ESC Shoot Out: RCX H2208 1800kv HQProp 6045CF
This time I tested with a lower KV motor to see how it affected the spread. As predicted the Atmel ESCs closed the gap a bit, but still couldn't quite keep up with their SiLabs cousins. The one exception was the Afro 20A HV, which was the only Atmel ESC I tested that has dedicated gate drivers. There seems to be a bug with this ESC though in BLHeli on both 14.1 and 14.2. It starts oscillating like crazy at high throttle on all timings, so I have no comparison data yet. Hopefully I can get a fix from Steffen and get some updated data for BLHeli.
Also in this test are some new ESCs I didn't have last time. People were asking for the ZTW Spider 18A, and a donor sent me one of those to test. Also have the Emax Nano 20A, thanks to another friendly donor. As I mentioned before I also have the Flycolor Fairy ESCs sent from 65Drones The biggie is the new UBAD 30A Angry Beast, thanks to Bruce over at UBAD.
I added results for the new DYS XM20A ESC (F396 MCU) as well as the ZTW Spider Pro 20A (F330MCU) to the spreadsheet. Unfortunately my batteries seem to have seen some degradation (one of them now has a weak cell) since I ran the tests 6 weeks ago I have new batteries coming in by Friday that should equalize that back up to a reasonable level. I'll rerun the tests then.
I re-ran the XRotor 40A on the existing setup (also flashed to 14.3, same as the others) to get a new baseline due to environmental and battery changes with the same batteries at an even lower charge level (I ran it right after the tests listed). Average peak RPMs were in the 21500 range compared to 21700 in the first test I ran 6 weeks ago. From those results, I'd say there's some other factor limiting the XM20A. My measured difference with the XRotor 40A between the good battery at 16.5v and 16.3v on the flagging battery was ~ 1%. If we adjust the new results by that margin it moves it up to the level of the Zeus, but still not to the level of the other F39X based ESCs.
In addition, this lower KV test is not an ideal indicator anyway for these ESCs, as it's not pushing the RPM limits of the MCUs. It actually surprised me when I ran the test initially that there was as much of a separation as there was here. Again, I'll rerun these two ESCs when my new batteries come in (or try to borrow another high capacity 4S from a friend before then), and I'll be testing as soon as I can on the 2700kv 2205 with 4" props. That should generate some significant separation.
A couple of things to note in this test data. The KISS ESCs were generating some weirdness on the current sensing. I had to compensate for a 6A rise across the entire test with both the 18A and the 30A. Not sure what's going on there, still need to debug. For now the max numbers are accurate, but there's still some weirdness going on in around the top end of the throttle ramp. I will try to change the grounding method and re-run them soon.
Also, there are a couple of sections in the data where the differences between the ESCs are basically within the margin of error, so performance should be taken as equal.
- The first 6 results (XR40, XR20 and UBAD30)
- The next 9 (Zeus through KISS 18A)
- The next 5 (Flycolor 30A - Flycolor 20A)
- The rest
You can see the spots where there are larger jumps in the numbers. The categories are pretty well differentiated by hardware. The top tear is fast processors and fast FET switching circuitry, the second is fast FET circuitry and medium speed processors. The third is a mix of medium speed processors and and slow FET circuits, and slow processors with medium speed switching circuits, and the bottom tier is both slow processors and slow switching circuits. There are a few anomolies like the Flycolor 30A where it seems to jump categories between 14.1 and 14.2. My guess is there was some other variable on that run that's causing issues. I may re-run it on both versions to see what happens.
I also changed the power setup for this test. I tried to use a bench supply, but it would basically trip the circuit protection immediately no matter what I tried. So I settled for 2 1800mah 75C packs in parallel. This seems to have limited the voltage varation accross multiple tests to make the results more consistent.
Also the final note is that my BearHug v2 seems to be having problems. I'll be trying to debug and see if I can't get some clean runs out of it, but it has lost it's ability to connect via the bootloader. I'll need to see if I can recover via the C2 interface.
The next motors on the bench are a prototype 2204 2700kv Blue Dragonfly motor (more info coming soon!) also kindly donated for testing by 65Drones, as well as the 2205 2633kv motor from MyRCMart. The high KV combined with a 14 pole count should really emphasize the difference in performance between the ESCs. This one will really be a thrash test!
Anyway, thanks all for your support and contributions! More test data will be coming soon!