FLIPSKY new 20s 100A tiny controller (vesc based)

[amberwolf]

transposon- thanks for that detailed description on adding the resistor. Though I believe if the whole 8 pin ADC jst plug becomes disconnected from the 75200 you're still in trouble...

Shouldn't be, because then you don't have the signal wire from the throttle itself attached.

What happens in cases that the ground fails to the throttle is that there is now no normal current path to ground for the throttle hall (or pot), so it becomes a simple pullup to throttle supply voltage, essentially sending a full throttle signal to the controller.

If it's not connected to the controller, there's nothing that can do this, especially if the MCU has an internal pulldown, or the PCB has an external pulldown. But even if it doesn't, there's now nothing on the controller that can act as a pullup like the throttle does, hence nothing to cause the problem.

I need to ask one of the dumbest questions ever of vesc tool users: where the hell is the download "button" on the vesc-project site to get the latest version, and what is that version #?

https://vesc-project.com/vesc_tool?destination=vesc_tool
assuming you don't want to donate using any of the more premium versions (which are the same actual program, according to this:

This edition of VESC Tool works in the same way as the others, but you don't send any donation to the VESC project when downloading it.

then scroll to the bottom of the page where ther'es a square with:

VESC Tool Free
Submitted by benjamin on Tue, 2016-12-27 16:02
€0.00
Price: €0.00
Read more

and click the words VESC Tool Free which then take you here
https://vesc-project.com/node/17
Then Add to Cart button near the bottom
That takes you here
https://vesc-project.com/cart
and you click Checkout that takes you here
https://vesc-project.com/user?destination=cart/checkout
where you login if you aren't already, or create an account, answer the captcha, and proceed from there. (I have nothing to use the tool on so didn't proceed further). If you already did all this and didn't get anything, you might check your user account there to see if it has messages, or a cart listing, etc., to see if you get the link that way instead of by email. This quote from a question on the site indicates this may be the case:

You can go to Purchased Files and simply download what you need

Optionally you can go here
https://github.com/vedderb/vesc_tool
and get teh source and compile it (but I have no idea how to do that sort of thing).

This page has a "windows optimized" version you can comple instead of the original
https://github.com/rpasichnyk/vesc_tool/releases/tag/v3.0.1

rpasichnyk January 21, 2022, 1:47pm #165

The reason is that official VESC Tool is compiled with GCC compiler which is not optimized for Windows. This is also why bluetooth does not work. You can try this version that I built with Microsoft compiler https://github.com/rpasichnyk/vesc_tool/releases/tag/v3.0.1, it starts much faster (6s vs 25s)

 

[Barncat]

Yes vesc tool is vesc tool whether you pay for it or not which I already did. Despite being logged in to vesc-project every available tab is a dead end with no specific way to force a download that I can see.

 

[amberwolf]

WHen you are logged in, do you have this page available
https://vesc-project.com/user/me/purchased-files
?
It should show across the top in these:
My account Messages My Comments Purchased Files Enter Code Log out


That is where it should show what you have put in your cart and checked out with.

For giggles I made a login and went thru the whole checkout process for the zero cost version of the tool, and then went to that page, and it shows the following list:

File downloads
Primary tabs
ViewEditFiles(active tab)OrdersFile browser
Purchasedsort ascending Filename Description Downloads Addresses
2023-01-17 vesc_tool_free_linux.zip (expires on 2023-07-17) 0/Unlimited 0/Unlimited
2023-01-17 vesc_tool_free_windows.zip (expires on 2023-07-17) 0/Unlimited 0/Unlimited
2023-01-17 vesc_tool_android.zip (expires on 2023-07-17) 0/Unlimited 0/Unlimited
2023-01-17 vesc_tool_mac.zip (expires on 2023-07-17) 0/Unlimited 0/Unlimited
2023-01-17 vesc_tool_BETA_ALL.zip (expires on 2023-07-17) 0/Unlimited 0/Unlimited
Once your download is finished, you must refresh the page to download again. (Provided you have permission)

Personally I find the whole thing really confusing, but it is just like a number of other download pages I"ve dealt with over the years for various music synths and effects, even when they are free downloads they can't be accessed except via this type of method.

 

[Barncat]

AW- for whatever reason, those link options were not available during the paypal based checkout process on the vesc-project site, but I just discovered them in an email sent early this afternoon after checkout. There was no notification that I had to approach it that way...... the 61MB download took over half an hour, but apparently worked. On to the next frustrations... your assistance is much appreciated.

