Bug Report: F-15 gun accuracy does not match sources

[SinusoidDelta]

First off, the angular relationship between 8 feet at 1000 feet vs 8 meters at 1000 meters is, of course, the same. You were exactly correct SinusoidDelta. I'm tired. :)

 

Second, thank you very much for finding that video! You've just given us the parameters of the test and a valid original source to present alongside our findings!

 

For a Test

 

First off, if we can somehow get bullet hole textures to show up on a target model, this will be much easier.

 

If we can't somehow get bullet holes, we're going to have to use a target that cannon shells will pass through without bursting, place the game camera next to the target, slow down game time as much as possible, make every canon round a tracer (or otherwise visible), elongate the tracer objects to keep them visible long enough as they pass through the target to record, record the "impacts" on/through our target to a video, and use video editing software to advance through the video frame-by-frame to mark each impact point on the target.

People we need:

 

 

1. Someone who can make and texture models for DCS and adjust them as needed
   
2. Someone who knows how to hack the game to modify the A-10C to make it a more stable shooting platform on the ground
   
3. (If we can't have bullet holes) Someone who can make every cannon round a tracer, slow the game time down more than currently possible, and possibly elongate the tracer models/effects to allow more time to mark them
   
4. Someone familiar with video editing software and willing to go frame-by-frame through a few minutes worth of slow motion impact footage to mark all of the outermost hits
   

 

Things we need:

 

1. A way to accurately measure distance in game.

 

 

• The most desired range would be 25.4 meters, as that is the shorter of the two distances used in the test video above. Other ranges could be used, we'll just have to convert the results so they can be compared with the ranges used in the video. Closer ranges should make for a more accurate test and are preferred.
  
• If the TGP is used for ranging, we'll have to adjust for the range error caused by the difference in range between the TGP mounted on a wing station and the gun's muzzle. This can be figured close enough using known dimensions and/or a little measuring with photoshop.
  

 

2. We need 2 indestructible, immovable, target models.

 

 

• If we can't have bullet holes, we need a target a bullet can pass through instead of exploding on.
  
• One 8 mil x 8 mil target with a "circle" texture that touches all four sides. (8 mil square with circle on it that touches each side = 8 mil circle.)
  
• One target about 16 mil x 16 mil. The bigger target would be placed behind the smaller target to record shots falling outsize the 8 x 8 mil target.
  

 

Once we know the range we're going to shoot the target from, we can then make the target dimensions match 8 mils at that range.

 

Since we can look up its dimensions, the tail of an aircraft parked next to our target in-game or in a model editor should be good to establish scale.

 

3. We may need to experiment with ways to immobilize the A-10C testbed. I don't know what's possible here, so I'll suggest hacks to increase mass and stiffen suspension or perhaps to over-weight the aircraft with ordnance and/or place a blocking object behind it to limit rolling and vertical stringing due to recoil.

 

4. If there is no way to get bullet holes, we'll need to record the firing with the in-game camera at a distance and angle that makes the points at which individual cannon rounds pass through the target as clear as possible. (No need to mark hits falling in the center 75% of the target circle.)

 

 

Or, someone with a connection to ED can simply tell us what real world values the numbers for the Da0 in the shell table lua equate to before I go stark raving mad, burst out of my apartment making airplane noises, and doing this in front of my neighbors: :pilotfly:

 

Haha! Thanks for the detailed reply. It is maddening trying to derive the units. I asked a few people and no one seems to know what they are. You described basically my image for testing in DCS.

 

1) I think placing the aircraft in mission editor should allow accurate enough distance measurement. In mission replay we can slow time down I forget the key binding atm) to the minimum and see if that allows us to see the round in flight.

2) This is outside my area of expertise. Maybe someone can chime in who is familiar with modeling. I could create a simple texture but I don’t have 3DS max to create new objects.

