Mad_Shell

Mmmmh, according to quaggles website quite a few parameters in the fligt model have been changed (https://github.com/Quaggles/dcs-lua-datamine/commit/cf37c477acf7370f0247867a1d7b71e7ba9609c7#diff-61004a5997f6aa0641f2e2225af9ea2c0d86d70f556ac5b672ebef28c3220bffR855) Even the center of mass of the aircraft has been shifted.

good to know the bug only affects a few IR missiles and no radar missile! Do you plan to modify the default value of 0.5 for some of those missiles? The Mistral in particular has a reputation for having good counter-countermeasure, but currently it almost always go for the flares.

You probably have your reasons, but wouldn't the "good coding practice" be to keep the code as clean as possible and remove the now useless parameters? Yes it should be carefully tuned, but I think the way the AI is flying really prevents it to effectively use the missile in many cases with such a low Fi_start value. Well, the document I attached in my second message in the thread mentions that the missile is put in rotation by the booster. BUT... after watching a few videos in slow motion the missile doesn't seem to spin after the booster ignition... and I agree with you that the 2 pairs of canards suggest a non rotating missile.

I agree with you, since day 1 of the new BVR AI I've criticized this behaviour, and a few others. - the AI always cranks and dive after firing its missile(s). But cranking AND diving only makes sense if the enemy is very high threat, and you suspect your'e close to the MAR of the enemy plane. Watching AIs F-18 firing Aim-120C from 30 miles and diving, while they're against a Mig-21 is ridiculous. They give up altitude and offensive potential for nothing. - the AI will ALWAYS go straight at the enemy until they fire their own missile. That means, we can see planes like Mig-21, Mirage F1, F5... happily go in a straight line to be slaughtered 100% of the time by superior planes and missiles. Against superior planes, they SHOULD either try to run away, or try to get closer by cranking, going cold then hot again, notching and hugging the ground... whatever it takes to try to preemptively trash enemy missiles and break radar lock to try to get closer and in range of their own missiles. - the AI defensive maneuvers against active Fox-3 are kind of pathetic. They always try dive and notch, dropping chaff, never pulling more than like 3 or 4 Gs. If the AI estimates that the incoming missile is high energy, then it is acceptable behaviour (no hope to kinematically defeat the missile). But if the AI estimates that the missile is low energy, just going in a notch and praying it works is a really bad decision... the AI should maneuver hard instead.

@Маэстро After a bit of testing, I found some problems with the new ED's version of the missile, so here are my suggestions: - X_back = 0 seems strange, since it should be the location of the exhaust along the X axis, and 0 should be the center of the missile. I propose X_back = -0.90, the value used for the nozzle_position parameter in the "march" section) - sigma = {10, 10, 10} is too high, and leads to about 15% of shots missing a static Ka-50 not launching any flares. Since MBDA advertises a 93-97% hit rate during exercises on moving targets smaller than a Ka-50, I propose sigma = {3, 3, 3}, which according to my testing should give very close results to MBDA's claims. - Fi_start is too small. I know it's realistic, but the AI Gazelles don't manage to use the missile on any target with some angular velocity because the "cone of fire" is too small. Based on my testing, Fi_start = 10° gives good results without making it a high boresight missile for the AI. - work_time = 0.2 is too long. On the real life videos, we can see the booster impulse is very, very short, and I propose to keep work_time = 0.048 like in Polychop's file. - ccm_k0 doesn't have any effect (see https://forum.dcs.world/topic/321195-possible-massive-bug-affecting-many-ir-missiles-and-possibly-radar-missiles-too/#comment-5171557), giving the missile a very, very poor resistance to flares (while IRL it is reputed as having very good counter-countermeasure). If ccm_k0 works again one day, I propose a value of ccm_k0 = 0.3 (a bit worse than AIM_9X, but better than the MagicII, which is older and of which the seeker was used to design the Mistral seeker). - we have cx_coeff = { 0.6, 1.6, 0.68, 0.55, 1.85 } in ED's file, and cx_coeff = { 0.65, 1.7, 1.6, 0.45, 0.01 } in Polychop's file. I really don't know which one is closer to reality, but given the importance of this parameter, sorting it out seems important. Maybe @Polychop Simulations and @Маэстро can give us some insight on how they calculated those parameters for this missile? - the ED's version doesn't spin on itself (it does IRL). The Polychop's version spins, I guess because it uses scheme="schemes/missiles/self_homing_spin_missile.sch", while ED uses scheme = "schemes/missiles/mistral_missile.sch" @btd this one is for you: as said above in Polychop's message, when the ED Mistral file is used, there is a delay of about 2 seconds between the missile launch and the sound of the missile launch playing (and no, I wasn't looking from 700 meters away lol). I attach the AA_missiles.lua modified with all the changes I suggest, excepting the "scheme" parameter (when I put scheme="self_homing_spin_missile" in ED's version, DCS crashes as soon as the missile is fired). The cx_coeff parameter is still the ED one too. AA_missiles.lua

