel it is well balanced.  If I am careful and wait for a rear aspect launch, I can almost always get 6 for 6 in the F-16.  The missile will loose lock against the sun or if a hot target passes through it's fight path.  I believe cold targets like the SU-25 have a better chance of defeating it vs hot targets like the Mig-29S, for example. 

You might want to try a mission like my PG Dogfight for testing.  Fly that a few times and you will see what I mean.

Cheers.

[Mad_Shell]

6 hours ago, Glide said:

I have a lot of experience with the 9x in DCS, and I feel it is well balanced.  If I am careful and wait for a rear aspect launch, I can almost always get 6 for 6 in the F-16.  The missile will loose lock against the sun or if a hot target passes through it's fight path.  I believe cold targets like the SU-25 have a better chance of defeating it vs hot targets like the Mig-29S, for example. 

You might want to try a mission like my PG Dogfight for testing.  Fly that a few times and you will see what I mean.

Cheers.

I know that the 9x is a potent missile in DCS. But, and no disrespect here, the goal of my testing was to go beyond the feels, and have some standardized test somawhat similar to the real life trial, and we can see an absolutely gigantic difference.

[Hobel]

vor 6 Stunden schrieb Glide:

I have a lot of experience with the 9x in DCS, and I feel it is well balanced.  If I am careful and wait for a rear aspect launch, I can almost always get 6 for 6 in the F-16.  The missile will loose lock against the sun or if a hot target passes through it's fight path.  I believe cold targets like the SU-25 have a better chance of defeating it vs hot targets like the Mig-29S, for example. 

You might want to try a mission like my PG Dogfight for testing.  Fly that a few times and you will see what I mean.

Cheers.

Same here, but I haven't flown with aim9x for a long time, maybe something has changed, the tests here are quite fresh.

[Glide]

7 hours ago, Mad_Shell said:

the goal of my testing was to go beyond the feels, and have some standardized test

I totally get it.  The thing to keep in mind is that this is a simulation, and the sim has to be fun.  On the one extreme you have missiles that never miss.  This is very easy to test, but not very realistic.  So, the devs have to add some level of randomness so you and your targets can sometimes defeat the missiles.  This add a huge amount of fun, but it's much harder to test.  Let's throw up some ideas.

Straight and level rear aspect.

Straight and level nose on.

Approach at 3 and 9 o'clock.

Approach into the sun.

Approach with the sun at your six.

Shooting high.

Shooting low.

Shooting with conflicting targets nearby.

I also find that the 9x is pretty deadly when fired just inside 1.5 miles, but depending on the merge dynamics, this short distance can quickly become a miss.  So, add to the above scenarios a variety of launch distances.

A good test mission would have a variety of drone targets to demonstrate each scenario, but drones do what their told.  After all of that testing there are the Red Flag tests against various AI skill levels. 

That's why I created that PG Dogfight mashup.  It's a ton of fun to fly, and I see all the behaviours of the AI and the missiles.  I know my mission is not very scientific. 

Edited November 4, 2023 by Glide

[Ramius007]

Guess AIM-9X in DCS is still better than IRL, at least missile not going stupid after launch

[Nealius]

2017 shootdown over Syria, didn't the 9X get decoyed by the Su's flares, requiring a follow-up shot with an AMRAAM?

[Hobel]

vor 55 Minuten schrieb Nealius:

2017 shootdown over Syria, didn't the 9X get decoyed by the Su's flares, requiring a follow-up shot with an AMRAAM?

 just need to know more about the situation, some Sus have ir jammer that + flares could sometimes be enough.  

But the whole thing is quite vague as it is

[Mad_Shell]

4 hours ago, Nealius said:

2017 shootdown over Syria, didn't the 9X get decoyed by the Su's flares, requiring a follow-up shot with an AMRAAM?

We don't know what happened there. Maybe the missile had a malfunction, maybe it was the flares. Not enough info to conclude anything. That's why I refer to the video of the trials, because we have a better understanding of the various parameters.

[TEDUCK]

On 11/16/2023 at 5:23 AM, Hobel said:

 just need to know more about the situation, some Sus have ir jammer that + flares could sometimes be enough.  

But the whole thing is quite vague as it is

IR jammers (I mean old ones, not the DIRCM) were meant against optically modulated seekers, they should give very little effect against FPA seeker. Flares should be almost useless because AIM-9X can reject them with ease (flare looks like a small spot to a FPA seeker while plane is much bigger object, the guidance system knows how the flares look and rejects them).