 

[amberwolf]

AW- for whatever reason, those link options were not available during the paypal based checkout process on the vesc-project site,

I didn't get any links during the checkout process either. I only get them when using the Purchased Files link across the top of the VESC site while I am logged in. (I hadn't checked the emails, but they are also in one of those, titled File Downloads for Order#, but I try to never rely on emailed links--sometimes email servers or providers filter such emails out as spam ).

 

[afzal]

transposon- thanks for that detailed description on adding the resistor. Though I believe if the whole 8 pin ADC jst plug becomes disconnected from the 75200 you're still in trouble...

What happens in cases that the ground fails to the throttle is that there is now no normal current path to ground for the throttle hall (or pot), so it becomes a simple pullup to throttle supply voltage, essentially sending a full throttle signal to the controller.

There is another aspect, if there is no pull down on the throttle signal pin, it is floating & software can read a random value at the throttle input. I think I faced an issue on VESC some time back that pulling out the throttle connector caused it go to high speed. Having the pull down on the other side of connector fixed it too.

But the safest would be to solder pull down on the PCB itself, so even pulling out on-board JST would not cause an issue.

especially if the MCU has an internal pulldown, or the PCB has an external pulldown.

On STM32F4, when a pin is configured as analog (done for throttle i/p), internal pull up & down resistors are disabled. And VESC hardware's normally doesn't carry onboard pull down on ADC lines (spintend as mentioned by chuyskywalker is an exception)

 

[Barncat]

AW- trust me I tried every possible tab on vesc-project repeatedly to no avail. So ironically, the free version worked and the pay version didn't directly... saw another account of that on that forum. And the email link was way slow. Pretty sure this is vesc-tool 6.00, will see if it works but I first need to update the firmware. There's a short Flipsky YouTube video on that. I believe the latest code is supposed to work better with that exotic TU-15 motor I have.

afzal- yes, I recollect a similar disconnected plug/ runaway throttle experience last year while bench testing on a Maker-x.

Anyway, vesc/esc is the most difficult part of ebike stuff for me.

[Update edit: the 75200 firmware is very easy to update if you follow the above mentioned short Flipsky video. If you're like me you'll need to watch it several times and take notes, and the backing music will drive you nuts... vesc-tool 6.00 seems to work great but I'm still tuning my TU-15 motor- will probably post in my Mongoose Blackcomb thread when successful. Edit: posting now]

 

[tom355]

I am trying to set up a Flipsky 75100 with a Gmac for quiet some time now. Wizzards are doing a job and as long as the wheel is in the air, it spins perfectly.
As soon as i sit on it, under load, it`s a total mess. Virbrating, stuttering and making noises.
When i accelerate very slow and carefully it runs smooth. When i pull the throttle hard it stuttering again.
FOC with sensors, Battery 14S9P

Anyone who is successfully running the 75100 with a ebike geared hub? Would you send me a config, PLEASE?

 

[addvalue]

I am trying to set up a Flipsky 75100 with a Gmac for quiet some time now. Wizzards are doing a job and as long as the wheel is in the air, it spins perfectly.
As soon as i sit on it, under load, it`s a total mess. Virbrating, stuttering and making noises.
When i accelerate very slow and carefully it runs smooth. When i pull the throttle hard it stuttering again.
FOC with sensors, Battery 14S9P

Anyone who is successfully running the 75100 with a ebike geared hub? Would you send me a config, PLEASE?

what kind? the boxy thingy or the alu version?

 

[scianiac]

I am trying to set up a Flipsky 75100 with a Gmac for quiet some time now. Wizzards are doing a job and as long as the wheel is in the air, it spins perfectly.
As soon as i sit on it, under load, it`s a total mess. Virbrating, stuttering and making noises.
When i accelerate very slow and carefully it runs smooth. When i pull the throttle hard it stuttering again.
FOC with sensors, Battery 14S9P

Anyone who is successfully running the 75100 with a ebike geared hub? Would you send me a config, PLEASE?

What current are you using for the "learn" mode? I've heard that running it at higher current levels yields much better results when you actually are applying a load. I think someone talked about it in this thread or maybe it was a different thread but in general if you're phase amps are set to 100A then a 2A learn mode is not enough and you want closer to 10-20A+. Probably safest to step it up and not set it to something crazy right off the bat but you can see if the motor parameters it generates are different and work better.

