Dshot1200! Is It better than Dshot600 and Why? - FPV Questions

Uncertain about what to buy? Check out THE ULTIMATE FPV SHOPPING LIST: www.fpvknowitall.com/ultimate...
SUPPORT ME ON PATREON at: bit.ly/2Ep5Sia
THIS IS MY FULL TIME JOB. Here are other ways that you can support me: www.fpvknowitall.com/support-me/
My main channel: / joshuabardwell

[@deice3]

Ok, I laughed when I realized the difference was in microseconds. Placebo is one hell of a drug.

[@IvanEfimovLimon]

Latest BF sceduler tries to update motors after the PId loop as fast as possible as far as i know. So if you are running 8k + Dshot600 motors will get an update every "tick". And there was lots of efforts from Steve to make sure there are no skips or anything like that. So the moment BF actually has a new value for motors it will send it. 8k times per second like a Swiss Watches it is sending a NEW value to the motors. So the frame is small:125 uS. Now Kiss running at 1k pidloop. Regardless of the dshot speed that's 1k new motor values per second only, that by itself is a 1 milisecond frame. Running faster DSHOT on Kiss maybe make sense to minimize latency added on top of 1ms frame. But speaking of gyro->motors latency it will never come even close to 8k pidloop on BF till they change PID loop to faster values.

[@ItsMeKevinFPV]

tnx. JB

[@martincreator2960]

i find it interesting that the going from dshot 600 to 1200 made my drone feel much different. That was the only thing changes a while back and i noticed the drone felt like it went from feeling like a 2207 motor drone(Dshot600), to a 2306 motor drone(Dshot1200). best way i can describe the difference was the throttle resolution was better/smoother on Dshot1200 yet the drone didn't quite feel the same as on Dshot600.

[@Dav2112]

I personally don't notice a difference between 600 and 2400 on my full KISS and FETtec quads, I just use 2400 on them because it's available. The guy who said you can hear the difference, if he's running full KISS or FETtec that will be the difference in sound he's hearing, KISS has a different sound to a BF/BLHeli32 setup regardless of Dshot speed.

[@Dav2112]

@@sullytrny you're probably right mate, I wasn't looking at flight times when I've tried out different Dshot settings, I was just curious whether I could feel any difference which I couldn't.

[@RobertLeclercq]

Coming back to this hypothetical test, it would have to be done in person. Audio recordings might work, but the differences I hear when running higher Dshot levels aren't massive. It very well could be placebo, but if I find a friend to help I'll try to post some blind tests.

[@Kajedkid]

Pretty sure KISS uses the sine-wave modulation among a few other differences ......

[@dmyers9230]

There is 1 basic truth that makes higher rates have no value, the motor can not change its speed that fast. It’s bell mass and the mass of the prop makes it physical impossible in the uS time frame. The ESC may be able to command it, but the Motor is not going to be able speed up or slow down to the new RPM that fast. It may however cause more current spikes, which is not good; more heat if anything.

[@garwol]

Its not about stick latency, its more about latency from gyro to motors in propwash situations

[@alexanderfedorov7890]

DS2400 performs better on ultra. Thats what we see from reality. When you get to 13% cpu load (you are closer to gyro read), you want to keep latency as LOW as possible. Most people feel the difference. If you load cpu 70% - you practically don't care. :) 1wire by definition slower then ds2400. Motors will have same latency, because all esc will process packet at the same time when it arrives (due to CS calculation). But what do i know, we still rocking 1khz pid loop on H7 :) haha :) Also, FYI, kiss/ultra DO NOT use anything but dshot. Its one way fire and forget. No bidirectional or anything like this. Also filtering not included any esc info.

[@lj516]

Motor signal speed should be at least double that of the control signal. If not you could be "waiting" for a motor to turn 120deg before it can fire again. This is the same concept that EE's use for measuring a signal with an o-scope in that the refresh speed of the scope needs reasonable headroom over incoming signal speed so that it can be "caught" within the right window.

[@JoshuaBardwellLivestreamClips]

The argument goes that the motor itself has about a 10-25 ms latency due to its own mass. This can be measured by watching the change in Dshot motor output vs. the change in motor RPM as measured by bidirectional dshot. This corresponds to about 40 Hz. So the 2000 Hz or 4000 Hz Dshot signal is more than sufficiently over-sampled. I know not everybody agrees with that argument but that's the argument that some people make. EDIT: I just double checked Chris Rosser's motor testing and he found about 30 ms latency for acceleration to 90% of RPM change. My 10-25 ms latency memory might be time to the beginning of change.

[@vonpwnz0r]

8K pidloop is 125 uS (microseconds). DShot150 is actually fast enough for this at 106 uS. The original recommendation when DShot was introduced was to use one level faster, that is Shot300 for 8K pid-loop (The original recommendation was from before DShot telemetry, I don't know if that changed the speeds).

[@deice3]

I believe in the video it was mentioned that that's per motor, so for DShot150 its 106*4=424uS. With DShot600 that becomes 26.7*4=106.8uS which is fast enough.

[@vonpwnz0r]

@@deice3 The original Github documentation says: DSHOT150: 4kHz max DSHOT300: 10,6kHz max DSHOT600: 16kHz max DSHOT1200: >32khz max So DSHOT300 should be enough for 8Khz according to the devs. Any caveats? Yes, I don't use telemetry myself and I don't know if that impacts the speed - the original documentation was written before DSHOT telemetry obviously. In the whoop-world where telemetry is rarely used and CPU/MCU resources are limited we use only what is exactly needed, I would never go higher than DSHOT300 for 8Khz

[@deice3]

Interesting, I wonder where the discrepancy comes from, it seems to make sense you'd have 4x 16-bit throttle values sent per pidloop. But I know nothing of the internals, so I'll certainly believe it if the devs say that's good enough.