Quote

2017 shootdown over Syria, didn't the 9X get decoyed by the Su's flares, requiring a follow-up shot with an AMRAAM?

AIM-9X uses reduced smoke motor and Su-22 has preatty crappy visibility outside of cockpit, IRL there is almost no chance for a Su-22 pilot to see the FOX-2 missile and react properly.

Edited November 20, 2023 by TEDUCK

[GGTharos]

On 11/15/2023 at 10:26 PM, Nealius said:

2017 shootdown over Syria, didn't the 9X get decoyed by the Su's flares, requiring a follow-up shot with an AMRAAM?

No, the Su did not user flares.   The missile likely suffered a fin failure or some other issue (possible launch out of parameters, though seems unlikely) that caused it to fail to steer.   The 9X has been demonstrated to be next to immune to flares IRL.  In other words if it's launched at you within parameters, hope for a very rare failure or eject.

Edited November 24, 2023 by GGTharos

[Dragon1-1]

Or defeat it kinematically. Unless launched from rear aspect, deep within the WEZ, the AIM-9X is subject to the same constraints as other missiles. A high aspect launch will be more likely to miss simply by virtue of being high aspect, thus requiring the missile to make large maneuvers in response to target's attempts to evade. High off-boresight shots are even worse, with the initial turn to target wasting a lot of speed. It's a great missile, but not a magic wand.

As for smokeless, well, this only helps you at altitudes where contrails don't happen. Also, it doesn't mean there's absolutely zero smoke, you can still see a launch in most conditions. If you have tally in heater range, you probably have a good idea when the other guy shoots.

I wouldn't discount flares, either, although against something like AIM-9X you ideally want more modern decoys. The "imaging seeker" is not the same as an infrared camera, and a flare can disrupt the image in various ways. Combined with maneuvering, they might change the image enough for the seeker to get confused. It was effective in tests, but everything is effective in tests, things that aren't don't pass the tests. In the real world, things usually aren't that simple. Recall the AIM-9M, which rejected high tech American flares no problem, but was fooled by dirty burning Iraqi flares due to their crappy rise time.

[Mad_Shell]

21 minutes ago, Dragon1-1 said:

Or defeat it kinematically. Unless launched from rear aspect, deep within the WEZ, the AIM-9X is subject to the same constraints as other missiles. A high aspect launch will be more likely to miss simply by virtue of being high aspect, thus requiring the missile to make large maneuvers in response to target's attempts to evade. High off-boresight shots are even worse, with the initial turn to target wasting a lot of speed. It's a great missile, but not a magic wand.

As for smokeless, well, this only helps you at altitudes where contrails don't happen. Also, it doesn't mean there's absolutely zero smoke, you can still see a launch in most conditions. If you have tally in heater range, you probably have a good idea when the other guy shoots.

I wouldn't discount flares, either, although against something like AIM-9X you ideally want more modern decoys. The "imaging seeker" is not the same as an infrared camera, and a flare can disrupt the image in various ways. Combined with maneuvering, they might change the image enough for the seeker to get confused. It was effective in tests, but everything is effective in tests, things that aren't don't pass the tests. In the real world, things usually aren't that simple. Recall the AIM-9M, which rejected high tech American flares no problem, but was fooled by dirty burning Iraqi flares due to their crappy rise time.

It's true that imaging seekers don't equal total flare immunity, and some specific tactics combined with some types of flares (different models of flares are not modelled in DCS sadly), have a chance to defeat them.

However, what we witness in DCS is that flares can quite easily defeat even an AIM-9X, without any particular maneuver, if the target is a few kilometers away. 

[Dragon1-1]

That's because flare modelling is extremely simplistic in DCS. It's basically just rolling the dice, no modeling takes place, although the dice roll is impacted by whether the afterburner is on or not. That's it. There might be other modifiers, but that's it as far as depth of the "simulation" goes in that area. You'll also notice that if the missile is decoyed, it will begin tracking the decoy instead, which is particularly absurd with chaff (implemented the same way).

Also consider how distance affects resolution. The CCD matrix on the AIM-9X isn't all that big, at a certain distance, a plane will also look like a single bright pixel (flares will likely show up brighter than the plane, too), so this has to be taken into account. I would expect that this distance would be out of parameters anyway (except perhaps directly head on, where ranges are long and planes look small), but the gist is that the further you are, the harder it is to tell a plane apart from a flare. 

