Gyro shootout: MPU6000 vs ICM20602 vs BMI270 and OSR4 mode for BMI Gyros! 

Hi everyone! If you have a BMI gyro board and would like to test OSR4 mode then I'm pleased to say its in BF4.3 by default now.

It is sooooo nice to see SCIENCE work. Doing similar, in-flight, tests a couple years ago (same quad and measured data with two different gyros at the same time) and got the same results and had the same recommendation. Love it! :-) LOVE IT EVEN MORE how this makes it so clear how an analytical approach leads to better product (BMI270 find).

Great work. I was worried that my FC maker changed the gyro from MPU to BMI. But now I am very confident my drone will fly just as well. Thumbs up.

This series deserves an award 👏

Excellent discussion on the how's and whys of these essential components! I'll be watching for more of your informative content.

Thanks Chris, This was amazingly interesting! Long live the MPU6000! and your awesome series!

In your last Video i really loved the Chapter Feature to find the most important Parts easily. Keeping them reoccurring in future videos would be splendid.

Awesome stuff, Chris!! Very well done as always, but this one was espescially interesting! :D

I love you for making these scientific videos! You are the best! This brings back all the great memories from being in university! Please keep your style, there is really nothing to improve!

I've been seeing your github activity lately. This may help me understand the point you were trying to make. Good video!

Thanks Chris! I learn something from everyone of your videos!

Dude you are a Genius !! thanks for presenting such details in a way thats easy to understand !!!!

Chris mate you never have to say it, but man I learned so much today! Cheers and hope you're having fun ripping around!!

Thanks you very much for your work. Very useful. Can’t wait for your video about damping greace.

thank you! We were all waiting for a complete and concise video on this hot topic.. while waiting for the return of the mpu6000s, even if it seems that the BMI works well, I'm still on configurations without BMI, and classic mpu6000s, I also have an FC with a dual ICM20602 gyro, and I had more trouble tuning this quad. but I swear by the MPU6000. unfortunately hard to find, but not impossible..

Amazing insight. Thank you, Chris!

Exceptional engineering work as always Chris! I wished you had included the ICM-42688-P gyro in your excellent comparison. I've been using this chip for a completely unrelated project concentrating on its very low frequency behavior (sub 10Hz). Its FSR (full scale range) of ±2000 °/s down to ±15.625 °/s, noise density of 0.0028 °/s/√Hz , extensive user configurable internal filters, and 16/19 bit native gyro data (as opposed to 16 for the rest) made it stand out for me. Utilizing two of these gyros setup at different FSRs and different internal filtering configurations to compile the final data may prove ideal for a flight controller.

thank you so much for this info! Can't wait for any updates given all the new gyro replacements going on in the TW world!

Very professional and great commentary

Oh wow! Im one thrid in the video and its a sub! Just right level of details vs explanation.

Great video as always!!

Great information, Chris, thank you!

I love these types of videos. Thanks:)

Great info! Thanks for deep dive.

Pretty amazing vid and glad someone dives as deep into the subject as you! This really a is adding a lot to where that d-term-noise might come from. Also a clever idea to run those props upside down! Now what I would wonder: - What are those peaks you see in those noise-charts? Especially that ICM206028k bump around 25hz that look almost like a resonance shape? - What's the actual response time of those gyros and does this differ? Like... have this controller move a rapidly between 2 angles or actual RPMs (=degrees/second) and watch the gyros follow the move. (Here also the 32khz mode comes into play: in certain applications you want fast response above all while not being interested in low frequencies or precise measurement values like crash-detection/emergency stop use cases. Also predictive maintainance FFT-outputting-sensors will use such a mode to analyse and trace peaks in higher frequencies - here without much interest in response time) - Is there a max G-Force such a Gyro can measure/withstand? If the signal peaks out in rapid moves the measurement value won't be correct and hence the pid loop will act all weird. Also would a crash physically damage the gyro? - Do they have long term drifts as they are measuring a relative value without a true reference? - Whats the noise behaviour of those gyros in an otherwise similarly tuned quad at similar flights? Would we see differences in the blackbox there? Thanks again for the vid and keep it up!

Very interesting descriptions of what makes our high performing models tick. I like listening to very smart people it makes me feel like I'm worth the oxygen I'm breathing.

Great explanation man! Thank you very much!

That was great Chris, thank you!

man, that's great! I think this is promising for the BMI gyros and opens more FC options.

