BLHeli32 100% Explained - Part 2 - PWM Frequency 

haha JB got me good. was ready for an hour. got 16min.

ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-bAqgPEW48Tc.html

Hello from the future! 2 years later and variable PWM is now possible as of Blheli_32 v32.7.4

A throttle curve would certainly work, but I also think you can just get used to it after a while. You can also try dropping the PWM rate back to 24 (which is what is used on BLHeli_S), and see if it changes things at all and if you develop any mid throttle issues. You can experiment with different values between and see if you can find a sweet spot. Give it all a shot and see what works best for you :D

Same here man.

@Quasi Maji Dude, my memory is unbelievable good... Well, except for names. I have difficulties in tests when I have to write down my own name. O.o

That's an interesting point. The faster the FETs switch, the hotter they will get. I think it would be very hard to control for environmental variables to measure this though, especially because the ESC would have to be driving an actual load.

Another issue with flir is it is very hard to get reliable and comparable data. Small things like surface reflectivity can make a difference in what temperature the sensor reads even if the actual radiated energy is the same. Flir is decent to compare temperatures against the surroundings, and as a broad strokes kind of measurement, but not if you're trying to look at specific data. I was looking at this as an additional data point for MQTB, but ended up abandoning it for those reasons. It would be interesting to see where the hotspots move as the switching frequency changes though, but I'm not sure if FLIR would be accurate enough to be able to detect a measurable difference. Or even if the difference in PWM rate is enough to dramatically change the switching losses. Higher switching losses implies lower efficiency at the motor output, but in my testing so far I haven't seen a dramatic difference in efficiency at different switching speeds. I did some back to back testing when BLHeli32 first came out. My guess is that the differences in switching losses is minimal compared to other losses in the system, but I don't have enough data to say that for sure.

The boundary conditions need to be carefully adjusted but it is a standard procedure in any industry .. why FLIR exists ... and it’s certainly much easier compared to maintaining equal condition for vTX testing :) But maybe a theoretical consideration (high level) should do it. I was just missing this downside of simply doubling switching frequency to be shared with the community because in can result in a burnt ESC. Double switching frequency means double switching losses. Assuming 50% duty cycle and 2/3 switching losses, the overall losses are increased by around 30%. Assuming you have 2W of losses in each MOSFET leading to 120C junction (depending on cooling conditions) temperature, a third more losses is a lot because it means a third higher temperature. The essence is, for those you have maxed out their ESCs with bad air flow for cooling, doubling switching frequency carries a risk that should not be ignored.

Definitely worth looking at closer. I have almost never seen an ESC fail due to over current or overheating in flight, even when they're stuffed in the middle of a build with no airflow, they typically barely get warm even in very heavy applications. Generally speaking the only ESC failures I have seen have been from either something else wrong in the system (bad tunes, etc), physical damage, or blocked/damaged motors. I've run some crazy setups on 20A aikon ESCs and can't get them to burn. I would think if going to higher frequency PWM increased the losses and heat saturation to a concerning point we'd see a corresponding rise in real-world failure rates at 48khz, but that hasn't been the case as far as the general community feedback goes. It's possible something else is mitigating the increased losses. Like I said, I measured no significant loss of efficiency from the motor output perspective at 48khz vs 24khz, which you would think would show up with a 30% change in switching losses. Maybe we're missing something? On the other hand it could be related to the low end torque losses we're seeing when using higher PWM frequencies?

Experience is also a good source:) In the same package ESC makers can get different Ohm-values which makes ESCs easy scalable at relatively low price adders. Seems there is quite some headroom by design which also makes sense ... an overheated ESC causes damage in reputation which the hole FPV community will get to know as soon as one of you guys have such a failure :) - Thanks for the quick feedback. I really appreciate this.

I hate to tell you but this is the exact opposite experience of almost everybody who tries it. If it doesn't work for you, no problem. But you are definitely the exception.

@@JoshuaBardwell I'm not really an exception, some other pilot give me the tips on emuflight discord. Looking for this wooble since weeks and others give me this Tips. Jump back to 24k and no more wobles. Maybe on beta but not on ému for me. And this is Real on my 6" / 5" and cinewhoop so 3 différents quads same issue same result. So why having this issue on 3 type of quad ? I mean wobble on 48K ?

I think you're thinking about the multishot frequency, which is between the FC and the ESC. We are talking about the PWM frequency between the ESC and the motor, which is totally different.

Thanks for the answer, I found the post, it was made by the Helio FB Account.- Saturday PRO-TIP: On blheli_32, If you’re getting weird oscillations you’re having trouble tuning out, make sure your PWM rate in blheli suite matches your pid loop speed. 48kHz PWM rate actually can create an EMI problem that can be resolved by lowering it back down to 32kHz, 24kHz(default), or 16kHz. ESC manufacturers, such as Aikon, are aware of the issue and have taken that into consideration while testing new 6s prototypes.

I have never tested a system with two separate transmitters before. It's possible that the external module is powerful enough to interfere with the weak telemetry signal coming back from the model.

Definitely not true. Higher PWM does not make it more likely to smoke the ESC. Maybe this was once true with older ESCs that were too slow to handle the faster PWM frequency but not any more.

@@JoshuaBardwell Actually, higher PWM does increase switching losses on FETs, so it's more likely to smoke the ESC. It depends on FET quality, PCB layout, etc. I think it's not an issue for these 24..48K speeds, it's more likely to get noticeable effect when you close to 100K switching frequency...

Use Dshot. Then min and max throttle don't matter.

I am not 100% sure but it sounds like you're right. 6S on 2400kv is a pretty unusual situation and it's plausible that you are exceeding some design limits. Are you flying on Dshot1200 and is it working for you?

Yes. I have a rig on 6S with DSHOT1200. Before BF3.5.1 I could only hit 50% throttle before stage one failsafe and it was horrible to fly. Now I can hit 80% throttle before small oscillations start showing up. Other than that, the quad flies like it's on rails. Feels great, sounds great, cold motors, cold ESCs, and the blackbox is fairly clean. I haven't tested full throttle as 80% feels like full throttle on 4S with what appears to be less amp draw. I've been out of the air for 6 months so I just wanted to fly. I started tinkering again (updated to BF4.0 dev build, tried different props, and looking at ESC settings) to see if I can clean it up. I just moved to DSHOT600 also. I'll see how it flies this weekend. My first 6S 2400kv build was perfect. This one is taking awhile. I like the 6S on 2400kv because I tend to use less amps and I don't notice any voltage sag. I also have extra punch if I need it. BTW - Thanks for all you do for the hobby. I started watching your channel when you first started. Well done.

#UpdateFirstAskQuestionsLast

Try something between 48 and 24.

@@JoshuaBardwell I will, thank you, although I think that the creamy feel will fade away

yes... and no, but yes.

No, it's not possible. The numbers in the blheli_s hex files related to the dead-time of the FETs. But if you flash the wrong one you'll possibly fry your FETs so don't fool with it.

Mhhhhh, Blheli_S doesn't have this PWM option... sorry....

Correct. This is one of the big reasons to go to blheli32.

MM75 blheli-64 😂