If we want better results, we need an overhaul of the CM simulation. That means simulating different seekers, different ECCM and various types of flares to go with that. On top of that, things like airframe heating (all heaters are all aspect when your target is doing Mach 2), angles from which engines and AB plumes are visible, that sort of thing. We have yet to see any sign of that being planned by ED. Basically, the options are turning it into a magic wand, or minor tweaks to what we have.

Edited November 24, 2023 by Dragon1-1

[ruxtmp]

As I understand it, the AIM-9x likely was an out of parameter shot with ground clutter being the primary culprit.

[TEDUCK]

On 11/24/2023 at 7:13 PM, Dragon1-1 said:

The "imaging seeker" is not the same as an infrared camera, and a flare can disrupt the image in various ways.

It is literally infrared camera. You can see one of the old videos on YT showing what the AIM-9X seeker is seeing.

On 11/24/2023 at 7:13 PM, Dragon1-1 said:

Recall the AIM-9M, which rejected high tech American flares no problem, but was fooled by dirty burning Iraqi flares due to their crappy rise time.

AIM-9M has optically modulated seeker, so totally irrelevent to the AIM-9X that uses imaging seeker. For imaging seeker every single flare will look like a small spot, and every single plane won't. At least when the shoot is in parameters and not from too far range when plane will look like a single dot from a distance (but then missile should recognize plane when it gets closer). The only way would be having so many flares close to eachother blending into some larger shape.

 

Keep in mind that Su-22 has very crappy visibility out of cockpit, so it is extermally unlikely for a pilot to see the missile launch if it's not a head on launch. In DCS AI planes seems to always see the missile launch, while players almost never spot the launch. Same IRL, pilot won't spot the launch unless it is specifically looking at that exact spot.

Edited November 26, 2023 by TEDUCK

[Dragon1-1]

17 minutes ago, TEDUCK said:

Same IRL, pilot won't spot the launch unless it is specifically looking at that exact spot.

Which he is, if he's any good. Lose the sight, lose the fight, if you're in WVR and lose tally, you'll die whether your opponent has AIM-9X or just a few rounds in the gun (unless you can run away before he gets in parameters, that is, which is not always possible). It's easy to hit when your opponent is not looking at you and doesn't know where you are.

17 minutes ago, TEDUCK said:

It is literally infrared camera. You can see one of the old videos on YT showing what the AIM-9X seeker is seeing.

It's not. Similarities with an IR camera end at the CCD matrix. The resolution is quite low and the algorithm is completely different from missiles that use an actual camera, such as the Maverick. A combination of flares, ground clutter, clouds, relative position of the sun (IR background, in other words) and a number of other factors influence the picture, and the missile is very limited in terms of electronics it can take to 

Also, one more thing. The seeker does not "recognize" a plane the way a human would see it when closing it. The seeker locks onto a target and guides the missile to that target. Target validation methods used by the seekers are obviously classified, but it's very unlikely they work right 100% of the time. It can recognize when the target looks like something it'd want to hit (plane, cruise missile) and will reject anything that does not. If the target manages to make itself look like it's not a valid target, the missile will go stupid. How could that happen? The Russians would most likely pay a lot to know that.

There's no such thing as a perfect seeker. Even sophisticated ML algorithms fail on occasion when dealing with real world images. Something you can cram into a missile will, by necessity, be much less robust.

[TEDUCK]

19 hours ago, Dragon1-1 said:

There's no such thing as a perfect seeker. Even sophisticated ML algorithms fail on occasion when dealing with real world images.

Another misconception. You don't need ML for such basic image processing. You just reject false targets procedurally. AIM-9X was developed before ML became a thing in real world use. ML will often fail, because it is sort of unreliable by the way it works. Procedural image recognition is very different thing. It is much more predictable, but less flexible. For flares procedural recognition is great, ML is not even considered any useful for such a simple solution.

19 hours ago, Dragon1-1 said:

It's not. Similarities with an IR camera end at the CCD matrix. The resolution is quite low and the algorithm is completely different from missiles that use an actual camera, such as the Maverick.

Maverick was developed before they started to use algorithms for image processing. In early versions they used simple contrast tracking with dedicated circuit to do the job, no software defined image processing for early Mavericks. AIM-9X uses software defined image processing.

The resolution is 128x128 pixels, not too bad. Keep in mind that the seeker has fairly low FOV and the missile is a short range missile.

FPA seeker is thermal camera by definition. Various cameras have various resolution of a seeker. The fact that it has 128 by 128 doesn't means it is not a camera.

