Open vs Closed Loop (steppers vs servos)

[beefy]

We were actually considering a carbon fibre gantry.

If you know what you are doing (I don't) you could do it in your home garage for about $150.

Rod, you are starting to give me ideas

[ScottRa]

Ok, good discussion here, I hope this isn’t too much of a tangent. I’ve seen some mention of dual drive gantries going out of sorts and twisting enough to do damage. I would imagine some of that would be outright electronics or motor failure, but can that happen if the z axis assembly is on one side of the table when it accelerates or decelerates and the extra mass on one side causes a stepper motor to slip steps? A servo would compensate because of the encoders.

With these two possibilities in mind does anyone go to the trouble of using a cable system under the table tied to the carriage on both sides to keep the gantry from going out of square if something bad happens. I ran across this years ago when I saw how some small flat bed plotters were designed. The plotters just used one motor and the cable was how the platten was kept parallel. In a plasma it would just serve as protection and to maybe even out the loading on the motors some. Benefits to just a stepper or servo as well? Any merit to that idea?

[Rodw]

These drives that momentarily increase current to a stepper motor sound good but doesn't that mean you need larger/heavier steppers from the start in order to take advantage of this feature.

Absolutely agree. But often the motor does not require to be driven at its maximum current so there is usually some headroom. Don't aske me how many motors we reviewed against our model.

Rod,
love your idea of a carbon fibre gantry. I've often thought of 4140 thin wall steel tube spaceframe for a gantry design. That gets used in aeroplane and racecar frames, but I think carbon fibre would be the ducks nuts.

Thanks Beefy. We got excited too when we found out how easy they would have been to make. Unfortunately, once the gantry forces were determined, there was no net benefit over an aluminium gantry beam so we chickened out!

Looking forward to seeing your tables on the market, I think they will be quite special.

Its starting to come together. I saw first cut of control box layout today and the Z axis is just about designed. Just a couple of limit switches and the like to go. Its like nothing you've ever seen. Kinda like form over function but still with the goods under the crazy skin.

All I can say is that its amazing how you can come up with much better solutions when the right people are working together.

[SegoMan DeSigns]

My gantry will have a removable air drill attached to it capable of 400 pds of uplift pressure so I had to go big or go home. I looked at the largest 80 / 20 profiles by the time I paid the extra length / weight charges along with t-nuts a <10' section would have been in the $800 range. I bought a 20' section of 4x6x1/8" rect tubing for a tad over $120. The weight per ft is the same as the aluminum profile and I have a shop full of drill presses and a tapping machine.

Heavy gantries are a give and take, I do mostly 1/4 to 1/2" materials so a speed cutting demon is not a must . My theory is determine the maximum cutting speed on your machine and adjust amperage / consumables around it. The rapid traverse will save travel time on the material loading process of a 6x12 table.

[robertspark]

Sometimes Robert you have no idea. We were actually considering a carbon fibre gantry.

I'd like to see the pressure vessel that the carbon fibre gantry is going to be cured in. I've done a bit of work with a few companies who manufacturer carbon fibre.

You still have no idea. No pressure vessel required. If you know what you are doing (I don't) you could do it in your home garage for about $150.

I do actually.... (Maybe a little about a lot )
https://www.hindawi.com/journals/amse/2013/356824/

It does depend on the quality of what carbon fibre you are producing / intending to produce

If you are doing it in your garage, you are effectively producing a sort of fibre glass where you have replaced the glass fibre with carbon fibre. It will neither be the strongest nor lightest that it can be given the cure needs to be done under pressure and temperature.

Good luck, I'm sure it will be a great table, I am not rubbishing the idea, it sounds good, will look great and will no doubt be a learning experience.

If you design it with something that uses fea, you will no doubt consider that it is a composite and fea presumes a homogeneous material so it can be difficult (impossible at present as I have not found anything yet) for fea to simulate the grain direction and also bond effectiveness of the carbon fibre and other such composites or additive manufacturing processes (eg 3d printing). (If you know something that does composite / additive fea I would appreciate the pointer)

I was handball calculating moments of inertia for structures 25 years ago and deflections etc.

I am actually a consulting engineer when covid19 is not putting everyone off living life normally and being scared to death of meeting anyone else and doing business in the UK.... Tough times ahead I fear.

