WARP Project Forums - Wireless Open-Access Research Platform

I tried to do random numbers (PRNG) putting the current time as the seed, but it always give me the same random number. Is it possible to do random number on the WARP board?

You can call rand() in the C code to access the stdlib PRNG. This isn't a great PRNG but is good enough for MAC purposes. This is how we generate a uniform random value when choosing backoff counts in the DCF.

Also, I need to synchronise the time on the WARP boards through connection to Network Time Protocol. Is this feature implemented in the current 802.11 design? If not, do you think it will take a long time to implement on my own?

The 802.11 ref design does not implement an NTP client. I'm not familiar with the details of NTP, but I think adding an NTP client to the embedded C code in the 802.11 design would be a big project. Can you describe the experiment requirement that you're planning to address with NTP? It's possible you can achieve the same result without the complexity of building a full NTP client.


Keep in mind the reference code does not implement any "real" time tracking. The design implements "System Time" (microseconds since boot) and "MAC Time" (microsecond timer synchronized via beacons).

The wlan_exp framework supports setting the MAC Time via Python.

It also supports sending a command that creates a TIME_INFO log entry at the node. The TIME_INFO entry records the node's MAC Time, System Time and a real time value supplied from Python. This TIME_INFO entry allows you to synchronize log entries with real time when analyzing the log data by using the TIME_INFO entries as "markers" which periodically record the relationship between the various timebases.

Is it possible for the AP in 802.11 Reference Design to sense whether or not the current channel is being used, and return the value to CPU_HIGH? So, if the AP is on channel 1, and some other AP is also transmitting on channel 1, how would I detect if the channel is taken up?

In CPU_LOW, I did this:

u32 mac_hw_status = wlan_mac_get_status();
int channel_free;
if(mac_hw_status & WLAN_MAC_STATUS_MASK_TX_PHY_ACTIVE) {
channel_free = 0; // not free
}
else {
channel_free = 1; // free
}

However, I found that the channel returns 1 (free) all the time.

Regards,
loho2027

Hi,

Does the 802.11 Reference Design support 5.0GHz WiFi? If so, how do I change the channel to a 5.0GHz channel, and what channels are available from the 5.0GHz band?

Regards,
loho2027

I'm trying to detect when a channel is free using carrier sensing. I have the following code:

if(wlan_mac_low_get_check_channel_free_cmd()) {

u32 mac_hw_status;
mac_hw_status = wlan_mac_get_status();
int chan_free;
if(mac_hw_status & WLAN_MAC_STATUS_MASK_TX_PHY_ACTIVE) {
	chan_free=0;
}
else {
	chan_free=1;
}
xil_printf("chan free:%d", chan_free);
wlan_ipc_msg_t ipc_msg_to_high;
ipc_msg_to_high.msg_id= IPC_MBOX_MSG_ID(IPC_MBOX_CHANNEL_CHECKED);
ipc_msg_to_high.num_payload_words = 1;
ipc_msg_to_high.payload_ptr = (u32*)&chan_free;
write_mailbox_msg(&ipc_msg_to_high);
wlan_mac_low_set_check_channel_free_cmd(0);
}

This code is intended to detect the channel's state and return the information to CPU_HIGH, which I require for implementing a certain algorithm.

I have placed this code in wlan_mac_dcf.c, in frame_transmit. However, I'm not sure where to place it to ensure I get the desired result. If I place the code before the below code, it says the channel is free almost all the time (this might be because it says waiting there is actually rare, and so the channel is mostly free there):

// Wait for the Tx PHY to be idle
        // Actually waiting here is rare, but handles corner cases like a background ACK transmission at a low rate
        // overlapping the attempt to start a new packet transmission
        do{
            mac_hw_status = wlan_mac_get_status();
        } while(mac_hw_status & WLAN_MAC_STATUS_MASK_TX_PHY_ACTIVE);

And if I place the code before the big 'do' loop (partially shown below), I get the channel is not free most of the time:

// Wait for the MPDU Tx to finish
        do { // while(tx_status & WLAN_MAC_STATUS_MASK_TX_A_PENDING)

            // Poll the DCF core status register
            mac_hw_status = wlan_mac_get_status();

            // Fill in the timestamp if indicated by the flags, only possible after Tx PHY has started
            if (mac_hw_status & WLAN_MAC_STATUS_MASK_TX_PHY_ACTIVE) {
            	tx_has_started = 1;

            	if(req_timeout){
            		gl_waiting_for_response = 1;
......

