Thursday, April 24, 2008
Made in China
Last month, a little company that makes stepper motors in Shanghai posted a thread on the Reprap forums. Annie Fan from CW-Motor basically said that her company makes steppers and that they wanted to do business.
Of course, she got hammered for spamming.
I've lived in China and possibly have a better feel for how the small companies in China try to do business than most, so I emailed Annie and started a conversation. It turned out that they make NEMA 23's just like Darwin needs.
Now as I am sure you all know, I've not been overwhelmed with the Darwin design. The requirement for big, power hungry NEMA steppers has always seem to me to be excessive in the best of lights. I know, mind, that Darwin is a first design and that subsequent work will make much more efficient ones. I guess I just have a slight allergy to working with stepper controllers that pump multiple amps. I didn't scar but I did get a bunch of blisters when I got the diodes hooked up wrong on an L298N board I built last year.
All that aside, I started digging through CW-Motor's website and found a NEMA 23 that is causing me to have a serious rethink about Darwin and my objections to it.
Basically, their 57BYGH320 looks to be the answer to just about every objection I had to the Darwin design in terms of its power hunger. It's a 15v motor that delivers as much power at 12v as the original specification Nanotech that Adrian first used.
I ran the specification past Chris (nophead) since he seems to know loads about stepper motors and he confirmed pretty much everything that I had read into the specification.
This bad boy only draws 0.4 amp at full power. It does that by using a LOT of very thin wire giving it a phase resistance of 38 ohms. What that means is that you could build a stepper controller board to drive this monster around the SN754410 chip that puts out 1 amp that Simon originally specified for Reprap years ago. No need for putting auto headlights in series with this thing to up the resistance or control amperage via firmware. You just wire it up and it should work fine as it is. No special arrangements and the controller for it could be put together and tested by a clumsy 12 year-old without your having to lay in a stock of burn creme and safety classes.
About the only shortfall of this stepper is that it is a 1.8 degree step angle instead of the 0.9 degrees that Darwin wants, so you would have to half-step with it.
Annie will sell this model to you for USD$14.50/unit. I mentioned that it would be nice to have a 0.9 step angle and she said that they'd upgrade that model to 0.9 degrees for USD$21.50, no problem.
Chris, unlike me, had the presence of mind to Google the model number and discovered that CW-Motor are apparently manufacturing these steppers for Kysan. You can see a fuller Kysan spec for this model here.
Anyhow, Annie has turned my calculations for a Reprap design upside down. My little tin-can steppers, which also draw 0.4 amps at twelve volts deliver a small fraction of the torque that this Chinese stepper can deliver for the same amount of electricity. My big objections, electricity waste and high amperage circuitry have disappeared like morning dew on a warm day.
I've been over at Ian's BitsFromBytes looking at the costs of parts kits.
Anyhow guys, take a look at this stepper and see what you think. It could sure make Darwin a lot cheaper and safer to build and operate.
Just to assure you, I have no commercial or consulting connection with this firm.
Of course, she got hammered for spamming.
I've lived in China and possibly have a better feel for how the small companies in China try to do business than most, so I emailed Annie and started a conversation. It turned out that they make NEMA 23's just like Darwin needs.
Now as I am sure you all know, I've not been overwhelmed with the Darwin design. The requirement for big, power hungry NEMA steppers has always seem to me to be excessive in the best of lights. I know, mind, that Darwin is a first design and that subsequent work will make much more efficient ones. I guess I just have a slight allergy to working with stepper controllers that pump multiple amps. I didn't scar but I did get a bunch of blisters when I got the diodes hooked up wrong on an L298N board I built last year.
All that aside, I started digging through CW-Motor's website and found a NEMA 23 that is causing me to have a serious rethink about Darwin and my objections to it.
Basically, their 57BYGH320 looks to be the answer to just about every objection I had to the Darwin design in terms of its power hunger. It's a 15v motor that delivers as much power at 12v as the original specification Nanotech that Adrian first used.
I ran the specification past Chris (nophead) since he seems to know loads about stepper motors and he confirmed pretty much everything that I had read into the specification.
