Re: LNAV Tracking issues: A favor to ask...

[acbent123 78]

I was doing some reading on issues people were having getting the autopilot to work on the default 787 and A320. There was a recommendation to set a null zone of ~5% on your yoke/controller aileron axis as not having one was interfering with the autopilot. I tried this and I am currently in a flight with the CRJ and it is making the turns and tracking perfectly. The caveat is that I am currently flying east to west - not north or south. 

 

Could a few people with the tracking issues try putting a null zone of 5 - 10% on their aileron axis and see if it improves things? I am also going to try a small null zone on my elevator axis.

 

This may have been mentioned before and maybe it has been tried, but I did not see any reference to it in these forums. I am very cautiously optimistic. I will report back on how the rest of my flight goes.

 

Update: The rest of the flight was perfect. The plane seemed to anticipate the turns better and rolled out right on course. None of the turns were very sharp, but there was no wobbling back and forth. Tomorrow, I'm going to do a flight with a more northerly heading. Will report back. 

[acbent123 78]

Seems to be working.

 

I deliberately picked a flight with northern direction and a SID with a sharp turn. The plane executed the turns with smoothness and precision.  - absolutely no hunting. The plane is also much more stable in flight. I found before that it would rock a bit back and forth a wee bit even when flying a straight line. I thinking that the controller is always sending some form of signal that is messing with the A/P. The null zone seems to have eliminated this. Here are a few pics of my first on course turn. I will do a video later.

 

 

 

 

Please login to display this image.

 

Please login to display this image.

Please login to display this image.

[MarcusIsnard 0]

One way to confirm this would be to view the raw joystick input - I can look into the debugger and see if I can find a way to monitor these paramters. Can you give me the route you flew? 

[Richard Dastardly 108]

I will do the short test route I used to test a hold ( the route is in the LNAV thread along with the original results ). Will update this post with findings. I've often thought the gains in the controllers were off ( I have in the past done a fair bit of work with basic controllers ) but if it's ok for some people & not others then that is not the entire story. My current working theory there is the controllers depending on timing data from the sim which should be a given either rate or interval, but isn't. That one's hit me in the past too.

 

Or alternatively it's people putting deadzones in their roll axis out of habit...

[acbent123 78]

12 hours ago, MarcusIsnard said:

One way to confirm this would be to view the raw joystick input - I can look into the debugger and see if I can find a way to monitor these paramters. Can you give me the route you flew? 

That sounds like a great idea. The route that you are interested in is the one in my most recent post? If so, it was an FS generated high altitude airways flight from CYYC to CYMM. I don't recall the arrival and departure runways, unfortunately. I used the generated SID and STAR and I believe added in the ILS. Thanks.

[CRJay 428]

I have had the swerving around even with Xbox controllers with pretty massive deadzones set up… I don’t think a simple dead zone will be the solution. 

[acbent123 78]

1 hour ago, Richard Dastardly said:

I will do the short test route I used to test a hold ( the route is in the LNAV thread along with the original results ). Will update this post with findings. I've often thought the gains in the controllers were off ( I have in the past done a fair bit of work with basic controllers ) but if it's ok for some people & not others then that is not the entire story. My current working theory there is the controllers depending on timing data from the sim which should be a given either rate or interval, but isn't. That one's hit me in the past too.

 

Or alternatively it's people putting deadzones in their roll axis out of habit...

That's great! I look forward to your results. I have done another two flights and in both cases, the plane tracked the course very well. It did overshoot one turn on the STAR a bit in one case, but my speed was a little high and it was a 90 degree turn. One thing that I have noticed is that if you do overshoot, the plane might end up flying parallel to the FMS flight plan as opposed to working its way back to the track. In this case it put me right enough of the path/runway heading that it would not grab the ILS and I had to manually land. In this case, I had used an RNAV approach and switched to ILS for additional guidance. We'll call this one 'pilot error' LOL. 

[acbent123 78]

2 minutes ago, CRJay said:

I have had the swerving around even with Xbox controllers with pretty massive deadzones set up… I don’t think a simple dead zone will be the solution. 

I'm not sure if it is a fix, hence the reason to ask people to help by checking it out. It has made all the difference in my experience with this plane, but it has been just a few flights. I did an experiment with the default A320 and B787. If I removed my aileron axis null zones, I could not get the two default planes to fly on autopilot at all. Heading sel and nav were completely ineffectual, as though they had not been engaged at all. The flight director would move toward the direction of the flight path, but the plane would simply continue to fly and ever tightening arc in the opposite direction. 

