# Machine controller - design suggestions

##### New Member
Hi,

I want to design a machine controller. This controller shall be used in an industrial environment. The controller consist out of one central control unit and one or more interface units. The power supply comes from a 400VAC to 24AC transformer. The communication between each unit is done over CAN Bus. Machines can communicate with each other as well via an additional CAN Bus. The separation is mainly done because of security reasons.

The central control unit is located inside a switch box together with a contractor or a contractor in star delta connection and a motor protection. The user can start the machine, make settings or can be notified of faults via buttons and an LCD display.

A CAN Bus goes from here to each interface unit. In addition the central unit has one relay to signal faults and one galvanic input contact for remote control.

The interface units get the power supplied from the central control unit. The central control unit and interface units are linked together with a 4 wire cable for CAN bus signal and for power supply.

An interface unit can have analog and digital inputs and outputs. The data is processed by a microcontroller and fetched by the central control unit.

Please take a look at my concept drawing.

Due to the EMC environment, I thought of galvanically isolating the supply voltage and the CAN bus to the interface modules, since interference can couple in via the supply line and I also want to avoid ground loops.

During my research on the internet I found at miscellaneous IC manufacturers applications notes for PLC modules. There the connection of analog and digital inputs and outputs to the micro controller is often galvanically isolated.

Now to my questions:

Galvanic isolation
Is an expensive galvanic isolation of the inputs and outputs to the microcontroller necessary? The analog and digital inputs and outputs must of course be protected from transients. The same question arises for the central control unit.

Principally I would use a galvanic isolation if I need to connect circuits with different power supplies . Are there other reasons when a galvanic isolation should be used?

Power supply
What's the better approach? An AC/DC converter on the control unit that supplies the control unit and all interface modules with a DC voltage. Whereby the interface modules need DC/DC converters to realise the galvanic isolation. Or individual AC/DC converters for each module.

The advantage of AC/DC converters would be that the number of modules could easily be increased by changing and increasing the transformer. Especially as I would have less waste heat at the control unit, because I don't need such a powerful AC/DC converter.

Many thanks for the help.

Best regards
Paul

#### dknguyen

##### Well-Known Member
Hi,

I want to design a machine controller. This controller shall be used in an industrial environment. The controller consist out of one central control unit and one or more interface units. The power supply comes from a 400VAC to 24AC transformer. The communication between each unit is done over CAN Bus. Machines can communicate with each other as well via an additional CAN Bus. The separation is mainly done because of security reasons.

The central control unit is located inside a switch box together with a contractor or a contractor in star delta connection and a motor protection. The user can start the machine, make settings or can be notified of faults via buttons and an LCD display.

A CAN Bus goes from here to each interface unit. In addition the central unit has one relay to signal faults and one galvanic input contact for remote control.

The interface units get the power supplied from the central control unit. The central control unit and interface units are linked together with a 4 wire cable for CAN bus signal and for power supply.

An interface unit can have analog and digital inputs and outputs. The data is processed by a microcontroller and fetched by the central control unit.

Please take a look at my concept drawing.

View attachment 115324

Due to the EMC environment, I thought of galvanically isolating the supply voltage and the CAN bus to the interface modules, since interference can couple in via the supply line and I also want to avoid ground loops.

During my research on the internet I found at miscellaneous IC manufacturers applications notes for PLC modules. There the connection of analog and digital inputs and outputs to the micro controller is often galvanically isolated.

Now to my questions:

Galvanic isolation
Is an expensive galvanic isolation of the inputs and outputs to the microcontroller necessary? The analog and digital inputs and outputs must of course be protected from transients. The same question arises for the central control unit.

Principally I would use a galvanic isolation if I need to connect circuits with different power supplies . Are there other reasons when a galvanic isolation should be used?

Power supply
What's the better approach? An AC/DC converter on the control unit that supplies the control unit and all interface modules with a DC voltage. Whereby the interface modules need DC/DC converters to realise the galvanic isolation. Or individual AC/DC converters for each module.

The advantage of AC/DC converters would be that the number of modules could easily be increased by changing and increasing the transformer. Especially as I would have less waste heat at the control unit, because I don't need such a powerful AC/DC converter.

Many thanks for the help.

Best regards
Paul
What is a module? Are you referring to an interface unit?

I think that they just always have galvanic isolation because even though it may be expensive, it is not much given the cost of the machinery that is being controlled and that if it does end up being required and causes a problem because it's not isolated, then there is the huge cost cost of downtime and lost production due to troubleshooting to actually know that it is required (which might be an intermittent problem), repair, and additional installation.

Not to mention that the system might work fine at the time it was installed, but then later on a giant motor is installed nearby which ends up causing problems because the control system is not isolated.

I guess you could always have isolation modules that could be attached to only the I/O that require it, but each module and it's packaging and connectors will probably just end up being more expensive than just including it to begin with.

##### New Member
What is a module? Are you referring to an interface unit?
This can be an interface unit as well as the control unit since both have inputs and outputs.
I think that they just always have galvanic isolation because even though it may be expensive, it is not much given the cost of the machinery that is being controlled and that if it does end up being required and causes a problem because it's not isolated, then there is the huge cost cost of downtime and lost production due to troubleshooting to actually know that it is required (which might be an intermittent problem), repair, and additional installation.
I was hoping to get a more technical answer when exactly to us galvanic isolation. A PLC control must be suitable for the most different applications. It is only reasonable to protect the control against all possible contingencies. But when I take the trouble to design an application-specific controller, it's different.

#### KeepItSimpleStupid

##### Well-Known Member
Galvanic isolation
Is an expensive galvanic isolation of the inputs and outputs to the microcontroller necessary? The analog and digital inputs and outputs must of course be protected from transients. The same question arises for the central control unit.
The need for isolation, I think, is basically inadvertant connections to mains or other detrimental signals that would damage the entire system and it's so much easier to connect to isolated systems.

4-20 mA or 0-20 mA is generally not isolated, but it helps eliminate ground loops.

In a research environment, the instruments were 0-10 and 0-5 V. USB really hadn't caught on yet. The A/D convertes were mainly pseudo-differential meaning they had a connection to ground and would create a ground loop, so the ground to + and ground to - voltages were measured.

If the interfaced gizmo was controlled by an analog voltage, it's best to control via 0-20 mA. The 0-5 or 0-10V needed can be converted locally with a resistor.

Only one 0-5V instrument had the ability to move the ground reference.

I tried to utilize a home made display that worked in a manual mode and forgot about the ground loop issue.
There were six mass flow controllers connected to a single 5V and +-12 V supplies. The +5 ran the displays and created the setpoints. The setpoints were derived from potentiometers at each MFC, so it worked fine in manual mode. There were lots of power and signal grounds available from the MFC.

Your PT100 sensor has to be at least 3 wires of the same AWG. It can also be 4. It should not be 2. Wire gauge is generally larger.

Against my wishes, I was TOLD TO make the system to use NuBUS modules.

So, the 5V setpoints needed current outputs or needed a differential amplifier for each channel. Management bought another controller. 4 vs 6 channels. 1 display instead of 6. So, I goofed.

Another comment from the US perspective. In order to not use conduit for wiring, the wires have to be power limited to 100 W and generally 24 V or less. See Class 2 wiring.

When I designed a system using strobes, the strobes were PTC limited to 100 W. Soome strobes (same model) didn't work.
it was an inrush current problem that the manufacturer fixed.

#### rjenkinsgb

##### Active Member
You don't really need low voltage isolation at both ends of the interconnection cable.
The main unit can be grounded & run directly from the transformer (or a switch mode PSU).
The interconnect cable can be grounded at that end.

The remote using can just use a small DC-DC converter for its isolated supply, they are smaller and cheaper than low frequency, low voltage isolation transformers - and you do not have to build rectification and smoothing etc.

You can get an isolated CAN transceiver off-the-shelf for that end of the link.

Optically isolating 24V I/Os is fairly simple, just optos at the MCU side of each and power to the isolated sections via customer terminals alongside each bank of inputs or outputs. That's general industry practice with PLC I/O isolation.

The same external supply concept is sometimes used with analog I/O, or you could use additional isolated DC-DC modules in the devices to provide isolated supplies internally.

#### dknguyen

##### Well-Known Member
This can be an interface unit as well as the control unit since both have inputs and outputs.
I was hoping to get a more technical answer when exactly to us galvanic isolation. A PLC control must be suitable for the most different applications. It is only reasonable to protect the control against all possible contingencies. But when I take the trouble to design an application-specific controller, it's different.
The thing about industrial environments...I found that a lot of the time no one seemed to care about the really technical technical answer because in order to use the technical answer you have to put time and effort (\$) into analyzing or testing the system. So you just use a rule of thumb whenever possible to speed things up which often matters a lot more.

Example: In Canada, we can direct bury Teck armoured cable. No conduit required. So you just specify Teck cable everywhere so that the engineer wouldn't have to waste time (that you're being paid for) sizing and specifying conduit, and so the installers wouldn't have to waste time cutting, bending, and installing conduit. If part of the cable needs to be buried, the entire run is specified as Tek1. We couldn't do this for American projects though and when I asked why, I was told because the conduit fitters' Unions blocked the legislative approval because not needing to size and bend conduit meant they would lose their jobs.

But when I take the trouble to design an application-specific controller, it's different.
Oh, this is application specific? Or do you mean facility specific? Facility specific might makes things a bit different...because then you can design to the equipment layout and cable runs ahead of time. But you'd still have to go through the analysis so you might still be going by rules of thumb, just on a signal by signal basis. There's also still the possibility of future equipment changes mucking up the works.

If it's application specific rather than facility specific then I don't think it would change much since it would still have to be general enough to handle whatever facility it was in.

Last edited:

##### New Member
The need for isolation, I think, is basically inadvertant connections to mains or other detrimental signals that would damage the entire system and it's so much easier to connect to isolated systems.
So when I take greate care about the ground loop then I could get around the extra isolation for the inputs?

You don't really need low voltage isolation at both ends of the interconnection cable.
The main unit can be grounded & run directly from the transformer (or a switch mode PSU).
The interconnect cable can be grounded at that end.
Ok noted
The remote using can just use a small DC-DC converter for its isolated supply, they are smaller and cheaper than low frequency, low voltage isolation transformers - and you do not have to build rectification and smoothing etc.
I just took a quick look. These cost a lot for 20W. My plan is to build 200 pcs. I thought about using a push pull DC/DC converter and do the circuit for it myself.

Optically isolating 24V I/Os is fairly simple, just optos at the MCU side of each and power to the isolated sections via customer terminals alongside each bank of inputs or outputs. That's general industry practice with PLC I/O isolation.
What about the lifetime of optocouplers? I actually wanted to use current-limiting inputs. Principle I have no problem with the glavanic isolation. But galvanic isolation of the analog signals would increase the costs quite a bit.

Oh, this is application specific? Or do you mean facility specific?
Well, specific for one machine. It's gonna be a small series. But basically they are all the same.

#### KeepItSimpleStupid

##### Well-Known Member
Ground loops and isolation are really two different things,

What happens if your mains should suddenly connect to a sensor. With isolation very few things go poof.

Isolation, does generally cure ground loops.

Ground loops are best eliminated by design. They are expensive to eliminate "after the fact".