This bad boy only draws 0.4 amp at full power. It does that by using a LOT of very thin wire giving it a phase resistance of 38 ohms. What that means is that you could build a stepper controller board to drive this monster around the SN754410 chip that puts out 1 amp that Simon originally specified for Reprap years ago. No need for putting auto headlights in series with this thing to up the resistance or control amperage via firmware. You just wire it up and it should work fine as it is. No special arrangements and the controller for it could be put together and tested by a clumsy 12 year-old without your having to lay in a stock of burn creme and safety classes.
About the only shortfall of this stepper is that it is a 1.8 degree step angle instead of the 0.9 degrees that Darwin wants, so you would have to half-step with it.
Annie will sell this model to you for USD$14.50/unit. I mentioned that it would be nice to have a 0.9 step angle and she said that they'd upgrade that model to 0.9 degrees for USD$21.50, no problem.
Chris, unlike me, had the presence of mind to Google the model number and discovered that CW-Motor are apparently manufacturing these steppers for Kysan. You can see a fuller Kysan spec for this model here.
Anyhow, Annie has turned my calculations for a Reprap design upside down. My little tin-can steppers, which also draw 0.4 amps at twelve volts deliver a small fraction of the torque that this Chinese stepper can deliver for the same amount of electricity. My big objections, electricity waste and high amperage circuitry have disappeared like morning dew on a warm day.
I've been over at Ian's BitsFromBytes looking at the costs of parts kits.
Anyhow guys, take a look at this stepper and see what you think. It could sure make Darwin a lot cheaper and safer to build and operate.
Just to assure you, I have no commercial or consulting connection with this firm.
Comments:
<< Home
Sounds good. Is there an issue with the max speed of this motor? I'd guess that those extra windings will introduce some extra inductance, slowing down the max stepping rate, but I have no idea if Darwin is getting anywhere near the speed limits of its current steppers. Probably not I would have to guess.
In my twisted imagination, adding more windings is similar to gearing down the output shaft - more torque, but not necessarily more power. As long as we're not limited by the power output or speed, no problem.
I'd buy a set myself, but coming from a background of kA IGBT chopper drivers, I'm leaning more towards overkill on the stepper/controller side of things.
Good find!
Wade
In my twisted imagination, adding more windings is similar to gearing down the output shaft - more torque, but not necessarily more power. As long as we're not limited by the power output or speed, no problem.
I'd buy a set myself, but coming from a background of kA IGBT chopper drivers, I'm leaning more towards overkill on the stepper/controller side of things.
Good find!
Wade
The inductance is higher than the Nanotec ones, but so is the resistance so the time constant (L/R) is the same. I think that means with constant voltage drive they will be just as fast (or slow depending on you point of view).
We should definitely look into this. Can she let us have 3 to try? Just as important, anyone got the time to build them in and try them? I'm afraid I haven't.
Basically we need a stepper-savvy volunteer who's building a Darwin and fancies some cheap/free steppers that may not work, but almost certainly will.
Basically we need a stepper-savvy volunteer who's building a Darwin and fancies some cheap/free steppers that may not work, but almost certainly will.
Well, I haven't started work on my Darwin yet, but I've ordered everything except the steppers and the McMaster parts, they should be good to go by the time I get back to Canada. I was waiting for the RRRF to get some more steppers in stock, but what the heck - I was looking for an opportunity to help out. Might as well order a set from Annie.
Do you guys think I should get the 1.8 deg steppers or the 0.9 version?
I've got a small amount of experience building things with steppers, I should be able to sort these out.
Do you guys think I should get the 1.8 deg steppers or the 0.9 version?
I've got a small amount of experience building things with steppers, I should be able to sort these out.
There was one question. IIRC, Darwin needs one of the stepper motors to have a two-ended shaft? Annie would have to know about that in that I know she can do it, but hasn't quoted on it.
OTOH, didn't Vik do a workaround where that two-ended shaft wasn't necessary any more?
OTOH, didn't Vik do a workaround where that two-ended shaft wasn't necessary any more?
yes, vik came up with a workaround.
unfortunately the RRRF just ordered a new batch of steppers. the plus side is that this gives me time to order some of these for testing.
i'll order a few, try them, and if they are acceptable i will probably switch over to these for the RRRF.
good find! cheaper/better stepper motors are great!
unfortunately the RRRF just ordered a new batch of steppers. the plus side is that this gives me time to order some of these for testing.
i'll order a few, try them, and if they are acceptable i will probably switch over to these for the RRRF.
good find! cheaper/better stepper motors are great!
