Elliott000
Member
Robert/gary, when could a guy put an order in.... How muchy
Sent from my GT-I9100M using Tapatalk
Sent from my GT-I9100M using Tapatalk
New DJI F550 hex with 2-axis belt drive gimbal...
- Thread starter GGoodrum
- Start date
GGoodrum
Member
Sorry for not keeping this thread updated. I've been busy testing some last minute changes/improvements, and I wanted to thoroughly test them first. Basically, what I've been working on is better vibration isolation. With some help from Chris, at Vortex Aerial, we found that the pool noodles I was using, were causing a low-frequency oscillation, mainly due to prop wash. I've now removed these, and have added some LED strips, to help with orientation.
We also figured out that we need a different scheme for the first-level isolation. We found some silicon 2-piece grommets that are extremely pliable. Anyway, these have really cut down the vibrations. We did a test where Chris hovered in front of me. I could touch the F550 legs/arms, and you can feel a strong vibration, almost teeth shaking.
Anyway, now when you touch one of the landing gear legs, or one of the skids, you just feel a very slight vibration. When you touch the mounting plate/battery tray, you get nothing.
In the videos below, I hard-mounted one camera to the top F550 plate, and i'm using the GoPro yoke with a GP2. It was pretty windy, and the WK-M ws struggling to hold position. The gimbal video is raw, no stabilization at all. In the first two, I'm using an XA v1 FC/AHRS as a standalone gimbal controller, mainly because the gimbal output rates can be set to 2ms/500Hz, vs the 10ms/100Hz used by the WK-M/Naza. The difference is that small movements are smoother with the faster update rate.
The second one is a test flight that I did shortly after the ground test. You can see how effective the vibration isolation is now, as there's a fair amount of jello in the hard-mounted GP video. None of it comes through to the gimbal GP, though, which is a good thing.
You can also see how aggressive the WK-M is in trying to hold position in the face of some pretty gusty winds. If you just watch the gimbal GP video, you really don't see that.
About halfway through there's a really big hawk that does a "drive-by". This guy has a wingspan over 5 feet, although that is hard to tell in the video, due to the GP's ultra-wide FOV. This Hawk, and its mate, have a nest nearby, so we see one or both of them occasionally. Once the one here sat on our fence railing. That's when I saw just how big he was. Anyway, once he figured out he couldn't mate with or eat the hex, he was on his way.
The next two videos are using the WK-M's outputs to drive the gimbal servos. The gains are are not "spot-on", so there's a little "twitching", which looks like a lag.
Because of the WK-M's slower output rate, small movements do appear a bit "rougher". For big movements, though, I can't tell any difference. In both videos, the winds were very gusty, especially once above the roof line. You can see and hear the WK-M's aggressive moves, trying to hold position. This typically has been a complaint of the WK-M system when used in windy conditions for video, that the movements are not smooth enough. To me, this always seemed to be an indication of a less than "perfect" gimbal setup. Everything is always a tradeoff, and yes, you can setup most "good" gimbals so that they have extremely smooth performance, but the tradeoff is that this "extra" smoothness is achieved at the expense of being truly platform independent. Hence the complaints about the WK-M's "overly aggressive" position hold function. Without "platform independence", you are forced to fly smoothly.
This irked me into this "holy grail" quest I've been on to achieve true platform independence, while still having a quite acceptable level of smoothness. What I've found is that there is an interrelated set of factors that include servo speed, servo resolution, servo travel, servo deadband timing, belt drive reduction ratios and gimbal controller output rates. You have to start with high resolution, 12-bit servos, the faster the better, but at a minimum less than .10s/60 degrees. Most digital servos have 8-bit control electronics. There are a number of "hires" mini-sized digitals, like the Savox SH-1357 and the Hitec 7235, plus some from Futaba, JR, Align, etc., but these all use 10-bit. The only true 12-bit/4096-type of high resolution servos all are full-sized. so that is why we are using the in both axis.
