Flight model issues - acceleration/top speed discrepancies esp. at 6+km altitude, T/D discrepancies at sea level [RESOLVED]

[rossmum]

All charts are sourced from SFI AJ 37 del 3, 1979 print. Test flights were flown in ISA, using weights as close as possible to those described for each loadout (remembering that the AJS is somewhat heavier than the AJ), either start weight or flight weight depending on what the chart in question described.

LEVEL ACCELERATION

Tests were run at 0km, 6km, and 11km in ISA conditions only. Each test included two runs (one by myself, one by Airhunter) with clean config/rent FPL. In addition, I ran a second set of 0km tests with a group 3 loadout (4x ARAK + XT), and 6km tests with a group 4 loadout (4x SB + XT). Two tests were run at 11km with the loadout stipulated in the manual, one by myself and one by Airhunter.

Airhunter provided me with tacview files and graphs which I then charted against the manual (obviously only for the clean config 0km and 6km tests, and the 11km test). On my own end I forgot to save tacviews, but did save trackfiles, which are attached as well.

The results show what I'd found myself:

At sea level, mil thrust is maybe a touch slow to accelerate to M 0.9 but is still relatively close, and zone 3 is tuned almost on the dot. Zone 1 and 2 show discrepancies, particularly zone 1, although they aren't very large at this altitude.

Excuse the rather unscientific graphs, but unfortunately just overlaying the ones from tacview wasn't really workable and I'm working with what I've got here:

At 6km alt, mil thrust gradually diverges from the chart, ending up a little slower than expected. Zone 1 and 2 show large discrepancies. Zone 1 follows zone 2's curve, zone 2 follows zone 3's curve, and zone 3 is doing something else entirely.

At 11km alt, only zone 3 was tested. The results were very far off the expectation: according to the chart, level acceleration from M 0.9 to M 1.6 should take about 5 minutes and 45 seconds. In the sim, this was achieved in just one minute (!), while acceleration from M 0.9 to M ~1.63 (as far as the line is drawn) should take about 7 minutes and 15 seconds, while in DCS it took 1 minute and 15 seconds. This is a very large discrepancy even compared to the aircraft's own charts, but also compared to acceleration profiles of other aircraft known to have phenomenal performance in this area (F-104, MiG-23, MiG-29). Something is clearly causing an issue here but I don't know what.

Big thanks to Æck for spotting this one during a MP session and bringing it to light - I wouldn't have thought to test it otherwise.

Finally for now,

THRUST/DRAG EQUILIBRIUM SPEEDS AT SEA LEVEL

This has been something I've been meaning to report for a while, but invariably by the time I got trackfiles, DCS would update and break them. As a result I'll just post the results of my testing now and drum up the trackfiles when I get time. Please note that this is separate to, but was compounded by, the now-fixed drag issues with Sidewinder launch rails.

Each test was run using the circled loadout group from the AJ 37 loadout tables - so rent FPL was clean, group 1 was KA-24-XT-24-KB, group 2 was KA-24-XT-blank-04, group 3 was RA-RA-XT-RA-RA, group 4 was SB-SB-XT-SB-SB.

All loadouts except group 4 (bombs) reach thrust/drag equilibrium above the intercept of MAX ZON 3 lines and thrust/drag lines, some significantly so. Group 4 is slightly slower than expected, the clean airframe flies beyond even SAAB's estimate for a world speed record using a stripped and polished modified aircraft, groups 1 and 2 are beyond the aircraft's Vne and fall off the chart, and group 3 (which should be the slowest according to the chart - just barely subsonic, likely due to the rocket pods generating enormous amounts of transonic drag compared to anything else tested) is sitting slightly above the aircraft's Vne. It seems drag values need looking at.

I don't have trackfiles handy for these tests (the ones I did have are now several DCS patches old), but I do have tacviews for them. I can get tracks again if necessary.

I haven't checked climb performance yet, but hopefully there won't be anything to add for that.

OWN TRACKFILES.zip Tacviews from Airhunter.zip drag tests.zip

Edited December 17, 2022 by rossmum
typo fix

[Spirale]

Thx for this tests Rossmun

[renhanxue]

Excellent work, this must've been a ton of effort! Weird to see that the top speed is still this extreme, it's been well established that it's too high for years now. The clean aircraft being extremely fast at sea level is one thing, but doing well over M 1.1 with the draggiest loadouts in the arsenal is just wildly off. Some of the weapons are clearly not modeled correctly, and the 11km test is just nonsensical.

