Usb failed packets

I think it is MARSet tried like this

void USB_Reset (void) {

#if USB_DMA
  LPC_USB->USBUDCAH   = USB_RAM_ADR;
  LPC_USB->USBDMARClr = 0xFFFFFFFF;
  LPC_USB->USBEpDMADis  = 0xFFFFFFFF;
  LPC_USB->USBEpDMAEn   = USB_DMA_EP;
  LPC_USB->USBEoTIntClr = 0xFFFFFFFF;
  LPC_USB->USBNDDRIntClr = 0xFFFFFFFF;
  LPC_USB->USBSysErrIntClr = 0xFFFFFFFF;
  LPC_USB->USBDMAIntEn  = 0x00000007;
  LPC_USB->USBDMARSet = 1 << EPAdr(0x83);<--------------
  //LPC_USB->USBDMARSet = USB_ENDPOINT_IN(3);<------------
  //DMA_REQ_SET = 1 << EPAdr(0x83);<------------------
//LPC_USB->USBDMARSet = 1<<EPAdr(USB_DMA_EP;);<------------

  DDMemMap[0] = 0x00000000;
  DDMemMap[1] = 0x00000000;
  for (n = 0; n < USB_EP_NUM; n++) {
    udca[n] = 0;
    UDCA[n] = 0;
  }
#endif
}

After I go back I misunderstood what you were saying it just does not seem to want to work properly in Slave mode,

The firmware handles the isoc IN endpoint (0x83) in "slave mode" (ie. not in DMA mode).
USB_SOF_Event() processes this endpoint.
So, don't register this EP at USB_DMA_EP

usbcfg.h
#define USB_DMA_EP          0x000000C0 // <--- 0x00000040
Tsuneo

Can it not be set up for the in endpoint to work in DMA MODE I have been trying with no success can you help.

anything like this?? I see in the orignal code in the SOF_EVENT endpoint 3 out is call in slave mode, and then in the USB_EndPoint3 (uint32_t event) it is then put in DMA MODE

If I just leave it how you orignal set it both in and out endpoint 3 will both be called in slave mode first then if I can figure out how it should be written I can call them both in DMA MODE
at void USB_EndPoint3 (uint32_t event) {

Becasue like I said the IN Endpoint 3 just does not want to work properly with out adding it into void USB_EndPoint3 (uint32_t event) { in DMA MODE after being first called in SLAVE MODE at the SOF EVENT

 uint32_t last_timer_count;
#if USB_SOF_EVENT
void USB_SOF_Event (void) {
#if USB_DMA == 0 <----- don't need this??
uint32_t current_timer_count;
  uint32_t feedback_value;

  uint32_t pclk;

                                                        // feedback
  current_timer_count = LPC_TIM1->TC;                   // capture current SOF timing on the Timer1
  if ( USB_AltSetting[1] == 1 ) {               // When interface 1 / alt 1 is enabled,
                                                // calculate master/SOF frequency ratio in 10.10 (10.14) format
    feedback_value = ((current_timer_count - last_timer_count) << 14) / (pclk/DATA_FREQ);
                                                // and send it to the feedback IN EP
    USB_WriteEP( 0x83, (uint8_t *)&feedback_value, 3 );
  }
  last_timer_count = current_timer_count;       // update the last SOF timing

void USB_EndPoint3 (uint32_t event) {
#if USB_DMA
  USB_DMA_DESCRIPTOR DD;
  uint32_t current_timer_count;
  uint32_t feedback_value;

  uint32_t pclk;

  if (event & USB_EVT_IN_DMA_EOT) {
    /* End of Transfer */
current_timer_count = LPC_TIM1->TC;                   // capture current SOF timing on the Timer1
  if ( USB_AltSetting[1] == 1 ) {               // When interface 1 / alt 1 is enabled,
                                                // calculate master/SOF frequency ratio in 10.10 (10.14) format
    feedback_value = ((current_timer_count - last_timer_count) << 14) / (pclk/DATA_FREQ);
                                                // and send it to the feedback IN EP