 

[chuyskywalker]

What current are you using for the "learn" mode? I've heard that running it at higher current levels yields much better results when you actually are applying a load. I think someone talked about it in this thread or maybe it was a different thread but in general if you're phase amps are set to 100A then a 2A learn mode is not enough and you want closer to 10-20A+. Probably safest to step it up and not set it to something crazy right off the bat but you can see if the motor parameters it generates are different and work better.

Might have been me as I've had this experience. It'll work fine on the bench after detection, but going out for a ride and you get very odd behavior under actual loads. Re-run the FOC tuning (there's a page for it) with much closer to max amps and this problem goes away. The auto-tune is just conservative, which is a good thing, but seems to have this effect on high amp systems.

 

[Tommm]

So from skimming over this thread. it looks like if you want FW, you have to go to 5.3 or newer, and then you can only use the unlocked firmware, where there are no limits including not able to limit controller/motor temperature against overheating?

 

[scianiac]

I believe all current and temperature limits can be set to whatever you want on whatever firmware, an "unlocked" firmware though just allows you to set them to potentially unsafe levels, as in you could set the current limits way above what it's actually capable of.

 

[tom355]

Still working on the problem, with support from a guy who sets up VESC controller all day. Higher Amp settings helped a lot but was not the final solution. Meanwhile if fried one controller with settings 160 Amps for LEARNING. No comment. Yes there are multiple ways to brick the controller with wrong settings.
Main mistake was to set pole pairs to 80 as this is written on the Grin homepage (this did not brick the controller), but you use this number only when programming a phaserunner/baserunner. The gmac has 16 pole pairs. Going to buy a 2nd hand home bike roller trainer to strap the bike on to make things much easier.

 

[tom355]

I got this home bike roller trainer where i put the bike on and it was possible to work much faster on the problem. I found settings which are not super, but ok for now. On BLDC (no FOC) with hall sensors. I get 100% power from the battery to the motor, only the launch from zero is still heavy stuttering. With 50% battery (52V 14S9P ) the BMS app stated 3200w on full throttle.
After hard acceleration or going uphill fast, power is dropping fast. Symptoms like voltage sag on battery, but battery is fine (new, genuine samsung cells). BMS showed a drop from 52 -> 49 volts under full throttle and no cut off. So the problem here seems to be the controller and i still have to figure out what is going on here. The controller gets pretty warm/hot and i really don`t believe that it is able to do the 72V 100A what it`s written on, without being toast after 10 seconds.
Does anybody knows which value i have to tweak for that stuttering launch problem?

 

[scianiac]

When you it's cogging at startup I assume you mean immediately so that would mean it's still in hall sensor mode. Does it still happen if you turn the phase current down? It could be there is too much noise on the hall sensor lines, that can happen sometimes on hub motors where the hall lines travel right next to the phase wires. If I recall there is a setting to take multiple readings of the hall sensors that might help.

It sounds like the loss of power could be from the controller thermal throttling, these controllers are known for heating up very fast. Should be easy to test by watching the temperature when it happens if you have a bluetooth module, back in this thread there are a number of people who have modified theirs to improve cooling.

 

[tom355]

Do you have any idea what`s the name of the parameter for the " multiple reading of the hall sensors"?
I made tiny little steps forward with adjusting the duty cycle. Also made quick dirty ghetto fix for the thermal problem.

 

[scianiac]

"Hall Sensor Extra Samples" in FOC > Hall Sensors. Is the cogging only at the very start, as in does it happen if the motor is spinning even super slowly? HFI is another option, just requires careful tuning.

 

[tom355]

The cogging happens under high load what means at the start from zero, and/ or going full throttle from slow cruising. It`s a bit better since i played with the duty cycle and i am not yet done with it. I stopped adjusting because i also wanted to make that cooling solution today.
I quit FOC and changed to BLDC a couple of weeks ago because under FOC it was cogging all the time and i had less than 50% power.

 

[scianiac]

It really would be nice if there was a ton more documentation on VESC tuning but here is what I've got so far from my recent tuning. I'm certianly no expert and would love to be corrected by anybody who actually knows more if I have anything wrong.