[Richard Dastardly 108]

Given there's been posts since I'll just post the first test result here.

 

EGDY/26 DCT IPZOL DCT ATWEL DCT TIVER
Hold EX NW/287 107/L 185kts 3500ft 1.0 min
EGTE ILS26 via EX1 
CRJ-900, 0 passengers, 0 cargo, 2480kg fuel

 

This is with 5% deadzone on the roll axis only

 

Please login to display this image.

 

If you want a visual of the flight plan refer to the post in the LNAV thread ( the approach is in the post above it ). I am going to stop testing holds - there's obviously something horribly wrong there, and it managed to screw something up right through the following procedural approach, so much that the a/c never picked up the GS & I went round.

 

The second time round it made a much better attempt of the procedural approach than usual - still cut the turn a fair bit but it didn't do the characteristic left-right-left it's always done in the past. If my PC can stop crashing all the time I'll try it at 10% - I *really* don't want to leave a 10% deadzone (  I don't want any deadzone! ) but if it works, I guess it's better than nothing. It might also be worth trying a 5% deadzone on pitch for people having VS issues.

 

Will be back with a run at 10% & without the hold...

[acbent123 78]

50 minutes ago, Richard Dastardly said:

Given there's been posts since I'll just post the first test result here.

 

EGDY/26 DCT IPZOL DCT ATWEL DCT TIVER
Hold EX NW/287 107/L 185kts 3500ft 1.0 min
EGTE ILS26 via EX1 
CRJ-900, 0 passengers, 0 cargo, 2480kg fuel

 

This is with 5% deadzone on the roll axis only

 

Please login to display this image.

 

If you want a visual of the flight plan refer to the post in the LNAV thread ( the approach is in the post above it ). I am going to stop testing holds - there's obviously something horribly wrong there, and it managed to screw something up right through the following procedural approach, so much that the a/c never picked up the GS & I went round.

 

The second time round it made a much better attempt of the procedural approach than usual - still cut the turn a fair bit but it didn't do the characteristic left-right-left it's always done in the past. If my PC can stop crashing all the time I'll try it at 10% - I *really* don't want to leave a 10% deadzone (  I don't want any deadzone! ) but if it works, I guess it's better than nothing. It might also be worth trying a 5% deadzone on pitch for people having VS issues.

 

Will be back with a run at 10% & without the hold...

Okay. Thanks for that. I am going to go back into my sim shortly and check my null zone. I did some tweaking, so mine may be at a percentage greater or less than 5. I have not tried a hold yet. I agree that having dead zones is less than ideal. I have cheap controllers - CH Products, so with no null zone on the roll axis, I tend to get some banking. The null zone actually helped my straight and level hand flying. My intention is to get some better controllers relatively soon. As an aside, I've noticed that MSFS planes don't seem to have much of a dihedral effect. Seems to me that planes in FS9 would self-level better if you let go of the yoke/stick. 

[Richard Dastardly 108]

10%, same plan, no hold. Both flights done with custom weather, so there's no wind to speak of ( it was about 2kts ).

 

Please login to display this image.

Does not look any better than 5%, but that is *much* better than 0% ( this is an approach I'm quite familiar with ). Good bit of deduction, well done!

 

This is the plan with the hold in the DA62 with the WT garmin & the performance mod ( I only know of one ) and no deadzone. I don't know if the WT garmin uses the stock internals underneath it all, I thought it might though?

Please login to display this image.

[acbent123 78]

27 minutes ago, Richard Dastardly said:

10%, same plan, no hold.

 

Please login to display this image.

Does not look any better than 5%, but that is *much* better than 0% ( this is an approach I'm quite familiar with ). Good bit of deduction, well done!

 

This is the plan with the hold in the DA62 with the WT garmin & the performance mod ( I only know of one ) and no deadzone. I don't know if the WT garmin uses the stock internals underneath it all, I thought it might though?

Please login to display this image.

Thanks! Just so I understand, the first pic in this response is an ILS approach and the plane flew it correctly? Before the added null zone, the plane would not fly this same  ILS approach, even without the hold prior to the ILS? Sorry, a bit brain dead here, LOL. Late night last night. I'm not familiar with that other aircraft that you flew the hold with in your second pic, but it does seem to demonstrate that the CRJ has problems with a hold. I wonder how a default 787 or A320 would handle the hold. 

