[jcg1541]

What does it do if you leave the normal PID active in the dive?  Oscillate?  Wobble around as if the servo isn't strong enough when you know it is strong enough?  Come straight down and require more pull out pitch than normal?  Is the new servo rated to have at least as much torque as the old servo (maybe the strength is marginal)?

I have a video footage of using normal PIDs in the dive using the new servo. I will post it next. It wobbles around.
The new servo,  eflight s60, is more beefy than my prior servo, emax 9051.  I have developed a "servo saver" modification to prevent gear strips in crashes. The mod weakens the maximum torque of the servo arms, but I am pretty confident that the mod does not impair the precision of the servos. See attached pictures. I use this mod in all my videos, old and new.

[jcg1541]

What does it do if you leave the normal PID active in the dive?  Oscillate?  Wobble around as if the servo isn't strong enough when you know it is strong enough?  Come straight down and require more pull out pitch than normal?  Is the new servo rated to have at least as much torque as the old servo (maybe the strength is marginal)?

The first dive was with the normal PIDs. The second dive was with 194% high gain mode. The 2 dives were in the same flight back-to-back.
I spun the rotor with consistent constant RPM by using BLHeli MULTI software with PI governor. Both dives had exact same RPM.
The first 5 seconds of the first dive was OK, but I had to abort it after 5 seconds when it swung so much I was not sure what it would crash into.

[TheOtherCliff]

I saw that same circular oscillation (pitch/yaw with roll too) in fast normal flight with normal PIDs.  I am guessing that the fact that the oscillation has a yaw component (with different servo, linkage, servo loading, PID) is saying something?

I would start with Manual mode, making copter adjustments so that it flips (roll and pitch) and pirouettes in alignment with the frame or at very least in alignment with the FC.  Then check that Rate mode flips are also in alignment.

I might try tuning with ZN method like this:  Set I and D to zero and adjust P up till it oscillates.  Cut P in half and adjust I up till it oscillates and reduce I a little.  Leave D set to zero.

[jcg1541]

Just an observation: Measuring servo error with a fixed servo signal is not the same as measuring the heli leveling error with a stabilizing FC running.  Any slop in the servo or even the linkage is removed by the FC.  Imagine a roll in manual mode with the stick released when it is level.  Ignore inertia and assume that the only issue is slop.  Imagine that with the slop the swash plate never comes all the way back to neutral and so you get a continuing slow roll.  Now imagine what the servo does with the FC in Rate mode when you release the stick when the copter is level.  It continues to drift past level and the FC sees that the rate is not zero and drives the servo far enough past neutral to both counteract the slop and the inertia.  Servo and linkage slop doesn't really matter because the FC takes the slop out.

I see your concerns with the slop play. I want to rest your mind that the problem was investigated and resolved with my 250-gram craft as attached first picture with the 0.1mm patching films (tape). The slop play was a very serious destabilizing problem even for simple hovering without any acrobatics because the airfoil CG shifted as explained in the second and third pictures. I have built the same 250-gram craft 10 times, and I can consistently build stable crafts recently.
I know that symmetrical airfoils have a stable center of pressure CP (but not zero CP shift), while cambered airfoils have a positive feedback CP shift, meaning that a CP shift resulting in more angle of attack change resulting in even more CP shift - an inherent unstable system. 
So, what I learned was
helicopter + small slop = stable
helicopter + large slop = unstable
airplane + any slop = unstable
.

[jcg1541]

I might try tuning with ZN method like this:  Set I and D to zero and adjust P up till it oscillates.  Cut P in half and adjust I up till it oscillates and reduce I a little.  Leave D set to zero.

The ZN method is used by the hobby industry to make the expensive "cyclic servo". They cut the instantaneous output from CC3D for swash in half and reallocate it for I gain in the servo's digital board. I-integration needs time to close the error gap, and for common helicopter flying the half gap is closed very tightly by integration over time, but there is no time when the craft dives at terminal speed. Every millisecond spent on integrating the error is a millisecond the craft spend on swinging like my last video's first dive. That explains why I needed 194% P gain, almost exact doubling, in CC3D as the "dive mode" PID when I used the precision cyclic servos. We are in "nested" PID loops when it comes to the swash. See this video, starting at 1:30 for the nesting control,


The ZN method is absent in the, also expensive, "helicopter tail servo". I installed 3 tail servos on the cyclic swashplate today. And my "normal mode" worked for both diving and cruising. As this video,

The "tail servo" needs the exact same PIDs for the swashplate as the lowest cost, simplest F3P servos, EMAX 9051, that I started this project with.   