    if (USB_DMA_BufAdr(0x83) != ((uint8_t *)&feedback_value, 3 )); {

    }

}
  last_timer_count = current_timer_count;       // update the last SOF timing

  }
  if (event & (USB_EVT_IN_DMA_EOT) | (USB_EVT_IN_DMA_NDR)) {
    /* End of Transfer or New Descriptor Request */
    DD.BufAdr  = ((uint8_t*)&feedback_value, 3)); /* DMA Buffer Address */
    DD.BufLen  = P_C;                       /* DMA Packet Count */
    DD.MaxSize = 0;                         /* Must be 0 for Iso Transfer */
    DD.InfoAdr = (uint32_t)InfoBuf;            /* Packet Info Buffer Address */
    DD.Cfg.Val = 0;                         /* Initial DMA Configuration */
    DD.Cfg.Type.IsoEP = 1;                  /* Iso Endpoint */
    USB_DMA_Setup (0x83, &DD);              /* Setup DMA */
    USB_DMA_Enable(0x83);                   /* Enable DMA */
  }
#else
  event = event;
#endif
}

Could would please try to get it work like this even if you don't think it is the problem.

EP3 IN and EP3 OUT works independently.

usbcfg.h
#define USB_DMA_EP          0x000000C0 // <--- 0x00000040

This setting means
- EP3 IN: slave mode
- DP3 OUT: DMA mode

You may run EP3 IN in DMA mode, too.
But DMA for just 3 bytes transfer is too much.

I tested this modification of USBAudio on Olimex LPC1766-STK and MacOSX, monitored on a hardware bus analyzer.
The EP3 IN correctly responds to the transactions from host, but its data was unexpected one.
Maybe, Windows upset with this unexpected feedback value.

In your USB_SOF_Event() code, pclk is used in right side without filling any value.

void USB_SOF_Event (void) {
  uint32_t pclk;

    feedback_value = ((current_timer_count - last_timer_count) << 14) / (pclk/DATA_FREQ);

I fixed it, but the feedback value over EP3 IN is still strange.

Almost same modification works on MCB23xx board without any problem.
There are some difference(s) on LPC17xx, on which I'm not yet aware of it.
Now, another test is going on.

Tsuneo

For test of EP3 IN transaction, put a fixed value to the feedback_value, as follows.
And then, you'll see Windows also run the device as we expect.
This is the clue that wrong feedback_value makes Windows upset.

void USB_SOF_Event (void) {
uint32_t current_timer_count;
  uint32_t feedback_value;
  uint32_t pclk;
                                                // feedback
  current_timer_count = LPC_TIM1->TC;           // capture current SOF timing on the Timer1
  if ( USB_AltSetting[1] == 1 ) {               // When interface 1 / alt 1 is enabled,
                                                // calculate master/SOF frequency ratio in 10.10 (10.14) format
//    feedback_value = ((current_timer_count - last_timer_count) << 14) / (pclk/DATA_FREQ);

    feedback_value = 32 << 14;    // test value for 32k sampling

                                                // and send it to the feedback IN EP
    USB_WriteEP( 0x83, (uint8_t *)&feedback_value, 3 );
  }
  last_timer_count = current_timer_count;       // update the last SOF timing

Tsuneo

This code did not seem to do anything when trying on windows vista

void USB_SOF_Event (void) {
uint32_t current_timer_count;
  uint32_t feedback_value;
  uint32_t pclk;
                                                // feedback
  current_timer_count = LPC_TIM1->TC;           // capture current SOF timing on the Timer1
  if ( USB_AltSetting[1] == 1 ) {               // When interface 1 / alt 1 is enabled,
                                                // calculate master/SOF frequency ratio in 10.10 (10.14) format
//    feedback_value = ((current_timer_count - last_timer_count) << 14) / (pclk/DATA_FREQ);

    feedback_value = 32 << 14;    // test value for 32k sampling

                                                // and send it to the feedback IN EP
    USB_WriteEP( 0x83, (uint8_t *)&feedback_value, 3 );
  }
  last_timer_count = current_timer_count;       // update the last SOF timing

I try to mod it but can not get it to work

feedback_value = ((current_timer_count - last_timer_count) (32<< 14) / (pclk/DATA_FREQ);

OK, I found a bug of the original Keil example.