3) I can try immobilizing the A-10 or F-5E suspension config lua and increase the braking force / tire friction coefficients. A hackish way to increase the mass maybe to alter the weight of stores to some extremely large value. . Also I think it may be possible to make every round a tracer through the shell lua file.

 

I’ll give it a shot tonight. No pun intended.

Edited December 19, 2017 by SinusoidDelta

[GGTharos]

Again.... F-5Es can be chocked. You can do so without playing around with LUAs etc.

[SinusoidDelta]

:smilewink:

Again.... F-5Es can be chocked. You can do so without playing around with LUAs etc.

 

Roger that.

 

So we just need something to shoot at ;)

 

Edit: Does the F-5E have a ground safety or breaker? I’ll have to look into that first.

Edited December 19, 2017 by SinusoidDelta

[GGTharos]

Unfortunately that I don't know. There's a CB panel right in front of the stick though.

[Shadow KT]

if you play around with the rpm(engine) on the Hog, you can probably make it stay still while firing and holding the brakes

[Aries144]

So the dispersion is wrong, but how wrong.

If it's not 80% inside the specified radius and 20 % outside, I haven't seen anyone say what the proportions are ... (as 20% are allowed to be outside, and some can be quite a long way outside)

 

The above-linked video shows an m61 cannon fire a burst with a dispersion smaller than 8 mil. The video goes on to state that, at a range of 1000ft, 95-99% of rounds were within the 8 mil bullseye. We can confidently say that 8mil, 80% is a minimum standard. In other words, 8 mil, 80% is what the gun would shoot when it's completely worn out and on the verge of needing parts replacement.

 

The easiest way to handle this is to just go with 8 mil and throw out the 80% entirely.

 

Or, if that isn't precise enough, there exists a formula that can estimate how far outside a given dispersion area rounds in that 20% could be. I've seen it but can't remember where. It isn't very far outside the circle. Maybe 4 mils for the outermost 0.5%. I'll try and dig it up.

 

In the mean time, unless someone can talk to ED and find out the what real-world figure "Da0 0.0022" in the shelltable.lua represents, we're going to have to rig up a test that's going to take a lot of hacking and work. I'm still not convinced that it isn't directly related to mils.

 

Something else that might work is to use the gunsight in the F-5. If it's accurately modeled, it has a diameter of 40 mil.

 

Again, by far the easiest way to handle this would be to have cannon rounds spawn "bullet hole" textures on the spots where they touch our target, similar to what's done in first person shooter games when a player shoots a wall.

[Ironhand]

...find out the what real-world figure "Da0 0.0022" in the shelltable.lua represents...

Is it 0.0022 or 0.00022? I ask because 8 mils is 0.000203 meters. And 0.00022 meters is 8.66 mils. If there's an extra zero it's close, though not exact.

[GGTharos]

Wait a sec, 0.000203 meters at what distance? The mil is angular. And yes, it is 0.0022.

[Aries144]

• The F-5's and F-15's guns both use the 0.0022 value.
  
• The A-10's gun uses a value of 0.0017.
  
• The Mirage is 0.0008
  
• the Harrier's gunpod is 0.0005
  
• the Mig-29/Su-27 is 0.0005
  
• the Su-25 and Su-25T are 0.0005
  
• the Mig-21 is 0.0007
  
• the Ka-50 is 0.0005 for HE and 0.0004 for AP.
  

 

You can modify the values in the file yourself and see the change in-game, it just won't pass integrity check for multiplayer.

 

The file is called "shell_table.lua" and is located in Eagle Dynamics\DCS World\Scripts\Database\Weapons

[Aries144]

Is it 0.0022 or 0.00022? I ask because 8 mils is 0.000203 meters. And 0.00022 meters is 8.66 mils. If there's an extra zero it's close, though not exact.

 

You mean deviation at the muzzle? I know the Operation Flashpoint/Arma games use a similar method for dispersion value.

[Ironhand]

Wait a sec, 0.000203 meters at what distance? The mil is angular. And yes, it is 0.0022.

