After losing my Eachine Trashcan (https://intofpv.com/t-lost-my-trashcan) and enjoying the GepRC Phantom (https://intofpv.com/t-geprc-phantom?pid=85007#pid85007) I decided I should build my next kwad myself.

I wanted something similar in size to the Phantom and I had read only positive things about Armattan frames so I chose the Tadpole  

All pictures and component details can be found here: https://rotorbuilds.com/build/22505

I was struggling to put the components in the best place but I'm happy with the result now. I really like the T-motors, they look very strong. The camera and antenna give me a much better image than I ever had, but I just wish there was some more contrast and detail in the green.

I'm flying it with Betaflight 4.2 (RC1) and I'm still trying to figure out the best PIDs because it doesn't feel super locked in yet. Luckily the Succex FC has some Blackbox flash so that should help. The PID defaults were too wobbly but with the P and D Gain sliders at 1.2 it's much better, though I'm seeing lots of high frequency vibrations in the logs (full image attached):



I think it's caused by the D gains being too high, so I'll try to lower those next time, but if anyone has any other suggestions I'm all ears.

Here's the first flight:

Didn't make much progress on that PID tuning... But had some fun with a tree...

What do your PIDs and Feedforward look like?

I tried the three profiles pictured below. The first one is the one I tried earlier and I liked most. I stated that the P and D Gain slider was at 1.2 but that's wrong; it was at 1.5. Perhaps that's simply too much and I should go back closer to the defaults, so I'll try 1.2, 1.3, ...

For the second profile I lowered the PD balance, which (in Betaflight 4.2) lowers the D gains. I didn't like it because there was some overshoot wobble after every move.

The third profile results in equal P and D gains. This felt weird and loose and showed propwash even for average turns.

All other PID and feedforward related values are at their defaults. I'm using RPM filtering and the filter sliders are all the way to 2.0, the dynamic notch filters at width 0, Q 200, min 90, max 350. Oh and I set up dynamic idle to prevent the motors from ever spinning too slowly.

Perhaps I should try the method outlined by UAV Tech (https://tiny.cc/pidtuning) where you disable D min and minimise I, then find the perfect P/D balance first and slowly improve from there. I'm starting to get a grip on the theory but real tuning still seems an art that just requires practice and experience...

You might already know this but for most of your flying, you'll be at your "D Min." You only kick into "D Max" at the apex of sharp moves based on your D Gain settings.

Try going back to the profile you liked best but bump up your pitch and roll feed forward to 150 each and go in increments of +25 till you get to where you like. This will give you the riding on rails feeling.

This is 2S? You're gonna think I'm crazy but theoretically you should be at around:
D Min = 60
P and D Max = 90

For ishts and giggles you might just wanna pop that in there (after the feedforward) just to see how it flies. Put it through some flips and rolls and hard cornering and check motor temps.

If it flies much better, your motors haven't exploded, and you're feelin' adventurous then +5 D Min if you still get prop wash on sharp cornering or -5 D Min until you do. This whole time set both P and D Max to 1.5x D Min. When you get D Min dialed in, start tweaking P until that feels right to you without any oscillations. Now remember this P-to-D ratio. Whenever you need to change D (for propwash) or P (for handling) either because you bash a prop or whatever, you just need to maintain this ratio.

If you're feelin' super adventurous, go back and start turning off your filters. Really, you only need RPM Filtering, Dynamic Notch, and Lowpass D 1. Again, check motor temps *OFTEN*.

Thanks for all the tips ronin!

I'll definitely try some higher feedforward values. Those high P and D values are a lot scarier though, and I was afraid the vibrations in the blackbox logs would only get worse...

