Beta Testers requested to fix Compass Issue
- Thread starter terry.one
- Start date
- Status
ianwood
Taco Wrangler
Terry asked me to add a few more people to test the beta firmware. I've sent PMs out with instructions to a handful of you. Please report back here with your observations and results.
OK, so I'd like everyone to remember that Project Drunken Sparrow has still not accomplished its main goal.
Yes, I realize that DJI listened to us and has essentially made the fix, but beyond the lucky beta testers we still don't have anything. I am not calling to re-engage the full court press just yet, but if DJI takes 4 months to push this out the door, where does that leave us?
I say we put a clock to it, kinda like we did in giving them a deadline to respond to us the first time. I think three weeks from the time the beta was sent out is a reasonable amount of time to put it through the required steps for public release.
Bottom line: We are close, but our goal was to get a fix. As of now, there is a fix... but 99% of us don't have it yet.
June 21st is the day. If we don't have a firmware update fixing the issue by then I call for a full re-engagement of the Drunken Sparrow.
Yes, I realize that DJI listened to us and has essentially made the fix, but beyond the lucky beta testers we still don't have anything. I am not calling to re-engage the full court press just yet, but if DJI takes 4 months to push this out the door, where does that leave us?
I say we put a clock to it, kinda like we did in giving them a deadline to respond to us the first time. I think three weeks from the time the beta was sent out is a reasonable amount of time to put it through the required steps for public release.
Bottom line: We are close, but our goal was to get a fix. As of now, there is a fix... but 99% of us don't have it yet.
June 21st is the day. If we don't have a firmware update fixing the issue by then I call for a full re-engagement of the Drunken Sparrow.
landonkk said:
Agreed. I mean, how hard would it be for Terry to just post a message here giving us an update, and a timeframe. That's just basic customer service failure. ENGAGE your customers.
My Phantom 2 Vision has no problems in Malaysia, flies well and feels great. But, now, I have moved to New Zealand, and face this problem big time... Please help!!
ianwood
Taco Wrangler
AnselA said:
I very unscientifically selected 5 more people based on their participation in the issue, their declination and what I believe they can contribute overall to getting the best resolution for all of us. I would choose more if I could as there are many of you who could add a lot of value to the process.
For everyone still suffering, I know it's not easy to wait any longer but DJI's engineers have asked for more feedback on the beta and that's what we're doing. From what I can tell, they're doing their homework and we need to be supportive of that effort. The Drunken Sparrow is not dead. Keep your avatars and stay vigilant until everyone has the fix.
ianwood said:
is required IOSD mark II? may be the core2 flytrex?
I hope to participate as a beta in the future, I live in Brazil.
my dec. is -23
Ianwood asked me to test and comment on the Beta software designed to fix the declination problem. I updated my Phantom this AM and headed out to a local soccer field.
I am slightly southeast of Seattle Washington USA
The local declination is 17 degrees.
I have a Phantom 2 Vision
I use NAZA mode & IOC as needed
I had noticed the declination problem on some of my first flights, my son and I commented when we saw it but we just ignored it as we were low on the learning curve. After ianwood published his intial testing, analysis and video it all made sense. I used a local soccer field since the lines and layout gave a good frame of reference. I was seeing significant TBE, left drift, and JHook that would improve with time. Using the soccer field as a calibration reference I determined my intial error was twice my declination. This also shows up in Course Lock and is most dramatically seen in Course Lock.
My testing today went great. Light winds and threatening rain clouds but it stayed dry. I did a compass calibration and took off.
Right from the start there was NO TBE. You could see some initial minor motion, I am talking about less than a foot, but it didn't swirl at all.
I lined up with the sideline as best I could. The inital run had some left drift and just enough JHook to be noticed. I estimated that it was about 30% (very approximate) of what it was before. On the return, just by pulling back on the stick, I could tell it was dramatically improved. PreBeta even with more JHook away from me the return would JHook back over my head necessitating some control input to get it stopped. With the Beta software today it flew almost straight down the center of the field with almost no JHook.
My second run appeared to have almost no drift or JHook, likewise the return. Same on subsequent runs.
I then switched to Course Lock to see if the initial lock showed any effect. My impression was that there was still some Course Lock error proportional to the overall reduced effect. I then lined it up and and reset the course lock reference in flight. Course Lock then worked correctly and showed little to no error and could have been somewhat pilot induced (my error).