No idea. :) I keep a bunch of conversion tools on speed dial and one of them is a mil to meter converter. I know it's not the same thing but, just for chuckles, I input "8" and out popped 0.000203. At first I had misread Aries144's post as having the extra zero and got all excited. Then I decided to toss this out there, anyway.

 

Be interesting if adding an extra "0" to the F-15's number, gave the correct spread and it turned out to be a typo in the LUA.

[Aries144]

Be interesting if adding an extra "0" to the F-15's number, gave the correct spread and it turned out to be a typo in the LUA.

 

Entirely possible. Typos in code can cause hours of hair pulling frustration! It could make sense too that Belsimtek would simply copy that value for the F-5 without checking it, as sources indicate that the M39 cannons on it have the same dispersion as the M61A1 of the F-15. It doesn't explain the A-10 gun's value though, as that is also too large.

[GGTharos]

Going with the idea that this may have been a weird way of writing mils (should be 0.008) I input 0.0008 into the M61 shell definition.

 

The result was a laser; spread was so minimal you could theoretically snipe a single soldier standing a meter away from another. I say theoretically because the moment you pull the trigger

you get 3000lbs of recoil yawing the aircraft.

 

Based on non-terribly scientific testing I concluded that 0.0015 was close enough to the 8 mils, but it was hard to control for all the variables.

0.0022 definitely looks like it is too much. 0.0008 is far too little. I suspect the other guns need to increase spread, while this one and the M-39, as well as GAU-8 needs to decrease it.

 

No idea. :) I keep a bunch of conversion tools on speed dial and one of them is a mil to meter converter. I know it's not the same thing but, just for chuckles, I input "8" and out popped 0.000203. At first I had misread Aries144's post as having the extra zero and got all excited. Then I decided to toss this out there, anyway.

 

Be interesting if adding an extra "0" to the F-15's number, gave the correct spread and it turned out to be a typo in the LUA.

[PeaceSells]

I have no idea how far the in-game spreads are from the real thing, but I find it considerably easier to hit the enemy using the F-15 gun than the Su-27 gun. It feels like the Su-27 gun in-game has way too little spread.

[esb77]

All sources I can find indicate that the F-15's gun has a minimum accuracy of 8 mil. It is currently much larger in the sim.

 

1 mil equals a 1 meter diameter at 1km, or 10cm at 100m. Thus, 8 mil would be an 8 meter circle at 1 km or an 80cm circle at 100m.

 

If the lua file entries equate the way they appear to, 0.0007 for the Mig-21, 0.0005 for the Mig-29/Su-27, etc, then the 0.0022 entry for the F-15 is clearly too big and would seem to indicate its current performance is about 22 mil.

 

Can we please get this fixed?

 

In the style of math that I was taught in school, when performing arithmetic operations in base 10

 

1000 * 0.0022 = 2.2

 

Not 22.

 

By the LUA numbers you cite the F-15's gun has 3 to 4 times the dispersion of the Russian 30 mm guns, not 40 times.

[Aries144]

In the style of math that I was taught in school, when performing arithmetic operations in base 10

 

1000 * 0.0022 = 2.2

 

Not 22.

 

By the LUA numbers you cite the F-15's gun has 3 to 4 times the dispersion of the Russian 30 mm guns, not 40 times.

 

Right. 22 mil would be 3-4 times larger. The Russian planes seem to have accuracy of 5-7 mil, if that interpretation of the lua number is correct.

 

Where are you getting 40 times larger from?

Edited December 20, 2017 by Aries144

[esb77]

If the LUAs are in radians, which they may or may not be:

 

0.0005 radians * 1000 milliradians/radian = 0.5 milliradians

 

0.5 milliradians is equivalent to one half mil, not 5 mils.

 

22 mils, which is what you were citing, is 44 times 0.5 mils, which is what your interpretation of the LUAs being in radians would yield if the math is done correctly.