[Rodw]

Ok, good discussion here, I hope this isn’t too much of a tangent. I’ve seen some mention of dual drive gantries going out of sorts and twisting enough to do damage. I would imagine some of that would be outright electronics or motor failure, but can that happen if the z axis assembly is on one side of the table when it accelerates or decelerates and the extra mass on one side causes a stepper motor to slip steps? A servo would compensate because of the encoders.

With these two possibilities in mind does anyone go to the trouble of using a cable system under the table tied to the carriage on both sides to keep the gantry from going out of square if something bad happens. I ran across this years ago when I saw how some small flat bed plotters were designed. The plotters just used one motor and the cable was how the platten was kept parallel. In a plasma it would just serve as protection and to maybe even out the loading on the motors some. Benefits to just a stepper or servo as well? Any merit to that idea?

I've twisted mine a couple of times but it always squares up when homed with Linuxcnc. My machine is quite floppy really when not powered up then its like a rock. I have seen a destroyed gantry end on a commercial table. It makes you wonder why though. It was a servo system so the following error from the encoders should have triggered an alarm. But things sure happen fast! There is some poor design out there.

It would not be hard to run a single shaft to both sides and use one motor. But if something goes wrong, there would likely be more damage...

The mass of my Z axis assembly is really insignificant.

[robertspark]

Ok, good discussion here, I hope this isn’t too much of a tangent. I’ve seen some mention of dual drive gantries going out of sorts and twisting enough to do damage. I would imagine some of that would be outright electronics or motor failure, but can that happen if the z axis assembly is on one side of the table when it accelerates or decelerates and the extra mass on one side causes a stepper motor to slip steps? A servo would compensate because of the encoders.

With these two possibilities in mind does anyone go to the trouble of using a cable system under the table tied to the carriage on both sides to keep the gantry from going out of square if something bad happens. I ran across this years ago when I saw how some small flat bed plotters were designed. The plotters just used one motor and the cable was how the platten was kept parallel. In a plasma it would just serve as protection and to maybe even out the loading on the motors some. Benefits to just a stepper or servo as well? Any merit to that idea?

The servo or stepper drives should stall before permanent damage occurs..... Operative words on should.

The feedback loop from the servos does not normally go back to the CNC controller. It is normally sent to the servo drive which errors out when a following error occurs. The CNC controller is not normally fast enough to keep track of say 4 or 6 servo motor encoders because they are very high resolution (I only have 1 servo drive, so my depth of knowledge is limited and it is presently installed as a lathe spindle and it's encoder has 1280000 pulses/rev)... Hence they use gearing to relate input pulses or position to the actual encoder position.... They can provide a simulated encoder output if you require it but this is just normally a visual indication of position and not used as a feedback given the servo drive has much higher resolution


Better quality stepper motor drivers will error out with an overcurrent / stall condition

[robertspark]

Heavy gantries are a give and take, I do mostly 1/4 to 1/2" materials so a speed cutting demon is not a must . My theory is determine the maximum cutting speed on your machine and adjust amperage / consumables around it. The rapid traverse will save travel time on the material loading process of a 6x12 table.

Well thought out design ... Function first .... Doing 600ipm + is not always a benefit.... If it does the job and earns the money / spits out the parts

You could always change the consumables and lower the amperage for thinner materials to use on a slower table

[ScottRa]

Thanks Rod and Robert,

The build I’m doing is a Bulltear (Starlabsj kit (long overdue build) 3x3 80/20 with rail and rack bolted to the side. With the z axis I’d guess about 80+ lbs. The arms are 3/4” aluminum with steel slides, maybe 15 lbs each. Steppers are 620’s is from canandcnc. Pretty stout build so maybe I shouldn’t worry too much. I expect the weakest point would be the connection between the 80/20 and the arms. I’m probably just a bit paranoid when putting all the work into it.

[robertspark]

Thanks Rod and Robert,

The build I’m doing is a Bulltear (Starlabsj kit (long overdue build) 3x3 80/20 with rail and rack bolted to the side. With the z axis I’d guess about 80+ lbs. The arms are 3/4” aluminum with steel slides, maybe 15 lbs each. Steppers are 620’s is from canandcnc. Pretty stout build so maybe I shouldn’t worry too much. I expect the weakest point would be the connection between the 80/20 and the arms. I’m probably just a bit paranoid when putting all the work into it.

