New ServoBlocks-based GoPro mount...

[GGoodrum]

After deciding that my mkTR Pro setup was just a bit heavy for my 600-size X8, I decided to try an optimize the performance of a new mount for my GoPro2. Based on my previous testing, I knew that if I could get the angular ratio between the servo and the camera as close to 1:1 as possible, for both pitch and for roll, that would maximize the responsiveness of the mount to changes in the MC's gimbal outputs. To do this, however, I now also know that you need to use high speed and high resolution digital servos. Most of the newer higher-end servos now have 12-bit resolution. This is key, or you end up having a lot of "jitter", as the servo bounces around between two positions.

Anyway, I then started looking around for what I had in my "pile", that I could use, and found one of Andrey's mid-sized landing gear. This is the perfect size for my X8 and all I had to do was drill four new holes. This mount also has an isolated cross-member, with a tab sticking out the front. Next, at Andrey's suggestion, I checked out the new ServoBlocks that ServoCity.com now sells. These are very cool, and are very reasonably priced. Quite a clever design. Anyway, I ordered up a couple sets, along with with two of the tube clamps and a 2" piece of 5/8" tubing. I also ordered two Hitec HS-7940TH ultra-fast (.06s@7.4V...) titanium digital servos. These are normally quite expensive, at around $150 each, but ServoCity sells them at $119. Still pretty pricey, but the blocks are cheap enough that it is worth spending the "extra" money on really good servos. What I don't have to do is spend any money on a G10 or CF-based mount. The blocks are the mount.

The only other bits I needed were some pieces to actually mount the GoPro2, and its case. Again, I went back to my "junk" box, and found a bunch of parts left over from the Raptor electric conversion kits I used to sell, that ended up working pretty well.

Here's some pics:

View attachment 1445
View attachment 1446
View attachment 1447
View attachment 1448
View attachment 1449


Once I got it all setup, and had everything plugged into the MC, I found that because of the extra side load on the tilt servo, there was a bit of noise and high frequency jitter. I fixed this by adding a small RC car shock, which eliminates the loading on the servo. I was absolutely amazed that once I did that, there was exactly zero jitter, in either pitch or roll, and there's not even any noise. They are completely quiet, which is a good thing. Next, I hooked up the CastleLink, and reset the 10A BEC to 7.4V.

Here's a short test I did, before doing any "tweaking" of the gains, etc.:

All-in-all, I'm pretty pleased with how it is working. I need to get a Hitec servo programmer, so I can dial it in a bit better. I want to play with the speeds a bit and see if I can get the roll and tilt operation to match even closer.

Tomorrow, I'll also do a test flight, which will be the real acid test, but so far, the performance is quite good. I just checked, and my total bill from ServoCity was $490. Not bad when you consider the cost of a "quality" mount, like the popular AV-130. Also, I saved about 1/2 pound, over the mkTR Pro setup. That will make my motors a bit happier.

-- Gary

 

[Tahoe Ed]

Gary, I like your design. This is an industrial strength compact roll/tilt unlike the one that I had for my first Gaui mount. If you have the time, could you post your parts list. I am thinking that this might be what I am looking for on my next build. Thanks!

 

[GGoodrum]

Gary, I like your design. This is an industrial strength compact roll/tilt unlike the one that I had for my first Gaui mount. If you have the time, could you post your parts list. I am thinking that this might be what I am looking for on my next build. Thanks!

Sorry, Ed, I missed this. Anyway, I just finished a minor "redo", which is simpler, and geometrically more accurate. The first version worked fine, as long as the camera wasn't pointed down. When that happens, the roll axis becomes more of an arc. To correct this, the tilt axis should also be centered on the camera sensor, or focal plane. To accomplish this, I fabricated a plate that moves the tilt servo out to the right, and then standoffs are used to move it forward a bit, to allow the tilt range of motion.

View attachment 1469


The rest of the servo block assembly simply slides on and attaches with four screws.

View attachment 1470

View attachment 1471


This balances much better as well, and I was able to eliminate the RC car shock.

Operation is even smoother than the first version, with the camera staying fixed, even with large movements of the platform.

I'll have to wait until tomorrow to give it a test flight, but already, I can tell this is working better than the first version.

-- Gary

 

[Tahoe Ed]

Gary, That is sweet! When you get a chance, after you post your video, please give us your final parts list and if you have a template for the roll to tilt mount that would be great. I would rather do your DIY mount that pay the same for an inferior mount. Thanks for putting this together.

 

[GGoodrum]

I've made a few more tweaks, mainly playing with the gains. One thing I did do to the mount itself was double up and use two of the old Raptor CCPM plates I'm using as part of the L-shaped mount off the tilt servo. This makes the camera mounting more rigid, and less susceptible to vibrations.

I started playing with the gains as well, and found that with a 1:1 ratio this sort of "direct drive" has, the gimbal gains need to be dialed down from the "standard" value of 20. At that value, the servo over-corrects. I ended up setting both tilt and roll values to 17, and it seems to be pretty close. Geared, or belt-driven mounts need to have higher values, based on how much of a ratio change the belt/gear drive induces.

Here's my latest test video:

This concept works pretty well, with a small, compact and fairly light camera, such as a GoPro/GoPro2, but I don't think this would scale up to something larger. I'm thinking would work well for a larger setup is having a belt drive in each axis, but probably no more than about 1.5:1 reduction (to minimize the speed hit..), and then still use blazingly fast, high-resolution digital servos.

-- Gary

 

[Gunter]

Gary, that's looking very good. I think when flying, it would be even better, because you are moving the heli a lot more than what it would move when flying.

