A3985/A3986 5A/50V Bipolar Stepper Design

[NE5534]

There was a long series of posts regarding the A3986 controller in a previous thread. I recently did a design using the A3985 controller. The A3985 is the fully functional / fully programmable version of the pair. Our requirement was precision operation over an extremely wide range of RPM with controlled acceleration, for which the A3985 was ideal.

Microstepping controllers require much higher dynamic range and very careful circuit layout than simple Full/Half/Qrtr step designs. The design shown here is high performance and requires less than 2x2 Inches of double sided. The layout supports both the A3985 and A3986, and only requires changes to the MCU firmware. It can also be readily modified for 10A or 20A designs. For this 5A design, convection air cooling is adequate for the TO-220's.

Regards, Chris / BSEE

[pminmo]

Actually the 3985 while an improvement over the 3986 because of the more flexability of timing, still has the same problem as the 3986, current control issues.

[NE5534]

Just out of curiosity, what microstepping controller IC do you like?

I guess my point would be that there are not a lot of choices. Comparing these parts to an L297 or some other Full/Half IC is apples and oranges. As I said microstepping requires much higher dynamic range. It's a tall order for any PWM power monolythic IC to handle.

To illustrate consider the following:

Assume a motor with R=0.5 and L=2.5mH
Assume we want to run it at 5A on 50V supply

Max Full Torque Impedance = 50V/5A = 10 Ohms
Max Full Torque Frequency = 10/(2*Pi*L) = 640Hz
Dynamic range for Impedance = 10/0.5 = 20:1 which is 26dB
Dynamic range for Amplitude 1/16 microstepping is another 24dB
Total dynamic range of current = 26+24 = 50dB

Now say we throw in another 10dB just for some linearity, glitches, blanking time, comparator response time, etc, and we have a PWM system that needs 60dB dynamic range. In simple terms that is a resolution of 1000:1

To implement this kind of dynamic range on an analog switching power circuit is far from trivial. The circuit design and PCB layout is critical, and must be extremely clean and well designed. Any glitches/spikes/noise can easily destroy the performance.

Alternatively, in the form of a digital PWM system assume we have a 50kHz PWM switching frequency. To provide 1000:1 resolution, the PWM edge resolution would need to be at least 10bit and the clocking frequency would need to be at least 50MHz. Producing a high voltage (50V) CMOS IC with those capabilities is also far from trivial, or impossible.

The A3985 has about 500% more duty cycle range than the A3986. So while it is correct to say that they both have limited dynamic range, one is certainly much larger than the other.

I agree that there are some logic issues in the A3985/6 involved with its comparators. Few chips are perfect. There are also workarounds. However the A3985 has some remarkable flexibility and works very well in our application. What it does do, very few chips can do. A great deal depends on how it is driven, quality of the circuit layout, and the design requirements.

Everything is a compromise, and the A3985/6 is no exception. To produce a high performance microstepping controller design really requires a dual chip solution. The high voltage driver portion needs to be separated from the controller portion, which can then be a low voltage high speed process. Unfortunately, nothing appears to be on the market in that form yet.

All things considered, as a single chip solution, the A3985/6 provides about the smallest, most flexible, and simple microstepping design at present.

Chris.

[LazyBee]

Hello,
your project is interesting for me.
Do you have more informations. What is the size of the board. Do you have already a functional sample?
Is it possible to connect the board direct to the bob and can I drive it with Mach3?

Wilfried

[maMar]

Hi.
your project is great
can you tell us with the result of this board
i'm very interesting

thanks.

[NE5534]

Actually I designed this, but never built it. The overall performance we required could not be met by this type of design. I talked to Allegro about various issues, and all of their chips have been designed for specific large OEMs. Everything they do is designed to order, and then they offer the same part to the general market as-is afterwards. There are design defects in many of the chips and they never get fixed if it met the original needs of the initial OEM client.

I came to the conclusion that there were no off-the-shelf IC microstepping controllers that would meet our needs. I decided to roll my own controller out of individual parts. That was a job, many of the so called "motor control" type MCUs had serious bugs. No single chip type solution could do it. I ended up using a Xilinx 9572 CPLD with an ADI ADU7021 ARM7 MCU, a pair of ADI AD8216 diff amps for true floating current sensing, 4 LM5109B drivers, a MCP6562 comp, and 8 FDP20AN06 FETs. This was a 5A/48V bipolar design

That's what it took. It's a relatively expensive design, not simple, but it gave us exactly what we needed. It works fantastic. Extremely high performance over very wide range RPM, and with 16 microsteps. We also run controlled accelleration/decelleration algos and need precise motor control.

I also implemented a different type of bidirectional current control scheme. There are no Fast/Slow decay time constants at all. Rather, the current through the coils is set continuously as required by the sine table, and the FETs are driven to force the current both up and/or down to maintain that current.

I don't know if this has been done before, I don't see it used by any IC controllers. In effect, it provides for optimum drive at any RPM automatically, because both the rising and falling currents are both regulated by the feedback loop. I also implemented a special alternation scheme that forces sharing of all current through all 8 FETs for equally shared dissipation.

Most controllers just regulate the rising current, but let the falling current decay by either the Fast or Slow RC time constants. This is not optimum. As a result they cannot follow the sine across wide frequency. This controller follows the sine perfectly up to the current slew limits imposed by the motor R/L and Vcc, and then automatically changes to square wave drive as needed at high RPM automatically. This is all handled by the natural bidirectional feedback control. There are no resonance issues because the motor is under full current control at all times.

Perhaps more info than you wanted, but that was my solution. Wish I could roll it into a single chip solution and offer it for sale, but it's not yet possible. I initially tried that with the PIC33F, but it had too many serious bugs. I tried several other similar MCUs as well. Same problem. Could only do it with separate high quality parts that all work. Current control requirements demand clean high dynamic range and fast logic control.

Chris.

[maMar]

thanks for your kind reply
you know .......i ordered 5 allegro a3986 chips yesterday,and i wase so excited to work with this chip, but you make me disappointe.
can you send some schematics to explain more what you just wrote above.
please can you explain more about the weak spots of a3986.
many thanks for your support.

[NE5534]

The A3986 chip has logic bugs as described on this forum elsewhere.

Giving you just a schematic would be of little help. You would also need my Verilog code for the CPLD, the C code for the MCU, and a proper PC board layout which is almost more important.

[RomanLini]

Originally Posted by NE5534 ... I also implemented a different type of bidirectional current control scheme. There are no Fast/Slow decay time constants at all. Rather, the current through the coils is set continuously as required by the sine table, and the FETs are driven to force the current both up and/or down to maintain that current. I don't know if this has been done before, I don't see it used by any IC controllers. ...

Yes, that's what the Linistepper does.

Full real-time linear control of the phase currents. Which is one of the reasons it out-performs chopper/bipolar driver chips etc in that 1A motor current range.

[NE5534]

The other controller may be doing the same kind of drive method, but it does not have the resolution or speed we needed. I wish it did, it would have saved me a lot of time.

Here's an example of my DAC current programming voltage waveforms on the left, and the actual coil current waveforms on the right. This is 7A peak and the motor was around 100 RPM.

[doorknob]

I am having some difficulty reading the scales on the DAC voltage and coil current traces in the photo - can you provide more info or a higher resolution image?

Thanks...