I'm confused about how we can determine anything about the software if it is running on different hardware. wouldn't we need the same hardware running the different software to see any difference between software? I fail to see what was accomplished here. I appreciate the time but, I'm not sure that this is a good way to see differences in software.
@user-sh6pc3on4f I understand that you are a fan but it makes no difference. This is still not a accurate measure for software comparison because it's running on different hardware.
The problem I see with using a flywheel for testing is that you load the motor at very low rpm where a prop would still have almost no load, this could be where the difference between firmware is showing up, with the blheli 32, it may be, for instance, that the load at low rpm is just not being detected as a normal operating parameter.
I am kind of curious how the old tgy code compares these days! It sounds like the major change is variable PWM frequency ("ByRPM"), which would be fairly simple to add. Meanwhile, dshot would not be fun to implement -- I avoided it as it meant many more interrupts. I did test ludicrous update rates by furthering the "onshot125" concept, so maybe it would be fine. Anyway, good to see that things are still changing and improving, and measurement is key!
I'd really like to see a video over driving these escs with higher kv motors run like a strong lower pitch prop on 2300kv maybe even a 2450kv on 6s to actually push the ESCs before they desync or whatever, maybe even using a 3000kv 2004 motor with a 3.5" prop on 20mm escs. I'm not so much concerned about 5" drones, but getting max performance for smaller light weight drones since RID had kind of forced us to look at smaller drones more
I'm actually waiting for a 40a f435 at32 aio to get here on the slow boat from china to run 1505 motors 3450kv 4s motors on 6s. Not sure how it will do, i tried it on a f411 40a aio with blheli S (ghf411aio)and it just cuts out on the high throttle, the f435 aio was much cheaper then a f7 aio with bl32, if the at32 escs can't take it then i will get the f7
I would have liked to see a more even playing field with as close to the same esc's as possible. Skystars makes a 55a in Blheli-s, blheli32 and am32 flavors I'm guessing they all use the same fets and board layout. They could probably provide the best results as there are definitely some differences coming from the fets, board layout, copper thickness, heatsink, etc.
i got 122 mph on a azure power 5048 with supernova motors. ALSO with real flight I noticed a MAJOR difference on my Speedybee bluejay and blheli the bluejay sagged the battery faster and not hitting top speed and the blheli went super fast BTW speedybee f405 stack has that nice esc for 70$
In practice, you aren't going to be at zero airspeed at 80% throttle, so you probably wouldn't see the same behavior as CR. That's one of the problems with bench testing at zero airspeed.
@@aviphysics that's why I am starting to do real world reviews and making my own charts. I do use all of Chris Rossers data to select my products but I have notice some differences that's really vital and important that's not caught on the bench
Interesting vid Chris! thanks for doing this work. I have one question about the torque vs acceleration results. Since torque is proportional to angular acceleration, how does blheli32 do worse in the torque test, but then do so much better in the acceleration test?
The reason for the poor blheli32 performance in the torque test may have to do with the "Low RPM Power Protect" setting. I tried to power a quadcopter with 15" props with T-Motor 3510, 630 kv motors and the T-Motor F55A Pro ESC with BLHeli32, and I got abysmal performance, thrust was pretty much flat above 25% throttle. I am fairly new to all of this, I spent two years (!!!) pretty much replacing every part of my setup, figuring that each component individually was working fine, but the combo of motor and ESC just didn't work. Then I set "Low RPM Power Protect" to "off" and, "wroooom", I suddenly got full power!! Seems that at low RPMs, blheli uses a heuristic to determine whether the props are stuck by looking at "unusually high" current draw. And for a firmware written for small FPV quads, the current drawn by a low kv motor with large props, and probably by a flywheel, is unusually high. I am trying to spread the word about this as I find it hard to believe I'm the only one who get stuck with this...
Yeah, that's a very good point. I was surprised that deadtimes were not considered at all. For BLHeli32 vs. AM32 easily the same ESC could be used, one stock on BLHeli32, the other one flashed to AM32 to at least level the playing field. This way only two platforms would be needed, one for 32Bit, one for 8Bit - ideally they should also have the same dead-times.