 

 

[acbent123 78]

 I am looking at this pic - of the attempted hold - and it looks like the CRJ did the first pass appropriately. Then it tightened its circle significantly. Then you left the hold and it flew toward the ILS/GS and flew inbound toward the airport. That's where I get confused. There are tracks as though it flew back and tried to repeat the approach. What was going on there?

 

 

Please login to display this image.

[Richard Dastardly 108]

29 minutes ago, acbent123 said:

 I am looking at this pic - of the attempted hold - and it looks like the CRJ did the first pass appropriately. Then it tightened its circle significantly. Then you left the hold and it flew toward the ILS/GS and flew inbound toward the airport. That's where I get confused. There are tracks as though it flew back and tried to repeat the approach. What was going on there?

 

 

Please login to display this image.

 

That one was a bit confusing, I was in two minds whether to submit it at all but I don't want to be flying this all afternoon.

What's going on there is -

* Passes ATWEL & TIVER ok

* Enters hold at EX ( as per chart in other thread ) - turns left as it should but the outbound track is something between the reciprocal of it's TIVER-EX track and the programmed one.

* Turns left at the end of the outbound leg of the hold & never stops turning left
* Armed hold exit 2nd time around the circle, it turned *right* out of the hold, forgot all about overflying EX to start the procedural approach & went for D112D

* At D112D turned the wrong way initially, and then turn left towards CF26, overshot, turned right again and then overflew CF26 and out of the other side of the ILS lobe... failed to capture GS so I performed missed app procedure, overflew the airport & turned right back to EX.

* arriving at EX again, A/C turned right to D112D perfectly, overflew D112D & turned left for CF26, overflew that while turning left properly for whichever WP it picked next ( I forget, it does skip some sometimes ) and captured GS & the localiser properly.

 

It *would* fly the ILS approach before, but it tended to turn early & cut corners & then overcorrect. You can see some of that in the other thread, it's 100% repeatable. If you compare the 5% track for the procedural approach with the 10% one they're pretty much the same, so I don't think going any higher helps.

 

The Longitude with the WT upgrade flies the approach just like the DA62 does. CRJ is the biggest A/C I fly which has an AP that can do that ( I fly the DC-6 with just the VOR gear ) so I don't know about either of the 320s. EGTE is a bit small for 787s

[acbent123 78]

9 minutes ago, Richard Dastardly said:

 

That one was a bit confusing, I was in two minds whether to submit it at all but I don't want to be flying this all afternoon.

What's going on there is -

* Passes ATWEL & TIVER ok

* Enters hold at EX ( as per chart in other thread ) - turns left as it should but the outbound track is something between the reciprocal of it's TIVER-EX track and the programmed one.

* Turns left at the end of the outbound leg of the hold & never stops turning left
* Armed hold exit 2nd time around the circle, it turned *right* out of the hold, forgot all about overflying EX to start the procedural approach & went for D112D

* At D112D turned the wrong way initially, and then turn left towards CF26, overshot, turned right again and then overflew CF26 and out of the other side of the ILS lobe... failed to capture GS so I performed missed app procedure, overflew the airport & turned right back to EX.

* arriving at EX again, A/C turned right to D112D perfectly, overflew D112D & turned left for CF26, overflew that while turning left properly for whichever WP it picked next ( I forget, it does skip some sometimes ) and captured GS & the localiser properly.

 

It *would* fly the ILS approach before, but it tended to turn early & cut corners & then overcorrect. You can see some of that in the other thread, it's 100% repeatable.

 

The Longitude with the WT upgrade flies the approach just like the DA62 does. CRJ is the biggest A/C I fly which has an AP that can do that ( I fly the DC-6 with just the VOR gear ) so I don't know about either of the 320s. EGTE is a bit small for 787s

Haha, thanks for the explanation. My head is spinning, lol. I haven't done a hold in ages. You have to set the leg lengths, correct? I wonder too if the CRJ is too big a plane for those procedures. But, overflying CF26 was unfortunate. Do you recall your approximate airspeed and configuration approaching CF26? I know that even at big airports, I don't like to be in excesss of 180 kts, 3 notches of flaps otherwise it might overshoot, but that's with a 90 degree turn to final approach course. You had a nice shallow intercept there. 

