[.xXACEXx.]

this is a dumb question i guess,but i gotta ask...

you mention a 1000 line count encoder,...so my question is are you using an off the shelf typical stepper motor with a encoder attached .in the same way a servo has?.if so will the end user need to buy the same stepper motor ,encoder,and stepservo drive to have motion control? or will this stepservo somehow get feedback from a typical stepper in the end with out the need for an encoder?


im (as always) a lil bit consused

[ger21]

Any double shaft stepper with an encoder mounted on the back will work. If you already have the steppers, you'll need the new drives and encoders.

[Mariss Freimanis]

A 1,000-line encoder is being used on step servo "development board" which also includes a tricked-up 20 microstep mixed-mode drive. The idea is to over-design every part of the servo so that hardware deficiencies are not a question during the development of the system. The primary design goal is to learn the underlying principles necessary to make to make an idea work.

Once everything works in a pleasing and respectable manner, the second phase of the design process kicks in; minimize and optimize:

Minimize means studying every aspect of the design for what is redundant, underused or overly complicated. It's trim the fat time; the design emphasis changes to component reduction without compromising performance. Many resistors, capacitors and other parts bite the dust in this phase.

Concurrent is optimization; it's the horse to minimization's carriage. Are the remaining circuits and their component values the best possible? What can be changed to get the best possible performance from a circuit function block? What can be moved from a digital implementation to analog or visa versa? What resources are freed if a move is made; can a freed resource be utilized to add a feature from the wish-list?

All our design projects have a "must accomplish" list and a secondary "would be nice to have" list. The members on the first list have confirmed seats, the members from the second list fly standby. Example: The pulse multiplier in the G320X was a "standby passenger" who got on-board because of a "seat" opened by the minimize and optimize design process.:-)

Mariss

[TOTALLYRC]

All our design projects have a "must accomplish" list and a secondary "would be nice to have" list. The members on the first list have confirmed seats, the members from the second list fly standby. Example: The pulse multiplier in the G320X was a "standby passenger" who got on-board because of a "seat" opened by the minimize and optimize design process.:-)

Mariss

Sounds like Gecko airways is the only way to fly.

Of course it depends on price, but it sounds like you will obsolete all of your other stepper drives, at least for the hobby market.
If I can swing the funds, my router would benifit greatly fron these new drives. As luck would have it, all of my motors are dual shaft and I have some amt-102 encoders on the shelf. :banana:How serendipitous.:banana:

Mike

[Meduza]

This sounds great, if i understand you right we are propably talking about less than a year before i can buy this wonderful drive?

I will propably not have finished my new build before that, so maybe i should just sell my G201 to someone and wait out for this one, it will be soo much better

Good thing i bought those 500oz/in dual-shaft motors ^_^

[Mariss Freimanis]

OK, an update on this subject:

1) I have a working PID step motor servodrive and I've had it for several months now. I wasn't entirely happy with it because it's too complex and certain aspects (inductive phase lag compensation) were solved using open-loop techniques. It didn't meet the "beauty is simple" template.

2) I was aware of the Field Oriented Control (FOC) algorithms in a general way. They are primarily applied in top of the line 3-phase BLDC servomotor designs. Using Clarke-Park math transforms, they convert from a rotating reference to a stationary one to allow a PID filter to work on a baseband signal and then using the inverse transforms, convert the signal back to a rotating reference. Field weakening is vital using a step motor and the FOC algorithms allow for that in an elegant way.

I had ignored these transforms because the standard implementation requires the use of a DSP (digital signal processor). The FOC algorithms are computationally intensive (sin, cos and 10 multiplications). DSPs can do the job but are expensive themselves and this expense is compounded by the numerous signals that require ADC and DAC conversion (analog to digital conversion and digital to analog conversion).

Then a three things occurred to me: Why not keep everything in the analog domain (no conversion necessary) and how to make a simple, accurate (+/-0.25&#37 and fast (<2uS) semi-analog 4-quadrant multiplier. These two ideas made it possible to revisit the FOC algorithms. A simplification was the Clarke transform and its inverse is unnecessary because a step motor already operates in the quadrature reference frame.

I drew up an analog FOC controller schematic and had some prototype boards fabricated. I populated a board Monday, commissioned it Tuesday and began to take data from it yesterday and today. Let me say FOC with Field Weakening is absolutely amazing!

a) The motor is amazingly silent (like you can't hear it) from 0 RPM to 1,500 RPM. It still sounds very quiet for a step motor out to past 10,000 RPM.

b) Unloaded, the 2A per phase, 2mH test motor draws <4W from the power supply at 1,500 RPM. The motor stays cool to cold unloaded. Loaded, it gives 48W mechanical at 24VDC.

c) Once I qualified the board as being true torque-mode amplifier, I did a quick circuit lash-up for a P-type servo (proportional mode servo) and took some data. I commanded the motor to turn at 1,500 RPM and reversed the direction twice a second. The motor did this in less than 180 degrees of travel; a net 3,000 RPM change in speed in less than 1/2 revolution of the motor shaft. Can't do that with an open-loop stepper.:-)

The design is remarkably simple; 4 quad op-amp packages, 10 SOT-23 size "multipliers", a dual comparator and a single 64-macrocell CPLD. That's it. Board area consumed is just over 1 square inch minus PID trimpots. Even the CPLD isn't stressed with only 50 macrocells used. All this will fit with room to spare in a G202 package.