If it's unstable or does not run well in FOC it's probably because the parameters in detection are off. Re-running the detection several times with various current limits (motor current max) and duty cycles (in the FOC > General > "D" setting in the test panel at the bottom) and seeing how the values shake out, some may vary quite a bit but the ones that are closer to your intended current limit and the duty cycle where the issue occurs but not so high you are saturating the motor are probably best. (maybe D 0.20-0.60?)

Personally I found lowering the observer gain significantly helped a lot in dealing with the observer loosing track of rotor position when switching from halls to sensorless when under high load and high acceleration.

What also helped was reducing the time constant to recalculate the current P and I terms (hit "calc apply old" after you change the "T").

The other thing that helped was using the realtime data capture and the sampled data but both can be a bit annoying to use, ideally with a laptop with bluetooth you can do the following. Connect to the vesc with bluetooth (assuming you have a bluetooth module), start realtime data with the little "RT" button on the right, do a short test accelerate, then stop the realtime data and see what you have. The sampled data is a bit more annoying to use on a bike but this is what I did, connect with USB (bluetooth will not work here it seems or is unreliable due to some bug), go to sampled data and use the "motor TN" or "bug TN", then you can disconnect, do the test, reconnect via usb and press the up arrow to download the data. Use more samples or "Dec" to capture more and it may help to turn down the absolute max current down to the point that it triggers a fault when it cogs so that the "bug TN" will capture right there.

 

[mxlemming]

The cogging problem has been a problem for years on VESC and relates to the phase difference between the observer and the hall sensors at the transition point. I had this problem 3 years ago when I first ever plugged in a flipsky 4.2 and used it on my ebike.

This phase difference is there for one of several reasons...
Hall sensor hysteresis
Wrong resistance setting (try increasing/decreasing by 50%)
Salience of the motor.

The salience is the hardest one to deal with, although the characteristic effect is quite similar to wrong resistance values. Effectively the observer (either version) tracks the total magnetic field rather than the permanent magnet field. This means the observers inherently track the MTPA angle which isn't the same as the true rotor angle (at higher currents the torque comes from the reluctance of the magnetic loop as well as the permanent magnets). When the switch occurs, there is a step change in angle and therefore q and d axis currents and the current controllers fight to keep up. This is why faster current control bandwidth can help.

The step change also has an impact on the pll and the torque output and it's possible to get an oscillation between hall and sensorless.

Benjamin recently started work on a gradual transition from hall to observer. It's very promising.

I got this home bike roller trainer where i put the bike on and it was possible to work much faster on the problem. I found settings which are not super, but ok for now. On BLDC (no FOC) with hall sensors. I get 100% power from the battery to the motor, only the launch from zero is still heavy stuttering. With 50% battery (52V 14S9P ) the BMS app stated 3200w on full throttle.
After hard acceleration or going uphill fast, power is dropping fast. Symptoms like voltage sag on battery, but battery is fine (new, genuine samsung cells). BMS showed a drop from 52 -> 49 volts under full throttle and no cut off. So the problem here seems to be the controller and i still have to figure out what is going on here. The controller gets pretty warm/hot and i really don`t believe that it is able to do the 72V 100A what it`s written on, without being toast after 10 seconds.
Does anybody knows which value i have to tweak for that stuttering launch problem?

 

[afzal]

I quit FOC and changed to BLDC a couple of weeks ago because under FOC it was cogging all the time and i had less than 50% power.

Is the motor back emf trapezoidal ?

 

[tom355]

Is the motor back emf trapezoidal ?

I don`t know

 

[IsaacLarsen]

I’ve read the entire thread—a lot of technical stuff that went over my head. I still can’t understand if one can make this work as a daily driver with a cheap Chinese ebike. I’ve read people complaining about motor noise, poor acceleration from 0km/h and other issues, but they all use big, powerful motors. I have a 20” fat bike with a 750W geared hub motor, 48v battery, 20A max, 10A rated controller. I’m looking to replace the original controller with something that will give me more power and a quieter and smoother ride.

 

[MorbidlyObeseKoala]

I think you can make it work if you want to get into all of the settings and learn the vesc software interface. It will increase your output dramatically if your battery is able to safely output more current, definitely something to confirm before changing to any different controller. There are also additional features that you might gain if you continue the journey.

If you are looking for a simple power upgrade, this is probably over complicating it and you can likely find a cheaper replacement with a higher output and less customizable features/irritating headaches.