| Figure 1: Simple BIM-UHF to COM
Port Interface Data Packet Format |
|
| Preamble This is essential to settle the data slicer in BIM-UHF at the correct slicing point by sending a period of preamble prior to the actual data, 3 ms is the minimum requirement, we recommend that 5 ms of preamble is optimum. Either 55h or AAh can be used as preamble. But 55h gives better balancing compared to AAh. |
|
|
|
| UART SYNC Byte (FFh) This is used to stabilize the UART, so it can detect a clean start bit after the numerous confusing transitions in the preamble. UART always waits for 1->0 transition then start reading the next 8-bits, (9 bits if parity is used) and stop bit. If it started on the wrong transition, it will not read the correct data, parity and stop bits. This will result in parity and framing errors. When the receiver is turned on it may not land in the proper start bit of the 55h. So, when errors occur the software forces it to ignore it. By putting an FFh, the UART will be made to think there is a break after the stop bit of the last byte. So, it waits for the next 1->0 transition which will be the actual start bit of the next byte. |
|
Start Byte This is used to identify the start of the data and to start the software to decode the next incoming data stream. When in Receive Mode, the software looks for a valid 01h. Sometimes, noise signal or interference from other sources could contain a wave form equivalent to 01h and trigger the software decoding. To prevent this an additional 7Fh is used as a secondary check. These two bytes are purposely made almost complementary to each other so that they maintain the data slicer in the middle and maintain the mark to space balancing. |
|
Data Length Byte The least significant 7 bits is used to identify the number of data bytes or Manchester encoded words in the packet. This value is used by the software to read in the next number of bytes/words into an memory array. MSB can be used to identify whether the data is encoded or not. |
|
Data Bytes or Manchester Encoded Data Words A maximum of 128 data bytes or data words are sent one after another from the transmitter to receiver. If it is not Manchester encoded, the data bytes will be in exact form as original byte. But on transmission, LSB is transmitted first and MSB last. Some data streams may contain long string of 00h or FFh, these are averaged by the slicer capacitor to either Low or High level. i.e. the slicer will not slice symmetrically. Because the data transmitted over radio is ac coupled, constant 00h or FFh behave like dc voltage and these will be attenuated. This could be clearly be observed on the Audio (pin 13) output (Test Point). Because there is always a transition in the start/stop bit, these transition may be missed by the comparator after a long Low or High. Comparator has an input offset which have to be overcome. Manchester or Bi-phase coding avoids this problem by forcing symmetry (50:50 mark:space) in the data stream. Data can also be encoded into 12-bit code like the RPC code. Each 8-bit byte is mapped into a 12-bit RPC code (RPC Data Sheet). |
|
16-bit Cyclic Redundancy Code (CRC) This is used to check for any errors in the data packet. 16-bit (2 bytes) CRC is generated from the bytes in the data packets and appended to the end of the packet. On the receiver, another 16-bit (2 bytes) CRC is generated from the bytes in the received packet. If both CRC word are same, then the packet received is assumed to be correct. If Parity is used, CRC acts as a Secondary Error Checking (or Double Check). |
|
Test Packets These are purposely made to test the BiM under unbalanced, worst data conditions. These data packets have continuous Low bytes (00h), continuous High bytes (FFh) and a few transition of the following types: |
|
 Figure: 2 Worst possible transition |
|
Packet Error Rate with Mark to Space The balance of the code is the ratio of 1's to 0's in a given period. (I.e. there should approximately equal number of bits which are 1's & 0's over 3ms time period for BIM). To test the effect of balancing we sent Test Packet 2 (please refer to UART_send_test_packet2 function). Number of 1's in the byte (8 bits) was varied from 0 to 8. The graphs below show how Packet Error Rate (PER) varies with Baud Rate and Mark to Space. The graph clearly shows that better performance is obtained with 50:50 mark to space. It also shows that there is a small imbalance in the performance between two extremes of mark to space. (I.e. between 0/8 for 00h and 8/8 for FFh). This is due to inherent property of internal components in the BIM |
Fig.3 |
Fig.4 |
|||
Fig.5 |
Fig.6 |
|||
Fig.7 |
-80dBm is a strong signal -104.2dBm is a weak signal near the threshold
of sensitivity of the BIM receiver unit. |
|||
|
||||
| Figure 3 clearly shows that Unbalanced Data will not pass
through radio at all at 38400bps however stronger the signal is. Figure 4 shows that unbalanced data will not reliably pass through radio at 19200bps, if long strings of FFh are present |
||||
Percentage Error with Parity Fig.8 |
||||
|
||||
| It clearly shows that worst error performance is obtained
for Even Parity. This is due to the sudden transition to 0, just before the stop bit. After a long logic High (e.g. 111111112) the capacitor in the slicer circuit is biased towards a Logic High, however, Even Parity for this will be 0 (Logic Low) while Stop Bit is Logic High. Therefore, this sudden logic transition of the parity will be missed at high speed (high baud rates). The graph proves that for reliable transmission through BIM, Odd parity should be used. |
||||
|
||||
| Simple Protocol It is better to implement a protocol to send back an acknowledgment to transmitting DCE. If no acknowledgment is received by the transmitting DCE, it can be made to retransmit the data packet again. This process can be repeated 20 times until an acknowledgment is received. If no acknowledgment is received after 20 tries, the transmitting modem can abandon the process and give an error message to the host DTE. Sometimes the receiving DCE may have sent back an acknowledgment and this may not have arrived at the transmitting DCE properly. Therefore, transmitting DCE will assume that its data was not received properly and it will try to retransmit the same message again. Receiving DCE will then receive a repeated packet which could be mistaken for another packet. This problem can be avoided by putting a counter or some sort of identifier to indicate that it is a repeated packet. Therefore, the receiver will know that it is a repeated message and it can ignore it if it has already received it correctly. It can also retransmit the acknowledgment The second start byte can be used for this purpose. |
||||
|
||||
| Conclusion Note: All these results were obtained for a set of unbalanced, worst data packet conditions. BIM will perform better than these results for ordinary data packets. However it is very important to balance every data packet for reliable data communications over radio. Three important conclusions from the experiment are |
||||
| 1. Data packet should be balanced to have a mark to space
of 50:50 over a 3ms period. Unbalanced data cannot be sent at 38400bps however stronger the signal is. 2. Optimum Baud Rate for RS232 transmission through BIM is 9600-19200bps. 3. If Parity is to be used, Odd parity should be used. |
||||
|
||||
| References | ||||
| [1] Radiometrix BIM-UHF Data Sheet [2] Herbert Schildt, C: the Complete Reference, Third Edition, Osborne McGraw-Hill [3] National Semiconductors, PC16550D Universal Asynchronous Receiver/Transmitter with FIFOs [4] Maxim Interface Products http://www.maxim-ic.com |
||||
 |
||||
| c | ||||