I mentioned 3 things that occurred to me? The third is there is no significant difference between a 3-phase BLDC motor and a 2-phase step motor except for the pole-count and the winding phase excitation. What works for a step motor servo will work for a BLDC motor servo. I now have two parallel projects going on; a step motor servo and BLDC motor servo. Both will be FOC and Field Weakening designs.

My hat's off to the FOC math transforms authors. I only wish I had thought of them. They are very elegant and solve some complex riddles in a simple and beautiful way. I would like to think I took these wonderful equations and applied them in an entirely analog way ($3 worth of parts) while the authors had intended them for a digital domain solution. Either way, they work brilliantly.

Mariss

[CJL5585]

Thanks for the update Mariss.
Sounds great.

I am looking forward to future updates, especially on the BLDC servoed unit.

[joeybagadonuts]

Any idea on what it will cost?

[raymondg]

Mariss,
Thanks for the update. Still hoping to use your step motor servo in my next conversion project. Any projections on dates, costs, and basic specs like power and motor sizes supported?

Thanks,
Raymond

[veteq]

Nice to that everything is going great!!!
How about the encoder that is used, is it also possible with a 500Cpr encoder?

Kind regards,

Roy B

[jmkasunich]

Interesting approach.

In a former life I've used the Analog Devices AD2S100 chip to do the vector transforms using a mix of digital and analog. The rotation angle is supplied as a digital word, but the rotating and stationary vectors are both analog. One chip was used to transform current feedback from the rotating frame to stationary, then we had analog PID loops for Id and Iq, then the resulting voltage vector went thru another chip to take it back to the rotating frame, before going to an old-school PWM modulator (with an actual analog triangle and comparators).

The AD2S100 implementation uses the digital angle to look up sin(angle) and cos(angle) in tables, then drives two multiplying DACs with sin and two with cos. The source vector is used as the reference input to the four DACs, and the outputs are summed (analog) to produce the result vector.

Are you doing something along these lines? I'm guessing you might be replacing the multiplying DACs with some PWM'ed analog switches, where the duty cycle is sin or cos?

[svenakela]

I was at the Tech Fair a couple of weeks ago and there was company showing a setup with steppers and encoder feedback. They claimed to be the fastest, I don't know about that but it was pretty good speed on the motors. According to the price, they better be fast...

[Mariss Freimanis]

jmkasunich,

I considered using sine and cosine PWM signals but I rejected that approach as too complex. It requires a 2-pole active LP filter and creates only a 2-quadrant multiplier if open-drain CPLD outputs are used. External analog switching is required to generate a true 4-quadrant multiplier. There is also the issue of phase lag introduced by the LP filters. I threw up my hands when the op-amp count passed 32 just to do the a,b/d,q and d,q/a,b transforms. Parts count estimate was >250 when allowing for an average of 8 passives per op-amp.

It then occurred to me that a potentiometer is a natural analog multiplier. Digital versions are small (SOT-23), fast and cheap. To test this idea, I designed a board which I have running now.

The signals stay in the analog domain from the phase current sense resistors all the way to the bridge PWM comparators. The main signal path uses only 8 op-amps to generate the 8 required product terms, their 4 sums and the 4 PI (proportional-integral) terms. A small CPLD does quadrature decoding, maintains the encoder position angle counter 'theta', generates 8-bit sin/cos theta serial data streams, forms a moving average filtered PWM output proportional to motor velocity and outputs a 12-bit serial data stream of the position error counter. A 12-bit serial DAC (also SOT-23) feeds the analog PID filter which unbalances the q term sum to generate motor torque. A field-weakening algorithm uses the motor velocity PWM signal to unbalance the d term sum to allow motor operation in the inverse torque region.

Long and short, the motor operates beautifully to 1,500RPM and soon out to 8,000 RPM once I resolve some issues related to field-weakening. Past that things fall apart due to the sin/cos theta multiplier update time of 1.8uS based on a 5MHz CPLD clock. In my judgment, that is fast enough (500kHz step pulse frequency using a 1,000-line encoder).

Presently I'm just happy to see the motor reverse direction at 1,200 RPM in less than 90 degrees of shaft rotation. I'll post pictures later today.

