You are not logged in.
Hi,
In my previous postings, you may notice that I'm still not able to
estimate the channel. One option is to implement physical layer
details in WARPLAB such as frame synchronization, CFO compensation,etc.
The other option that I can think is to use MIMO-OFDM reference design,
and change the core so that one designated receiver keeps sending UDP
packets conveying channel estimation results like H matrices.
Or do you have any good reference design for the precise channel
measurement? I think that the first thing you might have done with this
new testbed should be a campaign of channel measurements :-)
Tae Hyun
Offline
We don't have any reference designs for precise channel measurement. The MIMO OFDM reference design does estimate the channel, but adapting this model for just channel estimation is probably not the best route.
I still suggest using WARPLab for this application. You will have to implement the receive processing to establish synchronization (i.e. lining up training symbol boundaries, estimate and remove carrier frequency offset, etc.). But developing this code in MATLAB will be much faster than extracting and adapting the hardware designs from the MIMO OFDM model. Remember you can prototype all of this first without using WARP hardware. Using MATLAB's comm functions (awgn, rayleighchan, etc.) to simulate channels, adding CFO (just multiply by a complex exponential), you can create arbitrary channel conditions that your receive could should be able to handle. Once that works, then you're ready to run the signals through a real channel with WARPLab.
Offline
Yes, actually I'm planning to do that. But even with the PHy details to estimate the
channel, current WARPLAB core does not give me a control on the sampling rate.
This is important for me to track the channel variations; By my experiments,
one cycle of sending a frame using WARPLAB is roughly 100ms. This means that
I can only track channels which are quite static such as indoor environment.
One more question:
Is using the simulink in MATLAB the only way to implement a custom core for
channel measurements in MIMO-OFDM? Do I have more options?
Tae Hyun
Offline
Remember that each over-the-air burst in WARPLab lasts 409µs. This is a *long* time on the timescales of channel variations at 2.4GHz. 802.11a/g uses a cyclic prefix of 800ns, which corresponds to the maximum delay spread it can tolerate. So with one WARPLab Tx/Rx cycle, you could probe (and track variations in) a worst-case 802.11 channel many hundreds of times.
livekth wrote:
Is using the simulink in MATLAB the only way to implement a custom core for
channel measurements in MIMO-OFDM? Do I have more options?
We chose Sysgen for our PHY designs; this is definitely not the only way to design for the WARP FPGA board. If you want to implement a channel measurement block inside the FPGA, you can use any FPGA design flow that can target a Xilinx FPGA (hand-coded HDL, System Generator, etc.).
Offline