@Flappie @Маэстро I tag you both because that's possibly a very important one. While doing testing on the Mistral missile, I found out, as I have suspected for some time) that the ccm_k0 parameter has absolutely no effect. For those not aware, lower ccm_k0 means better resistance to countermeasures (flares or chaff, since both IR and radar missiles use this parameter). Even setting ccm_k0 to a value very close to 0, the missile almost always goes for flares. So I go check the Igla and FIM92C (stinger) files, and the ccm_k0 is not even present! The conclusion is that the ccm_k0 parameter is not used anymore for IR missiles using the new seeker modelling. That leaves the seeker field of view as the only parameter effecting the resistance to countermeasures. Since ED chooses realistic FOV for their missiles (between 2° and 4°), that means that those missiles are modelled as if they haven't ANY algorythm to reject flares. Which is of course not realistic at all, and gives all those missiles a terrible resistance to flares. I don't know if the radar missiles are also affected, but I think it would be worth to check for those too, especially the ones using the new radar seeker modelling.

+1, would be a great option, especially to create more realistic saturating attacks!

I asked the Gazelle developper about it, and here is his answer: "Hey there, In the current version, the gazelle still uses our own Mistral. This has to do with the fact that the ED mistral has had a sound issue that was only resolved last week. We plan to switch to the ED missiles in an upcoming update." And here is a pic of the upcoming updated launcher tubes he attached to his message:

Weight is definitely in the equation for terminal velocity of an object. As said above in the thread, terminal velocity is achieved when air resistance is equal to the force of gravity applied to the object (this force is also called... weight). And the weight is propotrionnal to the mass of the object (9.81 * mass, to be exact). That explains why 2 objects having the exact same shape, but different masses, will have different terminal velocities. To come back to the bombs, they generally have a high density and a better aerodynamic shape than bullets, being more elongated, so it really wouldn't surprise me if their terminal velocity is supersonic.

I believe it's normal behaviour, if you pull fast on the collective, it'll increase the G-load on the rotor system, inducing a stronger cone effect (blades angling up) and creating a transient high rpm effect for a few seconds because of conservation of momentum. After a few seconds, the increased G-load, demanding more power from the engines to keep the rotor spinning, takes over and the rotor rpm will deacrease again (to the initial 101% value, or lower if you're asking too much from your engines because you carry a heavy payload)

@Маэстро do you have news about the issues I mention in my last message, in particular the Gazelle using another file than the one you thought?

I'll provide tracks later. If you want to test it on your side, you just need to put 2 AI against each other at long range.

The improved BVR AI is nice, my only grip with it is it never anticipates missiles launches. It only begins to crank once it has fired its own missile. If the enemy plane has longer range, or even same range missiles, it will always defend too late and die. The AI should take into account what the enemy plane is, what missiles it may carry, and plan accordingly (try some preemptive defending, crank, low alt, going cold and hot again...)

Thank you for the feedback! I think there is a misunderstanding though. According to the lua exporter by Quaggle, there are 2 different versions of the Mistral missile in the DCS files. The file "./CoreMods/aircraft/AircraftWeaponPack/AA_Missiles.lua" indeed uses a value of 4 degrees for the FOV, and a value of 2.2 seconds for the burn time. HOWEVER, the Gazelle in DCS uses the file located in "./CoreMods/aircraft/SA342/SA342_Weapons.lua", where the FOV is 14 degrees, and the burn time is 3 seconds. A few more values are different between the 2 files. That's why in my first post I said that I used the "./CoreMods/aircraft/AircraftWeaponPack/AA_Missiles.lua" file as a basis for my own lua version of the Mistral. EDIT: also, I have a strong suspicion that the parameter of resistance to flares "ccm_k0" isn't working, and has no real effect. No idea if this could be specific to the Mistral, or if it's the case for all IR guided missiles.