 

I discovered this because I wanted to convert to MOA, and I couldn't get your source numbers to generate a MOA reading that was equal to 22 mils.

 

I got as far as differential equations in math, so of course I no longer really trust myself to remember simple things like arithmetic correctly, but for what it's worth, my expensive graphing calculator agrees with me that you're multiplying by 10000 when you should be multiplying by 1000.

 

The math as done by a Ti-89

 

0.0005 * 1000 = 0.5

0.0007 * 1000 = 0.7

0.0022 * 1000 = 2.2

 

 

You are being consistent though, so when you start multiplying by the correct 1000 conversion factor when going from radians to milliradians, you should start getting consistently correct results.

 

 

You didn't specify in the first post, so I though you had just misplaced the decimal when looking at the F-15, I didn't realize you were doing it for all of them. So I compared a correct value of 0.5 mils with your incorrect value of 22 mils to get a roughly 40 times difference in dispersion.

 

 

As far as the LUA values go, in whatever unit they really are, 3-4 times more dispersion compared to a round that's 50% greater diameter is within the realm of reasonable, especially when you get out to 1000 m or greater ranges, but a lot of that is also gun construction and ammunition quality, so a dispersion for a 20 mm that's half of a 30 mm's is also possible.

 

I was just trying to very indirectly point out that if your theory that the LUA values are in radians is correct, then if the math is done correctly, a value of 2.2 mils is well within the expected amount of 8 mils for the F-15's gun.

[SinusoidDelta]

According to that logic then, a 30mm (1.2ish inch) cannon .5 mil 100% circle would mean all rounds fly through a 6 inch hole at 1000 feet?

[GGTharos]

I believe the sight and the eagle's better high AoA control may have a little more to do with that, but that's just a feeling.

 

I have no idea how far the in-game spreads are from the real thing, but I find it considerably easier to hit the enemy using the F-15 gun than the Su-27 gun. It feels like the Su-27 gun in-game has way too little spread.

[esb77]

According to that logic then, a 30mm (1.2ish inch) cannon .5 mil 100% circle would mean all rounds fly through a 6 inch hole at 1000 feet?

 

Approximately yes, though if you're measuring from the center of the round the width of the rounds might make it about 7 inches.

 

It works out to something like 1.7 MOA, and a good sniper or target rifle will yield around 0.5 MOA, (or if you're lucky and spend a lot of money maybe even as low as 0.25 MOA).

 

So it's not superb accuracy in terms of what's technically possible, but 1 - 2 MOA is achievable for mass produced small arms, and 20 and 30 mm are still close enough to that size so that it wouldn't be a particularly astonishing level of precision.

 

In most cases though, on the aircraft between engine, aerodynamic, mechanical from auto-fire, and other sources of vibration and flex, I'd really expect that stability of the platform is going to be more of a limit on how precise your groups are than the quality of the gun and ammo once the manufacture of the weapon is reasonably high quality.

 

 

Edit: Honestly, up to a point especially with a high rate of fire gun like the Vucan, bigger less precise groups are potentially an advantage. When shooting down birds on the wing there's a reason shotguns are preferred over rifles. ;)

 

The Ka-50 has good reason to want a real laser gun of a cannon to put rounds into armor where it hurts, but for the planes 0.5 mil precision may be a bit too much of a good thing.

Edited December 20, 2017 by esb77

[Aries144]

Edit: Esb77 I see what you're saying, but those figures don't match what I'm seeing in the game. The dispersion is certainly much larger than 2.2 mils.

 

A larger spread being an advantage or not, the fact is that the sources all state 8 mil is the precision of the gun. This is a simulation, so we want it to be as accurate as possible.