With some short hovering tests in the house I now figured out that the vibrations can be slightly reduced if I lower my D terms, but I already know it'll fly worse like that outside. So I thought it really must be caused by bad mechanical vibration dampening. Now I've replaced the 20mm screws for the flight stack with 22mm screws so I don't have to tighten them as much as before and I could add the little pieces that I cut off the bottom standoff rubbers (see my comments at https://rotorbuilds.com/build/22505) between the frame and the nut on the bottom:



The result seems quite promising! This is a hover before changing the screws:



And this is after:



(full images attached)

Hopefully this means I can keep the filtering to a minimum and increase the P and D values the way you described

I'm currently flying with 2S 450mAh batteries but I intend to go to 3S 450mAh, the delivery is just taking a long time due to the pandemic. I was wondering if the P/D balance can or should remain the same, and the P+D gain should be lowered about one third? Or is it best to re-tune everything from the start when I get those batteries?

I did some flight testing today and concluded that the default P/D balance is fine, and the 1.5 PD gain is still the best. Those crazy high suggested PIDs just introduced crazy high oscillations so that's not an option.

The extra feedforward definitely improved handling though, and in the blackbox log I can see that with a very hard left roll the gyro (blue) follows the setpoint (orange) very well thanks to the feedforward (purple):



So these are the PIDs that work best now:



Even with these moderate values the oscillations in the blackbox logs are still there, so the changes I made to the flight stack screws didn't make the difference I hoped for. During the flights I also noticed that the low battery alarms came up fast (even though I have the warning at a low 3.2v and minimum at 3.0v) and that flight time was pretty short. So either I've battered these batteries too much already or these PIDs demand too much from the motors. On the other hand the motors were never hot at all.

Now I've done more testing at home:
- without props: still lots of oscillations in the blackbox logs, and I can actually feel lots of vibrations while holding the frame
- without RPM filtering: massive oscillations in the blackbox logs, so the RPM filtering definitely works great!
- with filter sliders at 1.0 (instead of 2.0 or 1.5): much less gyro and D term oscillations in the blackbox logs

So I must conclude that the motors (T-motor F1103) are simply very noisy and it's best to leave the filter sliders at 1.0. For my next flight tests I'll disable D-min and try to find the actual perfect D values without touching any other values. I'm also curious if propwash will be worse with the increased filtering, and if it has any effect on the flight time.

(25-May-2020, 01:54 AM)L0stB1t Wrote: I did some flight testing today and concluded that the default P/D balance is fine, and the 1.5 PD gain is still the best. Those crazy high suggested PIDs just introduced crazy high oscillations so that's not an option.

What does that mean exactly? I would actually try it without D Gain disabled so you have just a single D variable to deal with. So P=90 and D=60 as a starting point. At 1.5x P-to-D, I'm not sure why you would P term oscillation. What I was afraid you would come back with was that your motors were running way too hot especially with all that filtering.

My concern is that with a lower than appropriate D, the oscillations you are seeing are not actually from P but rather from the I term kicking in because your P is actually *too low*. These are "I term induced oscillations" because I cannot react as fast as P but absent sufficient P gains, I is second on scene to push the gyro to better track setpoint. Hope that makes sense.

D terms are generally in the 20-40s for 5S-6S; 40s-60s for 3S-4S; so it would stand to some reason that 2S should see D in the 60+ range.

Thank you for taking the time to follow up

(25-May-2020, 04:12 AM)the.ronin Wrote: What does that mean exactly?  I would actually try it without D Gain disabled so you have just a single D variable to deal with.  So P=90 and D=60 as a starting point.

You mean with D-min disabled, right? I completely agree that's better to find the right P/D balance, with both D-max and D-min it's quite impossible to know which value the system uses at which point (though earlier tests I've done with my GepRC Phantom showed me that it's generally much closer to the D-min value and very very rarely goes up to the D-max). That's why on my next flights I'm definitely going to test with D-min disabled.

The problem is simply that with the high D gains (the term which is most sensitive to vibration) I got tons of high-frequency oscillations in Blackbox:

(25-May-2020, 04:12 AM)the.ronin Wrote: At 1.5x P-to-D, I'm not sure why you would P term oscillation.  What I was afraid you would come back with was that your motors were running way too hot especially with all that filtering.