Nice table, I'm sure it will be fine as its well proven

motor info here
viewtopic.php?t=22452#p135329

and more likely these
https://www.automationtechnologiesinc.c ... 570-oz-in/

[beefy]

Better quality stepper motor drivers will error out with an overcurrent / stall condition

Rob,

how does that work ?

A stepper drive normally supplies a constant current.
The drive only sends out the step pulses (at the set current) to the motor coils and has no idea if the motor has moved or not. There is no feedback unless it's a closed loop stepper system.

Servos on the other hand have their current adjusted based on position error.

Just checking if you know something I don't.

[Rodw]

Seeing I had the manual open...
LAM DS series drives have a FAULT output you can send back to the controller and an LED on the driver to report the fault. These are the errors

Screenshot 2020-06-06 at 5.57.09 AM - Edited.png

[robertspark]

Better quality stepper motor drivers will error out with an overcurrent / stall condition

Rob,

how does that work ?

A stepper drive normally supplies a constant current.
The drive only sends out the step pulses (at the set current) to the motor coils and has no idea if the motor has moved or not. There is no feedback unless it's a closed loop stepper system.

Servos on the other hand have their current adjusted based on position error.

Just checking if you know something I don't.

I only know my drives and what the manual says....

http://www.leadshine.com/UploadFile/Down/AM882m.pdf

PDF page 18/20

I presume it will detect a short circuit for the overcurrent condition, given it actually lists the coil resistance in the EEPROM parameters you can view.
It does detects stall.... as that I've done.

Stall can be detected through current flow and voltage measurement, given if you are pulsing very fast and the voltage is very low you have a problem as the motor has no inductance (but I've never designed such a circuit so can only surmise)

I've never fed them overvoltage to trigger overvoltage protection.

And it does detect phase errors when changing between stepper motors (I had a wiring error! not hot swapping motors ! that will kill them) and driving a drive with step and direct signals without a motor connected.

The same functions are offered on a number of their other drives.

I know some have had problems with leadshire drives, I am not one of them but I've only ever bought a specific model AM882H (AC/DC fed), and I have 9 of them now without issue.

These would be their newer drives with the same features (but I'm sure upgraded somehow for better performance)
http://www.leadshine.com/productdetail. ... del=em882s

Provided with
Protection: Over current; over voltage; motor cable error; stall detection

[beefy]

Hi Rob,

that's interesting, never heard of stepper drives being able to do that. Guess you learn something new every day.

I was one of the "lucky" ones with Leadshine drives, but maybe they've developed somewhat since then. The LS drive I bought could barely turn my steppers more than a few rpm before they'd squeal and stop, and that was with no load.
Tried a Gecko G203V and it ran my cheap steppers beautifully, cool, and fast. Had them on the table now for quite a few years and still going good.

Some years ago, if I remember correctly, I heard a rumor that LS was copying the drive circuitry from Gecko drives. Who's to know.

[tcaudle]

The simple physics formula for acceleration is :

A = F/M
Acceleration = force / mass

Where force is rotary torque converted to linear force (i.e. ft-lbs ) and Mass is the gantry total weight in lbs divided by 32 (slugs)

This is a number that ignores friction , and other loses in efficiency. No system is 100% efficient .

Though (lots) of empirical testing over years we found the overall efficiency to be between 30 and 40 % . The actual number you get needs to be multiplied by .33 to get the real number. Its based on testing multiple setups with a varying increase in acceleration settings until lost steps start to appear. As you can see the only way to get more acceleration is to either increase the force (bigger motor, higher ration transmission) OR reduce the mass (weight) of the load.

How do you profile that for steppers? You consider the torque rating of the motor (in oz-in converted to ft-lbs) at a normal CUTTING speed . Ratings are in Holding Torque not not rotating toque. Since steeper torque curves drop off at about a linear rate up to roughly the mid point of max RPM, you can use that number to plug in as the acceleration number. So if a stepper motor can do 1000 RPM with the voltage applied it drops off in a linear fashion to abut 500 RPM. After that its no longer linear and is much faster up to the stall speed of the motor.