In addition to being 12-bit, you also need to have a programmable throw/travel range. Most servos are set to operate with 90-degree throws. That's +/- 45 degrees, but after a 2:1 reduction, you only end up with +/- 22.5 degrees at the camera. So far, I've only found two servos that are fully programmable, several models from Hitec, and also from Hyperion. These let you open the throws back up to at least 120 degrees, which means +/- 60 degrees, or +/- 30 degrees at the camera. The Hyperion models let you open them to about 178 degrees, which would be close to +/-45 degrees, at the camera, but I've found that even with very aggressive moves, +/- 30 degrees is more than enough. Some "expensive" gimbals get an increase in throws by using an external pot that is mechanically attached the yoke/camera platform, but that is mainly because they are using very high reductions, like 4:1, or higher. These gimbals have such high reductions in order to get smooth operation, even using lower resolution servos. The tradeoff, however, is reaction speed. If the reduction is too great, and/or the servo is not fast enough, the gimbal won't keep up with platform movements. Another reason for using external pots is so that you can have a wider adjustment range, for controlling the down angle, for instance. This adds a lot of complexity, and cost, which I think is not needed for most setups that aren't going to use a separate camera operator. With a little shot planning, it is quite easy to adjust the down-angle on this mount, simply by loosening one set screw. Andrey is working on a larger full-DSLR variant that will probably end up having external pots, but mainly because these will likely be used on larger platforms that will be used in more "pro"-type environments, where it is common to use a separate camera operator. These will also have the 3rd/pan axis.
Anyway, after many weeks of trial and error, I've finaly got what I believe is an optimized "ZenMuse"-like setup that has true platform independence and smooth operation, even for small movements. This is using 12-bit servos with a speed of .06s (on 7.4V...), set to 120 degrees and with zero deadband. The belt reduction is 2:1. Based on other tests, I think the reduction can be increased a bit, to about 2.8:1, and this will allow the throw to also be opened up to the max, 178 degrees.
As I said, everything is interrelated, so to get the "best" performance, the controller's gimbal ouput rate needs to be higher than the 100Hz limit on the WK-M. I'm hopeful that I will be able to convince DJI to add a faster rate, maybe as a setting, like XA does in there configuration software.
-- Gary
We also figured out that we need a different scheme for the first-level isolation. We found some silicon 2-piece grommets that are extremely pliable. Anyway, these have really cut down the vibrations. We did a test where Chris hovered in front of me. I could touch the F550 legs/arms, and you can feel a strong vibration, almost teeth shaking.
Anyway, now when you touch one of the landing gear legs, or one of the skids, you just feel a very slight vibration. When you touch the mounting plate/battery tray, you get nothing. In the videos below, I hard-mounted one camera to the top F550 plate, and i'm using the GoPro yoke with a GP2. It was pretty windy, and the WK-M ws struggling to hold position. The gimbal video is raw, no stabilization at all. In the first two, I'm using an XA v1 FC/AHRS as a standalone gimbal controller, mainly because the gimbal output rates can be set to 2ms/500Hz, vs the 10ms/100Hz used by the WK-M/Naza. The difference is that small movements are smoother with the faster update rate.
The second one is a test flight that I did shortly after the ground test. You can see how effective the vibration isolation is now, as there's a fair amount of jello in the hard-mounted GP video. None of it comes through to the gimbal GP, though, which is a good thing.
About halfway through there's a really big hawk that does a "drive-by". This guy has a wingspan over 5 feet, although that is hard to tell in the video, due to the GP's ultra-wide FOV. This Hawk, and its mate, have a nest nearby, so we see one or both of them occasionally. Once the one here sat on our fence railing. That's when I saw just how big he was. Anyway, once he figured out he couldn't mate with or eat the hex, he was on his way.
The next two videos are using the WK-M's outputs to drive the gimbal servos. The gains are are not "spot-on", so there's a little "twitching", which looks like a lag.