The 320 motor is 76mm long the Darwin standard design specifies 51mm not an issue (except weight overhang on Y) for the X & Y axis but cuts your (already limited)Z by 25mm. While this motor draws less amps I can't see that it will take less power.
Nophead will verify this I'm sure, but for a given output power you need a given input power the difference being the efficiency of the system (motor & driver) I can't believe they have a fantastically more efficient motor and would expect the power to be equivalent to what we currently use. Just higher voltage less current same power.
While they look cheaper a very good point I'm not sure how much cheaper after freight and duty and is this price for small quantity? Zach buys his existing motors at a good price so would be interesting to see exactly how the economics work out.
Finally it is a Bipolar design which is fine for existing driver circuits but the 8 wire motor gives the option to work with all driver types.
Nophead will verify this I'm sure, but for a given output power you need a given input power the difference being the efficiency of the system (motor & driver) I can't believe they have a fantastically more efficient motor and would expect the power to be equivalent to what we currently use. Just higher voltage less current same power.
While they look cheaper a very good point I'm not sure how much cheaper after freight and duty and is this price for small quantity? Zach buys his existing motors at a good price so would be interesting to see exactly how the economics work out.
Finally it is a Bipolar design which is fine for existing driver circuits but the 8 wire motor gives the option to work with all driver types.
The 320 motor is 76mm long the Darwin standard design specifies 51mm not an issue (except weight overhang on Y) for the X & Y axis but cuts your (already limited)Z by 25mm. While this motor draws less amps I can't see that it will take less power.
Nophead will verify this I'm sure, but for a given output power you need a given input power the difference being the efficiency of the system (motor & driver) I can't believe they have a fantastically more efficient motor and would expect the power to be equivalent to what we currently use. Just higher voltage less current same power.
While they look cheaper a very good point I'm not sure how much cheaper after freight and duty and is this price for small quantity? Zach buys his existing motors at a good price so would be interesting to see exactly how the economics work out.
Finally it is a Bipolar design which is fine for existing driver circuits but the 8 wire motor gives the option to work with all driver types.
Nophead will verify this I'm sure, but for a given output power you need a given input power the difference being the efficiency of the system (motor & driver) I can't believe they have a fantastically more efficient motor and would expect the power to be equivalent to what we currently use. Just higher voltage less current same power.
While they look cheaper a very good point I'm not sure how much cheaper after freight and duty and is this price for small quantity? Zach buys his existing motors at a good price so would be interesting to see exactly how the economics work out.
Finally it is a Bipolar design which is fine for existing driver circuits but the 8 wire motor gives the option to work with all driver types.
Ian,
The torque is given as 18Kg.cm but they are 15V motors and they will only get about 9V so call that 10.8Kg.cm. The Nanotech motors are 1.06 Nm which is almost the same but takes ~800mA @ 9V compared to about 230mA.
So it looks like these are more efficient, probably because they are bigger, so have more room for wire and possible have better magnets.
Static torque is proportional to ampere turns, not power.
Somebody would have to try these of course to be sure.
The torque is given as 18Kg.cm but they are 15V motors and they will only get about 9V so call that 10.8Kg.cm. The Nanotech motors are 1.06 Nm which is almost the same but takes ~800mA @ 9V compared to about 230mA.
So it looks like these are more efficient, probably because they are bigger, so have more room for wire and possible have better magnets.
Static torque is proportional to ampere turns, not power.
Somebody would have to try these of course to be sure.
nophead: we've basically switched away from the nanotec motors. they're way too expensive.
would you mind breaking down the differences between this motor and the Keling motor described on this page: http://www.reprap.org/bin/view/Main/StepperMotor
would you mind breaking down the differences between this motor and the Keling motor described on this page: http://www.reprap.org/bin/view/Main/StepperMotor
For the Keling stepper wired in bipolar serial you get 1.1Nm for 1.7A at 6V so same torque for a lot more current .
The time constant L/R is 3.6ms compared to about 4ms for the other two so slightly faster when driven with a constant voltage drive of 6V. Faster again when driven by 12V through a chopper and even faster when wired bipolar parallel if you have a driver that can give 3.4A.
Ed told me the existing motors will run up to 2000pps but that stresses the mechanics. That is 200mm/s I think so with the Darwin belt drive there is no need for faster motors.
