WARP Project Forums - Wireless Open-Access Research Platform

I have been using the hardware design given in warpmac_labs at  http://warp.rice.edu/bigFiles/warpmac_labs.zip.

I am trying to explore ways to improve transmission rate and so would like to try all possible methods- modulation scheme etc.

1. Could you suggest me methods and possible C files by which I could manipulate the transmission rate?

2. Moreover is the RSSI value 1023 a valid number as on the internet, I have found much lower ranges of values of RSSI

[ I am using the following code to print the RSSI value whenever any packet is received i.e. in receiveGoodPacket() and receiveBadPacket():   

     short y = WarpRadio_RSSIData(XPAR_RADIO_CONTROLLER_0_BASEADDR,FIRST_RADIO | SECOND_RADIO);
     printf("RSSI = %d    \n",y);
]

3. I believe that the at the present, the code is such that the radio transmits at a constant rate. Is that correct? Are there any adaptation codes built for WARP which can adjust the transmission rate based on error rate or rate of time-outs (on ACKs)?

The new OFDM PHY with support for per-packet modulation selection is part of the new OFDM Reference Design.

In our reference designs, packets are transmitted whenever they arrive on the Ethernet interface. The wireless MAC will also re-transmit a packet if no ACK is received. You'll have to customize the MAC code to implement whatever packet transmission timing you want. You can use an opb_timer or usleeps to time & trigger events, if needed.

Reading RSSI isn't a good metric, since it's an instantaneous measurement of received energy. If you're reading it before transmitting or after a PHY interrupt (likely, since the PowerPC doesn't know when the medium is busy), it's also a measurement of energy during an idle period. As you found, these readings won't correlate with network throughput.

The PHY (as of reference design v04) provides a register that reports the AGC gain values per-packet for exactly this sort of application. This is our best proxy for average SNR on packet timescales. Use this function to read the gain register in the good packet interrupt handler:

ofdm_TxRx_mimo_ReadReg_Rx_Gains(OFDM_BASEADDR)

This will return a 15-bit number, composed of:

bits[14:0] = [ AntB_RFG[1:0] AntB_BBG[4:0] 0 AntA_RFG[1:0] AntA_BBG[4:0] ]

For SISO systems, you only care about the AntA numbers. You combine the two gain values according to this formula:

TotalGain = 15*(RFG-1) + 2*BBG

The result is roughly inversely proportional (in dB) to a received packet's SNR.

How you use this is an open research question and depends entirely on what questions you're trying to answer. One possible use is a protocol like one we have developed at Rice known as Opportunistic Auto Rate (OAR). In any case, please keep us updated on what you find.

ofdm_TxRx_mimo_ReadReg_Rx_Gains is a macro, defined in the OFDM core's driver (in /drivers/ofdm_TxRx_mimo_opbw_v1_25_a/src/ofdm_TxRx_mimo.h). This function is new in v04 of the OFDM Reference Design.

OK - so masking and bit-shifting is the way to go. Ignore my prior inquiry.

We have been trying to make some measurements.

In our experiments, we have been trying to measure total gain and RSSI with varying distance. What we have found is that the total gain does not vary much and RSSI shows a wavy trend. The distances have varied from 50 inches to around 850 inches and we have measured at every 50 inches. The experiment was conducted both outdoors and indoors.

Could you please share some of your results with us and the details of your experiment?  How have you changed the channel conditions? Have you changes the distance between the sender and receiver? We would like to hear from  you.

Thank you.

First, here is our code:

int uphandle() {
//    warpmac_setTargetPlus();
    usleep(100);
    unsigned int rxgain = ofdm_TxRx_mimo_ReadReg_Rx_Gains(XPAR_OFDM_TXRX_MIMO_OPBW_0_BASEADDR);
    // pull off the SISO Values
    // from the forums
    /*ofdm_TxRx_mimo_ReadReg_Rx_Gains(OFDM_BASEADDR)
    This will return a 15-bit number, composed of:
    bits[14:0] = [ AntB_RFG[1:0] AntB_BBG[4:0] 0 AntA_RFG[1:0] AntA_BBG[4:0] ]
    For SISO systems, you only care about the AntA numbers. You combine the two gain values according to this formula:
    TotalGain = 15*(RFG-1) + 2*BBG*/
   
    // create the masks
    int AbbgMASK = 31; //0b0000000000011111 = 31 in decimal.
    int ArfgMASK = 96; //0b0000000001100000 = 96 in decimal.
    int ArfgSHIFT = 5;
   
    int Abbg = (rxgain&AbbgMASK);
    int Arfg = (rxgain&ArfgMASK)>>ArfgSHIFT;
   
    int tgain = 15*(Arfg-1) + 2*Abbg;
   
    xil_printf("\r\nAverage RX Gain: %d", tgain);
   
    short rssidat = WarpRadio_v1_RSSIData(FIRST_RADIO);
    xil_printf("\r\nRSSI: %d", rssidat);
   
}

As far as results go - I'll have to get them a bit later today as the workspace is being used to prepare for a demonstration this morning.

So we ran a slightly basic updated experiment, just using a signal generator to blast noise (no distance readings) - the signal generator was about 10 feet away from the node's antenna:

RSSI Value Readings:
(Using 18dBm noise on the signal generator)

Noise off: ~100-200 units
Narrowband Noise On (~12.5 kilosamples/second): 200-300 units (on same channel)
Wideband Noise (12.5 Mega-samples/second) On: ~800 units (on same channel)
Wideband noise 3 channels away ~400 (due to energy leakage)

A few things:
-The TotalGain is expressed in dB, so you want to subtract, not divide
-One definition would be SNR_dB = RxPower_dBm - NoiseFloor_dBm, where RxPower_dBm is the power received at the antenna and NoiseFloor_dBm is the inherent noise. But you can only digitally measure the voltages driven into the Rx ADC, after the two amplifiers. For accurate results, I would suggest transmitting a signal of known bandwidth and power, driving this into the Rx radio, settings Rx gains so no clipping occurs, and measuring the magnitude of the signals received digitally. Do this for a few points and, knowing the Rx gains for each, you should be able to calculate the Rx power.