The high P-term is not a problem in itself, but to (more or less) keep the 1.5x ratio the D-term(s) must go up also, and since the frame has a lot of vibration I got pretty crazy D-term oscillation. Usually this results in hot motors but these motors seem pretty immune to that. But next to the Blackbox logs the high battery usage is probably also an indicator that something is wrong...

Instead of "with all that filtering" you really mean "with the filtering sliders high" - and actual filtering low, correct? I'm quite sure now that this kwad is just not as clean as one would wish for, and requires some extra filtering on top of the RPM filtering. Though this will cause some delays and the kwad will never fly ultra-precise, I guess it's a much better base to re-start my tuning. In some hover tests I'm now seeing much cleaner lines so that's promising:

(25-May-2020, 04:12 AM)the.ronin Wrote: My concern is that with a lower than appropriate D, the oscillations you are seeing are not actually from P but rather from the I term kicking in because your P is actually *too low*.  These are "I term induced oscillations" because I cannot react as fast as P but absent sufficient P gains, I is second on scene to push the gyro to better track setpoint.  Hope that makes sense.

I'm quite sure the I-terms are no problem at all. In the logs I only see I term going up and down very slowly. I can imagine they could cause a slowish wobble if they would be too high but nothing compared to the very high-frequency oscillations in the gyro, D-terms and motors.

I did actually try a short flight with the I terms at 10. This resulted in what I expected; a kwad that "slides around" and never really goes exactly in the direction I send it, it needs much more corrections. In the logs the general picture was still the same with lots of high-frequency oscillations.

(25-May-2020, 04:12 AM)the.ronin Wrote: D terms are generally in the 20-40s for 5S-6S; 40s-60s for 3S-4S; so it would stand to some reason that 2S should see D in the 60+ range.

Good to know, thanks!

I hope to fly again soon, with filter sliders at 1.0, D-min disabled, and pushing P/D towards 90/60

Thanks Lost (cool name lol) for all the work you put into the feedback. Please take those general starting D's with a massive grain of salt. They are just rules of thumb based on what UAV Tech has found appropriate over time. But then I looked at his and another youtuber's whoop and toothpick tuning video and he violates those guidelines.

Oh ok so you are already running really aggressive filtering. Yes, for me the I term and yaw PIDs are generally after thoughts and only if something really seems off. In particular the effect of wind on I. Yaw I basically set and forget (it's inertia not prop thrust so its super slow and will never overshoot anyway).

Hmmm ... sorry I'm still obsessing over this D term. It just kills me that your motors are running cool on high D and low filtering. That's like found cash lol. You know, even with the gnarly D traces, how does it fly? Like even if you cranked down P below D. My Acrobrat ended up still showing some oscillations in blackbox but motor temps were fine and it flew how I wanted them so I kept it.

Don't worry I won't go straight to 90/60, I'll start around 60/40 and work my way up from there.

I made the mistake to use 1.5 and 2.0 filter sliders without checking 1.0 first. I was just blinded by the (I think it was UAV Tech) opinion that with RPM filters you can usually go straight to 2.0 on a "clean" kwad. I figured that since the T-motors are built like a tiny tank (bigger bearings, 2mm shaft (only 1.5mm at the top), sturdy casing) they are probably well balanced also. Well perhaps they were when I got them but I've crashed quite a few times and some motors ate some yummy grass and sand so...

Now with Joshua's Troubleshooting Quadcopter Vibration video I learned that everyone should check their motor noise, and I found out that my motor 1 is really smooth, 2 and 4 not too bad, but motor 3 is very noisy. You can see, hear and feel the difference. So guess who just ordered 2 extra motors on Banggood

Even though I had all that vibration and wiggly D traces the kwad actually flies really great if you ask me, and the high-frequency oscillations are not visible in the FPV feed. I'm not worried about all the vibrations in itself, but I am afraid it lowers performance by wasting energy and killing the poor 2S batteries.

By the way there's finally some movement in the shipment tracking of my 3S batteries so I kinda hope to get them next week. I'm sure the Taddy will be very happy with the power boost!

