[H500]

The thread below presents the best design I'm aware of. It's not complete, but an EE will have no trouble doing the minor mods.

http://cnczone.com/forums/showthread.php?t=81125

[yamukha]

Hello everyone,
I can manufacture 2 player PCBs at home (Eagle license), but I'll be happy to buy an existing design. Any suggestions?

Hi,
I can help you.
What is requirements?
I have proven design (scheme and prototype using L6219 and CPLD for BLDC/stepper 0,750A full/half step mode).
Currently I am going to run project with A3986 for running 1KWatt BLDC at very low speed.
As well as I continue to develop my own project with CPLD and galvanic isolators published on:
http://dev.emcelettronica.com/brushl...r-control-cpld.
If you interested please contact me yamukha at ukr dot net.

[MechanoMan]

Microstepping looked interesting more for a quiet operation than for accuracy. I can manufacture 2 player PCBs at home (Eagle license), but I'll be happy to buy an existing design. Any suggestions?

No, it's not. Quiet is "nice", but we're primarily worried about it working. And we don't need the accuracy of microsteps, it's kind of irrelevant because the mill's backlash error will be much larger than that so it's unlikely to actually guarantee more accuracy.

Microstepping is needed because the pulses of jumping from one state to the next cause vibration which can cause the motor to fail to catch a step and thus stall. It reduces the possible speed, it reduces the torque available to push into material, and just makes the machine wholly unreliable. Mach3/EMC will not catch a stall unless you have encoders (which are rarely put into stepper setups)- and it will not only result in a failed run and ruined stock, it can cause the table to run the leadscrew off the nut, break off your cutting tool, or run the cutting head right through the table! Once it stalls, it'll generally stay stalled until the code brings the motor to a stop, at which point Mach3/EMC will believe the table's in a wholly different position then it actually is.

The A3986's microstepping will NEVER work with a higher voltage supply used for CNC. The minimum on-time creates a minimum duty cycle, and thus a minimum current in the coil. Many of the microstep step phases require lower current it can't create, so you get a badly distorted waveform which causes a bunch MORE harmonics. In fact, in some cases the minimum current output can't even create the target current for a fullstep! The A3986 would only be able to resolve the microsteps if you use a much lower voltage power supply than we prefer- and that'll limit your speed to something far too slow for a mill.

There is no way to modify the A3986 to make it "work" for CNC. Changing transistors or adding capacitors or whatever, nothing will fix or lessen the basic problem.

[morbak]

HI,

I'm looking for a n SMD design with the a3986. Is componnent is still relevant ?

What SMD mosfet can i use? (2-3 Amps motors)

Thx,

[MechanoMan]

HI,

I'm looking for a n SMD design with the a3986. Is componnent is still relevant ?

What SMD mosfet can i use? (2-3 Amps motors)

Thx,

You can read through the million pages of this thread if you like, but the bottom line is this:
The A3986 is critically flawed and WILL NOT WORK PROPERLY for CNC work. Nothing you can do with it will make it work in a CNC application. End of story.

[morbak]

Hi,

Ok, in this case which component to use to replace the A3976? (A component that works correctly?)

Thx

[MechanoMan]

I haven't researched the topic in some time, but there just wasn't anything which an amateur could just drop in to make a practical CNC driver. Sure, any CNC driver is made of electronic components, but that involves major design work, which is advanced enough to exclude amateurs, and in fact it's simply too much work to design and build them for one machine even if you DID have the knowledge.

The Geckodrives really aren't that expensive in the long run. The G540 is a really excellent design and will not disappoint. The midband resonance compensation really does work and no other design uses that.

There are a few people giving plans, but they're not really practical for CNC use. Some are selling kits, but again, not practical for CNC use, and by the time you've bought 3 or 4 of them you're a good bit of the way up to what a G540 would have cost and you'll probably soon find they're not gonna cut it in the long run. And at worst, they could fail or stall during a run and end up driving the milling tool into the fixture or table.

Yeah, I wasted a lot of time myself with the idea of building a stepper driver. This is what I ended up concluding.

[Elco01]

After a lot of reading it is clear that there are design limits in the A3976 which make it uninteresting. I bought - to get started - some 2 A stepper controllers (for peanuts) that work fine.

The good solution looks to be microprocessor control that allows to integrate PID speed control. There are several designs out which I will look at. But my first goal is to make a PCB NC drill, so that I don't have to drill the holes of the next controller board...