I would say that my initial testing today is totally consistent with the report and neat diagram that ianwood posted on June 6th.
I switched batteries, therefore a complete power cycle. My impression was that the second set of tests was identical and went through the same process.
My initial impression is that it isn't perfect. But it is so dramatically better that it will seem perfect in normal use. If I was to takeoff and fly a normal mission, instead of a test, I think I would hardly notice any error. I would feel comfortable taking off in a confined space as the TBE is totally gone. The only action I think is required is for Course Lock. When I am going to be using Course Lock I would reset the Course Lock reference in flight about a minute or so after takeoff.
Fun footnote. When I fly at the soccer field people take notice but no won has stood and watched. Today while I was testing I had two instances. The first was a mother and about six children who came over and were watching. I could hear her commenting to the kids so I flew over and hovered in front of them and put on a little demo. After I landed they came over to look at it - all were very interested in the details and thought it was great. Later three people out walking similarly watched an chatted. The gentleman suggested to the ladies that it might make a good father's day gift.
I am slightly southeast of Seattle Washington USA
The local declination is 17 degrees.
I have a Phantom 2 Vision
I use NAZA mode & IOC as needed
I had noticed the declination problem on some of my first flights, my son and I commented when we saw it but we just ignored it as we were low on the learning curve. After ianwood published his intial testing, analysis and video it all made sense. I used a local soccer field since the lines and layout gave a good frame of reference. I was seeing significant TBE, left drift, and JHook that would improve with time. Using the soccer field as a calibration reference I determined my intial error was twice my declination. This also shows up in Course Lock and is most dramatically seen in Course Lock.
My testing today went great. Light winds and threatening rain clouds but it stayed dry. I did a compass calibration and took off.
Right from the start there was NO TBE. You could see some initial minor motion, I am talking about less than a foot, but it didn't swirl at all.
I lined up with the sideline as best I could. The inital run had some left drift and just enough JHook to be noticed. I estimated that it was about 30% (very approximate) of what it was before. On the return, just by pulling back on the stick, I could tell it was dramatically improved. PreBeta even with more JHook away from me the return would JHook back over my head necessitating some control input to get it stopped. With the Beta software today it flew almost straight down the center of the field with almost no JHook.
My second run appeared to have almost no drift or JHook, likewise the return. Same on subsequent runs.
I then switched to Course Lock to see if the initial lock showed any effect. My impression was that there was still some Course Lock error proportional to the overall reduced effect. I then lined it up and and reset the course lock reference in flight. Course Lock then worked correctly and showed little to no error and could have been somewhat pilot induced (my error).
I would say that my initial testing today is totally consistent with the report and neat diagram that ianwood posted on June 6th.
I switched batteries, therefore a complete power cycle. My impression was that the second set of tests was identical and went through the same process.
My initial impression is that it isn't perfect. But it is so dramatically better that it will seem perfect in normal use. If I was to takeoff and fly a normal mission, instead of a test, I think I would hardly notice any error. I would feel comfortable taking off in a confined space as the TBE is totally gone. The only action I think is required is for Course Lock. When I am going to be using Course Lock I would reset the Course Lock reference in flight about a minute or so after takeoff.
Fun footnote. When I fly at the soccer field people take notice but no won has stood and watched. Today while I was testing I had two instances. The first was a mother and about six children who came over and were watching. I could hear her commenting to the kids so I flew over and hovered in front of them and put on a little demo. After I landed they came over to look at it - all were very interested in the details and thought it was great. Later three people out walking similarly watched an chatted. The gentleman suggested to the ladies that it might make a good father's day gift.
I just finished some testing on Beta 3.05
Headed home to get fpv and tripod mounted video uploaded.
Headed home to get fpv and tripod mounted video uploaded.
Hi,
Do you have any beta testers in Brazil? Here in my city the declination is -22 degrees and the TBE, J hook and drift is really accentuated. I am flying in Naza mode.
Do you have any beta testers in Brazil? Here in my city the declination is -22 degrees and the TBE, J hook and drift is really accentuated. I am flying in Naza mode.
Just completed a couple of initial 10 minute test flights with 3.05.
Location: New Mexico
Declination: + 8.5˚
Elevation: 7500 ft
Winds: None
Equipment: P2, Lightbridge, 14SG, GoPro H3+
Gains: Default
Satellites: 8
Advanced IMU, compass and sticks calibration after update. Previous firmware produced mild TBE but quite noticeable J-hook and a CL error of around - 2 x declination.
