The principles

Definitions

First of all, don't mix Internet and Ethernet, it looks close, but reflect different meaning. Internet, everyone knows what does it mean, but ethernet is different. Ethernet (also here a definition) is a hardware layer for networking (i.e. exchanging data) between devices. Internet protocol generaly includes the ethernet physical layer (This is not the case when one use a PPP or Point to Point connections).Ethernet can be used to transport many kind of protocols like TCP/IP, UDP/IP, IP/ICMP, IPX/SPX, etc etc ...We will use this technology to exchange data between a PC/Workstation and a device (this one will always behave as a server (or slave) with respect to the PC), the protocols supported will be IP/UDP for data exchange and IP/ICMP for ping requests to check that our embedded communication board is still alive.

State of art

In good old days, it was very usual to use the PC's parallel printer port as a physical layer to transfer data from a PC to a CCD camera (for instance). The PC was sending commands through the parallel printer port interface to a microcontroller, or even worse was driving directly the clocks of a CCD camera or a motor stepper. The microcontroller executed the command, and send back the data to the PC. This concept was straightforward to settle, the software was also quickly-dirty done (no classes, no driver) under DOS or Win95/98.

The main drawbacks of this concept are :

• The short length between the PC and the CCD camera
• The data are coming to the PC very slowing (even though with the use of EPP/ECP modes)
• Only one device can be attached to parallel port (LPT1), the use of LPT2 might overcome a little to the problem ;-)
• Software is difficult to make for Win NT/2000 Operating systems, because it needs the access to low level resource.
• High sensitivity to electric disturbances (a parallel bus running on the floor !!).
  

Old scheme for controlling devices such as a CCD camera by a PC. This system is employed by almost all the CCD amateur CCD camera manufacturers.

A new concept

So, let's give up this concept and let's move on to a new concept : the LAN ethernet link. For that purpose, we add a new board (called here "Our asynchronous communication ethernet board 10BaseT).This board talks to the PC using ethernet IEEE802.3 as a physical layer and UDP/IP protocols. On the other side, the ethernet board talks to the CCD camera microcontroller through a 8 bit parallel bus that can run up to 50Mhz.

New scheme for controlling devices such as a CCD camera by a PC, the bidirectionnal 8 bit bus wide is very close to the microcontroller.

The advantages of this concept are the following :

• The DATA can be transferred in a full duplex mode up to 1Mo/Sec leading to a transfer (including overheads) of 600Ko/sec.
• The length of the RJ45 can be up to 200m, extended to 500m with ethernet hubs and kilometer if fiber transceivers are used.
• Software is very easy to make, lots library with classes are existing concerning IP/UDP protocols for Win NT/2000/LINUX Operating systems,and also because it doesn't need to have ANY access to PC's low level resource.
• UNIX Workstation can be used, no printer parallel port is required any more.
• Low sensitivity to electric disturbances (differential duplex signal).
• Many computers and ethernet measuring/imaging devices can be tied up together. The next drawing shows that an infinite number of computers can be attached to an ethernet hub (50DM for 8 ports) and also infinite slave ethernet devices can be either. Each ethernet device has its own IP address.
  

A nice way to chain many king of devices with an ethernet network, the PC computers behave all as clients, the devices (CCD camera, Telescope drives, Dome and weather station) behave as servers with our embedded communication ethernet board. The distance between all the devices can be very large.

A closer look inside our embedded communication ethernet board

The master of this board is the SX52 scenix microcontroller running at 50Mhz, this one controls the CS8900 ethernet controller, the 2K SRAM and the two FIFOs. How does it work ? An ethernet frame is received by the CS8900, an interrupt is raised to the SX52. The later reads the whole ethernet frame and store it into the 2K SRAM (because SX52 has only 262 bytes of paged memory). The SX52 decodes the frame and stripe out the MAC, IP and UDP destination addresses. If those address match to the SX52's internal set of addresses, it processes the message as an order which is written to the FIFO SX52. The device microcontroller (let's say the on from the CCD camera) polls the FIFOSX52, and if this one is not empty, read the order (such as for instance wipe the CCD n times) and proceed it. Once the order is completed, the device microcontroller sends back an acknowledgment the the FIFOmicrocontroller. The SX52 pools the FIFOmicrocontroller and again sends back the acknowledgment to the CS8900 ethernet controller as an ARP/IP/UDP frame. The later sends finally the ethernet frame to the cable and to the PC computer.

This board responds also to PING (IP/ICMP request).

How does it look like ?

As a 101x68mm 2 layers board :

Top layer of the board, the input is the RJ45 connector is located at right bottom of the image. The dip 12 connector at the middle of the board is the link to the device microcontroller.

Bottom layer of the board, RJ45 connector is located at top right of the image. SX52 microcontroller is visible.

Board running with the SX key KIT used to program the SX52 microcontroller (underneath).