Since linear force is a function of the mechanical transmission the ratio determines the final number . A direct drive motor to rack loses torque based on the radius of the final drive gear The speed increases by a factor of the final drive gear circumference (Pi times diameter) Its always on a stepper a trade off of the speeds versus the final linear force . since you always trade speed for torque and resolution you need a ratio that works as the best compromise. Turns out its a ratio that gives you one inch of linear motion for one rotation of the motor. So with a 1" final pinion it works out to a 3:1 reduction between the motor and the pinion shaft. The rate of torque decrease best matches the torque increase from the step down fatio.

The target for acceleration for older MACH based systems is in the 20 to 35 ips/sec acceleration at cutting speeds. The reason to have high acceleration is due to the fact plasma cuts differently at different speeds so keeping that speed and having short acceleration (and just as important short deceleration times) is critical to prevent torch dive in corners and sharp turns. MACH used CV (constant velocity) that factored in keeping velocity (cut speeds) as high as possible and letting the toolpath vary to keep it from losing position . Consider a car on a race course where the important factor is speed so it lets the car move outside a strict path . If the turn is sharp it lets the torch round the corner or in some cases take the turn at a lesser angle. So the term "tracking" and accuracy come into play.

If you optimize for accuracy you have to let the control software change the feedrate (slow down as you approach a curve or corner ) at a rate that keeps the path to the tolerance . Its that approach that some control software uses where tracking tolerances are tight. In routing, slowdown is not a big issue but it is in plasma. It results in torch dive and the faster your THC can react the worse it becomes. There are several methods of compensation. The voltage based anti-dive that senses the rapidly rising voltage and stops down motion works on slower response systems but as the Z speeds and acceleration go up for thinner material and non-level cuts it reacts too slow to be of use. In short the Z response is faster than the antidive

Velocity based anti-dive is the answer that controls the Z moves using the real time speed of the XY plane. That requires that eveything happens in REAL TIME. there can be no delay between the sensed feedrate changes and the lockout of the Z THC response. So the THC feedback has to be in sync with the motion Any Windows based control software has problems because motion is not real time unless the hardware (pulse engine) can sync with the THC enable /disable .

You can brute force it and use cut rules in SheetCAM to turn off the THC as it approaches corners and angles but that can make detailed cuts cut with the THC locked off entirely .

One thing LINUXCNC offers is Real time Control meaning there is minimal delay between when the code says move and the table actually does it and and any input feedback is in sync with the motion. LINUX has less overhead and handles multitasking better than Windows at any speed.

A well designed Velocity Anti Dive with LINUXCNC makes a very effective cutting machine especially for detailed cuts and more accurate toolpath. In LINUXCNC toolpath tracking tolerance can be set with a G-code and it will "drive" the toolpath and give preference of accuracy over speed

Candcnc switched to LINUXCNC in early 2016 for this and other important factors that come with open source OS and open source control software. The fact the OS is free and that it is actively being developed and does not phone home to Bill all of the time is a plus.

Now: Stepper VS Servo has been a "religious " argument for many years . With modern stepper drivers and motors the stark differences start to diminish. Biggest thing is the torque curves where the stepper drops with RPM while the servo remains constant across the full RPM range. In short steppers are high torque low RPM motors that when operated inside their torque curves do not lose steps. Servos are high speed low torque motors (using motor size as the gauge) that have positional feedback and can increase torque up to the max when needed. Servos need different transmission matching to gain the torque back they lack at cutting speeds. The argument that servos are more accurate comes from the resolution gain out of the transmission ratio . They can still lose position but the controller should shut down all motion if it exceeds a certain amount. If your mechanics cannot hold the tolerance having 5 rimes higher resolution does little good . A stepper system with microstepping can hold better accuracy than the mechanics it drives. Running down square tube or angle iron with bearings is far more inaccurate than the motor resolution.
with either motor RPM comes form voltage and torque comes from current (AMPS). Stepper drivers are current limited per pulse that makes them torque limited . because torque drops with RPM just increasing the current does some to help marginal lost steps but not nearly as much as slowing the RPM down moving it back up the torque curve. Sounds easy until you throw in two or more axis moving in corroboration to move along a path so slowing one and not the others won't work. The trajectory planner in the software controls the speed of each axis and changing one externally will result in extreme toolpath deviation. Bottom line if a stepper starts to lose steps (like a car in a skid) just adding in more steps does nothing. In theory increasing current could help but if the motor is rated for higher current why not just run it at that higher current to make it less of a problem with lost steps to begin with? Never use peak numbers to predict RMS (average ) performance .