OBSERVATIONS
I'll try to get more comprehensive testing done tomorrow, but overall my initial conclusion would have to be that this is not a robust fix. In particular, if declination were being properly corrected then I do not see why the AP would still need a couple of learning passes, or why CL would be in error as before. It appears that DJI have simply increased the learned corrections as the AP compares intended course with actual track.
The CL error, in particular, would seem to confirm that at startup it still does not correctly interpret its orientation; instead of correctly measuring it as magnetic bearing plus declination it still apparently records and flies it as magnetic bearing minus 2 x declination. Note that as it takes off it does actually turn (yaw) to fly the wrong heading.
I have both onboard and ground video that I'll take a look at and upload if it shows these effects clearly enough, but the flights were right at sunset and the light was not optimal.
Location: New Mexico
Declination: + 8.5˚
Elevation: 7500 ft
Winds: None
Equipment: P2, Lightbridge, 14SG, GoPro H3+
Gains: Default
Satellites: 8
Advanced IMU, compass and sticks calibration after update. Previous firmware produced mild TBE but quite noticeable J-hook and a CL error of around - 2 x declination.
OBSERVATIONS
- Flight is stable, with little or no TBE.
- CL is still off by the same as before, but note that I did not reset it during the flights.
- J-hook still present on first pass in both directions - somewhat less than before but not much.
- J-hook mostly disappeared by the 3rd pass, but was not consistent - it reappeared a couple of times later in both flights.
I'll try to get more comprehensive testing done tomorrow, but overall my initial conclusion would have to be that this is not a robust fix. In particular, if declination were being properly corrected then I do not see why the AP would still need a couple of learning passes, or why CL would be in error as before. It appears that DJI have simply increased the learned corrections as the AP compares intended course with actual track.
The CL error, in particular, would seem to confirm that at startup it still does not correctly interpret its orientation; instead of correctly measuring it as magnetic bearing plus declination it still apparently records and flies it as magnetic bearing minus 2 x declination. Note that as it takes off it does actually turn (yaw) to fly the wrong heading.
I have both onboard and ground video that I'll take a look at and upload if it shows these effects clearly enough, but the flights were right at sunset and the light was not optimal.
In general I agree with sar104's comments regarding that it still has to go through the learning process but I do believe the initial pass has less error than before. I am also planning on doing more testing in the morning if the rain holds off. Besides trying to refine my observations I curious about the initial yaw that sar104 mentioned. I didn't notice it but I am sure I could have easily missed it. I do find it interesting. I don't remember seeing that in the CL testing I did on earlier versions. So I'll try it with and without Course Lock.
John Shaw said:
So when your P2 flew the wrong heading in CL, it was still facing its original direction and applying roll as well as pitch?
sar104 said:
I went back and looked at my old CL test video on the soccer field. I lined up with a sideline, the track took directly off at an angle but the P2V didn't seem to change heading (no yaw). During the flight it was a little hard to tell but as slowed down and JHooked it was still almost perfectly perpendicular to the end of the field - looking directly at the goal. That matches my mental impression - as you say it pitched and rolled but didn't yaw.
I very unscientifically selected 5 more people based on their participation in the issue, their declination and what I believe they can contribute overall to getting the best resolution for all of us. I would choose more if I could as there are many of you who could add a lot of value to the process.[/quote]
New Zealand.
My results with the 3.05 beta.
The weather today is very cloudy with an 8kmh wind.
First flight, I went straight up, no TBE at all. A small amount of drifting.
Pushed forward stick and it went straight forward, no drifting.
Slight right hook at the end.
Pulled back and it came straight back with a slight left hook at the end.
Repeated this a few times and the hook reduced.
I tried the IOC and noticed it was drifting off track by about 20-30degrees.
I’ll charge my batteries, wait for the weather to clear a bit, and go out and do some
more tests.
New Zealand.
My results with the 3.05 beta.
The weather today is very cloudy with an 8kmh wind.
First flight, I went straight up, no TBE at all. A small amount of drifting.
Pushed forward stick and it went straight forward, no drifting.
Slight right hook at the end.
Pulled back and it came straight back with a slight left hook at the end.
Repeated this a few times and the hook reduced.
I tried the IOC and noticed it was drifting off track by about 20-30degrees.
I’ll charge my batteries, wait for the weather to clear a bit, and go out and do some
more tests.
- Status