This might be because the AP senses its own transmission?

Can you tell me where I should place my code so that it can sense if other APs are transmitting at the same frequency?

Regards,
loho2027

Referring to Post #2, I'd like to synchronise the access points so it performs a function at the same time. Am I right to believe that without using Python this is not possible with the current implementation, since the timing information inside the WARP board are just the microseconds since boot up?

And even with Python providing the time, is that just for the logs, ie. only looking at the data retrospectively?

Regards,
loho2027

Hi,

I have the following code in the frame_transmit function in wlan_mac_dcf.c:

        
if(wlan_mac_low_get_check_channel_free_cmd()) {

			u32 mac_hw_status;
			mac_hw_status = wlan_mac_get_status();
			int chan_free;
			if(mac_hw_status & WLAN_MAC_STATUS_MASK_CCA_BUSY) {
				chan_free=0;
			}
			else {
				chan_free=1;
			}
			xil_printf("chan free:%d", chan_free);
			wlan_ipc_msg_t ipc_msg_to_high;
			ipc_msg_to_high.msg_id= IPC_MBOX_MSG_ID(IPC_MBOX_CHANNEL_CHECKED);
			ipc_msg_to_high.num_payload_words = 1;
			ipc_msg_to_high.payload_ptr = (u32*)&chan_free;
			write_mailbox_msg(&ipc_msg_to_high);
			wlan_mac_low_set_check_channel_free_cmd(0);
		}

I understand that I cannot have real-time channel information using CPU_HIGH, but having one or two seconds latency is fine since I want to determine whether the currently selected frequency channel clashes with another AP's frequency channel. I want the code to detect when it's time to check the channel (using wlan_mac_low_get_check_channel_free_cmd()), checks the channel, then sends a mailbox message to CPU_HIGH to say whether or not the channel is free.

The algorithm I want to implement works like this: it starts off with random frequency channels assigned to each AP, then if it detects a clash of frequency channels, it assigns the channels again with a particular set of probabilities. If there is no clash then the channels should remain the same, and is said to have converged to a solution.

However, I cannot get the channels not to clash, even though in some cases the selection of frequency channels clearly indicate it shouldn't clash. For example, when four WARP boards are in operation, and the frequency channels are 1, 11, 40, 48, then there should be no clash of channels. However, it still detects a clash and switches to another set of frequencies.

I have raised the threshold for carrier sensing in wlan_phy_util.c to a high number (I tried -40dbm) but to no avail. The frequencies still switch. The thing is that it doesn't always sense that the channel is busy. Sometimes it senses the channel is free and so the channels don't change, but they change eventually, even though the channels should be free.

I can think of three reasons:
1) The interference from existing Wi-Fi makes the channels busy. I cannot stop those Wi-Fi networks, but using a Wi-Fi analyser from my laptop the signal strengths appear to be below -50dbm.
2) It is sensing its own transmissions. So whatever channel I choose it will eventually sense itself and switch to another channel according to the algorithm I am using.
3) Related to (2), the code doesn't sense the channel in the way I want it to do. But the channel sensing is used for DCF, so it should work, right? So maybe I have it in the wrong place? But if so, I wonder why it would sometimes detect that the channel is free, and sometimes it detects the channel is busy?

Regards,
loho2027

Thanks. So does that mean I should check the list variable 'bss_info_list' in wlan_mac_bss_info.c to see if it's empty or not? If it's empty then no APs are communicating to it, and if it's non-empty then at least one AP is able to connect to it.

How would I lower the sensitivity so that it does not detect APs that are not in close vicinity to the AP? Do I still use wlan_phy_rx_set_cca_thresh from wlan_phy_util.c, or is it not possible to do it?

Regards,
loho2027

Hi,

Actually I'm not just trying to detect if the APs are on the same channel. If, for example, an AP is on channel 1, and another AP is on channel 2, then there could be interference if the APs are close enough to each other. I want to be able to detect this interference as well.

I'm guessing that bss_info_list does not include APs of adjacent channels?

Also, I found that even though another AP is on the same channel as an AP, it doesn't show up on the bss_info_list all the time. Sometimes it would show that AP, sometimes it doesn't.

Regards,
loho2027