AWG-9 Does not have track memory.

[KenobiOrder]

So after some additional testing to figure out why the AWG-9 has been so temperamental over the last year (constantly losing tracks, generating false targets, unable to resolve not so closely spaced targets, etc) I am fairly certain there is a bug or oversight where the AWG-9 has no track memory. I had a friend fly a 300knot 3G turn in a constant circle. As the target enters the notch very briefly, the radar loses the contact dumps it either immediately or near immediately. So either the track memory of the TWS-A mode is nonexistent, or it is very very small. This is probably the main contributor to the issues plaguing AWG-9 performance in DCS because the radar doesn't retain contacts long enough to associate them with new contacts. The target is re-detected almost immediately after the radar dumps the track, but the radar does not associate this is the last track and instead generates a new one.

 

I dont have documentation on the AWG-9 but it seems highly improbable that the TWS radar modes would not have track memory, because it is essentially necessary to how such a system works. It would take at least one frame after last detection before the radar would flag the track file as missed, and it would then enter a track extrapolation mode with a correlation gate based on the possible target maneuvers. Based on other systems I do have knowledge of (all the other teen series fighters), this would last between 4-7 seconds. As it takes about 2.5 seconds to complete a frame, a target would have to avoid detection for 6.5-9.5 seconds before the radar would dump the track file and no longer associate a detection in the correlation gate as being the same contact.

 

@IronMike

 

 

 

It may take some time for the video to be full HD.

 

 

 

Edited August 22, 2021 by KenobiOrder

[Nozzle]

You didn't mention if your experience was online or in single-player but the DCS AWG-9 must contend with something the real one never did; internet lag.  I suspect the modeling of the radar track-file build logic is deprived of the required return hit-rate to maintain consistent tracks.

 

ED seems to have fixed this issue with their TWS radars (F-15, F-16 and F-18) but I don't know if the third-party vendors are provided the same insight.  This is my working theory at any rate.

[KenobiOrder]

15 minutes ago, Nozzle said:

You didn't mention if your experience was online or in single-player but the DCS AWG-9 must contend with something the real one never did; internet lag.  I suspect the modeling of the radar track-file build logic is deprived of the required return hit-rate to maintain consistent tracks.

 

ED seems to have fixed this issue with their TWS radars (F-15, F-16 and F-18) but I don't know if the third-party vendors are provided the same insight.  This is my working theory at any rate.

It was done on a private server. With just a necessary aircraft needed to test in the mission. I doubt latency influenced this at all, but I totally agree that its an issue in multiplayer in general.

[captain_dalan]

This could be a feature. Not an expert myself, but interviews with people who flew with or against it, seem to imply simple "check" turns were enough to "break" the TWS "lock". 

[KenobiOrder]

4 hours ago, captain_dalan said:

This could be a feature. Not an expert myself, but interviews with people who flew with or against it, seem to imply simple "check" turns were enough to "break" the TWS "lock". 

Possibly, but I would like to get dev confirmation of this if possible. It is hard to believe this is how the system worked because basically nothing else works like that. The tactical problem it creates would render the system next to useless and its fairly easy to correct for a huge (in fact necessary) performance improvement. Also if the system did not do this in general it would cause huge problems since you can have targets fade from scan to scan for any number of reasons. It is also strange that it fades so fast and then the radar seem to initiate a new track file immediately when the contact unfades. You wouldnt establish a track file on the first scan, it would be logged in memory and would be registered as a tentative track until the radar confirmed it by hitting it again.

 

There have been claims about check turns but we really dont know the context of these.  There are any number of factors that could cause this, in particular the duration and range of the turn. For example, if the turn is done far enough away, the reduction in SnR could result in a target fading long enough to escape track. Same with simply notching for more than about 6 seconds. But simply notching for a half second seems a bit absurd.

[dundun92]

28 minutes ago, KenobiOrder said:

nothing else works like that.

There is no TWS system of the era to compare it to though. All the other TWSs are much newer and built upon the AWG-9s original TWS

[Naquaii]

On 8/22/2021 at 3:53 AM, KenobiOrder said:

So after some additional testing to figure out why the AWG-9 has been so temperamental over the last year (constantly losing tracks, generating false targets, unable to resolve not so closely spaced targets, etc) I am fairly certain there is a bug or oversight where the AWG-9 has no track memory. I had a friend fly a 300knot 3G turn in a constant circle. As the target enters the notch very briefly, the radar loses the contact dumps it either immediately or near immediately. So either the track memory of the TWS-A mode is nonexistent, or it is very very small. This is probably the main contributor to the issues plaguing AWG-9 performance in DCS because the radar doesn't retain contacts long enough to associate them with new contacts. The target is re-detected almost immediately after the radar dumps the track, but the radar does not associate this is the last track and instead generates a new one.

 

I dont have documentation on the AWG-9 but it seems highly improbable that the TWS radar modes would not have track memory, because it is essentially necessary to how such a system works. It would take at least one frame after last detection before the radar would flag the track file as missed, and it would then enter a track extrapolation mode with a correlation gate based on the possible target maneuvers. Based on other systems I do have knowledge of (all the other teen series fighters), this would last between 4-7 seconds. As it takes about 2.5 seconds to complete a frame, a target would have to avoid detection for 6.5-9.5 seconds before the radar would dump the track file and no longer associate a detection in the correlation gate as being the same contact.

 

@IronMike

 

 

 

It may take some time for the video to be full HD.

 

 

 

 

 

15 hours ago, KenobiOrder said:

Possibly, but I would like to get dev confirmation of this if possible. It is hard to believe this is how the system worked because basically nothing else works like that. The tactical problem it creates would render the system next to useless and its fairly easy to correct for a huge (in fact necessary) performance improvement. Also if the system did not do this in general it would cause huge problems since you can have targets fade from scan to scan for any number of reasons. It is also strange that it fades so fast and then the radar seem to initiate a new track file immediately when the contact unfades. You wouldnt establish a track file on the first scan, it would be logged in memory and would be registered as a tentative track until the radar confirmed it by hitting it again.

 

There have been claims about check turns but we really dont know the context of these.  There are any number of factors that could cause this, in particular the duration and range of the turn. For example, if the turn is done far enough away, the reduction in SnR could result in a target fading long enough to escape track. Same with simply notching for more than about 6 seconds. But simply notching for a half second seems a bit absurd.

 

As far as we know that's how the TWS in the AWG-9 was. Saying nothing else works like that isn't fair, the TWS in the other aircraft in DCS are 10-20 if not 30 years newer. As far as we know the AWG-9 had no way to re-establish an already lost track with a new radar returns. That's not to say it hasn't got a track memory, it can miss a few returns and it will keep a track as dead reckoned until time-out. But it won't correlate that track to new tracks appearing after it's marked as lost.

 

I've said it before but I'm leaning towards this being the source for the AIM-54 "being useless against fighters" circulating on the Internet. The missile itself has no problems engaging fighters, it's the radar that has.

[KenobiOrder]

Thanks, guess that clears that up then.