[Rodw]

Our overriding design premise was to harness the high low down torque of steppers to maximise acceleration. And yes, gearing ratio and pinion diameter are two different tuning parameters, both of which contribute to energy loss. I agree that you need to account for energy losses in the design model and also found that 3:1 reduction hit the sweet spot but there is a range. eg. for example we worked with a 30mm pinion cos that was what I had. But I have also got good results from a 5:1 reduction.

The maths around stepper motors is well defined (the first ones came out in the 1910's) but there is a lot of incorrect information out there in so called authoritative white papers. Its not enough to stop at an estimate of efficiency at the pinion, your design model needs to take that down to an estimate of the amps used and the minimum required voltage for a given motor. I agree you need to design for your cutting speed. But at the same time you need to refer to your torque charts to make sure there is enough left in the tank to get up to the desired rapid speed.

Its good to see that others see the advantages of Linuxcnc for plasma controllers. One of its biggest advantages is that you can extend its functionality with custom written components which once installed are treated as if they are part of the Linuxcnc core and are executed in real time 1000 times per second.. This makes it trivial to build a Velocity antidive component. The first one I wrote back in 2017 was all of 20 lines of C code. However, one difficulty is that Linuxcnc has no memory of the feed rate set in gcode when running and only knows the commanded velocity set by the trajectory planner once it accounts for the allowable G64 position errors, the laws of physics and the constraints the machine applies on velocity and acceleration. BUt this has changed recently and I am working on extending some of the newly developed features of Linuxcnc to provide additional features that plasma cutter integrators could use. Imagine a plasma cutter controller that knows when it is cutting a hole and take automatic steps to improve hole quality. What if the controller knew the material thickness and referenced that as a control parameter when cutting holes?

As far as acceleration goes, 35 ip/sec^2 (900 mm/sec^2 )today is on the low side. We are planning for 0.5 m/s^2 (0.5g) at the low end and note that some commercial machines achieve up to 0.7g.

With a solid model of motor performance, the model will predict the maximum required amps (peak and RMS) and even go down as far as predicting power supply specifications. Despite profiling hundreds of motors against out model, I still don't have a clear idea about what constitutes a good motor. They all need to be profiled in our model. Rotational inertia definitely is a factor and is part of the rason why few NEMA 34 motore are a gooe

But what I do know that say a 2 amp stepper motor might only need 1.2 amps to achieve predicted performance. And while you have a 72 volt power supply, it might only need 55 volts. Plus when you consult the torque charts for the motor, in many cases at higher revs, the torque output of the motor flattens so that tits identical to what can be achieved at lower revs. So under these conditions, there is no need to run higher amps at 100% duty cycle. That just wastres power and make the motors run hit,

So with careful design, you can get value out of a boost component.

Of course this is all well researched theory right now until I get the last of my upgraded motors on site and installed. But right now the boring end of financial year chores are distracting me from this for another week or so.

[acourtjester]

I have not be a fan of the dual motor table set up, all the table I have built had one motor and the shaft to transfer drive to the other side of the table. I have built both 4X4 and 4X8 table used to plasma and or router, never slipped out of square. Image shows my setup with another 4X8 table next to it

4x4 table.JPG

[robertspark]

I have not be a fan of the dual motor table set up....

Tom,

I've been wondering for a while now, how do you get on with the use of chains and sprockets and edge finish.

never though that chains were very cnc compatible, do you just have a lot of pretension on it to compensate for the natural play between the rollers on the chain and the links + sprockets?

[acourtjester]

I use a number 35 roller chain to move the X axis, I support the chain for the full length of the X rail, and is acts like a rack with no up-down movements. This image shows the support under the chain, each side is adjustable to align the squareness of the Y rail to the X rails. The other image shows sign vinyl cutting with fine detail when attaching a drag knife to my plasma table.

DSCN1646.JPG

DSCN3026.JPG

[Rodw]

I looked at a similar chain driven setup when I first started but in the end moved on because of the number of problems reported on forums. Just google "roller chain backlash" and you'll see what I mean.

I was also advised about problems with getting a gantry out of square with dual motors. This may be an issue where motors are slaved but with Linuxcnc, each side is under independent control. Linuxcnc squares the gantry every time it homes. When you initially set up your machine, there are homing offsets in the .ini file that allow you to fine tune squareness to account for different positions between the two home switches.