BUG: USB_SOF_Event() is repeatedly called in the interval of around 3us
It should be called in 1ms interval.

Solution: Clear SOF (FRAME) interrupt flag in ISR, as follows

MCB1700 Sample Code Bundle for LPC17xx Peripherals using Keil's MDK-ARM V2.00 (Jan 10, 2011)
ics.nxp.com/.../mcb1700.code.bundle.lpc17xx.keil.zip

keil_examples\USBAudio\usbhw.c

void USB_IRQHandler (void) {
  ...
  ...
#if USB_SOF_EVENT
  /* Start of Frame Interrupt */
  if (disr & FRAME_INT) {

    LPC_USB->DevIntClr = FRAME_INT;   // <---- bug fix

    USB_SOF_Event();
  }
#endif

Tsuneo

Nope did nothing, apart from hang the pc after a couple of seconds of audio.

What do these definitions do anyway where is the documentation. the same errors are still happing

usbreg.h

/* Device Interrupt Bit Definitions */
#define FRAME_INT           0x00000001
#define EP_FAST_INT         0x00000002
#define EP_SLOW_INT         0x00000004
#define DEV_STAT_INT        0x00000008
#define CCEMTY_INT          0x00000010
#define CDFULL_INT          0x00000020
#define RxENDPKT_INT        0x00000040
#define TxENDPKT_INT        0x00000080
#define EP_RLZED_INT        0x00000100
#define ERR_INT             0x00000200

Hmm,
On my side, the mods work as we expect.
Maybe, I didn't touch to all of modifications I made.
I posted current modification, here.
www.8052.com/.../MCB1700_USBAudio_Async_32k_r0.zip
Try this one.

To run above firmware on MCB1700, touch this line.

usbdmain.c

int main (void)
{
  ...
  ...
#define OLIMEX_LPC1766STK   // <--- comment this line for MCB1700

At this stage, we have examined basic implementation of the feedback endpoint for Async Sink.
To brush up above implementation, these issues still remain.

1) Feedback refresh interval
In above implementation, Windows read out the feedback EP every 11 or 12 frames, Mac does every 8 frames.
But our firmware refreshes it in every frame.

We have to change the refresh interval and bRefresh value on sync EP.

USB Audio Spec, www.usb.org/.../audio10.pdf
3.7.2.2 Isochronous Synch Endpoint
The sample clock Fs is always derived from a master clock Fm in the device.
P is related to the ratio between those clocks through the following relationship:
Fm = Fs * 2^(P-1)

As indicated earlier, a new Ff value is available every 2^(10 – P) frames with P ranging from 1 to 9. The
bRefresh field of the synch standard endpoint descriptor is used to report the exponent (10-P) to the Host.
It can range from 9 down to 1. (512 ms down to 2 ms)

2) Variable number of samples per frame
Async Sink results +/- 1 samples per frame occasionally, when the frequency ratio of
the local sampling clock and SOF is apart from an integer number.
But original Keil implementation assumes that the sample number per frame would be constant,
and this number is power of 2, ie. 1, 2, 4, 8, 16, 32,...
This assumption is fine for Adaptive/Synchronous Sink, but it is not approved for Async Sink.

- I touched to the mods for Keils's "mask" method, in this post
http://www.keil.com/forum/19557/
Posted 30-Sep-2011 03:43 GMT

- The mods on above post has to be modified more for variable transferred data size on each frame.

3) Pop noise on Windows
Windows irregularly put Set_Interface(0) / Set_Interface(1) pair to the device, while it plays sound.
This makes pop noise. I'm not sure if this behavior is caused by the value of feedback or not.

Tsuneo

Which one of these should I be changing the feild bInterval and bRefresh for. If you could explain how they both work together would help.