Flare will always look like a small dot, that's it, unless you have whole lot of flares close to eachother.

https://sites.google.com/site/samsimulator1972/home Here you can find a PDF called Histoy of the Electro-Optical Guided Missiles. It describes many IRCCM methods used, it also describes differences in seeker technology.

Quote

. It can recognize when the target looks like something it'd want to hit (plane, cruise missile) and will reject anything that does not.

It most likely does the opposite. It knows how the flare will look like and rejects anything that corresponds to the scheme. Targets are very complex, while flares are extremally simple to recognize. Here is a quote from the paper:

During mid-course, the tracking FPA will typically have 7 × 7 or 9 × 9 pixels on the target 
aircraft. Note that the target presents a complex pattern of pixels receiving energy. The hot 
flare is physically small and therefore puts (lots of) energy into a single pixel. The gray-body 
decoy puts the amount of energy equivalent to a valid target into multiple pixels, as this type 
of decoys uses rapidly oxidizing foil pieces that bloom to fill a large volume. However, the 
shape of the energy pattern is changed from the spatial energy distribution of the target. 

The key is that the shape does not have to look like an a priori stored image of what an 
aircraft should be. Rather, the tracker can reject the decoy because it does not correlate 
with the energy distribution seen a short time before.124

 

Quote

The Russians would most likely pay a lot to know that.

Su-22 was developed before Russia even existed. It was developed in Sovlet Union well before first FPA seeker based air to air missile.

Edited November 26, 2023 by TEDUCK

[Dragon1-1]

58 minutes ago, TEDUCK said:

Su-22 was developed before Russia even existed. It was developed in Sovlet Union well before first FPA seeker based air to air missile.

You missed the point. If the Russians knew exactly what conditions cause the AIM-9X to miss, they'd know a whole lot about how exactly the IRCCM on the missile works, and they could incorporate that knowledge. That's why this stuff is kept under wraps. This might as well have been an IRCCM false positive, with things like flares (many of which are usually dumped when defending from a missile launch) aircraft maneuvering and terrain clutter all causing a change in energy distribution of sufficient magnitude that the tracker rejected the real target. Again, procedural algorithms are not magic, and they make mistakes. If you've ever done image processing you'd know. And with a 128x128 pixel image, processing you can do is limited.

Also, things this paper talks about do not constitute the whole picture. There's a good chance that the IRCCM technique that actually got fooled was not described in the paper, because, again, the actual way of doing those things is pretty classified.

[TEDUCK]

We don't even know if the pilot knew about Hornet in the area. If he knew he would probably have aborted the mission.
If he didn't knew about the hornet he was likely unaware of the threat and the missile being launched.
Ground attack pilots are usually more concerned about ground fire, especially in Syria where AAA and MANPADS were common threat. It was not very common for US aircraft to shot at Syrian planes.

2 hours ago, Dragon1-1 said:

If the Russians knew exactly what conditions cause the AIM-9X to miss, they'd know a whole lot about how exactly the IRCCM on the missile works, and they could incorporate that knowledge.

Aircraft and pilot were Syrian. I wouldn't expect Russia to be upgrading old Syrian Su-22 with some countermeasures against AIM-9X. The only method known to work well against that missile (block 1 variant of AIM-9X) would be DIRCM (like one mounted on Ka-52).

 

Aircraft maneuvering wouldn't look like a flare. It wouldn't be rejected. No matter what maneuvers you do the aircraft won't be looking like a flare, unless the distance is very long (which would mean that the shot would probably be out of range anyway). It would be theoretically possible for a missile to go for ground clutter, but almost impossible for it to just reject the real target as a flare. Ground clutter is much harder to reject than flare. Because it is not uniform and easy to predict. I would bet on a ground clutter rather than a flare.

Edited November 26, 2023 by TEDUCK

[Dragon1-1]

46 minutes ago, TEDUCK said:

Aircraft and pilot were Syrian. I wouldn't expect Russia to be upgrading old Syrian Su-22 with some countermeasures against AIM-9X. The only method known to work well against that missile (block 1 variant of AIM-9X) would be DIRCM (like one mounted on Ka-52).

Again you're fixating on this single engagement. Of course they're not going to be upgrading an old Syrian Su-22, they'll be designing a countermeasure for Su-57 which would reliably replicate the conditions which trashed that shot. It's much easier to develop a countermeasure when you know what a given seeker does to avoid countermeasures. That is what they would need it for. It's not about upgrading Su-22s, it's about developing highly effective new CMs.