I just now realized that when I dive vertically, my main rotor stands up straight, exactly the same as a tail rotor. Maybe that is why they mean to sell me tail servos for my main rotor.

[jcg1541]

I have been trying out a new set of heavier rotor blades. And somehow the PIDs ratios between the 2 sets need to be the square power of the blade weight ratio.
The lighter set weighs 14.6 grams,
 
, and the heavier set weighs 17 grams,

.
The 2 sets of blades have the exact same geometry, shape and dimensions.

[TheOtherCliff]

Some wild guesses. 

I can imagine that pitch flexibility (twisting) or loose blade grips or blade grip pivot hole farther back in airfoil or blade airfoil being more tail heavy could cause the blades to overreact to (cyclic) blade pitch changes... and thus to need smaller PIDs.  Are either set of blades weighted?  Weight is usually added close to blade leading edge and that would be more stable and need more throw and more PID to move it?

[jcg1541]

Everything geometry wise is identical between the 2 sets. Pivot holes, bolts, etc. The airfoil is the popular NASA0015 .
The airfoil is pictured on this page,
https://nocomputerbutphone.blogspot.com/2018/10/oxy-210-blade-characteristics.html
The lighter set is simple nylon. The heavier set is nylon reinforced with graphite.

The only difference is the material.

Grip is tuned by the same procedure as this video,

Some wild guesses. 

I can imagine that pitch flexibility (twisting) or loose blade grips or blade grip pivot hole farther back in airfoil or blade airfoil being more tail heavy could cause the blades to overreact to (cyclic) blade pitch changes... and thus to need smaller PIDs.  Are either set of blades weighted?  Weight is usually added close to blade leading edge and that would be more stable and need more throw and more PID to move it?

[jcg1541]

Mystery solved. It was my grip setting procedure allowing too much error margin.

I allowed the blades to drape between 45 degrees and 90 degrees, and the 2 extreme ends had very different rotor stiffness.  The looser the grip, the higher the PIDs were needed.

Now my new procedure is to set the grip to allow the blades to drape at 80 to 90 degrees,

, and the PIDs for diving is almost the same for either the 17-gram blades or the 14.6-gram blades.

.

[TheOtherCliff]

Glad you got it sorted.  I always enjoy watching your flying straight down.

On my .30 size helis I just went by feel.  Equal and fairly stiff.  I figure the force pulling out on the blades will straighten almost anything out. 

[jcg1541]

Appreciate that. The last video was an example of slight rotor imbalance. Today I got the well balanced rotor with the same PIDs and everything.

The 2-blade system is actually equivalent of a string music wire tensed up. And my RPM happens to be close to the natural resonance frequency.
I am balancing it with pieces of 1 square cm scotch tape, less than 0.01 grams each. And each piece counts.  Yesterday's flying was 1 piece off balance.
Today's dives are here,

[karla]

Really admire your perseverance  You been working on it for a long time not giving up.

[TheOtherCliff]

I am balancing it with pieces of 1 square cm scotch tape, less than 0.01 grams each. And each piece counts. 

For my low RPM (1000KV@3S) plastic quadcopter props (after trimming on magnetic balancer) I use black plastic electrical tape of about that size to balance them running.  Also the short way (thinner strips of tape) of the prop vs. the long way.

I also balance motors with several layers of the tape pieces.  One out of ten might need it.  I might add a little CA around the edge of the tape square, especially for motors that get warm and the tape flies off.

Even glued, the black electrical tape sometimes comes off.  I have always just trimmed a blade in the long direction (research one bladed props if you think trimming a blade tip is bad).  I now use hot melt glue on the hub for the short prop balancing direction.  It works well, at least on my 10000RPM motors.  None has come off.