Thank you for your good work!

woah! that osr4 mode sounds just like what we need. NICE

THANK YOU Chris, you busted a few myths with this one. 👍👌

Thanks for great deep dive!

Nice content and thanks for the info about my Esc

Love the content... Wish these were a class at Uni

Hi Chris. One thing I have been very wary of in the last 4-5 years as we have transitioned from 4s to 6s is the gyro supply voltage regulator and the spec frequencies and voltage limits they state that they can handle! Less noise would be awesome.

just found this video because i am having a rare issue with my Betafpv pavo 25,(and now RU-vid only suggest related stuff) this quadcopter when i am flying if i try to descend (put throttle 0%) the motors start spinniing fast and i cannot descendk quadcopter keeps at same level of even fly higher! i been trying differrent pids, filters... and i found they changed the gyro on the fc! i am very impressed with the improvements you give to 95x and pavo 30, betafpv must send you a pavo25 to give the best settings and fix this problem

Thank you Sir👍👍👍

Thank You for making this video Chris! Recently I purchased an IFlight Beast AIO thinking it came with the MPU6000 But instead came with the BMI270. Which only supports a 3K Gyro and PID loop in BF which affected its flight performance negatively 😠. Great video as usual 😂

I love how when I watch a Chris Rosser video I fill like I got some college credit towards my engineering degree.

Hi Chris! Great video, I'm always impressed by how well you explain everything at such a level of depth. You say in the video that the MPU is still in production. Just curious, how did you verify that? It's interesting that the oft-repeated "common knowledge" is that the MPU6000 is out of production and we're getting "inferior B-stock" now on FCs with that gyro. It would be kinda amusing if that was just a myth...

For Drone application, you should look at the BMI088 instead of BMI270. With BMI088, you are able to resist the vibration frequency hook to system over the cut off frequency (230Hz with 2000Hz ODR).

why the recommendation of biquad over pt2 for the bmi (25:25) ? Im curious as I dont really understand the differences

My favorite gyro is the icm20601 (not 602), why, it never gets confused on any of my aggressive maneuvers. Even if I lower my rates to 1900 (on the 602 gyro), I can still get the gyro to flip out, in these situations it switches the sign of the output data, that puts the control loop in a unstable positive feedback loop that the quad in a full thro spin, only disarming is the only option at that point. It's a form of spin to the moon. And enabling the anti-taz (hate that name since I created the tasmanian devil trick), basically disables many of my trucks (yaw launch, throw launch, to name a couple). Another reason is the ACC is good for 32G's, great for performance testing.

Thank you for this awesome video, Chris. I really doubt that BMI270 undocumented OSR4 feature will be used in any “production” release of the FC FWs, as it may disappear in the future production batches, which may cause a lot of problems.

In my years of flying quads at the highest level of Freestyle performance, I can say that the MPU6000 for me has been a total winner. I tend to do alot of on-purpose building tapping, rail riding and I subject the quad to alot of blunt force, tapping stop signs and other things. I fly hard. ALL the flight controllers that I own with ICM 20602 gyro have all had gyro failures after 4-6 months of flying. For me, the ICM 20602 is a waste of time and money because they lack durability and reliability for my flying style. The ICM 20602 flies very nice in the beginning, but after a few months they get twitchy for me and make the quads fly like junk. I mostly fly Holybro flight controllers and all their new models have switched away for the ICM 20602 to the MPU6000 and that made a huge difference in gyro durability & performance for me. I do have some un-used New Radix 2 flight controllers sitting on the shelf that I will be using for the first time this spring. Looking forward to testing out the BMI270 gyro for sure. Chris, I truly appreciate all the time you put into your testing and production. Thankyou for this video.

Hey! thanks for your work!!! helped a lot!!! there is one question: what usually makes noise at 40khz

oO ...seriously need to change my shorts now, just an swesome piece of work, as everytime :)

Hi Chris thanks for the amazing videos you do for the community. I’m having a hard time finding mpu6000 in uk at the moment. I have fc here with the bmi 270 and mpu6500. Are these any good for a race quad? Been struggling with bmi for months with oscillations,trilling on my quads but can’t seem to stop it. What do you recommend please.

Please more videos like this

Great video Chris. Is the OSR4 mode now available for the BMI 270 in the latest versions of Betaflight? I'm wondering whether I should get a FC board with MPU6000 rather than the BMI 270.