Rob

[morbak]

and what about the design Microcontroller+TMC428(17€)+TMC249 (8€ Farnell),

[Elco01]

and what about the design Microcontroller+TMC428(17€)+TMC249 (8€ Farnell),

That is pretty cool stuff indeed. There is a lot of documentation, need the Easter weekend just to read.

Rob

[fakukac]

You can read through the million pages of this thread if you like, but the bottom line is this:
The A3986 is critically flawed and WILL NOT WORK PROPERLY for CNC work. Nothing you can do with it will make it work in a CNC application. End of story.

Hi!
I am very disappointed ! I've just started to make a four-axis driver with
A3986! I've already ordered chips. Nothin' hope?
FA@

[pminmo]

You can read through the million pages of this thread if you like, but the bottom line is this:
The A3986 is critically flawed and WILL NOT WORK PROPERLY for CNC work. Nothing you can do with it will make it work in a CNC application. End of story.

While I've been quite critical of the 3986, saying

"Nothing you can do with it will make it work in a CNC application"

is too strong of statment and factually incorrect. It is flawed in position accuracy in many cnc applications. Lower accuracy conditions such as a wood cnc router with screws that require multiple turns to move a full unit won't notice the accuracy issue. They may notice a "jerkyness" at really slow rates depending on the motor and power supply voltage.

I did post a "white paper" some threads back that supposedly came from Allergo on PWMing the enable signal to solve the low speed accuracy. I received it from an unknown source so I didn't know if it was relavent. A few weeks ago I received it from my Allergo Field Engineer that did verify it was work that came from Allegro.

[fakukac]

Hi!
Good news!!
I'm a joyner, and use mainly wood! My CNC has not too stable superstructure and is driven with cogwheel. (Except the new Z axis ) The accuracy I expect is around 0,05mm at best.
( due to backlash error)
FA@

[dandumit]

I hope that you didn't throw away all of your boards.

http://www.allegromicro.com/en/Produ.../4989/4989.pdf

Allegro just released a new IC pin compatible with a3986.

Reading datasheet doesn't seems to be any difference. T Blank it's still fixed...
Maybe someone it gives another try to this.

[riko]

Looks like they fixed it if you look at table 2.

In the old chip slow decay is always used (ie overrides the PFD settings).
If you look at table 2 it has 11% fast decay (mixed) for the first 5 microsteps which is where the problem is. Have to calculate what happens on the sixth as this will give a threshold for how high the DC bus volatage can be vs the motor voltage. With a high enough voltage it would be still be possible to eventually have a problem on the 6th step.

But for all practical purposes the chip is fixed.

I think though that the plots on page 13 have a typo according to Allegro.

The first one shows the regulation problem with PFD0 and pfd1 = 0 which selects all slow decay.
The first plot incorrectly has pfd0=1 and pfd1=1.

The second one shows how table 2 corrects the regulation with PFD1 and PFD0 = 1. According to table 2 other mixed combos are possible.

So as long as you avoid column 1 which selects all slow decay, then the new chip should work fine.

I never did solder up a board after all the posts. So still have 3 to try out with the new chip.

So I guess eventually the chip makers fix the flaws.

With all the bad luck with the A3986 I guess a new thread is in order.

[JohnDH]

They say timing is everything, and yesterday my technician handed me the first prototype of a new driver board with the A4989 installed. Today when I finish with e-mails and similar stuff, I get to power it up and start software debugging.

The board has a NXP LPC1768 microcontroller with built-in encoder electronics.

The drive will always run the A4989 at 16X microstepping, but works from fullstep or halfstep inputs.

A 32-bit 100MHz processor allows you to run some very fancy interpolation algorithms.

When I'm not designing motorized microscope stages, or similar lab instruments with step motors, I design step motor drivers and controllers for a living. So mostly I get paid to have fun.

By this time next week I should have some waveforms that I can post.

[pminmo]

Allegro notified me last week, I should be getting some parts to evaluate.

[eSilviu]

I see no trace of this new IC in Europe. At least for now.

[rs5088]

Hello, I'm interested to tile traits of the A3986 can now be equipped with the A4989, or have some elements of change?

[rs5088]

Hello, I'm interested to tile traits of the A3986 can now be equipped with the A4989, or have some elements of change?
scheme under which the PCB is made: