What does USB3 mean for the contol industry?

[The Plc Kid]

Maybe like the attachment?

Here are some similar links

http://www.riaconnect.com/news_detail.asp?news_id=17

http://www.rjfield.com/

http://www.conecspec.com/IP67RJ45/HTML/IP67RJ45 Product Overview.htm

http://www.optimization-world.com/details/prodid/230.html

 

[The Plc Kid]

For plugs i use these but the have a crimper.Still easier to assemble.

http://www.telephoneparts.com/product/EZRJ45/EZ-RJ45+Crimping+System

These are tooless http://cableorganizer.com/outlet/CTG-27581.html

 

[Brian123]

Beckhoff already have a USB Adapter for their IO blocks. I guess the intended usage is with a PC being the controller.
However, the limitations with USB are so significant, I can see zero justifaction for it in an industrial environment. Maybe in a lab, maybe not.
No, we dont crimp ethernet cables any more. In an industrial application we use either preconfectioned cables, or use industrialised connectors that can be assembled without crimping tools.

Jesper's mention is the first I've heard of "industrialized connectors". What gives?

I've heard of those. His post was the first time I have heard of preconfectioned cables though. Do those go well with coffee in the morning?

Brian

 

[surferb]

It gets hits in Krone manuals. Really, I have no idea...I'm giving Jesper the benefit of the doubt on that one as well...

I've heard of those. His post was the first time I have heard of preconfectioned cables though. Do those go well with coffee in the morning?

Brian

 

[Brian123]

It gets hits in Krone manuals. Really, I have no idea...I'm giving Jesper the benefit of the doubt on that one as well...

Hmm... Could be. Just ran a Google search on the word. Not very common at 105 hits total. This thread was on page two of the results.

Brian

 

[surferb]

That's what I did too - and wikipedia. But this manual was convincing...

Hmm... Could be. Just ran a Google search on the word. Not very common at 105 hits total. This thread was on page two of the results.

 

[Peter Nachtwey]

Pete - do you have reference material?

Reference manuals on what? The part about the slow 8 bit interfaces? Our 3rd generation products have an USB interface for configuring the controller. USB works well under ideal conditions over short distances. Once the system is configured it is best to use Ethernet. However, USB is much better than serial ports.

1 GB USB will not be that much faster than 2 MB USB unless there is 500 times more processing power available to the USB protocol stack. You guys are talking about wire time and rates. What about processing time and rates?

 

[Brian123]

That's what I did too - and wikipedia. But this manual was convincing...

Wiktionary has a couple of interesting entries.

http://en.wiktionary.org/wiki/confect
http://en.wiktionary.org/wiki/confection

Brian

 

[Brian123]

Reference manuals on what? The part about the slow 8 bit interfaces? Our 3rd generation products have an USB interface for configuring the controller. USB works well under ideal conditions over short distances. Once the system is configured it is best to use Ethernet. However, USB is much better than serial ports.

1 GB USB will not be that much faster than 2 MB USB unless there is 500 times more processing power available to the USB protocol stack. You guys are talking about wire time and rates. What about processing time and rates?

On the face of it, ethernet would be subject to similar restrictions (need more processing power to use the increasing speed.) Is Ethernet better because it has a lighter stack (less complicated/processor consuming), or there are dedicated co-processors, or ? Just curious more than anything.

Brian

 

[Peter Nachtwey]

Is Ethernet better because it has a lighter stack (less complicated/processor consuming),

No

or there are dedicated co-processors, or ? Just curious more than anything.

Yes in our case. New Ethernet chips are often 32 bits and use DMA to directly put packets into memory. The microcontroller we use has a built in Ethernet processor that sorts packets in parallel with the motion processing. The IP,UDP,and TCP part of the Ethernet stack must still be done using the micro controller but at least it isn't wasting clock cycle stuffing or extractint from the wire 8 bits at a time with lots of wait states.

I have the same complaint about Profibus DP. The profibus chips were made for 8 bit micro controllers back in the 80s. Accessing the Profibus DP chip 8 bits at a time with lots of wait states is painful and Profibus DP runs at 12 Mbs. I want 32 bit zero wait state Profibus interface chips or many will lose interest and it will go the way of Modbus Plus.

 

[surferb]

Pete - let's take a step back from motion controllers and refer to PLCs or other devices with embedded Ethernet vs USB 3. Won't the device have a dedicated USB controller/processor. Why would you be in a wait state much at 3 Gb/s, even with an 8 bit controller? I understand DMA, but don't get what differences there would be between an Ethernet and next gen USB controller that would strongly favor the former. I don't see how your comment about the layer 3/4 portion of the TCP/IP stack is even relevant here. According to Intel's UHCI rev 1.1 (2.0 spec), the host controller generates a start of frame (SOF) packet every 1 ms.

It sounds to me like you may be carrying assumptions about disadvantages that legacy USB 1.1 implementations into the new technology.

 

[JesperMP]

Sorry about "preconfectioned". It is a term that seems doesnt have it english counterpart. I meant finished cables.

As to the industrialised connector, we started with a Harting type, but have switched to this Siemens type (6GK1901-1BB10-2AA0). With the proper stripping tool, it takes less 30 seconds to finish one cable+connector.

 

[Peter Nachtwey]

Pete - let's take a step back from motion controllers and refer to PLCs or other devices with embedded Ethernet vs USB 3. Won't the device have a dedicated USB controller/processor.

It may if you want to pay for it, but then how quickly with the PLC CPU be able to read from the dedicated USB CPU?

Why would you be in a wait state much at 3 Gb/s, even with an 8 bit controller?

Wait state have nothing to do with whether the data is moving at 1G or 3GB over the wire. It is about how quickly you can read data from the wire and write data to the wire.

What about the time to access the USB chip? If there is a dediciated USB processor then there must be some hand shaking between the PLC CPU and the USB CPU. This is where the wait states occur. Most 8 bit micro controllers do not access memory or I/O nearly as fast as 32 bit micro controllers so the 32 bit micro controllers must wait for each byte to be transfered.

I agree that the wire time will be very short with 1GB USB but you must also take into account the time it takes to shuffle the data into and out of the USB chips. You have ignored it. Then there is the application layer that interpets the USB data. That all takes time. Even if you shrink the wire time to 0 you still have the data transfer time and the application layer time.

It sounds to me like you may be carrying assumptions about disadvantages that legacy USB 1.1 implementations into the new technology.

It isn't about 1.1 vs 2.0 vs 3.0 technology. I think you are missing the point. USB 1.1 would be relatively fast if the data could be accesses with 0 wait states 32 bits at a time. USB 3 will be slow if one must read or write data to the USB chip 8 bits at time and each access has a lot of wait states.

Show me a USB interface chip with 32 bit and at no wait states. You do know what a wait state is in hardware terms don't you? If not then you need to study wait states and the ready line. These are hardware topics.

 

[surferb]

Jesper - the word is Danish? I like that word - it may slip into my vernacular

Why would you want your Ethernet cable to only have 4 wires?

Sorry about "preconfectioned". It is a term that seems doesnt have it english counterpart. I meant finished cables.

As to the industrialised connector, we started with a Harting type, but have switched to this Siemens type (6GK1901-1BB10-2AA0). With the proper stripping tool, it takes less 30 seconds to finish one cable+connector.

 

[surferb]

OK, I get the point - the basics, anyway. But I don't get any feel for how significant the effect is. It sounds like a millisecond/sub-millisecond wait, which could be significant for high speed/realtime devices like motion controllers and PLCs. Aren't you always going to have a similar delay, or can you get closer to achieving a "straight to the wire" I/O connection with any modern high speed connections?

Wait state have nothing to do with whether the data is moving at 1G or 3GB over the wire. It is about how quickly you can read data from the wire and write data to the wire.

It isn't about 1.1 vs 2.0 vs 3.0 technology. I think you are missing the point.