Have you managed to do a test flight yet?

Gunter.

 

[GGoodrum]

No, sadly, I didn't make it out today. :shame: I got tied up doing some other work and looked up to see that it was getting dark already. I hate that it gets dark so earlyy in the winter, such as it is. Today it was about 75F/24C and sunny all day, but the "day" portion is just too short. I'll try and do better tomorrow. :tennis:

-- Gary

 

[GGoodrum]

Okay, I finally got my act together, in the daylight hours, and managed a quick test flight. The mount works quite well. I do see, however, that I have a basic vibration issue, which shows up as a typical "jelly roll" in the video. This is being caused, I'm sure, because after my first "drop-from-the-sky" episode, I ended up with three of the props with nicks in them, so they certainly aren't balanced anymore. :shame: I have some brand new Graupner 11 x 5s, to go on here, but I wanted to wait until I had all the power issues sorted out. I thought I was there, but in today's flight, it turns out only one of my two 4s-3300 packs were being used. The packs, which i got at my LHS, came with Deans, and I got lazy, and didn't change them out. :shame: That will change before the next flight.

Anyway, the mount did its job quite well. I didn't really yank it around too much, mainly as I was also trying to test the voltage monitoring function, so I didn't want to stray too far. With only half the capacity I was expecting, I'm glad I stayed close.

I'm editing the video right now, and I'll post it a bit later. I'm doing two versions in the same video, one "raw" and one that has stabilization applied, to see if it will clear up the jelly roll.

-- Gary

 

[GGoodrum]

The video is done. The stabilization didn't do much good, and it has a bit of a quality hit I'm not pleased with, so I'll avoid that in the future. Also, I'm thinking now that most of this jello is due to servo jitter. You can here it in the video. Also, when big movements are commanded, the jello goes away, which again makes me think it is jittering around the zero point. I did some research, and I now think I know why. When I placed the order with ServoCity, there is a selection that you have to make on whether or not you want 180-degrees of travel. The default is 180-degrees. In reading more about the Hitec servo programmer I have coming, I found that the high resolution only works when the servo is set for 120-degrees. Forcing the servo to the 180 degree mode resets the servo back to the lower resolution. What threw me was that these come set from the factory defauted to 120-degrees/high resolution, but apparently ServoCity reprograms them. Great. :shame: Anyway, I have a servo programmer coming, but I still don't have a shipping email from Tower Hobbies, and I ordered it on Monday. I don't know why it takes them so frickin' long. HitecRCD is down the road from me, in Poway, CA, and Tower has these in stock anyway. Why do I keep ordering from them? This happens every single time. :livid:

In any case, here's the video:

 

[GGoodrum]

My servo programmer finally showed up, so I was able to finally set the servos to the hires/12-bit mode, and I was able to widen the deadband. Movement now is very smooth, without the "stairstepping" I was getting before. Unfortunately, I've run into a showstopping problem that is going to kill using a direct drive setup with the WK-M gimbal outputs. :upset: It seems that at the level/neutral position, the outputs are making small movements, 2-3 "clicks" in one direction, and then back. This does this in both pitch and roll. I thought I might be able to dial this out, by opening up the deadband, but it didn't work. Here's a short video, showing the problem:

This means I will have to use a mount with at least some reduction, in both pitch and roll, so that these small movements get "absorbed". Back to the drawing board, I guess. :dread:

-- Gary

 

[GGoodrum]

Okay, so Sydney sent me a PM, over on RCGroups, that they will have someone look at the "drift"/hunting problem, that occurs around the neutral point, on Monday (tomorrow...). I did another flight test yesterday, and it is clear that the only jello I'm getting at all is right around the neutral point, which unfortunately, is most of the time. :dread: Another interesting point is that this hunting/drift problem doesn't happen until the system is seeing enough GPS sats that the red LED starts flashing twice, instead of three times. It doesn't matter whether you are in Manula, ATTI or GPS mode, it works exactly the same. You can see this in the video, at the very beginning, when it it is clear that the red LED is flashing three times, that there is no hunting/drifting. You can then see the effects, quite clearly, when I moved it to the grass.

It was really windy yesterday, when I did this test, but that's okay, because my real purpose was to see if I needed to change the gimbal gains again, and the GPS mode does a very good job of keeping this put. I purposely cranked around quite a bit, and was very happy with how the mount is responding. Once DJI comes up with a fix, for the hunting and pecking, I will be quite happy with this as a nice platform for doing GoPro2 videos.

-- Gary

 

[GGoodrum]

For those looking for a parts list for this setup, Here it is:

The main pieces are built around two of these: ServoCity Hitec ServoBlocks.

I added two of these 0.770" 5/8" bore tube clamps and a 2" length of 5/8" tubing.

The key to getting the best performance is to use fast, high resolution (12-bit...) digital servos. I used two Hitec HS-7940TH Titanium digitals, which can be run on 7.4V and at that voltage are blazingly fast at .06 sec/60-deg. These are programmable, and you need one of these to make sure the high-resolution mode is set, and to widen the dead band space.

The rest is mostly just some hardware bits'n pieces, like 6-32 countersunk screws (3/8" and 1/2"...) and four 1/2" long nylon spacers.

There are also two other custom pieces required, an L-shaped platform, for the camera, and a flat piece about 3-1/2" long by about 1" wide. These are fairly easy to fabricate, as long as you pay attention to the geometry. As soon as I verify that DJI has a fix for the drift problem, I'm thinking of getting some of the two unique pieces made.

-- Gary