Keep the info coming, Chris! At 4:07 I'm seeing that at 10% throttle, a single motor with a 5" prop generates 800g of thrust. If that were the case, wouldn't all our quads just take off the instant we arm, with 4 motors generating roughly 1.6kg of thrust at 5% idle throttle? It also seems really strange that 100% throttle is only 2.25x the thrust of 10% throttle. Something about this data doesn't seem right to me.
Chris is just using proper SI units for Force (= mass x acceleration). The unit of the thrust on the y-axis is gF or gram-force which is a product of the actual grams of thrust by the acceleration due to gravity. So to get actual grams you'll need to divide by 9.8 m/s^2, i.e. 800 gF = 89.88 g of thrust
The ESCs differ also by the dead times that the FETs need. How much of an influence does this have on the parameters? Ideally (I know it it a wish) we would compare the ESC firmwares with HW / settings realizing the same boundary conditions.
I think this is a pretty important consideration - I know that some ESCs have very long dead times compared to others. I think some of the Holybro Tekkos have very short dead times and some the AM32 targets / ESC have much more conservative dead time ratings, in some cases at least. It would be nice if BL32 exposed the dead time setting like BL_S, Bluejay, AM32 (assuming you install the firmware) so you knew what they were set at.
What 32 bit 20x20 esc would you recommend for the flywoo Explorer lr4 that's lightweight, i am considering the flywoo goku g45m running am32 but i have not seen any videos on it. I also will be using the Holybro H743 mini fc and arducopter or inav and flywoo 1404 2750kv motors. This drone will be for long distance flying and cruising. Thank you from America 🇺🇸
There's a current limit and rampup power, both of which affect that max torque. I'd really like to see some of these comparison tests with that set to different ones - eg make sure the tests are done without the cirrent limit enabled, then try the default rampup power and compare it to the torque with max rampup power selected, and then compare those two results to the other ESCs' results. (And do the same thing with the other firmwares.) Otherwise what we're really comparing here is the ESC firmwares' default settings instead of the ESC firmwares themselves (which you kinda touched on in the video). Also, how is the acceleration/deceleration for Blheli_32 so high when its torque is lower? I would've thought those two would corellate?
I think this is a topic you should continue investigating. Though it may be impossible to nail a winner, we as a community are aware that there is no such thing as a free lunch. We just want to know what might suit our flight style.
I swear by BLHeli_32 for "3D" flying. I am really curious about Motor Rampup and other settings to get acceleration on AM32/BlueJay ESC's to increase. Personally running Motor Rampup of 100% (default is 50%) and it makes a noticeable difference in the air.
atm price dictates what i buy. speedybee is killing it with their prices, besides they end up destroyed. And ty for throwing in skystars, another brand i can just afford lol.
Not sure if you are kidding but he did test a Skystar ESC. I am also in your boat, I have been buying parts here and there, slowly building up all the components to make a new drone. The Speedy Bee is definitely on my radar as is the Skystar stack. I do like the Skystar set, not sure if it is just me rooting for the lower price brand to be just as good as the premium ones... I do have to also wonder if I would even notice the difference between any of these in a blind test as I am a new pilot and my skills are not so great.
Please test the ESCs for desyncs, I have been getting desyncs because of high KV (1950kv F60IV) motors, I tried Mamba F40, F50, F55, AIRBOT Furling32 (even got them smoked just from giving a throttle), I tried Oscar Liang's recommendations with a bunch of different rampup power and demag comp-s settings, changed all set of the motors, nothing ever helped, always got the desyncs until I tried a high end expensive ECS which I won't name because this is not an advertisement. I am really curious what other ECS-s can avoid desyncs on such a high KV. The main scenario how I got the desyncs was - I gave a full throttle for 1-2 seconds, then throttle to 0, and immediately Roll to the side on 1100deg/s rate. At the end of the roll I have been getting the desyncs. I think because of back EMF and from cutting the throttle from 100 to 0 and from acceleration of motors to stop the roll.
i have desyncs, too skystars am32 esc guess a fet is bad, already ruled out motor as cause i like reliability and would like to know the name of the esc as well
This is by far not a "high" kv by today's standards! Racers and even loads of freestyle pilots have been using motors above 2000kv for years already. Trust me, your 1950kv (which actually is lower than that) is not the cause of desyncs you're having. And it all just points in the direction of shitty motors. I challenge you to try another motor with the same or bigger kV. I personally have 3 freestyle rigs with 1900+kv and 2000+kv motors, with different ESC, different firmwares, and I am YET to get a desync. Been rocking for 2 years.