Not sure if P < D will be the way to go but who knows, I'll figure it out soon enough.

Did you check my Memorial Day deals thread? Emax is having a sale I just picked up a set of motors for $4 each lawls.

I really like the whole thread.   This is good for me learning how to tune    Thanks

@the.ronin $4 for a motor is really amazing. Sadly I'm from Belgium and we don't have many good FPV shops and even less promotion stunts. I get most of my stuff from Banggood and AliExpress, or with our neighbors at droneshop.nl when I'm less patient  

So yesterday I went for a tuning session (4 batteries, 2 x 8MB blackbox logs) hoping to figure out the perfect P/D balance. The filter sliders were at 1.0, D-min disabled, FeedForward disabled, I-terms at the default all the time. I've dropped the blackbox results here: https://drive.google.com/drive/u/1/folde...v3Mj_leVem

Sadly it seems like I have to choose between overshoot and oscillations (high P/D ratio) or lots of higher-frequency vibrations (low P/D ratio where D overpowers everything). The best ones according to the blackbox logs seemed to have low P values around 50-60 and fairly low P/D ratio's of about 1.3. So today I went for a short session during lunch break with 3 profiles; the first with all slider defaults (except the FF slider all the way to the left), the second same but with the PD balance slider at 0.8 and the third with the PD balance slider at 1.2. The results were similar; with higher D values it follows setpoint better, but there are much more oscillations also. During flight the higher D values always felt better unless both P and D got too high.

All this confirmed I simply have a hardware issue so I checked the troubled motor 3 again and the problem is not just some imbalance, if I wiggle the bell back and forth in a certain direction I can feel it's a tiny bit loose. At this scale we're probably talking about a gap of 0.01 mm somewhere in the bearings, but it's definitely there and the other motors don't wiggle at all. Until I receive my new motors I'm not going to continue trying to tune the PIDs because it's pointless with all this noise...

By the way I also checked the motor temps after every test flight and never ever were the motors hot, they barely get warm. I actually think it's because smaller motors have proportionally larger surfaces that cool them down (if you calculate it you'll find that a 1103 cylinder has a 1.03 surface to volume ratio and a 2207 cylinder has a 0.47 surface to volume ratio). So my guess is that motor heat rarely becomes an issue for tiny whoops and toothpicks.

Soooo... Turned out Banggood didn't actually have the F1103 motors in stock so I canceled my order and got them much faster from droneshop.nl. I also got some soft mounts off eBay (https://www.ebay.co.uk/itm/Armattan-Tadp...2886561513) so a bit of tinkering later it's ready to rock and roll again!!!



By the way it turned out that the problem with motor 3 wasn't the bearings but the motor bell was actually loose on the shaft - it could rotate freely and it wobbled a bit! When checking my other motors it seemed motor 4 was nearly as bad, so I also replaced it with the extra I got. After this the kwad was much quieter and blackbox shows much less vibrations.

I also received the 3S batteries I was waiting for and went for a little test flight launch



This little thing rips soooo hard now!!! The power and speed is incredible, and it's actually easier to control because you can correct mistakes faster. The throttle blips I do before a roll or flip also finally make an actual difference, this will definitely help me to learn some more advanced freestyle tricks. The only problem is still the flight time, I'm getting less than 3 minutes out of it.

I tried some different P/D balances but I can't really draw meaningful conclusions from the Blackbox logs (https://drive.google.com/drive/folders/1...j11bsexed4). I think with the 3S batteries the default P gain is too high so next time I'll start with lower P values. I was still flying with a fair bit of filtering so I'll also check if lowering it is a good idea. The flight precision and propwash handling was never bad but I mostly hope that a better tune will increase flight time. My GepRC Phantom easily gets 5 to 6 minutes out of the 2S batteries so the Tadpole should at least get 4 minutes from the 3S batteries in my opinion.

Now bad weather is spoiling the fun though and it looks like I'll have to wait a full week before I can fly again...