(24-Oct-2022, 09:50 PM)MomoBrut Wrote: OK, it's very late here, hopefully it won't rain tomorrow and I'll try it.

Perfect. I'll check back.

Not more than 2 MB is allowed to upload ...

Upload to where?

I tried to upload the log in here.

You know how you upload video to YouTube (or some other hosting platform) and
then put the  link in a post ??  You might have to do the same thing with the log.
Upload it to YouTube and then add the link to the post here.

Zip up the log file using something like 7zip and then attach the (much smaller) ZIP or 7Z file.

Alternatively upload the files to an external file hosting provider like Google Drive, Microsoft OneDrive, Dropbox, MEGA etc. and then post a link to the file back here.

(24-Oct-2022, 08:47 PM)Recursion Wrote: Hey bud. I just saw your message, and I wish I'd seen this thread earlier. Your build looks great, and I'd be happy to help you get this thing flying perfectly. I have 3 of them built on the board you have right now, so our builds would be identical outside of the motors, so I'll start there.

Motors: Do you have bi-directional dshot and rpm filtering enabled? How many poles do you have configured? Most motors of that size have 12 poles, but the BH 1504.5 have 14 poles, so make sure that is configured correctly. 

Tune:

Make a backup of what you have (diff all) and save it somewhere. I found that there were a few changes needed from stock to make smaller builds fly a lot better. Copy and paste this in and see if this helps since this is the current tune I'm using:

Code:

set vbat_sag_compensation = 100
set anti_gravity_gain = 4000
set throttle_boost = 4
set p_pitch = 44
set i_pitch = 83
set d_pitch = 43
set f_pitch = 118
set p_roll = 42
set i_roll = 79
set d_roll = 38
set f_roll = 114
set p_yaw = 42
set i_yaw = 79
set f_yaw = 114
set d_min_roll = 28
set d_min_pitch = 32
set thrust_linear = 50
set feedforward_smooth_factor = 35
set dyn_idle_min_rpm = 40
set simplified_master_multiplier = 95
set simplified_i_gain = 105

set gyro_lpf2_static_hz = 850
set dyn_notch_count = 1
set dyn_notch_q = 400
set dyn_notch_min_hz = 130
set gyro_lpf1_dyn_min_hz = 0
set gyro_lpf1_dyn_max_hz = 850
set mag_hardware = NONE
set baro_hardware = NONE
set align_board_yaw = -90
set current_meter = ADC
set battery_meter = ADC
set ibata_scale = 256
set ibata_offset = 80
set pid_process_denom = 2
set simplified_gyro_filter_multiplier = 170
save

Let me know if that helps or not. If not I'd love to get a copy of the blackbox log.

This was not very nice ... set align_board_yaw = -90 ... it flips my quad 3 times before I realized, what is happening, next time I will check what I copy to CLI ... you should too ...

So I used your setup and deleted the killer gyro repositioning you had there.
I first flew a full throttle increase, then several sharper throttle additions, then sharp turns gradually in all axes and finally a lot of sharp turns into the propwash. So everything can be beautifully explored.

With your settings the quad is quite loose, the propwash is even bigger than with my settings. Dyamic idle 40 leads to too floaty quad, does not want to go down at all when the throttle is pulled down. I looked at the blackbox log, the spectrum in the Roll axis looks just as bad as with my settings, which is what I expected and what I also wrote yesterday. If you look at the Roll axis over time, you can see how noisy it is, while Pitch and Yaw are smooth and follow the setpoints beautifully.

Nothing was resolved.

Log:
https://drive.google.com/file/d/1-14vvim...sp=sharing

Sorry about that command. I was quickly pulling the config where I only had a few minutes so I was rushing. These boards have different versions with different alignments (for whatever reason), so that's not great.

So a few things. The first is that I don't think you have a bad gyro. Looking at a specific section where you punch the throttle and kill it + flips the filtered noise on roll doesn't seem abnormal, and I'm not seeing any gyro data spikes that I've seen with bad gyros.



That doesn't mean the tune is good. The tune I provided was done for the T-Motor F1404 3800kv, which performed the best of all the 1404(ish) class motors I tested. I can not speak to your motors since I have not tested them, but if they perform like some others I've tested they might have lower top end and slower spin-up times on these propellers, so the tune would need to be different (i.e. higher pids). The dynamic idle for me spins much faster than a 5", but I found this necessary to maintain stability on these smaller motors. That said, they are not floaty at all for me when diving with no wind, so that's interesting.

One thing we could do here to get better insight into that would be to do another test flight with Blackbox debug set to "DSHOT_RPM_TELEMETRY" to record actual motor RPMs. If you could do a few 100% throttle punches where you look down (to make sure you don't hit the ground) at zero throttle and provide no stick input (yaw/pitch/roll), that'd be some good data. If you could also do some hard 360 turns (prop wash) as well smooth fast cornering, that'd be nice to see as well.

It would be best if you can upload both the blackbox and goggle DVR data to Goggle drive for the flight, so I could sync them together to map the BB data to what is going on.

There's so much to say, I'm running out of energy, so briefly ... Displaying a spectrum independent of motor rpm is absolutely useless, you miss the information that the noise is independent of motor rpm and is everywhere. I've been trying to explain this for several times, it seems pointless and tiring.

Two pictures below are from the uploaded blackbox log. Spectrum in the first picture is great, this is the pitch axis before filtering and no problem, this will be easily filtered out by the rpm filter. Spectrum in the second picture is a damn problem, that's the roll axis before filtering, no filter will filter this without introducing a delay to the PID loop, because it would have to have a cutoff frequency of 10 Hz. 