Bottom line is these motors should work well even with the simple PIC driver provided you are on belt drive.
If you are on threaded rod then you need all the speed you can get so the Keling motors and chopper drive are better. But you then have way too much torque so you are probably better with smaller motors with bigger steps.
The time constant L/R is 3.6ms compared to about 4ms for the other two so slightly faster when driven with a constant voltage drive of 6V. Faster again when driven by 12V through a chopper and even faster when wired bipolar parallel if you have a driver that can give 3.4A.
Ed told me the existing motors will run up to 2000pps but that stresses the mechanics. That is 200mm/s I think so with the Darwin belt drive there is no need for faster motors.
Bottom line is these motors should work well even with the simple PIC driver provided you are on belt drive.
If you are on threaded rod then you need all the speed you can get so the Keling motors and chopper drive are better. But you then have way too much torque so you are probably better with smaller motors with bigger steps.
I think the chinese motor is much slower due to its high inductance. The speed at which the current rises through a coil is given by:
dI/dt = V/L
So it's proportional to the applied voltage and inverse proportional to the coil's inductance. The high resistance only makes it worse, because it causes an extra drop in voltage.
Compared to the Keling stepper for instance, the chinese motor's inductance is 10 times higher, and it reaches full torque at 0.4A compared to 1.5A for the Keling one. So (theoretically) the speed at which maximum torque starts to drop is 1.5/0.4 * 0.1 = 0.375 of the Keling one.
In practice, things are worse. At 0.2A, the internal resistance of the chinese motor already causes a drop in voltage over the coil's resistance of 7.6V, lowering the voltage left for the inductance significantly.
Power dissipation is not an issue as well, because the L297/298 pair used in the driver board uses pulse width modulation; no power is wasted in serial resistances. According to my calculations, the chinese motor uses (0.4)^2 * 38 = 6.08W, against 6.57W for the Keling motor, for instance.
At last, the Keling motor has a holding torque of 1.2Nm, against 0.56 for the chinese one.
A fast stepper has a low resistance, a low inductance and should not exceed the maximum current rating for the driver board. If power dissipation is an issue, it's probably best to use the coils in parallel and limit the current to 2A. It will save you some energy, at the cost of a lower torque, but you get the maximum speed out of the stepper.
dI/dt = V/L
So it's proportional to the applied voltage and inverse proportional to the coil's inductance. The high resistance only makes it worse, because it causes an extra drop in voltage.
Compared to the Keling stepper for instance, the chinese motor's inductance is 10 times higher, and it reaches full torque at 0.4A compared to 1.5A for the Keling one. So (theoretically) the speed at which maximum torque starts to drop is 1.5/0.4 * 0.1 = 0.375 of the Keling one.
In practice, things are worse. At 0.2A, the internal resistance of the chinese motor already causes a drop in voltage over the coil's resistance of 7.6V, lowering the voltage left for the inductance significantly.
Power dissipation is not an issue as well, because the L297/298 pair used in the driver board uses pulse width modulation; no power is wasted in serial resistances. According to my calculations, the chinese motor uses (0.4)^2 * 38 = 6.08W, against 6.57W for the Keling motor, for instance.
At last, the Keling motor has a holding torque of 1.2Nm, against 0.56 for the chinese one.
A fast stepper has a low resistance, a low inductance and should not exceed the maximum current rating for the driver board. If power dissipation is an issue, it's probably best to use the coils in parallel and limit the current to 2A. It will save you some energy, at the cost of a lower torque, but you get the maximum speed out of the stepper.
No Botpot the time constant, which defines when the current gets to about 2/3 its final value is L/R. So a bigger R actually makes the motor faster.
The final I is V/R, so as R gets bigger the final current is less, so the time it takes to reach it is less.
In your comparison you forgot to factor in the Chinese motor is 15.2V for 0.4A and the Keling is 6.12V for 1.7A.
I don't know where you get holding torque of 0.56 from. It's given as 18kg/cm, which I make 18 * 9.8 = 176 Ncm. Allowing for 9V instead of 15.2V that gives about 100Ncm compared to 110 for the Keling.
Keling power is 1.7A * 6.12V = 10.4W. Chinese at same torque is 9V^2 / 38R = 2.1W, much more efficient.