If you search the mpu6000, it comes up as discontinued from pretty much everywhere that may have sold them. The manufacturer may not indicate as such but it definitely appears they are, as most people are saying.

very nice :)

Thanks!

BMI270 6.4K ODR is already implemented in betaflight, but only for chip flash space >512KB. And in order to do that the BMI270 must be set to FIFO mode instead of "normal pulled" mode. In that mode there's not any DLPF at all. This is different compared to ICMxxxxx which still has DLPF in 32k ODR.

Hardware DLPF options are also available for the MPU6K series as well as the 6060X series … we have known about these functions for a very long time ..

Hi Chris. Recently I've built AOS 5 in the same config as you have in your AOS 5 video (SucceX-D Mini F7 TwinG Stack (Mini F7+45A ESC), iFlight XING2 2207 1855KV, ...). This is dual gyro FC with ICM20689 and I have to say I have only bad experience with this gyro. Quad twitched at BF 4.3 defaults and no filtering tweaking helped. BB showed random "unnatural" spikes in gyro data. I tried only first gyro, not good (spikes in roll and pitch axis), only second, not good (but now spikes in yaw axis). FC soft mounted. I tried hard mounting - even worse. I bought another FC (the same type, but different date of production, different batch). But I had the same issue! Power is properly filtered (big cap). However when FC is just hanging in the air (holding in hand) twitching is gone on both FCs. So I tried FC with MPU6000 (there are just a few in 20x20mm config with DJI plug, so I went with HGLRC Zeus F722 Mini). Problem solved! (everything else on the quad stayed, I just swapped the FC). => either I was extremely unlucky or ICM20689 are not good gyros. My guess, from all the clues I have is, that the tiny MEMS structure in ICM20689 is susceptible to "oversaturation" with the motor noise motion when we use it in our quads (all of these gyros are primarily designed for smart phones, right? We "misuse" them in quads). For me it's definitely only MPU6000 from now on! Thanks for your videos, Chris, you're one of the most important contributor to the FPV community nowadays! Your scientific approach has no match! Excellent!

Would it be worth oversampling at an even higher rate? For instance, what would happen if we had a target data rate of 2kHz and oversampled to that? At what point is there diminishing returns? Is filtering and oversampling within the gyro more efficient than filtering the gyro output in betaflight?

very interesting, I just checked out the new mamba board, it features two ICM-42688-P gyros. On the first look, they seem to only offer 32KHz of sampling but also a higher ADC resolution. I wonder what the frequency in Betaflight will be

So is this the preset settings found in some BNF's or premade drones. Like my iflight drone came with BF V4.3, but they dont have a preset/ custom tuned fw version for this on my drone yet. Is this the type of stuff ill have to look into if i update without a preset available for BF v4.4 (iFlight Nazgul5 XL5 V2), when i do the update? Or is it more the PID tuning and settings like that? Great video! lots of good information. i dont have a mpu6000 in my new FC and it feels pretty much the same, but when i get a chance i will watch this again to check and see if i need to change any filtering, etc like suggested..

Hello. Thanks for the great lecture video. I recently built a geprc cinelog30. I felt the screen vibrate, and looking at the black box log, the noise was very severe at about 268hz (motor noise). Even if I apply the low-pass filter, notch filter, and rpm filter, the noise around 268hz does not disappear... Is it a problem with the bmi270 gyro itself? Any solution? I used bf4.3.1 version.

I wish there was a way of lowering the 250Hz cut-off on the MPU/ICM gyros, say by half (similar to what's possible onthe BMI gyro), even at the cost of doubling or even quadrupling the noise spectral density....

Hi Chris, I'm having alil trouble with the filtering for the bmi270 I have a a single biquad lpf 2 of 250 like you suggest in this video and the quad flys better than it was but I'm still getting high throttle vibes and bounce back on flips and rolls.. I'm using the supafly preset and on an identical quad with the same stack (except the gyro) and it flys perfect. What do I need to do with the filtering and do I also need to adjust pids with the bmi270. Tia

Cool

What about the range settings? In addition how well do their onboard bias and scaling calibration perform? Please do a review for accelerometer as well. Thank you.