I've only ever done that once and even if the gantry is pushed out of alignment, it will right itself on homing.

[robertspark]

The other image shows sign vinyl cutting with fine detail when attaching a drag knife to my plasma table.

thanks Tom, no you can't fault that vinyl detail cut. impressed, I was like Rod and thought backlash would be an issue (rollers on the link pins and rollers to sprockets )..... shows how the internet of gossip and rumour can get it wrong.

I have individual Y and Y' motors, as far as I am aware quite a few CNC control applications allow for squaring of the gantry .... providing your Y and Y' home / limit switches are on different input pins and not all sharing the same pin ( that is the case with uccnc and mach3, and I presume mach4)..... however... with mach3 and 4 the squaring of the gantry is not a function of mach3/4 but a function that must be programmed into the motion controller firmware (such as an ESS / pokeys / etc....)..... and like always some motion controller firmwares do not have the code built into them as it is a probing application that runs the squaring of the gantry.... and just like probing (skip signal) it is a function that must be part of the motion controller firmware because of the latency of receiving the signal and doing something with it to record the probe position the instant it arrives and stop motion and latency is too long for it to be done in mach3/4 via the plugin.

it is quite common for 3d printers to share the same driver to feed two z-axis motors (if they have 2 motors).... this is obviously not recommended as the motors won't be generating the right torque due to recieving less current... less of an issue with a 3d printer where the z axis only rises 0.1, 0.2 up to maybe 0.8mm (1.2mm nozzle) at most per layer.

all learning and fun

[acourtjester]

For the backlash I added the support under the chains and have not had any problems. I have built 9 table and sold 6, I also sold plans and parts to a large number of people, inside the US and many countries (7 in the UK) outside the US. This system works just fine I have no complains.

[Rodw]

I have individual Y and Y' motors, as far as I am aware quite a few CNC control applications allow for squaring of the gantry .... providing your Y and Y' home / limit switches are on different input pins and not all sharing the same pin

Linuxcnc has this built into its homing sequence and its quite cool to watch if you slow the homing velocities right down. The first side to the home switch stops and waits for the other side to catch up. Then both sides move off the limit switch and approach again at a slower speed. Then the final position is set by a home offset setting for that side .ini file. So then when setting up, if the gantry is not square, you simply just need to change the offsets in the software to account for the difference. THere are generally no restrictions on input pins with linuxcnc if you buy the right hardware. My 7i76e has 32 inputs and 16 outputs so there is not really an excuse not to run separate home and maximum and minimum limit switches on every joint.

[rehoward]

First I would like to say that I am a newby to plasma cutting. But I did manage a robotic installation for Boeing and have an amatureish bacgkound in servos and steppers. At that time servos were the only way to go if you wanted high torque and quick response. My thoughts are that at one time steppers were a lot cheaper than servos. Has that price gap narrowed by now? And later steppers came out wth rotary encoders so no glass scales or other similar type encoders. And as an alternative, rotary encoders can be retrofit to servo motors. So it seems to me that for the budget minded retrofitting used servos with a rotary encoder or buying new steppers with an encoder already fit might be an economical way to do this. My comments are geared more toward the hobbyist than a machine manufacturer.

Randy

[Rodw]

I think if I was retrofitting a machine with servos, I would stay with servos and try to use the factory motors and drives. Servos are definitely easier to design for. Because Plasma has no cutting forces, the engineering is very predictable. We found the engineering is much more complex for Steppers and while they have been around for about 100 years, the engineering was forgotten. Steppers are much easier for a DIY guy or a fab shop without the engineering and electronics skills to tune servos.

I've still got a bit of work to do but so far I've proved the point that Steppers are up to the task with a properly engineered design matched to the forces. eg. I have a 2 amp NEMA 24 motor on my table accelerating at 8m/s^2 (or 0.8G). It gets to its rapid speed of 36 m/min (1420 in/min) on 75 milliseconds which is equivalent to 0-100 km/hr in 3.47 seconds which puts it up in the Tesla Ludicrous mode range. The next challenge is to make sure the table can handle the amazing inertial forces!

I personally don't think encoders add much value to steppers becasue they will only be triggered if the design limits are exceeded and a good designer will make sure that never happens and keep the steppers in their happy zone!