I'm almost certain the thrust-to-drag diagram doesn't considers fuel consumption, so in reality the aircraft should probably do slightly better than the diagram as fuel is consumed (the diagram is calculated at 100% fuel for loadouts with the drop tank mounted), but that doesn't come anywhere close to explaining the huge differences here - the benefits of slightly lower alpha are insignificant compared to the enormous transonic drag from the stores.

Edited May 29, 2022 by renhanxue

[rossmum]

I would've liked to draw up curves for fuel burn and distance covered as well, just haven't got time at the moment. I figure fuel burn in particular might help solve what's going on.

Do you know if there's much hope for actual notes from flight testing to double check the numbers against? I ask because in the case of the 21 it turned out the charts were quite conservative, though not by a large enough margin to explain what's going on here. More numbers is always better, even if it means more work 

[MYSE1234]

5 minutes ago, rossmum said:

I would've liked to draw up curves for fuel burn and distance covered as well, just haven't got time at the moment. I figure fuel burn in particular might help solve what's going on.

Do you know if there's much hope for actual notes from flight testing to double check the numbers against? I ask because in the case of the 21 it turned out the charts were quite conservative, though not by a large enough margin to explain what's going on here. More numbers is always better, even if it means more work 

Fuel usage is already reported being inaccurate internally.
You finished the acceleration stuff before I did (only got one chart done so far). Now I won't have to do that, so thanks for that; Saved me quite a few hours of work there (as I'm sure you are aware )

[renhanxue]

7 hours ago, rossmum said:

I would've liked to draw up curves for fuel burn and distance covered as well, just haven't got time at the moment. I figure fuel burn in particular might help solve what's going on.

Do you know if there's much hope for actual notes from flight testing to double check the numbers against? I ask because in the case of the 21 it turned out the charts were quite conservative, though not by a large enough margin to explain what's going on here. More numbers is always better, even if it means more work 

There's almost certainly data from actual flight tests in the national archives, the problem is finding the right place to look. As someone with a decent amount of experience in Swedish military archives, I'd say it could easily take a year to find something like that - weeks of digging through indexes to find likely places to look, then months waiting for declassification and weeks to dig through the results, then repeat because you probably looked in the wrong place the first time but now you know more about how the archive is organized.

I've been kinda itching to get back to archive diving, but unfortunately the archives have moved so they're way less convenient for me to get to. I used to be able to spend an hour or two on Wednesday evenings after work there (they were only open evenings one day a week), but now it's like 45 minutes just to get there.

[CaptCrash]

I'm sure @IronMike will appreciate the info. Nice work @rossmum

[Exorcet]

It's really crazy that the Viggen has overperformed like this for so long. When I found out about the issue I shelved the module, and that was years ago.

[Sideburns]

Neat, some really good "just the facts" investigation here and good to see this finally formally reported. Hopefully these drag and speed issues can be addressed as part of finishing the Viggen off and soon.

[StarLiner]

+1, hope this gets looked at soon.

[MYSE1234]

On 5/29/2022 at 5:13 AM, rossmum said:

.... group 3 was RA-RA-XT-RA-RA, group 4 was SB-SB-XT-SB-SB.

Just seen that you listed group 3, and possibly 4 slightly, incorrectly.
Group 3 should be: "KB-RA-XT-RA-RA"
Group 4 might be: "SB-SB-XT-24-SB"
So the results you got is in fact with the aircraft even draggier/heavier than the aircraft would be in the charts from the manual. 

Regarding group 4 it's both listed as "16x SB + XT", as well as "SB-SB-XT-24-SB" depending on where you look. I personally think the latter is what is used, but that is of course just a guess and not 100% certain.

[rossmum]

Good catch - I'll run those again over the weekend and see what the difference works out to be. My eyes go square when looking at tables.

[rossmum]

Revised figures for the T/D equilibrium - group 3 (now using correct KB-RA-XT-RA-RA) 1391km/h IAS or M ~1.14; group 4 (now SB-SB-XT-24-SB) 1325km/h IAS or M ~1.08. This means that group 4 is also too fast by a fair margin.