Excellent!! 👏👏👏 I just bought a FC BetaFPV Toothpick 20A V4 with gyro BMI270. The manufacturer recommends removing the GLP1 and activating a GLP2 with cut-off frequencies at 135 ... In some posts they recommend applying the official preset for BMI270 that I no longer find moreover there is another preset that shifts the PID frequency at 3.2khz ... from your video I understand that with OSR4 activated, none of this is needed anymore. I'm not understanding anything anymore and I no longer know which of these settings to apply .... So since I trust your opinion a lot I ask you: OSR4 is finally inserted by default in betaflight or do you have to do something to activate it? Once activated, do you still have to apply some filter such as the GLP2 at 135khz or is it no longer necessary? If there is OSR4, should the PID frequency be set to 3.2khz or not? thank you very much and sorry for the language but I use google translate

hey chris, i been having a few issues with 2year old mamba f405 gyros, i dont feel much jitter when i fly (or mistake it for my thumbs) but when i try to use rsgo or gyroflow it struggles to stabilize at all, lots of rejected flights. so i watched back raw footage and noticed quite a lot of regular wobble, then checked gyro in betaflight and saw that it was gaining -1 degree yaw every 6 or 7 seconds, plus roll and pitch were having periodic micro twitches. so i blamed my footage on a bad gyro and ordered a new fc from unmannedtech, i plugged it into bf and to my surprise it also has the -1 degree yaw thing, though a bit less frequently, not so much issue with roll and pitch, i checked on cli status and its an mpu 6500 on the new FC, is there anything i should do? is the new FC ok to fly? do all gyros lose a degree ? i should add that this loss is cumulative, its not back and forward, its always minus, so after 5 minutes its -10???

Chris I have a question for you I'm running a ultralight toothpick with 2004 2900 KV motors. When I fly with the 3's battery all up weight 233 G the noise plots look perfect there's no bad frame resonance and The Dterm filtering is clean. When I run the same quad on a 4S battery all up weight 250 grams I get a huge resonance problem on the pitch axis only right around 200 Hz and the Dterm has a huge hot spot. This is on stock filter settings. my question is is it the extra voltage somehow causing this or that extra 17 grams of weight somehow doing this?

I recently bought a iflight evoque f5x and my first one the esc went bad within the first 4 flights. The replacement that came didn’t fly the same with the same pid values and noticed that the first one I bought didn’t have the same crispness in turns so I went to play with the PIDs and noticed the PID loop was at 8khz on the original one I received and the new one is at 3.20khz. Because of this video I found that the first one I received the gyro was a MPU6000 and the replacement is a BMI270. Thankfully I was able to switch the FC before I sent the bad esc quad back. Am I crazy or is there that big of a difference in the feel of these gyro chips? I’m 3 months into this hobby so I’m still new at this, could I eventually have gotten the BMI270 tuned to fly like the MPU6000 or with the lack of refresh rate with the BMI270 hopeless? Thank you for the information in this video you got my subscription.

Ciao Chris ☺️ ho visto tutto il tuo video, è cambiato qualcosa? Su bf è stato già fatto un preset? Sarebbe bello semplificare il "passaggio" a questo nuovo chip per giroscopio in maniera più Easy ☺️

Hey Chris I've got a Mamba F722 with Icm42688-p. Any ideas how to get high throttle noise away. FC is soft mounted, bidirectional D-Shot, stock PIDs and filters from Betaflight 4.3. Motors are Geprc 1507 3600kv. Frame is Geprc Cinestyle 3". The quad flies good. But when I raise the throttle fast above round about 80% the quad is shaking like some motors stalls for microseconds. I bet for heavy gyro noise?!

How much delay does each gyro & mode introduce? I'd assume filtering will be done very time efficient this close to the hardware but it will introduce a delay nonetheless. Negligible?

Hey Chris, why do you still recommend old Biquad over PT3? PT3 promising less delay and same performance as Biquad... PS: thanks for this awesome comparison ✌

I'm wondering if OSR4 mode adds additional phase delay and if so how does it compare to a LPF @250hz in BF? BTW, this one video alone is worth the price of Patreon for a year in my book.

Great explanation why 32kHz mode in FC is a technical nonsense. 24DB more noise in usable bandwith is a lot too much. Now it is obvious why they like to fry motors even with slightly biten or bended props.

Is there anyway to test the accuracy of the Gyro against a known input - like using an audio speaker? Is the on chip filtering more accurate - or lower latency - vs the betaflight filtering... from the way you described it - it seemed like it was mechanical filtering - you have two sensors and you filtering out the noise between the two (am I correct?).

