Anyone familiar with this ML7280 clone?

[magicmicros]

I'm running a 3kW bike with a KT72V50A controller, but it's so-so.
I want to update, use a sinewave controller ( or better )
So, I found this, what seems to be a ML7280 clone.
Price is good, about $65. And after looking inside, it's SO much better quality than the KT crap, which cost even more.
I've spent 3 days on reverse engineering it and now have complete schematics, in addition to a detailed knowlegde of the design.
It's un-named, but looks like the ML7280 (capabilities were similar, 72V, 80A).

Does anyone here have any experience with this?

I have a "real" ML7280 incoming, so I'll soon know if there's any differences.
Anyway, the plan is to write my own firmware for this.
First of all because it's totally undocumented but also to understand better how stuff works.

Documentation is here:
https://github.com/magicmicros/ml7280_clone

/J

 

[mxlemming]

Looked over your schematics looks like you probably got it all right.

It's not a good controller or PCB layout or micro or... And it's pretty big for what it does.

Tbh, if you're going to build your own controller, I wouldn't start here. Up to you though! Bemf sensing on phases makes for fairly easy control and commutation.

You may well be able to run foc on this using the st motor control library. The pins are generally correct but... Are you able to measure the bus voltage?

 

[magicmicros]

For the size, yes, it's big. But it's cheap and also pretty capable.
But I would like to hear you expand a bit on why this is a bad layout and why it's a bad micro.
I'm still waiting for my original Sabvoton ML7280, but my bet is that they're completely similar.

This is not going to be the base of my own controller, but it's going to be my playground for initial code development.
I don't know about the ST library, will give it a look.
For now, I have verified all HW inputs and are trying some simple trapezodial and sine wave control with a small 24V motor.
Might give it a try on one of my 48V hover-board motors and have a look at the BEMF signals.
Current and voltage measurement work fine.

 

[mxlemming]

The MOS driving circuitry is set up for incredibly slow switching, discrete transistors not proper gate drivers.

The layout looks like it'll have huge parasitic inductance, therefore lots of ringing... That'll be why they've slowed the switching so much. Switching losses will be quite large, and low pwm frequency so you'll be limited on rpm.

F030 is the lowest stm32, slow clock and no floating point unit so you'll have to implement everything in integer logic which is just a bit tedious. Higher end stm32 also gets faster ADCs.

Stm32f3 and f4 is where you want to be if starting your own from scratch. The base f401 normally costs only a few$ (bear in mind there's currently a blanket stm32 shortage and prices aren't representative) and has vastly more capability than f030. If i was starting a high end thing I'd use h7 but you're looking for cheap so...

 

[stancecoke]

Quote from the github readme:

The board has a main shunt, but does NOT have any phase shunts.

So you will never do "classical" FOC with this controller . You can only calculate a theoretical advance angle from load and speed like @casainho does in his TSDZ2-project.

regards
stancecoke

 

[mxlemming]

Quote from the github readme:

The board has a main shunt, but does NOT have any phase shunts.

So you will never do "classical" FOC with this controller . You can only calculate a theoretical advance angle from load and speed like @casainho does in his TSDZ2-project.

regards
stancecoke

You can do classical FOC. Just have to strategically choose the sampling points. You can read the phase currents when one phase is pwm high and the other two are pwm low. The common shunt is then reading the high phase.

When two are high, one low, you know the current through the two high... Then subtract the first result and you have another phase current, which is enough to do clarke and park transform.

But this is pretty pointlessly hard and crappy since you're heavily constrained to low frequency pwm and get poor quality readings.

 

[stancecoke]

Ah, you are right I remember, a similar procedure is explained in the ST paper UM1052 also!

For my understandig: The current never goes backwards through the shunt. Is that right?! The pictures show the ripple on the average DC current?! Or are there really times during one PWM period, where the current trough the shunt is negative?!

regards
stancecoke

https://www.ti.com/lit/an/spract7/spract7.pdf?ts=1622324641002&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FTMS320F280025

 

[mxlemming]

Ah, you are right I remember, a similar procedure is explained in the ST paper UM1052 also!

For my understandig: The current never goes backwards through the shunt. Is that right?! The pictures show the ripple on the average DC current?! Or are there really times during one PWM period, where the current trough the shunt is negative?!

regards
stancecoke

Negative current would imply you're in a regenerative region otherwise you'd be rapidly oscillating current in-out of the dc link which would be very inefficient. The dc link/single shunt current should only ever be positive on motoring and negative on Regen unless your just very close to zero power.

That second last waveform is confusing. The single shunt won't ever have nice gradual transfers between pwm sections. That looks like an editorial error or it's meant to be phase current.

Regardless... Definitely don't design hardware to do this. It's a crappy way of doing it probably only applicable to things like computer fans and other simple, predictable operating regime widgets

 

[Jonspeed]

For those of us who have no idea what you all are talking about, is this a good controller for a beginner to upgrade to, I saw a similar looking one on amazon (80a) an would like to know if it would be a good buy for a 1500w ebike. Its the cheapest 80a sinewave i have seen.

 

[electric_nz]

I have one of these and am looking to mod it to get better throttle response etc. Attached is a pic of the rear of the connectors. I think the round circular connector is for a display, and the programming cable is part of the 30-pin connector. (The real mq7280 from the mqcon website has its programming cable branching off from the 30 pin connector.)

Next step is to hook it up to USB and try to connect to it.

 

[warden_imst]

I have one of these and am looking to mod it to get better throttle response etc. Attached is a pic of the rear of the connectors. I think the round circular connector is for a display, and the programming cable is part of the 30-pin connector. (The real mq7280 from the mqcon website has its programming cable branching off from the 30 pin connector.)

Next step is to hook it up to USB and try to connect to it.

did you manage to get it connected to a pc? I'm trying to get regen braking working, and so far no success...

 

[Lotharn]

anyone made any real progress with this?

 

[ganzey]

These controllers are not worth modding anymore when you can get a $90 vesc (makerbase 75200) that can do more, and is a fraction of the size. I used half a dozen or so for random builds, modded a few of them, wasted a lot of time. Got regen brake on one but it would randomly apply sometimes when hitting the throttle and was straight up dangerous ( imagine cruising, hit throttle thinking your gonna go WOT, but instead apply 100% brake)so that went away. There are several versions of them and no way to differentiate which one you are getting, even if buying through same supplier. Some have decent fets, some come with absolute trash assembly, laser etched mcu's and smaller caps . They are all clones of a sabvoton ml7280s, some are almost identical, some have quite a different pcb layout, heres 2 and a sabvoton, guess which is which?

 

[flez1966]

not very waterproof......

I never did get anything other than basic functions and poor slow performance, kinda glad to see it gone, most useful part is the heatsink.....

 

[E-HP]

not very waterproof......

I never did get anything other than basic functions and poor slow performance, kinda glad to see it gone, most useful part is the heatsink.....

They are good for use as a dumb controller. I bought 2 off Amazon. They came stock set to around 60A, but they’re shunt modded to 90A. I’ve done a ton of miles, and they’re regularly subjected to abuse and never get past lukewarm. As long as that’s all you need, and not a bunch of features, they’re fine. They’re basic raw power. Phase current is really impressive compared to the old KT I used before, but similar to the 7kW Powervelocity controller it replaced. Night and day compared to the KT.