Does anyone here understand?
Either it's a faulty gyro, or the noise is coming from the frame because it's the wrong shape. Please send me your blackboxlog, I'll take a look and see if I have a problem with the gyro or if it's a feature of the frame. Thanks.

Here you go



Best of luck to you.

I've been focusing my time lately on trying to mount a GoPro 7/8 as light as possible. I've settled on a screwless TPU case I've designed which has been the lightest solution I could find and has 'so far' survived some abuse.



167.8g AUW with a 550mah 4s battery.

Thank you very much for the pictures. The spectrums are exemplary. If the problem was in the frame, the resonances will be mostly concentrated around individual frequencies, never broadband noise, which I expected, now it's confirmed. So the problem is either in the AIO or the gyro on the AIO.

Going back to the gyro again, you probably have it turned 90 degrees. Which exact version of AIO with MPU6000 do you have? Could you, please, send a detailed photo of the AIO this blackboxlog is from? There are at least 4 versions of this AIO, they differ in gyro type and angle, at least two types with MPU6000, one with ICM42688-P and one with BMI270. It is very frustrating to find out which version works and it can even be purchased. Below is a photo of my AIO, it's version 1.1, which is marked next to the USB connector, the photo also shows the orientation of the gyro.

And one more thing, I've tried adding capacitors to Vbat, 5V and 3.3V rail before. None of this helped me resolve the noise. What capacitors are you using to make the spectra look like you sent them?

It is very very frustrating and time consuming.

I have a number of the v1.1 boards in builds right now which are identical to that one. It's the MPU6000 gyro which I have not had an issue with. I do think this design does make them susceptible to some electrical noise, which is less of an issue because the gyro is so forgiving. The ICM42688-P board should never have been released, and is untunable and unflyable due to the noise (which completely freaks the gyro out at higher RPMs), even with a large cap on the power input that they put into the design. They discontinued the MPU6000 boards when releasing the ICM42688-P board, but now they seem to be selling the MPU6000 again instead (likely because the ICM board was so terrible). I wonder where they are sourcing the gyros from now...

I have a 25v 470uf panasonic low-esr cap on the power input for the board.

Edit:

Scratch my first comment. Your board IS different. Looking at your picture again I realized your 'boot' button is a tiny membrane button, which is the same they put on the ICM42688-P boards updated design to fit a larger capacitor. All of my MPU6000 boards have a larger boot button, which means yours is an updated design. Your Gyro has different markings than mine as well. Here is a close up from mine.



Edit 2:

I just read the spec sheet and based on the markings it looks like your chip was manufactured in April 2018, where mine was made in January of 2020. So it looks like they've either been able to acquire an older batch, or they are salvaging older chips.

First of all, thanks for the photos and info.

Many things, so gradually...
1. The ICM42688-P gyro is absolutely perfect; it just has to be on the right PCB of right FC or AIO. I tuned the HAKRC AIO F722D so another 3.5 inch frame flies fantastic, no propwash. It is described in another thread how to do it. So definitely don't condemn the ICM42688-P gyro, it can work great. I, on the other hand, am beginning to distrust the JHEMCU company in general ... I will explain later.
2. If the sent settings and the log you sent are from the same quad, and the AIO photos are again from the same quad (I asked for a photo of the one from which the log is), then I don't understand your gyro orientation settings. Your gyro is oriented exactly the same as mine, so why did you rotate it by -90 degrees, can you explain it to me?
3. The used capacity is not very large and at the same time you have absolutely no noise, I used a significantly larger capacity and the noise was still the same.
4. I was told that the v1.1 they sold me is from a new batch. According to the gyro marking, it is clear that the gyro is from 2018, I see it as either the gyro is from some discarded stock, because it was 2nd quality. Or even worse, which the Chinese can do, it is a reused part, i.e. resoldered from older usage. And it's a MEMS component, so mechanical impact can simply damage it. The AIO manufacturer is responsible for these things, I don't trust the JHEMCU manufacturer at all.

To sum it up, I'm pretty pissed.

Any solution is not good, there is absolutely no guarantee that a new piece of AIO will be fine. RMA this with a Chinese is almost impossible. Another option is to get the Gyro itself, I have the equipment to solder such a chip. But you can't find it anywhere, the MPU6000 is simply sold out.

I wrote an email to JHEMCU, I am very curious to see what they will answer...

(26-Oct-2022, 05:37 PM)MomoBrut Wrote: 2. If the sent settings and the log you sent are from the same quad, and the AIO photos are again from the same quad (I asked for a photo of the one from which the log is), then I don't understand your gyro orientation settings. Your gyro is oriented exactly the same as mine, so why did you rotate it by -90 degrees, can you explain it to me?

I have a bunch of these boards and builds (around 10 boards total). The log was from a Scythe I've been flying recently to experiment and get a good configuration for the latest DJI Betaflight OSD (MSP-OSD) from FPV.WTF (which is amazing). The picture that I posted I took from my digital scope months ago when comparing the ICM board with a MPU board I had lying around.

The ICM Gyro is amazing on paper, but it has to be powered properly. JHEMCU obviously didn't take the time to do that (or even test it) as it looks like they just threw a large cap on the input and called it a day. I made a video about it here.


So yeah, I don't trust JHEMCU either anymore, especially since they continued to sell the ICM board after I showed them how bad it was. They promised me a refund and then never sent it after I released the video. They must have lucked out with the MPU6000 design since the size/price (was originally $45)/performance was amazing, especially since I was able to burn motors on 6s/5" racer without blowing an ESC. I wish I had bought more.

I'd love to hear how they respond... if they respond.