The final I is V/R, so as R gets bigger the final current is less, so the time it takes to reach it is less.
In your comparison you forgot to factor in the Chinese motor is 15.2V for 0.4A and the Keling is 6.12V for 1.7A.
I don't know where you get holding torque of 0.56 from. It's given as 18kg/cm, which I make 18 * 9.8 = 176 Ncm. Allowing for 9V instead of 15.2V that gives about 100Ncm compared to 110 for the Keling.
Keling power is 1.7A * 6.12V = 10.4W. Chinese at same torque is 9V^2 / 38R = 2.1W, much more efficient.
nophead, I had the wrong numbers. I'm sorry for that. Here's another comparison, with the correct numbers:
I get a time constant L/R = 3.67mS for Keling, 3.95ms for the Chinese one.
Now supply both motors with 12V. The current at infinity will be 3.33A for the Keling, and 0.32A for the Chinese stepper. So the Keling motor will reach 2.22A after 3.67ms, and the Chinese motor will reach 0.21A after 3.95ms. The time for both motors to reach half torque is, using linear interpolation, 0.85/2.22 * 3.67 = 1.4ms for the Keling motor, and 0.2/0.21 * 3.95 = 3.76ms, which is a significant difference.
The torque of the Chinese motor is indeed larger. For a torque of 0.5Nm, the Keling needs 0.77A, and the Chinese one 0.11A. The Keling reaches this point in the above example after 1.27ms, the Chinese one after 2.13ms. At that point, the Keling dissipates 2.13W, and the Chinese motor 0.46W, which is indeed much more efficient.
Put it in another way, although the time constants are approximately equal, 12V is for the Keling motor twice its nominal voltage, and for the Chinese one 80% of its nominal voltage. If you want to make the Chinese motor react as fast as the Keling one, you'll need to supply it a higher voltage, something like 25 or 30V.
Speed comes at an efficiency cost. The steppers I use are really fast, but even less efficient than the Keling ones. They have the following specs (bipolar serial):
Max torque: 0.5Nm
at current: 1.25A
Resistance: 1.52ohm
Induction: 3mH
Supplied with 12V they reach 0.5Nm after 0.469ms, dissipating 2.375W. They are made over here, in the Netherlands.
I don't see the power dissipation as a big problem, my guess is that most of the power goes to the heater in the extruder. A total of 7W for the steppers is probably even less than the power that is wasted in the PC power supply. Correct me if I'm wrong.
I get a time constant L/R = 3.67mS for Keling, 3.95ms for the Chinese one.
Now supply both motors with 12V. The current at infinity will be 3.33A for the Keling, and 0.32A for the Chinese stepper. So the Keling motor will reach 2.22A after 3.67ms, and the Chinese motor will reach 0.21A after 3.95ms. The time for both motors to reach half torque is, using linear interpolation, 0.85/2.22 * 3.67 = 1.4ms for the Keling motor, and 0.2/0.21 * 3.95 = 3.76ms, which is a significant difference.
The torque of the Chinese motor is indeed larger. For a torque of 0.5Nm, the Keling needs 0.77A, and the Chinese one 0.11A. The Keling reaches this point in the above example after 1.27ms, the Chinese one after 2.13ms. At that point, the Keling dissipates 2.13W, and the Chinese motor 0.46W, which is indeed much more efficient.
Put it in another way, although the time constants are approximately equal, 12V is for the Keling motor twice its nominal voltage, and for the Chinese one 80% of its nominal voltage. If you want to make the Chinese motor react as fast as the Keling one, you'll need to supply it a higher voltage, something like 25 or 30V.
Speed comes at an efficiency cost. The steppers I use are really fast, but even less efficient than the Keling ones. They have the following specs (bipolar serial):
Max torque: 0.5Nm
at current: 1.25A
Resistance: 1.52ohm
Induction: 3mH
Supplied with 12V they reach 0.5Nm after 0.469ms, dissipating 2.375W. They are made over here, in the Netherlands.
I don't see the power dissipation as a big problem, my guess is that most of the power goes to the heater in the extruder. A total of 7W for the steppers is probably even less than the power that is wasted in the PC power supply. Correct me if I'm wrong.
I have a suggestion related to the stepper power dissipation.
When the steppers are moving the axes, power is needed to accelerate and to overcome friction. But when the steppers aren't moving, all power is wasted into heat. Wouldn't it be better to lower the current through a stepper when it isn't rotating? There's not much energy needed to keep the axes in place.