[@vonpwnz0r]

@@deice3 DSHOT600 is good for 8K as well. If you are running a modern setup (non-whoop) there's no reason not to run DSHOT600

[@alexanderfedorov7890]

@@deice3 wrong. they run parallel and start at the same time for all motors.

[@MCsCreations]

You know... Perhaps we would be using only F7s and H7s today, if it wasn't for the pandemonium and chip shortage. But that's the reality we live in. In that case, perhaps we could use higher PID loops and DShot.

[@memed_sipalingfpv]

my drone makes a strange sound when using dshot600 and motor more hot. and it's been fine after i used dshot300.warm motor My pid loop 8k

[@IvanEfimovLimon]

Addressing the claims that someone "feels" the difference between KIss DSHOT600 and DSHOT2400: its a close source firmware, we dont get to know whats in. I surely can modify BF code that it will run worse on DSHOT600 than on DSHOT2400. But that does not proof that DSHOT600 is worse, it only will proof that i made it worse in my "Firmware".

[@pilotlawaran]

If i have esc with support dshot600, can i change protocol to multishot? It means support dshot so will support multishot too?

[@JoshuaBardwellLivestreamClips]

Yes, but multishot is in every way worse than Dshot. It's an analog motor protocol and needs calibration of your escs. - Blunty

[@pilotlawaran]

@@JoshuaBardwellLivestreamClips i think my build will fail before it fly, I want build x8 with matek f722-se, support 8 signal motor, but if using dshot some people say that DMA crashed will happen, so all motor will work if using multishot, n I don't know how to calibrate multishot esc 🤕

[@RCOverKill]

A friend demonstrated that dshot 1200 is noticably faster at arming(motors start sooner). Since bf doesn't support dshot 1200 anymore I don't know how you would see it now.

[@flapjackfpv]

Most likely it’s the advertising on the ESC when you buy it. And you can run emu flight which does have 1200 D shot

[@RCOverKill]

@Russell Phelan I'm sure he ran stock bf with stock settings on blheli32, and I did the same with blheli s. We armed at the same time, and his would start up 1st, I has just started, and he said he noticed this on his other quads, and that's why he showed me. I have no idea why it started 1st, just what I was told when I was shown. We were both using xm+ rxs, and same txs, it's not something I tried to repeat, just thought I'd mention it.

[@RCOverKill]

@@flapjackfpv it's just something I was told, 8 never tested it, or even really thought about it after that, his quad didn't fly any better than mine that I could tell, but I was new.

[@alexanderfedorov7890]

@Russell Phelan esc packet transmission time changes.

[@TimeFadesMemoryLasts]

In theory Dshot 2400 is a bit "better" than Dshot 600. BUT: tuning the filters will save you latency in the milliseconds. Going from Dshot600 to Dshot2400 will only save you microseconds (100x-1000x less latency improvement than tinkering with filters). So it's a question of "do the ever shrinking benefits of higher Dshot outweigh the exponentially growing problem of running higher looprates?". And at some point when going to a higher Dshot number, the improvements are practically 0. For BF that's considered to be around Dshot600.

[@TimeFadesMemoryLasts]

@Russell Phelan I never talked about "feeling delay" but actual delay improvements (and therefore better PID performance: more phase margin -> better PID stability). But I should make myself more clear. There's one important detail I forgot to mention, sorry: the motors on Dshot2400 can read/write 4 times in 125us, while Dshot600 can only read/write 1 time in 125us. That means you can increase your gyro loop x4 to 32kHz, which means the filters will have less delay. Why is that? (and now comes the important detail =>) Digital filters have practically the same delay as the analog counterpart BUT they have N samples of additional delay for an Nth order filter. So for example a PT1 filter (1st order) will have 31.25us of "extra delay" at 32k, while at 8k it will have 125us of "extra delay". This means running Dshot2400 and 32k gyro gives you 125us - 31.25us = 93.75us less gyro delay. So rather than what I said in my previous comment it's 0.1ms vs. milliseconds, so the benefits are bigger than mentioned by me. But still, this small improvement comes at a heavy price of having to run a stable loop 4 times as fast. So adjusting the filter cutoff just slightly will already outweigh the delay benefits of running the filters at 32k looprate. Of course it's debatable when "enough is enough".

[@martincreator2960]

even though the differences seem minimal on paper, in practice i noticed huge differences. One Time i Flew on Dshot600, then i went home to try Dshot1200 as i watched some vids and was like, maybe it will be better, after all who doesn't want lower latency i guess. I Changed the Dshot to 1200 then when i wen to fly, it felt much different. Best way i can describe the difference is like a 2207 motor drone(Dshot600), vs a 2306 motor drone(Dshot1200), where on dshot1200 the throttle resolution was better/smoother. But the change made me fly worse as it just felt different.

[@TimeFadesMemoryLasts]

@@martincreator2960 Observer bias is a hell of a drug

[@LukeRT]

Isn't Dshot 1200 removed from betaflight altogether now , 600 is the max

[@k.van_niekerk7228]

There is proshot 1000

[@k.van_niekerk7228]

It's on f7 chips

[@manuelsolano7525]

My God Joshua how much data can you put on that brain of yours very smart.