AJS 0km drag equilibrium GRUPP 4.trk AJS 0km drag equilibrium GRUPP 3.trk

[Hummingbird]

Yeah, bug reported this about 3 years ago when I noticed some pretty extreme climb performance:

 

 

Viggen is basically a rocket until its engine dies around ~15+ km height, not really losing any speed until then. Had the engine not died, it seems it would've happily gone well into the stratosphere

Meanwhile the F-15 is, as it should, losing speed gradually with altitude.

 

 

Edited June 4, 2022 by Hummingbird

[Harlikwin]

Well done report. 

[IronMike]

Thank you for the report, very well put together. We will most certainly look into, however I cannot promise how soon we can fix this. The Viggen code is our oldest, and in particular the engine is a fine balanced equilibrium of "trying hard" if that makes sense, for the lack of a better simplification. Time permitting, in our opinion, at some point a complete overhaul of the engine (no promise for now) would likely be better than bandaging the remaining issues. Either way it needs to be fixed of course, and we hope we can do so soon.

[Machalot]

5 hours ago, IronMike said:

Thank you for the report, very well put together. We will most certainly look into, however I cannot promise how soon we can fix this. The Viggen code is our oldest, and in particular the engine is a fine balanced equilibrium of "trying hard" if that makes sense, for the lack of a better simplification. Time permitting, in our opinion, at some point a complete overhaul of the engine (no promise for now) would likely be better than bandaging the remaining issues. Either way it needs to be fixed of course, and we hope we can do so soon.

But we have it on good authority that it only takes about 90 days to make a full fidelity model of a modern combat aircraft!  But to be fair, that doesn't include modeling vacuum pump failures.

[amalahama]

Any news on the potential FM rework?

[D4n]

Exactly. Come on HB, more and more servers are adding the feared "instantly explode if you accelerate above 1350 kph"-script... It's a terrifying development, and you know it best... It's rather shocking even, to be honest. And that's obviously DEFINITELY not gonna help you guys at Viggen module sales... Isn't that logical??

Edited September 8, 2022 by D4n

[D4n]

Whoa, I just discovered, Viggen can even pull over 11G (very very shortly), while often I'm strangely also stuck at 4-5G for several seconds even though I have stick full aft... (then it jumps up to 8G.). Strange

[renhanxue]

21 hours ago, D4n said:

Whoa, I just discovered, Viggen can even pull over 11G (very very shortly), while often I'm strangely also stuck at 4-5G for several seconds even though I have stick full aft... (then it jumps up to 8G.). Strange

11 or even 12 G should be attainable at low altitude in a fairly narrow high subsonic speed range (no guarantee the wings will stay attached, however). Above Mach 0.9 and/or above 8 km altitude or so, you start running into a complex combination of flight control system peculiarities that limit the max attainable load factor quite severely. At subsonic speeds the aircraft tends to run into 18° alpha limit first, but at higher speeds the flight control system itself can't always keep up. They fixed that on the JA 37, but for the AJ 37 it wasn't seen as a crippling flaw. See this old post for some more details:

 

Edited September 10, 2022 by renhanxue

[TOViper]

On 9/9/2022 at 11:39 PM, D4n said:

Whoa, I just discovered, Viggen can even pull over 11G (very very shortly), while often I'm strangely also stuck at 4-5G for several seconds even though I have stick full aft... (then it jumps up to 8G.). Strange

Just asking out of curiosity: During this situation, was your aircraft set to "Immortal"?

[D4n]

On 9/12/2022 at 2:31 PM, TOViper said:

Just asking out of curiosity: During this situation, was your aircraft set to "Immortal"?

Negative.

On 9/10/2022 at 9:22 PM, renhanxue said:

11 or even 12 G should be attainable at low altitude in a fairly narrow high subsonic speed range (no guarantee the wings will stay attached, however). Above Mach 0.9 and/or above 8 km altitude or so, you start running into a complex combination of flight control system peculiarities that limit the max attainable load factor quite severely. At subsonic speeds the aircraft tends to run into 18° alpha limit first, but at higher speeds the flight control system itself can't always keep up. They fixed that on the JA 37, but for the AJ 37 it wasn't seen as a crippling flaw. See this old post for some more details:

 

 

Okay! Thanks! Seems you linked 2 posts though, in your post. One from 11.9.2016, and also one from 6.12.2015.

[Mongolf]

Yes, The value on drag is far on external load, you have to change this, it is totally wrong now! Effect on engine si very close to original but external load drag value is totally wrong according to 4 viggenpilots that flew the origina. aircraft!!!