46 minutes ago, TEDUCK said:

Aircraft maneuvering wouldn't look like a flare. It wouldn't be rejected. No matter what maneuvers you do the aircraft won't be looking like a flare, unless the distance is very long (which would mean that the shot would probably be out of range anyway). It would be theoretically possible for a missile to go for ground clutter, but almost impossible for it to just reject the real target as a flare. Ground clutter is much harder to reject than flare. Because it is not uniform and easy to predict. I would bet on a ground clutter rather than a flare.

Read what I wrote. I never claimed the aircraft would look like a flare. I said it would stop looking like itself. Some of its pixels would be obscured by flares (Soviet designs often eject them up and forward), others would change with maneuvering. Ground clutter would add to that. If a large enough change happens in a short enough time, it can break the "compare the image from a moment ago to the one I see now" part of the missile's routine, which would make it reject the target as not being what it was locked onto. Note that unlike in DCS, the missile would simply go stupid in that case (and maybe self-destruct for safety), not chase after a flare. 

[TEDUCK]

40 minutes ago, Dragon1-1 said:

If a large enough change happens in a short enough time, it can break the "compare the image from a moment ago to the one I see now" part of the missile's routine, which would make it reject the target as not being what it was locked onto.

I don't think that Su-22 would be so good at maneuvers. Those sensors work at fairly high framerate. No chance the plane gonna maneuver so aggresively to fool it that way. If processing worked at 2 or 3 FPS or so it could have been a thing. Even for a very slow processing at 20 FPS the maneuvers will be looking very gradual. The missile was designed to be used against targets maneuvering much better than Su-22.

40 minutes ago, Dragon1-1 said:

they'll be designing a countermeasure for Su-57 which would reliably replicate the conditions which trashed that shot.

By time they finally finish the Su-57 and put it into service US will probably update the algorithm multiple times or even introduce AIM-9X block III. Su-57 seems to be stuck in development being too expensive to reliably produce and use in combat. You cannot evaluate the missile performance based on a single engagement and a short video released by US that doesn't show much (IIRC it only shows the AMRAAM engagement, not even the AIM-9X moment itself). Reality is not a TACVIEW where you can easily analyze the shot. Russians would need to know the exact position of both planes and the missile updated every split second. It is not possible, unless it is DCS where you can launch a replay and press F6 to see the missile

Edited November 27, 2023 by TEDUCK

[SickSidewinder9]

Anyway...

What really bothers me about the aim-9's not tracking is that the Russian heaters seem far more immune to flairs.  So mine will miss and theirs almost never do.  Really frustrating.  And there's no way the AIM-9X sucks in meat space as hard as it started to a few patches ago in DCS.

[Dragon1-1]

Flares not flairs. There might be a bug or misconfiguration in the files, they should be comparable. Although, keep in mind it's also a matter of kinematics. I don't know if you're fighting humans or AI, the latter is good at knowing its firing parameters. The better they see the nozzles, the better they are at tracking. Afterburner state also matters, if you're in full AB, don't expect flares to help much. DCS does not model any seeker intricacies, it's just a dice roll with a few modifiers.

49 minutes ago, TEDUCK said:

By time they finally finish the Su-57 and put it into service US will probably update the algorithm multiple times or even introduce AIM-9X block III. Su-57 seems to be stuck in development being too expensive to reliably produce and use in combat. You cannot evaluate the missile performance based on a single engagement and a short video released by US that doesn't show much (IIRC it only shows the AMRAAM engagement, not even the AIM-9X moment itself). Reality is not a TACVIEW where you can easily analyze the shot. Russians would need to know the exact position of both planes and the missile updated every split second. It is not possible, unless it is DCS where you can launch a replay and press F6 to see the missile

Could you please read what I'm actually saying? It's not about videos or even having engagement data, it's about knowing what was going on inside the seeker of the shot missed. This would be valuable data, that's why it's classified. Hence, our speculation about how exactly the miss happened is pointless, because we're missing important stuff about the operational details of the seeker (such as framerate). If you want to participate in a discussion, please use logic and address arguments that were actually made, instead of strawmen of your own making. Otherwise you're just making noise.

Su-57 is in service, BTW. Not many, but they launched missiles over Syria. That, however, is peripheral to the discussion. Having your enemy know exactly how your missiles work is a big liability regardless of how good that seeker might be.

[GGTharos]

The seeker framerate is out there somewhere (I forget the document now) and it is on the order of 100FPS with full multi-target processing...probably before the upgrade.

There's no point in talking about the seeker, it could have easily been a power failure, a fin failure, out of parameter shot or really anything else.