/* Endpoint - Standard Descriptor */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_OUT(3),                  /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS | USB_ENDPOINT_SYNC_ASYNCHRONOUS,        /* bmAttributes */          // <-------
  WBVAL(66),                            /* wMaxPacketSize */   // <-------
  0x01,                                 /* bInterval */
  0x00,                                 /* bRefresh */
  0x83,                                 /* bSynchAddress */    // <-------

/* Endpoint - Standard Descriptor - for feedback */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_IN(3),                   /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS,        /* bmAttributes */
  WBVAL(3),                             /* wMaxPacketSize */
  0x01,                                 /* bInterval */
  0x01,                                 /* bRefresh */
  0x00,                                 /* bSynchAddress */

If I set them like this

/* Endpoint - Standard Descriptor */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_OUT(3),                  /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS | USB_ENDPOINT_SYNC_ASYNCHRONOUS,        /* bmAttributes */          // <-------
  WBVAL(66),                            /* wMaxPacketSize */   // <-------
  0x01,                                 /* bInterval */   <------------------------ Interval for polling endpoint for data transfers expressed in milliseconds  0x01 = 1 packet every frame
  0x09,                                 /* bRefresh */   <----------------------------- 512ms until synchronization pipe provides new synchronization feedback data.
  0x83,                                 /* bSynchAddress */    // <-------

would that mean I would get 1ms delay for polling data transfers of 1 packet every frame, and then 512ms before the endpoint provides new synchronization feedback data.

what do you feel they should be set at.

What are you suggesting to be changed when you wrote these comments, are you talking about audio.h and desc.c WBVAL(66), and P_S Packet Size

2) Variable number of samples per frame
Async Sink results +/- 1 samples per frame occasionally, when the frequency ratio of
the local sampling clock and SOF is apart from an integer number.
But original Keil implementation assumes that the sample number per frame would be constant,
and this number is power of 2, ie. 1, 2, 4, 8, 16, 32,...
This assumption is fine for Adaptive/Synchronous Sink, but it is not approved for Async Sink.

/* Audio Definitions */
#define DATA_FREQ 32000                 /* Audio Data Frequency */
#define P_S       32                    /* Packet Size */
#if USB_DMA
#define P_C       4                     /* Packet Count */
#else
#define P_C       1                     /* Packet Count */
#endif
#define B_S       (8*P_C*P_S)           /* Buffer Size */

>> The mods on above post has to be modified more for variable transferred data size on each frame.

Could you give me details about what I am to do here

I could not get your code to work with MCB1700 after commenting out the line,

usb device is unrecognised

<<<<<<<<<<<<<<<<<<<<<<<

Hmm,
On my side, the mods work as we expect.
Maybe, I didn't touch to all of modifications I made.
I posted current modification, here.
www.8052.com/.../MCB1700_USBAudio_Async_32k_r0.zip
Try this one.

To run above firmware on MCB1700, touch this line.

usbdmain.c

int main (void)
{ ... ...
#define OLIMEX_LPC1766STK // <--- comment this line for MCB1700

<<<<<<<<<<<<<<<<<<<<

Do you know why

> Do you know why

I don't know :-)
There aren't so many differences between MCB1700 and OLIMEX_LPC1766STK for this example.
Just the analog port, to which a pod connects.
I put the #define into the source, to cover this difference.

a) Is the original Keil example runs on your MCB1700?
b) What is the version of Windows?
c) How doesn't it "work"? Is it enumerated? No sound?, etc

Tsuneo

Could you answer these comments first please. Much more important to me than getting the code you posted to work thankyou.

Which one of these should I be changing the feild bInterval and bRefresh for. If you could explain how they both work together would help.

/* Endpoint - Standard Descriptor */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_OUT(3),                  /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS | USB_ENDPOINT_SYNC_ASYNCHRONOUS,        /* bmAttributes */          // <-------
  WBVAL(66),                            /* wMaxPacketSize */   // <-------
  0x01,                                 /* bInterval */
  0x00,                                 /* bRefresh */
  0x83,                                 /* bSynchAddress */    // <-------

/* Endpoint - Standard Descriptor - for feedback */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_IN(3),                   /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS,        /* bmAttributes */
  WBVAL(3),                             /* wMaxPacketSize */
  0x01,                                 /* bInterval */
  0x01,                                 /* bRefresh */
  0x00,                                 /* bSynchAddress */