 

Anyway, hopefully the addition of a null zone, sends us down a path that helps to resolve these issues. I do really wish though that the developers would be more forthright about what's going on, if they are working on this or if they have given up. 

[Richard Dastardly 108]

3 minutes ago, acbent123 said:

Haha, thanks for the explanation. My head is spinning, lol. I haven't done a hold in ages. You have to set the leg lengths, correct? I wonder too if the CRJ is too big a plane for those procedures. But, overflying CF26 was unfortunate. Do you recall your approximate airspeed and configuration approaching CF26? I know that even at big airports, I don't like to be in excesss of 180 kts, 3 notches of flaps otherwise it might overshoot, but that's with a 90 degree turn to final approach course. You had a nice shallow intercept there. 

 

Anyway, hopefully the addition of a null zone, sends us down a path that helps to resolve these issues. I do really wish though that the developers would be more forthright about what's going on, if they are working on this or if they have given up. 

Speed would have been a bit above 170kts with... I can't remember the exact flap setting at that point ( the plane was very light, was giving me a Vapp of 121kts.. ). It's definitely not too big, that is the big boy version of the approach & the airport takes up to 767s ( and has been for a while, so they can't all have used the RNAV approaches ).

[acbent123 78]

2 minutes ago, Richard Dastardly said:

Speed would have been a bit above 170kts with... I can't remember the exact flap setting at that point ( the plane was very light, was giving me a Vapp of 121kts.. ). It's definitely not too big, that is the big boy version of the approach & the airport takes up to 767s ( and has been for a while, so they can't all have used the RNAV approaches ).

Thanks. Well, I am going off to do a flight. I may even try a hold. I'll have to read up on that as I don't think I've done once since the LDS767 in FS9. I appreciate the help here. It would be great if others would join us in troubleshooting this. If nothing else, maybe the devs will look at the issues again. 

[Richard Dastardly 108]

Did a full flight with a 5% deadzone.

Please login to display this image.

 

I'd post closeups of that but it tracked right down the line everywhere except for cutting the corner over the Irish Sea & having a slight wobble compensating.

 

I have *no* idea what happened on the approach - the wobbly line is the first attempt ( at one point when it was east of where it should be I watched it sequence through all the approach WPs & stop at the runway WP - maybe it was telling me something ) and the line that makes a really quite good job of it - possibly the best job I've seen this plane manage -is the second go at it after the go-around. Around 170kts / flaps 20 for all of it - was pretty turbulent & the speed ceiling is 185kts there. The onyl difference I can think of with the second attempt was I re-entered the approach & went direct to the start ( at EX ). Corrupted data at the end of the flight plan? is it flipping between an approach WP and a spurious corrupt one? don't know.

Please login to display this image.

[acbent123 78]

7 hours ago, Richard Dastardly said:

Did a full flight with a 5% deadzone.

Please login to display this image.

 

I'd post closeups of that but it tracked right down the line everywhere except for cutting the corner over the Irish Sea & having a slight wobble compensating.

 

I have *no* idea what happened on the approach - the wobbly line is the first attempt ( at one point when it was east of where it should be I watched it sequence through all the approach WPs & stop at the runway WP - maybe it was telling me something ) and the line that makes a really quite good job of it - possibly the best job I've seen this plane manage -is the second go at it after the go-around. Around 170kts / flaps 20 for all of it - was pretty turbulent & the speed ceiling is 185kts there. The onyl difference I can think of with the second attempt was I re-entered the approach & went direct to the start ( at EX ).

Please login to display this image.

I'm working all next week, but I plan to try your flight one evening next week. Sorry if you mentioned it, but do you use the default flight planner to get your procedures? I do and then

just enter the SIDS, WPs, airways and STAR into the CRJ FMS. 

[Richard Dastardly 108]

16 hours ago, acbent123 said:

I'm working all next week, but I plan to try your flight one evening next week. Sorry if you mentioned it, but do you use the default flight planner to get your procedures? I do and then

just enter the SIDS, WPs, airways and STAR into the CRJ FMS. 

I use Simbrief, attempt to load that into MFS itself ( just so I can at least attempt to use the ingame ATC ), set the start point as a gate & find MFS has changed the plan & then ignore MFS entirely & enter the Simbrief plan by hand. I could just load the thing into the CRJ from Simbrief export but I like setting it up.

 

That is not quite my usual ( or the real usual ) Glasgow-Exeter routing but the more usual route is pretty much a straight line anyway.