7:08 This efficiency graph looks a bit to jaggy, are you sure the measurement readings are synchronised? I've had similar things happen when the measurements weren't synchronised. Constantly auto-ranging measurement equipment could be the problem too. But it looks like you are using RCbenchmark, in my experience, that should get you good measurements.
..... this is one of the least scientific tests I've ever seen. The control is nonexistent, and the conclusion is inconclusive because of that. The pro 3 esc has better microprocessors than most. All this really proved was that faster microprocessors are better than slower ones 😂. Go figure.
The increased acceleration of BL_Heli32 is the most noticeable characteristic that I experience when trying any other firmware. It probably has a lot to do with my flying style but it's something I can't overlook when using other FWs.
How much is component choice and how much is firmware? Change only one variable at a time, hardware or software - not both. Also the KV setting in AM32 should allow more acceleration.
am32 you can set the kv slider lower to allow higher acceleration of the motors at the cost of over current protection.. personally I start at half your motor kv for the slider and take it lower till it wants to not run clean..
Look at the old dronemesh channel and the electrical noise testing he did - high levels of electrical noise would impact filtering tuning. Would recommend testing the Xrotor ESCs.
Research of ESC efficiency I would start with wiring optimization. 50A at the top of acoustic frequency just beg for using twisted Litz pair! Simple experiment: check efficiency and dynamics on one ESC, prop and motor connected with recommend gauge wire, twisted and not, then the same with electrically equivalent gauge Litz, twisted and not. This may have bigger effect than variety of ESCs gives!
This is great...i believe it may explain some of the issues I've had with AM32. The PIDs must be dialed down quite a bit. The low-end torque on my high KV motors is too much. (2207 RCINPOWER 2040kv). I would love a video on AM32 settings. I've tried several, and I'm just not getting the smooth performance i was expecting. Thank you for everything you do for us in this hobby!
12:50 I can second that, the ESC are operating at a pretty ideal point. They basically just switch between motor phases, there is not much loss here. However switching at a wrong or non ideal position will result in a less efficient motor, not necessarily a less efficient ESC. It's not unusual to have an ESC to be more than 95% efficient. The only major efficiency loss is because of the forward voltage drop because of Rds(on) of the FET.
When people talk about ESC efficiency, they don't just mean how efficient the ESC itself is at providing power down the wires. They usually mainly mean how efficient it makes the motor, which is what Chris Rosser is talking about here - which esc gets better efficiency in terms of thrust out of the same motor & prop combo.
Hi Chris, I appreciate all of the excellent bench testing, but you absolutely have to fly and try tuning a quad using the ESC to see if it actually works in a Betaflight quad. I have had two Skystars ESCs (the same Kamram 55A 20x20 and 30x30 AM32) and both have produced nearly unflyable and untunable quads. There was so much oscillation and also strange sounds coming from the motors. And I used your AOS EVO 4" and 5" frames, so we know that resonance was not the cause. Everything worked perfectly when I switched to iFlight BLHeli_32 ESCs. Someone, somewhere, said that AM32 doesn't work well because it produces more torque and suggested dialing down the d-term. But I couldn't see much, if any, difference when I tried that. Ultimately, I don't know if Skystars ESCs are junk, or if AM32 might not work well with quads. Perhaps I was incredibly unlucky. However, I have seen similar complaints online about the Skystars AM32 ESC, and also that some NeutronRC AIOs with AM32 were untunable. I'll be avoiding Skystars and AM32 for a while until they both build a decent track record.
I have been struggling for a while with Skystars AM60, I finally got good results but I had to disable variable PWM frequency and Bidir Dshot (I think we can keep bidir dshot, it was actually Dynamic Idle that was making it unflyable).