If I set them like this

/* Endpoint - Standard Descriptor */
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_OUT(3),                  /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS | USB_ENDPOINT_SYNC_ASYNCHRONOUS,        /* bmAttributes */          // <-------
  WBVAL(66),                            /* wMaxPacketSize */   // <-------
  0x01,                                 /* bInterval */   <------------------------ Interval for polling endpoint for data transfers expressed in milliseconds  0x01 = 1 packet every frame
  0x09,                                 /* bRefresh */   <----------------------------- 512ms until synchronization pipe provides new synchronization feedback data.
  0x83,                                 /* bSynchAddress */    // <-------

would that mean I would get 1ms delay for polling data transfers of 1 packet every frame, and then 512ms before the endpoint provides new synchronization feedback data.

what do you feel they should be set at.

What are you suggesting to be changed when you wrote these comments, are you talking about audio.h and desc.c WBVAL(66), and P_S Packet Size

2) Variable number of samples per frame
Async Sink results +/- 1 samples per frame occasionally, when the frequency ratio of
the local sampling clock and SOF is apart from an integer number.
But original Keil implementation assumes that the sample number per frame would be constant,
and this number is power of 2, ie. 1, 2, 4, 8, 16, 32,...
This assumption is fine for Adaptive/Synchronous Sink, but it is not approved for Async Sink.

/* Audio Definitions */
#define DATA_FREQ 32000                 /* Audio Data Frequency */
#define P_S       32                    /* Packet Size */
#if USB_DMA
#define P_C       4                     /* Packet Count */
#else
#define P_C       1                     /* Packet Count */
#endif
#define B_S       (8*P_C*P_S)           /* Buffer Size */

>> The mods on above post has to be modified more for variable transferred data size on each frame.

Could you give me details about what I am to do here

1) Feedback refresh interval
> Which one of these should I be changing the feild bInterval and bRefresh for.

Both of bInterval fields of the OUT and IN endpoints are set to 1 (ms)
I mean the bRefresh field of the feedback IN EP.

Host (Windows and Mac) polls the feedback IN endpoint in the interval specified by the bRefresh value, not in 1ms interval.
USB Audio spec tells the shortest interval is 2 ms (bRefresh = 1)
But there seems lower limit by OS audio driver implementation,
MacOSX: 8 ms (bRefresh = 3), and Windows: 11 or 12 ms

In above implementation, the firmware calculates the feedback value on every SOF interrupt.
We have to change this behavior.
When bRefresh = 3, the firmware calculate feedback value in every 8 ms.

#define FEEDBACK_REFRESH_EXPONENT  3      // bRefresh value for the feedback IN EP (1..9)
#define FEEDBACK_REFRESH_INTERVAL  (1 << FEEDBACK_REFRESH_EXPONENT)

uint8_t feedback_refresh_timer = FEEDBACK_REFRESH_INTERVAL;
uint32_t last_timer_count;
extern uint32_t sampling_interval;

#if USB_SOF_EVENT
void USB_SOF_Event (void) {

  uint32_t current_timer_count;
  uint32_t feedback_value;
                                                // feedback
  current_timer_count = LPC_TIM1->TC;           // capture current SOF timing on the Timer1
  if ( --feedback_refresh_timer == 0 ) {

    if ( USB_AltSetting[1] == 1 ) {             // When interface 1 / alt 1 is enabled,
                                                // calculate master/SOF frequency ratio in 10.10 (10.14) format
      feedback_value = ((current_timer_count - last_timer_count) << 14) / FEEDBACK_REFRESH_INTERVAL / sampling_interval;
                                                // and send it to the feedback IN EP
      USB_WriteEP( 0x83, (uint8_t*)&feedback_value, 3 );
    }
    last_timer_count = current_timer_count;     // update the last SOF timing
    feedback_refresh_timer = FEEDBACK_REFRESH_INTERVAL;
  }

WBVAL(66) (wMaxPacketSize of isoc OUT EP)

The original example sets this value to WBVAL(64), ie. 32 samples for 32kHz monaural 16bits.
But MacOSX doesn't put any feedback IN transaction for this value.
To receive feedback IN transactions, at least WBVAL(66) (32 + 1 samples) is required.