To achieve this, the stepper motor driver needs a little hack. Instead of using the trimpot to limit only the maximum current through the motor, we need a voltage that can be regulated from the arduino board. My idea is to use one of the arduino's PWM outputs, and convert the PWM signal into an analog value using a capacitor and a resistor. When the motor is moving, the arduino should supply a continuous high voltage, but when the motor is holding, it can serve a PWM signal with a pulse/pause ratio of, say 10%, effectively limiting the current through the stepper at any value you want.
When the steppers are moving the axes, power is needed to accelerate and to overcome friction. But when the steppers aren't moving, all power is wasted into heat. Wouldn't it be better to lower the current through a stepper when it isn't rotating? There's not much energy needed to keep the axes in place.
To achieve this, the stepper motor driver needs a little hack. Instead of using the trimpot to limit only the maximum current through the motor, we need a voltage that can be regulated from the arduino board. My idea is to use one of the arduino's PWM outputs, and convert the PWM signal into an analog value using a capacitor and a resistor. When the motor is moving, the arduino should supply a continuous high voltage, but when the motor is holding, it can serve a PWM signal with a pulse/pause ratio of, say 10%, effectively limiting the current through the stepper at any value you want.
The L298 drive circuit only gives about 9V when loaded at 1.7A so the difference is less marked.
At 9V the Chinese motor gives the same torque as the Keling at 6V. Current rise is an exponential curve so you can't do linear interpolation.
I = Imax(1 - e^-(tR/L))
For the Chinese to get to half torque it will take about 0.7 * 3.95 = 2.8ms. For Keling to get to the same torque the current will be 0.5 * 6 / 9 i.e. 1/3 which happens at about 0.4 into the time constant so about 1.5ms.
Keling is faster, probably around twice as fast, but with a lot more power and a chopper drive. Extra speed is not required for Darwin's belt drive, only Seedling's threaded rod.
The power saving for three axes is considerable: about 6W against 30W in the motors and further savings in the driver chip losses, which are significant with Darlintons. In contrast the heater uses about 12W. Yes PCs waste far more power but adding an SD card to the micro will allow it to run without a PC.
The advantage of low power is that a cheaper drive chip without a heatsink will do for the Chinese motors and it won't get as hot, which is a problem if you make the mounting bracket and coupling out of PLA or PCL.
Yes you can reduce power when not moving but building an object can take several hours without pauses so it doesn't really help except with the z-axis.
At 9V the Chinese motor gives the same torque as the Keling at 6V. Current rise is an exponential curve so you can't do linear interpolation.
I = Imax(1 - e^-(tR/L))
For the Chinese to get to half torque it will take about 0.7 * 3.95 = 2.8ms. For Keling to get to the same torque the current will be 0.5 * 6 / 9 i.e. 1/3 which happens at about 0.4 into the time constant so about 1.5ms.
Keling is faster, probably around twice as fast, but with a lot more power and a chopper drive. Extra speed is not required for Darwin's belt drive, only Seedling's threaded rod.
The power saving for three axes is considerable: about 6W against 30W in the motors and further savings in the driver chip losses, which are significant with Darlintons. In contrast the heater uses about 12W. Yes PCs waste far more power but adding an SD card to the micro will allow it to run without a PC.
The advantage of low power is that a cheaper drive chip without a heatsink will do for the Chinese motors and it won't get as hot, which is a problem if you make the mounting bracket and coupling out of PLA or PCL.
Yes you can reduce power when not moving but building an object can take several hours without pauses so it doesn't really help except with the z-axis.
Damn! I feel like I'm getting a graduate seminar on stepper motor dynamics. I've learned more from you two talking about the relative merits of the Keling and the CW Motors steppers here than I've been able to find over the rest of the whole web.
It would be brilliant if, when you guys could find the time, you could write up a short course for about stepper design.
It would be brilliant if, when you guys could find the time, you could write up a short course for about stepper design.
I'm building a reprap and am at the point where I need to get my hands on the steppers. I've emailed them asking about a 57BYGH320 with a 0.9deg step. If that goes anywhere and they're willing to send me a set I'll let you all know how it goes.
-Tyson
Post a Comment
-Tyson
<< Home