 

I just finished recording a track of the F-5 firing a single gun with all rounds being tracers. The aircraft is airborne, loaded with bombs to help stabilize it, and a mountain peak is aligned with the 6 o'clock pip on the gunsight for reference. All shots are fired in a single burst. This is very effective because a given round reaches the height of the sight line so quickly that there isn't much time for the aircraft to yaw from recoil. We also have the mountain peak to use as a reference if we want to be very precise.

 

By measuring only in the horizontal and only as the rounds' reach the sight line to eliminate error from the rounds' drop, it looks as if the Extreme Spread is, in fact, very close to 22 mil. Rounds can be seen going as far out as halfway between the center pip and the outer ring, both left and right. That's very close to 22 mil.

 

I'm going to attempt to get hold of someone with recording software so that they can advance the track footage frame-by-frame and mark the high point of each round's flight through the sight. This, combined with different values entered in the F-5's Da0 entry, and measured against the 40 mil gunsight ring, will give us a very accurate estimate of what the Da0 value means in relation to mil.

 

Next experiment will be to set the Da0 value to 0.0040 and see if the outer-most shots closely reach the outside of the 40 mil ring. If they do, frankly it's so close that further experimentation will just be academic.

 

Whew! That's a lot easier than the previous method I was looking at. I may try another experiment using the A-10C firing from the ground with the recording taken through the gunsight of an F-5, also on the ground, near the A-10C.

Edited December 20, 2017 by Aries144

[GGTharos]

I'm going to attempt to get hold of someone with recording software so that they can advance the track footage frame-by-frame and mark the high point of each round's flight through the sight. This, combined with different values entered in the F-5's Da0 entry, and measured against the 40 mil gunsight ring, will give us a very accurate estimate of what the Da0 value means in relation to mil.

 

I don't know how this can be accurate when your eagle is hauling butt towards the rounds ... ie. the perspective will be distorted.

[Aries144]

Just tested again with Da0 value set to 0.0040. It does indeed produce an extreme spread that very nearly touches the insides of the 40 mil sight ring.

 

Only a small percentage of rounds are going that far out. As an imprecise observation, I'd say that 50-60% are falling within 20 mils with the rest spread out between 20 and 40 mils.

 

Pending more precise testing, it's a safe assumption at this point that using 0.0008 would produce very close to the source value of 8 mil.

Edited December 20, 2017 by Aries144

[Aries144]

I don't know how this can be accurate when your eagle is hauling butt towards the rounds ... ie. the perspective will be distorted.

 

Can you explain to me why this would be?

 

The rounds velocity will be it's own plus that of the launching aircraft. The dispersion angle is still the same...

 

Regardless, this demonstrates the relationship between the lua value and mil. Multiplying the lua value by 10,000 produces an effect conspicuously close to the appropriate mil value, as measured using the F-5's 40 mil gunsight.

 

Modify the lua so that all fired rounds are tracers, go into the game, and see it for yourself.

Edited December 20, 2017 by Aries144

[Aries144]

So the dispersion is wrong, but how wrong.

If it's not 80% inside the specified radius and 20 % outside, I haven't seen anyone say what the proportions are ... (as 20% are allowed to be outside, and some can be quite a long way outside)

 

Please check the first post. Essentially, the 80% isn't important to model, as 8 mil, 80% doesn't represent the typical precision of a new gun, but instead the failure point beyond which a gun is considered too worn and requires repair. The linked M61 accuracy test video on the first post illustrates this, as the vast majority of the 100 shot burst seen on target in the video visibly impacts entirely inside the 8 mil circle, with the narrator stating later that only 1-5% of shots were striking outside of that.

 

Even without a more precise, scientific test, it can clearly be understood from my included Da0 = 0.0022 track (the current default) that far more than 20% of shots are exceeding 8 mil.

 

Also:

 

Tracks didn't work. Here's some raw video recorded with ShadowPlay instead.

 

Default accuracy, Da0 = 0.0022

40 mil accuracy, Da0 = 0.004

 

8 mil accuracy, Da0 = 0.0008

Edited December 22, 2017 by Aries144