Data loss in isochronous end point

Tip / Sign in to post questions, reply, level up, and achieve exciting badges. Know more

cross mob
Anonymous
Not applicable

Hi..

I have implemented a GPIF II interface using an FPGA spartan 6 (XC6SLX9) and a Cypress FX3 (CYUSB3013) controller.

An Isochronous USB endpoint is configured to trasfer the data from the Cypress FX3 to a host PC.

The average data generate rate of the FPGA is about 75Mbps which is generated as distributed chunks rather than as a constant rate data stream. However, I am losing a significant amount of data at the host PC allication.

Followng are some details on my sytsem setup.

PCLK = 60Mhz ; (PCLK is driven by FPGA)

FX3 system clock = 403.2 Mhz; (setSysClk400 = true)

GPIF clock = 100 Mhz ; (pibClock.clkDiv = 4)

USB end point type :- Isochronous

*FX3 Firmware

CY_FX_ISO_BURST   = 15;

CY_FX_ISO_PKTS    = 1;

epcfg.enable = CyTrue;

epcfg.epType = CY_U3P_USB_EP_ISO;

epcfg.burstLen = CY_FX_ISO_BURST;

epcfg.streams = 0;

epcfg.pcktSize = 1024;

epcfg.isoPkts = 1;

dma_cfg.size = ((1024 + 0x0F) & ~0x0F);

    dma_cfg.size  *= CY_FX_ISO_BURST;

    dma_cfg.count = 8;

    dma_cfg.prodSckId = CY_U3P_PIB_SOCKET_0;

    dma_cfg.consSckId = CY_U3P_UIB_SOCKET_CONS_2;

    dma_cfg.dmaMode = CY_U3P_DMA_MODE_BYTE;

dma_cfg.notification = CY_U3P_DMA_CB_CONS_EVENT;

dma_cfg.cb = 0;

dma_cfg.prodHeader = 0;

dma_cfg.prodFooter = 0;

dma_cfg.consHeader = 0;

dma_cfg.prodAvailCount = 0;

<ENDPOINT>

Type="ISOC"

Direction="IN"

Address="82h"

Attributes="01h"

MaxPktSize="15360"

DescriptorType="5"

DescriptorLength="7"

Interval="4"

<SUPER SPEED ENDPOINT COMPANION>

Type="SUPERSPEED_USB_ENDPOINT_COMPANION"

MaxBurst="14"

Attributes="00h"

BytesPerInterval="3C00h"

</ENDPOINT>

*PC application code to read usb

LONG length = 1024 * 15 * 8 * 64;

std::shared_ptr<uint8_t> image_buffer = std::shared_ptr<uint8_t>(new uint8_t[length]);

QueueSize = int(length / (iso_ep->MaxPktSize * 8));

if ((length % (iso_ep->MaxPktSize * 8)) > 0)

QueueSize++;

for (int j = 0; j < QueueSize; j++){

long len = length / (QueueSize);

std::unique_ptr<uint8_t[]> image_data = std::unique_ptr<uint8_t[]>(new uint8_t[len]);

iso_ep->TimeOut = 5000;

int pkts;

CCyIsoPktInfo * pktInfos;

pktInfos = iso_ep->CreatePktInfos(len, pkts);

if (iso_ep->XferData(image_data.get(), len, pktInfos)){

for (int i = 0; i < pkts; i++){

if (pktInfos.Status == 0 && pktInfos.Length>0)

{

memcpy(&image_buffer[success * iso_ep->MaxPktSize], &image_data[i * iso_ep->MaxPktSize], iso_ep->MaxPktSize);

success = success + 1;

}

}

}

delete[] pktInfos;

}

length = success * iso_ep->MaxPktSize;

0 Likes
1 Solution
Anonymous
Not applicable

Please try with the Cypress streamer application which uses asynchronous data transfer. (BeginDataXfer, WaitForXfer, FinsihDataXfer) instead of synchronous transfer (XferData API). The former is more efficient.

If you do not face any issue with Cypress streamer, please use the same code in your host application.

Regards,

- Madhu Sudhan

View solution in original post

0 Likes
1 Reply
Anonymous
Not applicable

Please try with the Cypress streamer application which uses asynchronous data transfer. (BeginDataXfer, WaitForXfer, FinsihDataXfer) instead of synchronous transfer (XferData API). The former is more efficient.

If you do not face any issue with Cypress streamer, please use the same code in your host application.

Regards,

- Madhu Sudhan

0 Likes