You are not logged in.
A new reference design has been posted to the same page. The physical layer has been upgraded to support dynamic modulation. In other words, a transmitter can choose the modulation order for the payload of the packet and place that information in the header (which runs at an agreed-upon base-rate). The receiver's PHY can process that header information and decode the payload of the packet automatically. This enables per-packet rate control without handshaking between the transmitter and receiver. To use this new functionality, see line 310 of csmamac.
The default fullrate speed for the reference design has been bumped up to 16-QAM, so there should be an immediate performance boost relative to the last reference design.
Offline
These question are regarding version v04 of ofdm design.
1 a. Is there any queue maintained for packets sent by the ethernet i.e. as the ethernet sends packets to the warpmac board, does the warpmac code maintain a queue for all these packets?
b. What happens when I use the warpmac_sendofdm() function? Is the queue bypassed?
2 ofdm_Tx_mimo.h is found in three places:
To make any changes to this file, do we need to make changes to all the three places or making changes at one place is enough?
The three places are:
a. OFDM_ReferenceDesign_v04\drivers\ofdm_TxRx_mimo_opbw_v1_25_a\src\
b. OFDM_ReferenceDesign_v04\ppc405_0\libsrc\ofdm_TxRx_mimo_opbw_v1_25_a\src\
c. OFDM_ReferenceDesign_v04\ppc405_0\include\
3. Regarding the case of dynamic modulation. We have tried to test various modulation rates. We have got the following results (To test the various modulation schemes, we changed argument of warpmac_setFullRate() function).
[Base rate was set to QPSK. The transmitter and receiver were about 1.5 - 2 meters of each other. A video was being played at the transmitter end. No other node was contending for the medium during the experiment.]
QPSK: The performance is best in this case i.e ratio of bad packets to total packet received is lowest in this case.
In about 26000 packets received at the receivers end, 17000 packets were good.
QAM_16: This has second best performance and has a higher ratio of bad packets to total number of packet received. In about 26000 packets, 1700 were good.
QAM_64: All the packets received were bad.
Are these results expected?
Last edited by toolbox (2007-Jul-05 10:47:26)
Offline
toolbox wrote:
These question are regarding version v04 of ofdm design.
1 a. Is there any queue maintained for packets sent by the ethernet i.e. as the ethernet sends packets to the warpmac board, does the warpmac code maintain a queue for all these packets?
b. What happens when I use the warpmac_sendofdm() function? Is the queue bypassed?
There is no queue in between the Ethernet MAC and the PHY (in at least the receive-from-wireless direction.. more details here), due to the use of DMA. In the Ethernet RX handler, the Ethernet payload is copied to the user's transmit buffer specified with warpmac_setTxBuffer. You could pretty easily modify warpmac and csmamac to allow for an array of transmit buffers. Bear in mind that changes would also have to be made to the code in the Ethernet interrupt handlers, since that code currently only executes when the PHY is ready to send.
toolbox wrote:
2 ofdm_Tx_mimo.h is found in three places:
To make any changes to this file, do we need to make changes to all the three places or making changes at one place is enough?
The three places are:
a. OFDM_ReferenceDesign_v04\drivers\ofdm_TxRx_mimo_opbw_v1_25_a\src\
b. OFDM_ReferenceDesign_v04\ppc405_0\libsrc\ofdm_TxRx_mimo_opbw_v1_25_a\src\
c. OFDM_ReferenceDesign_v04\ppc405_0\include\
As we described in this thread, never modify anything in the ppc405_* folders. After you make changes to anything in /drivers, choose Software->Clean Software in XPS, then update the bitstream.
Offline