Because of the WK-M's slower output rate, small movements do appear a bit "rougher". For big movements, though, I can't tell any difference. In both videos, the winds were very gusty, especially once above the roof line. You can see and hear the WK-M's aggressive moves, trying to hold position. This typically has been a complaint of the WK-M system when used in windy conditions for video, that the movements are not smooth enough. To me, this always seemed to be an indication of a less than "perfect" gimbal setup. Everything is always a tradeoff, and yes, you can setup most "good" gimbals so that they have extremely smooth performance, but the tradeoff is that this "extra" smoothness is achieved at the expense of being truly platform independent. Hence the complaints about the WK-M's "overly aggressive" position hold function. Without "platform independence", you are forced to fly smoothly.
This irked me into this "holy grail" quest I've been on to achieve true platform independence, while still having a quite acceptable level of smoothness. What I've found is that there is an interrelated set of factors that include servo speed, servo resolution, servo travel, servo deadband timing, belt drive reduction ratios and gimbal controller output rates. You have to start with high resolution, 12-bit servos, the faster the better, but at a minimum less than .10s/60 degrees. Most digital servos have 8-bit control electronics. There are a number of "hires" mini-sized digitals, like the Savox SH-1357 and the Hitec 7235, plus some from Futaba, JR, Align, etc., but these all use 10-bit. The only true 12-bit/4096-type of high resolution servos all are full-sized. so that is why we are using the in both axis.
In addition to being 12-bit, you also need to have a programmable throw/travel range. Most servos are set to operate with 90-degree throws. That's +/- 45 degrees, but after a 2:1 reduction, you only end up with +/- 22.5 degrees at the camera. So far, I've only found two servos that are fully programmable, several models from Hitec, and also from Hyperion. These let you open the throws back up to at least 120 degrees, which means +/- 60 degrees, or +/- 30 degrees at the camera. The Hyperion models let you open them to about 178 degrees, which would be close to +/-45 degrees, at the camera, but I've found that even with very aggressive moves, +/- 30 degrees is more than enough. Some "expensive" gimbals get an increase in throws by using an external pot that is mechanically attached the yoke/camera platform, but that is mainly because they are using very high reductions, like 4:1, or higher. These gimbals have such high reductions in order to get smooth operation, even using lower resolution servos. The tradeoff, however, is reaction speed. If the reduction is too great, and/or the servo is not fast enough, the gimbal won't keep up with platform movements. Another reason for using external pots is so that you can have a wider adjustment range, for controlling the down angle, for instance. This adds a lot of complexity, and cost, which I think is not needed for most setups that aren't going to use a separate camera operator. With a little shot planning, it is quite easy to adjust the down-angle on this mount, simply by loosening one set screw. Andrey is working on a larger full-DSLR variant that will probably end up having external pots, but mainly because these will likely be used on larger platforms that will be used in more "pro"-type environments, where it is common to use a separate camera operator. These will also have the 3rd/pan axis.
Anyway, after many weeks of trial and error, I've finaly got what I believe is an optimized "ZenMuse"-like setup that has true platform independence and smooth operation, even for small movements. This is using 12-bit servos with a speed of .06s (on 7.4V...), set to 120 degrees and with zero deadband. The belt reduction is 2:1. Based on other tests, I think the reduction can be increased a bit, to about 2.8:1, and this will allow the throw to also be opened up to the max, 178 degrees.
As I said, everything is interrelated, so to get the "best" performance, the controller's gimbal ouput rate needs to be higher than the 100Hz limit on the WK-M. I'm hopeful that I will be able to convince DJI to add a faster rate, maybe as a setting, like XA does in there configuration software.
-- Gary
Attachments
Last edited by a moderator:
Hi Gary,
I agree with your conclusions. I "follow" your experience with a build of my conception for my Gopro.
I used Hyperion servos for the raisons you explained with a not to bad result. Until now I have not yet finished my trials and errors but results are pretty good for me.View attachment 3088
I agree with your conclusions. I "follow" your experience with a build of my conception for my Gopro.
I used Hyperion servos for the raisons you explained with a not to bad result. Until now I have not yet finished my trials and errors but results are pretty good for me.View attachment 3088
Attachments
rod tenney
New Member
new here would buy f550 for first platform want to mount camera fpv thermal im sometime for search rescue