The whole jamming environment needs an overhaul in DCS. Burnthrough range is always the same for all aircraft and radars, no matter target RCS, radar power, jammer and radar technology... the AMRAAM never lofts even if the target in within burnthrough range, the AMRAAM seems to remain in hoj mode without trying to get a radar lock (or has no burnthrough range modelled)...

Oops, I forgot a doc yes, here it is! It is written: "Une fusée de proximité qui a pour but de donner l’ordre de mise à feu de la charge militaire soit à l’impact, en détection de cible à proximité, soit en autodestruction." Translated: "A proximity fuse meant to trigger the warhead, at impact, or if a target is detected at proximity, or for self-destruction" For the parameters "SeekerSensivityDistance" and "sensitivity", hard to propose a value without knowing what they do exactly. All I know is that in "Janes Land Based Air Defenses 1992-1993" (I posted screenshots of the relevant pages in my 1st post), it is stated that the Mistral should be able to lock on a non afterburning jet at at least 6km, and on a combat helicopter with heat reduction systems (IR suppressors) at 4km. It is also written that the Mistral seeker is 3.5 times more sensitive than the Magic II seeker. In the Magic II lua file, the value used is SeekerSensivityDistance = 20000 (vs 10000 in th Mistral lua). There is no "sensitivity" parameter in the Magic II lua since it doesn't use the more modern api with a modelled seeker). Last time I tested the Mistral the lock distance seemed quite close to what is specified in "Janes Land Based Air Defenses 1992-1993", so I suspect that it's more the Magic II seeker beeing too sensitive in DCS. Poste_de_tir_Mistral.pdf

Yes I confirm this has been fixed and works correctly in the latest open beta

I up this topic, since Casmo just posted a video showing what rapid fire looks like, and how quickly you you destroy targets. George could do that with what I propose above.

Wait, you mean that as soon as a target is jamming, the AMRAAM uses hoj mode only? And won't ever try to use active seeker?

Hi there. The Mistral missile suffers from several issues and inaccuracies, making it severely underperform. Here is my work, using solid sources, to correct the missile. Note that DCS uses the Mistral located in "./CoreMods/aircraft/SA342/SA342_Weapons.lua" (by Polychop) which is encrypted. So I used and modified the Mistral located in "./CoreMods/aircraft/AircraftWeaponPack/AA_Missiles.lua" (by ED, and with only a few different values compared to Polychop's Mistral) to make my new Mistral version. Here are all the changes I propose, and the sources I've used: - the seeker field of view used in DCS is 14°. That is a ridiculous value. It's 7 times what is used for the stinger or the Igla (2°). Since a larger field of view means that flares stay longer in the field of view of the missile, that also gives an atrocious resistance to flares to the Mistral. That also leads to situations where the Mistral will switch to another aircraft while it's quite far from the intended target... In real life, the Mistral is described as having a very narrow field of view, combined with advanced algorithms, giving it an excellent resistance to flares (Janes Land Based Air Defenses 1992-1993; SAFRAN document on Mistral seeker). I recommend a value of 1°, given the Mistral is described as having a better seeker than the Stinger and Igla in "ANALISIS DEL MISIL MISTRAL INFRARROJO COMO SISTEMA ANTIMISIL Y ANTIAEREO", and the best kill probability of all MANPADS in "MANPADS A Terrorist Threat to Civilian Aviation?" (page 39). - the proximity fuse has a radius of 5 meters in DCS. The document "ANALISIS DEL MISIL MISTRAL INFRARROJO COMO SISTEMA ANTIMISIL Y ANTIAEREO" cites a proximity fuse radius of 2 meters, more plausible for a missile with a 3kg warhead. - the value of K = 2 used in DCS in the autopilot part, is too low. For a short range missile, many articles explain that a value of K between 3 and 5 gives better results (see the attached article "Proportionnal Navigation Guidance"). The AMRAAM in its terminal phase uses K = 4 in DCS. Based on my tests, a value of 5 gives the best results for the Mistral (https://cdn.discordapp.com/attachments/925038843399376917/955940311354454056/K_effect.mp4 ). And I know it's not definitive and documented proof, but one would wonder why MBDA would not give the better guidance to its missile... - the flag_dist = 150 parameter is too high. It leads to the missile brutally diverging from its target if the target is flanking (see example K = 2 in the video). A value of 50, like the one used for the Igla, combined with K = 5, gives the best hit probability (see example K = 5 in the video). - the GimbLim parameter has a value of 30°. I propose a value of 38°, based on "Janes Land Based Air Defenses 1992-1993", which cites a gimbal limit of 38° for the Mistral 1, and the SAFRAN Mistral missile seeker document citing a >30° gimbal limit. - for omega_max, which is the maximum angular tracking rate of the seeker, I propose a value of 20°/second, as stated in the SAFRAN document about the seeker. - The rocket burn time should be about 2 seconds, rather than 3, based on video evidence ((https://youtu.be/KK9_Iq0AM9k?t=15 ; https://youtu.be/RY78aBlUkMQ?t=27 ). - I add that the missile should self destruct after 14 seconds of flight time (fuse confirmed to self-destruct in the document "poste de tir Mistral", and 14 seconds max flight time cited in "ANALISIS DEL MISIL MISTRAL INFRARROJO COMO SISTEMA ANTIMISIL Y ANTIAEREO". - As a bonus, it would be great to see tha shaped trajectory added. The Mistral is designed to have a shaped trajectory (it climbs a bit just after launch, see photo attached). This is especially important to engage enemy helicopters, as it allows the missile to continue to track them even if they try to terrain mask. You'll find attached a lua file, which is the ED's Mistral file with all the parameters corrected with the values I propose. Some things I haven't included in the lua are the self-destruction at 14 seconds, and modifying the "SeekerSensivityDistance" and "sensitivity" parameters to match a 3.5 times better sensitivity than the Magic 2 missile seeker (according to Janes Land Based Air Defenses 1992-1993. That's for the Mistral 1, but the Mistral 2 we have in DCS basically has the same seeker.) SAFRAN_Mistral_Seeker.pdf ANALISIS DEL MISIL MISTRAL INFRARROJO COMO SISTEMA ANTIMISIL Y ANTIAEREO.pdf Proportional_Navigation_Guidance.pdf New_mistral.lua