Great Video I want to ask which version of the Diatone Mamba basic h743 I should get the BMI270 or the dual MPU6000 version

It would be awesome if you could compare some flight controller cpu as well

Hey Chris I have the bmi270 and I run the gyro and d term sliders at 1.4 and it flys great. How will I know if there is noise issues I don’t have no twitches or glitches during flight

Chris, can I ask you a big question? With regard to the software processing that's happening on the gyro chip, do we have any way whatsoever of knowing what theyre doing under the hood? I've been very curious to know whether their algorithm makes any use of the onboard accelerometer during the gyro output processing. Perhaps they use it as a cross check during oversample averaging as a noise reduction method. How would we know if they did this?

Chris, Do you mind signing my attendance sheet. I would like to get some college credit from this lesson. Seriously this is great info that I have never seen anyone put in a way that can be understood. Reminds me of back in the day when Bardwell would create a slide show to explain how dShot motor works. Thank you for putting this together.

Very good explanation about how the gyro work. So, about the BMI270 on betaflight 4.2.11, do you have a preset filters, or where i can find a good one. Thanks a lot

sorry im quite new to fpv and i just got a fc that has a bmi 270 gyro in it , is it true that if you install BF 4.3 on ur fc, the program automatically integrates these low pass filter settings into the fc? so im assuming we dont have touch the filter settings on bf?

The SPRacing H7RF uses ICM42688P gyro. Any idea how that stacks against these?

Hi chris how about icm42688p is there any good?

Since BMI270 is locked at 3.7khz should we run dshot300 since BF devs recommend dshot600 for only 8K/8K?

I have the iFlight Blitz F7 V1.1, with the BM1270 gyro. I backed it up from new (Dump & Diff All) and updated to BF 4.3.1. Applied my settings, and picked the available BM1270 preset. Now I'm getting the 'NO GYRO' error. Flashing back to factory 4.2.11 and factory dump/diff all is not solving the issue. Is it dead? Please help me!!! :(

My FC has two Gyros. The MPU6000 (SPI) and the ICM20689. Should I use one or both?

I just got a fc with a bmi270 and I’m running betaflight 4.3.1, how do I enable osr4? Or is this already implemented automatically?

bmi270 on my quads have to push filters atleast 1.1 but the bmi270 seems much much smoother then the old mpu6000

Manufacturers do not leave a mode out of the datasheets for no reason. Sure, manufacturers do reuse logic and code so quite often there are common features between a range of products. But, chances are that there are or were issues with that particular feature on that particular product. There is also no guarantee at all that they won't disappear or change function between revisions. So, why would I want that OSR4 mode? It is unsupported, possibly unstable in the future, and the same can be accomplished using filters in software which we already have. Even better: those filters adapt to flight conditions as much as possible, something that hardcoded filters in the gyro (or any other piece of equipment) cannot do. On gyro choice: from a flight performance perspective I cannot feel the difference between MPU6000 and ICM20602. However: I tend to use my stuff until it is really beyond repair, and I have the repair skills to really stretch the life of quad hardware. Rewinding burnt stators, fixing broken ESC's, reanimating butchered flight controllers. Now, I have several MPU6000-equipped flight controllers that are so butchered that I had to solder wires to the CPU to replace PCB tracks coming loose. They still perform fine. ICM20602-equipped FC's never reach that age; they tend to develop a twitch long before that. Do you have any information about the durability of the BMI gyros?

I think there was an error exporting the video, there seems to be 2 summary slides at the beginning with duplicated audio. :)

What do you think about using a absolute orientation gyro's like the BNO085. Often wondered why they are not used.

Isnt oversampling the same as using a low pass filter?

ppl said dont conformal coat the gyro, only conformal the sides (legs).. is this true? willl conformal coat on top gyro causes noise/weird vibe?

I'm looking at a new Gyro H743 Mamba, it looks like they will sell two options a Dual MPU6000 or a Single BMI270, based on this video I assume Dual MPU6000 is the better choice?

Did someone try to oversampling 8khz => 1Khz in betaflight ?

Once u added those changes u referenced in the video on the bmi270 gyro u basically said it performs slightly better than the mpu6000, but then at the end I said the 6000 is basically better? Are u saying that based on stock settings alone but with the changes the bmi is better? Also those low pass setting u suggested are those the osr4 changes you meantioned or is that something else that’s still in the works?

Do you know how ICM20689 compares?? MPU6000 is discontinued so this research is very much needed these days, thanks for that! :)

re: oversampling page; is that mean averaging rather than mode average/median? Too few samples for mode but median?