 

Edit: note you asked about holds - holds need ( generally ) the course of either the outbound or inbound leg, whether it's a left or right orbit, and the speed, at min. Standard hold is just fly for a minute & turn around again. The CRJ is a bit wierd, and it takes a couple of goes to get the right heading in there sometimes & I'm never really sure what it really thinks it's meant to be doing despite drawing something appropriate on the FMS.

[Ha_Ma 195]

vor 20 Stunden schrieb Richard Dastardly:

I use Simbrief, attempt to load that into MFS itself ( just so I can at least attempt to use the ingame ATC ), set the start point as a gate & find MFS has changed the plan & then ignore MFS entirely & enter the Simbrief plan by hand.

How do you set the startpoint to a gate?

You must not do this by clicking a Gate/Parkposition in the map! You have to chose a gate from the scrolldownmenu below "Departure". This always works fine for me and the FP will not be changed.

[Richard Dastardly 108]

5 hours ago, Ha_Ma said:

How do you set the startpoint to a gate?

You must not do this by clicking a Gate/Parkposition in the map! You have to chose a gate from the scrolldownmenu below "Departure". This always works fine for me and the FP will not be changed.

Ah, that's how you do it. It doesn't really bother me too much anyway other than ingame ATC being a bit odd, but I'll definitely remember that one.

[LesOReilly 314]

On 3/13/2022 at 10:00 AM, Richard Dastardly said:

10%, same plan, no hold. Both flights done with custom weather, so there's no wind to speak of ( it was about 2kts ).

 

Please login to display this image.

Does not look any better than 5%, but that is *much* better than 0% ( this is an approach I'm quite familiar with ). Good bit of deduction, well done!

 

This is the plan with the hold in the DA62 with the WT garmin & the performance mod ( I only know of one ) and no deadzone. I don't know if the WT garmin uses the stock internals underneath it all, I thought it might though?

Please login to display this image.

The WT G1000Nxi puts the the Stock AP into Pitch and Roll Mode and then it does all of it's own calculations to drive it....  That said there are people that complain about "hunting" and "oscillations"  in NAV/GPS modes ... and they have a hard time recreating them OR find that there is something with controllers that keep sending in information.. ie dirty axis...

Now that said .. there implementation is refined and obviously improved over how they did the CJ4 which used HDG and VS mode under the hood.  They spent a lot of time implementing a system in the G1000 that including big changes in SU5 where they disconnected the AP modes from the AP functions.  That allowed them to put the AP into a fixed Pitch and Roll mode while "lying" to the sim and telling it which MODE they had the GFC700 in.  They also implemented the K Event capture that powered the ability to get to a system where sending the stock AP events could be ignored by the sim AP which would stay in Pitch and Roll while telling the sim Vars that it was in other modes...

So long story short -- all of the custom AP systems have to work with the Sim AP in some manner ...  Had Aerosoft not spent time on the CRJ until a year later maybe we would be in a more comparable place with all of the NEW capabilities... Some times being the first to market is not always the best thing....

[Richard Dastardly 108]

On 3/16/2022 at 3:17 AM, LesOReilly said:

So long story short -- all of the custom AP systems have to work with the Sim AP in some manner ...  Had Aerosoft not spent time on the CRJ until a year later maybe we would be in a more comparable place with all of the NEW capabilities... Some times being the first to market is not always the best thing....

There are a lot of inferences I'm taking away from that explanation I'm not liking at all ( about layers under the API, and behaviour I've noted from other a/c ). Going to have to study the API I guess, if it's all published to everyone & not just partners, anyway.

[LesOReilly 314]

On 3/24/2022 at 10:41 AM, Richard Dastardly said:

There are a lot of inferences I'm taking away from that explanation I'm not liking at all ( about layers under the API, and behaviour I've noted from other a/c ). Going to have to study the API I guess, if it's all published to everyone & not just partners, anyway.

It is ...  Also Working Title published their Framework they did for the NXi and the GFC 700 AP that is built into it.  The Massive change what adding the ability to decouple the AP A:Vars from the actual AP (Autopilot Avionics Managed) That sticks the AP in a PITCH and ROLL Mode then you can take in all the K:Events for the AP to Your custom AP run the math there and then tell the Sim A:Vars what state to be in but without that putting the AP into those modes.

 

The FD is now kind of backwards to how the real one works in that it is being driven not doing the driving..  But really the User experience is not aware.