With the huge difference in responsiveness with BLHeli-32 vs BLHeli-S/BlueJay, how is it that in the real world, these two are used pretty much interchangeably with no real need to retune? Folks that want to save some money just buy BLHeli-S ESCs and there doesn't seem to be any huge consequence to that.
Should probably point out from the data. That Blheli-32 using by rpm also decreases motor acceleration performance. Not a feature I want when flying freestyle. Best to keep parameters from its lowest PWM to its highest right PWM NOT including by rpm.
The torque-speed and acceleration results need digging into... Given torque = inertia * acceleration, there is a major inconsistency in the BL_HELI32 firmware achieving the lowest torque AND highest acceleration. Is the interruption in measurement data with BL_HELI32 fw causing a false acceleration measurement?
Great video, as usual. Thank you. With regards to the surprising amount of extra torque that BL_32 seems to provide: How about using a scope/logic analyzer to record the actual signals coming from an ESC microcontroller its the FETS during the acceleration, for BL_32 and AM32? This would also require recording a timing signal in order to get actual timing data of the signals. This could be done, for example, by placing a hall switch close to the bell. Or by making a reflective / dark mark on the bell and recording its passing. Graphing the pulse-frequency(for bi-rpm), -width and -phase with regards to the timing signal, you might be able to get some useful information as to. At 24Khz, that's a lot of data, but a single capture should not last longer that a second and with some coding, much of the analysis could be automated I presume.
Hi Chris- every test opens up more questions- would you get the same spread of results using a motors? Ditto for props? Keep testing, it’s great that someone is producing some facts rather than just opinions For a long range efficiency conclusion would we need a different test - efficiency at a stable 40/50/60% throttle level?
I appreciate you say you have not finalised your test protocol, but surely you need to use the same firmware on each of the ESCs in turn in order to provide a meaningful comparison. I would be interested to see if the noise levels generated by the ESCs are impacted in any way by the firmware used.
Definitely very interested in seeing if there is any extra performance that can be squeezed out of bluejay. That difference with bl_32 is quite considerable.
Looks like testing ESCs has some challenges. Kudos to you for discussing this openly and transparently. The responsiveness data is fascinating. Thank you for doing this testing. I was thinking of trying out AM32 but I think I'll hold off for now.
The question for me here is does that top performer efficiency actually translate to more thrust when the system becomes power limited by smaller cells and more motors
In my opinion testing different ESC's with different firmware's doesn't make much sense if you label them by FW type. First of all, everyone is labeling A ratings without a common rules. I believe that ESC's layout, stack up and component selection plays a much bigger role in the performance then the FW it runs. I would like to see first a test of current draw capabilities for continuous A vs some reasonable operating temperature. And only then label them by your own standard and compere the performance. Would be interesting to see the BL32 ESC tested before and after reflashing to AM32 FW.
Maybe. Get it on the scope with a rotor position timing mark. I'd like to see how these different programs change commutation to accelerate the motor. I suspect blheli_32 more aggressively advances the phase angle during acceleration.
Hardware doesn't really play a role. Basically, the only possible difference is that the MOSFETs do or don't smoke during testing. 'Component layout, stackup and component selection' can affect EMI but not performance to a non-negligible degree. Also any differences in FETs would be really obvious in efficiency testing (higher ON resistance means more losses), which is where all ESCs are basically the same in the tests.
If you compare fair, yes. Hardware determines the current rating. My concern is that manufacturers applies A ratings by different methodologies and cant be trusted or compered directly. A rating is used more for a marketing than a specification on ESC's. Alternative would be to compare by retail value. @@jiatan_fpv
I disagree with you..I think the actual components don't matter once you're buying over a certain level. I want to know how the difference firmwares compare
Now that you got us the beastiest beast of a motor you gotta figure out which ESC and Prop can actually handle that power xD I really would like to see how the foxeer ESCs score in these tests as these are the ESCs racers use for their reliability in frequent turtles and MCK also says it's the ones least prone to demag. I also found that bluejay is super chill on defaults, especially with noisier builds while BLHeli32 is much more prone to oscilations or other issues. Might that actually be the double acceleration-performance transforming that noise much more effectively into prop movement?