Mariss

[jmkasunich]

Clever technique. Your ability to do more with less continues to impress me. ;-)

I have a question about field weakening a PM motor - what happens when you are running very very fast and "something" happens that shuts down the drive? At high speed, isn't the open circuit un-weakened motor back EMF very high?

On a work project with a very high speed PM motor that did heavy field weakening, the back EMF could blow up the drive before the motor slowed down to a safe speed.
However, that was much larger motors, maybe the stored energy in a stepper isn't enough to worry about...

[.xXACEXx.]

wow! changing direction in less than 1/4 turn and able to get to 1200 rpm before direction change? thats pretty fast , those drives should "rock"

just imagining a router with 1/2 x 10 x 5 start trying to move that fast..there by shaking itself apart

would a bldc motor be any better than a step motor at those kind of speeds or at the torque range,or would there be any advantage to use a bldc motor?

watching with anticipation

[Mariss Freimanis]

jmkasunich,

The same is true for an open-loop step motor drive. Desynchronizing (stalling) a step motor at high speeds results in breath-taking BEMF voltages; hundreds of volts even from a 24VDC powered drive. The power section bridge design must handle the resulting insult by shunting the returned energy back to the supply's bulk capacitor or dissipate it in an overvoltage protection shunt across the supply bus. A full-bridge power section design greatly simplifies the solution because the intrinsic drain to source diodes form a full bridge rectifier from the motor winding to the DC supply bus.

The problem is caused by the mechanical energy stored in the motor and load moment of inertia; just by itself, a NEMA-34 motor spinning at 6,000 RPM stores about 35J of energy. Experiments have shown well over 20J of that gets converted back into electrical energy, the balance being dissipated mechanically.

xXACEXx,

It's pretty fast accel/decel all right; The NEMA-23 test motor will literally jump off of the lab bench if it's not secured. Keep in mind this is on an unloaded motor and the purpose is to see how closely the results approach the theoretical max accel rate of the motor.

This is a funny screw-up: I used a really large square NEMA-34 motor as a weight atop the NEMA-23 and even that barely held it. Yesterday the NEMA-23 bucked it off and the NEMA-34 fell onto the prototype drive. That brought all further testing to a halt until the board was repaired. I have the motor screwed down to the bench now.:-)

A BLDC servo is definitely in the schedule now, based on the recent FOC circuit results and the fact both require remarkably similar control circuits. The main difference is a BLDC also requires a Clarke transform and its inverse. The Clarke transform converts a 3-phase rotating reference frame to a 2-phase or quadrature reference frame. Its inverse does the vice-versa thingy.

Mariss

[amplexus]

Elegant stuff Mariss,
I don't understand what you are doing with the digital pot.
Amplexus Ender

[Xerxes]

"would a bldc motor be any better than a step motor at those kind of speeds or at the torque range,or would there be any advantage to use a bldc motor?"

A good BLDC or AC motor accelerates from 0 to 3000 rpm in 1/6 of revolution or even less.

I have done similar development with steppers and FOC about for about 2 years now but not putting all of my effort in it. This is because today you can buy a real BLDC motor with Nema frame sizes for price of a stepper (see this). A low pole count motor also runs quieter, cooler and smoother than stepper will ever run.

[Mariss Freimanis]

The main difference between step motors and BLDC motors is the pole-count. A typical BLDC is a 4-pole motor while a typical 1.8 degree stepper is a 50-pole motor. Everything else being equal, the more poles you have, the greater the low-speed torque.

Everything is not equal; a BLDC is a 3-phase motor while a stepper is a 2-phase motor so let's equalize them: Poles times phases is '100' for a stepper and poles times phases is '12' for a BLDC. This means a stepper should have 8.3 times more low-speed torque as a same size BLDC motor. Let's compare two motors:

Keling KL23BLS_115 BLDC motor has a rated torque of 0.43Nm and a rated 4,000 RPM
Keling KL23H2100 step motor has a low-speed torque of 3.5Nm

The stepper has 8.1 times more torque than the BLDC. This meshes nicely with the expected 8.3 number.

The BLDC motor power is 180 Watts at 4,000 RPM
The step motor torque at 500 RPM has to be 3.45Nm for the same power.

Motors attach to mechanisms. By extension from the above comparisons, a BLDC motor needs 8 times greater reduction gearing than an equivalent step motor. This makes a big difference if a stepper can run a load 1:1 (no gearing needed) while the BLDC requires an 8:1 reduction (gearing needed).

The BLDC motor mentioned above is the one I'll be using to test our upcoming BLDC servodrive.

Mariss

[veteq]

Once i`ve seen a nice guy standing on a THK linear actuator, he moved with rocketspeed in you`re little movie. Is he still alive? can we see him with the new drive including the quiet stepper motor. Would love to see and hear the drive working.....

Kind regards,

Roy B.