Yeah, would love to see George use LMC or IAT when own helo and/or target are moving so that he misses less in those situations.

We need more realistic stats for many ground units. Examples: in DCS, AI tanks can't hit anything when on the move (try to make them shoot at an enemy tank 2km away while moving at 40kmph...). Try driving a tank and do the same: really hard to hit too. In real life, during exercises modern MBTs have more than 95% hit rate in the same scenario! Another thing about tanks: reverse speed are all wrong! In DCS tanks reverse speeds are less than 10kmph. In real life they're between 30 and 40kmph! This is super annoying when you're trying to do real life tactics, such as shoot and scoot, or reverse in front of an obstacle. Finally, the artillery needs a global overhaul. The accuracy (AI or player controlled) is often way worse than in real life. When aying an arty unit, the fire control is super, super basic and could be improved to allow more accurate shots. Finally, the AI arty could be created (I say created because it doesn't exist) because AI arty never, ever fire on their own, even when friendly units have spotted enemies well within range...

Just did a test with an AI AH-64D block II: the AGM-114K it has fired continues to perfectly track the enemy tank, all while the Apache was destroyed only a few seconds after the missile was fired. No other unit lasing the enemy tank. Track attached. PS: I haven't tried if it works for Hellfires launched by players, if someone can test that it would be great! PS 2: to add a bit more details, once the AGM 114K loses the laser spot, the missile seeker should continue to point at the last laser spot location, ad the missile should align itself with the seeker to go straight toward the last laser spot location. That's the method used to reacquire the laser spot in case the missile climbs too high in the clouds and doesn't see the laser spot anymore. Source attached (public). bug_AI_hellfire_continues_track.trk hellfire_article.pdf

The main problem I have with the current BVR ai is that they always work the same way: Go straight to the enemy and climb at 11000m, no matter the plane (planes like the Mirage 2000 would fly higher realistically). Then either they have time to launch a missile, and always crank and dive, no matter what the enemy plane is (why dive and lose altitude if the enemy plane has no BVR missile? Or way inferior ones?) Or, the enemy plane launches a missile first, and the ai only begins to evade when the missile goes active, which is usually too late. They don't anticipate a missile is probably already inbound based on the type, aspect, altitude and speed of the enemy plane.

+1 During my last mission George needed 3 Hellfire to kill a non moving T-72 located 6 km away.... I was in forward level flight, as stable as you can be...