2) Variable number of samples per frame

This modification is required not just for Async Sink, but also for support of fractional sampling rate, such as 44.1kHz

For DMA implementation of the isoc OUT EP,
USB_EndPoint3() is called when packet counts (P_C) number of packets are received.
Packet Info Buffer (InfoBuf array) holds received size of each packets (LSB 2 bytes)
(refer to "11.15.6 Isochronous endpoint operation" of the LPC17xx manual)

For slave implementation of the isoc OUT EP, USB_ReadEP() in USB_SOF_Event() returns received packet size.

In these callbacks (DMA:USB_EndPoint3(), slave:USB_SOF_Event())
DataIn index increments by actual received packet size, instead of the constant value (P_C*P_S or P_S) of the original.
A problem lies in the end of the cyclic buffer. It moves around by the extra (or less) sample(s).
When DataIn reaches near the end of the cyclic buffer, this index is held in a current_end_of_buffer variable.
And DataIn wraparounds to the top of the buffer.

In the playback routine (TIMER0_IRQHandler()),
DataOut wraparounds to the top of the buffer, when it reaches to above current_end_of_buffer.

Tsuneo

With mod you posted for usbuser.c, it does not work until I have to change the bRefresh to 0x03 before it will work.

/* Endpoint - Standard Descriptor */                       // <-------- additional feedback EP
  AUDIO_STANDARD_ENDPOINT_DESC_SIZE,    /* bLength */
  USB_ENDPOINT_DESCRIPTOR_TYPE,         /* bDescriptorType */
  USB_ENDPOINT_IN(3),                   /* bEndpointAddress */
  USB_ENDPOINT_TYPE_ISOCHRONOUS,        /* bmAttributes */
  WBVAL(3),                             /* wMaxPacketSize */
  0x01,                                 /* bInterval */
  0x03,                                 /* bRefresh,  */     //<-------  every 8ms
  0x00,                                 /* bSynchAddress */

in vista for some reason not sure it is bug but it always trys for 2ms even when changes for 8ms, I do get 8ms transfers but for some reason some 2ms transfer do happen and fail:( and I made sure the vid pid were changed as well

USBlyzer Report

Capture List
Type    Seq     Time    Elapsed         Duration        Request         Request Details         Raw Data        I/O     C:I:E   Device Object   Device Name     Driver Name     IRP     Status
URB     24534-24533     8:29:28.673     17.530532 s     1.383 ms        Isoch Transfer                  in      01:01:83        86F4E6F0h       USBPDO-5        usbhub  86E69E28h       Unsuccessful (Iso Request Failed)
URB     24535-24532     8:29:28.673     17.530538 s     1.395 ms        Isoch Transfer                  in      01:01:83        86F57C80h       0000009e        usbccgp 86E69E28h       Unsuccessful (Iso Request Failed)
URB     32498-32497     8:29:34.364     23.222458 s     1.806 ms        Isoch Transfer                  in      01:01:83        86F4E6F0h       USBPDO-5        usbhub  86E69E28h       Unsuccessful (Iso Request Failed)
URB     32499-32496     8:29:34.364     23.222461 s     1.813 ms        Isoch Transfer                  in      01:01:83        86F57C80h       0000009e        usbccgp 86E69E28h       Unsuccessful (Iso Request Failed)

Now for win7 even with the bug fix you posted I get these us errors

URB     0047-0046       8:49:33.314     32.829 ms       53 us   Isoch Transfer                  in      01:01:83        85057030h       USBPDO-5        usbhub  94D72B00h       Unsuccessful (Internal HC Error)
URB     0048-0045       8:49:33.314     32.832 ms       59 us   Isoch Transfer                  in      01:01:83        867F9A80h               ACPI    94D72B00h       Unsuccessful (Internal HC Error)
URB     0049-0044       8:49:33.314     32.834 ms       66 us   Isoch Transfer                  in      01:01:83        8699BC68h       00000068        usbccgp 94D72B00h       Unsuccessful (Internal HC Error)
URB     0191-0190       8:49:33.364     87.787 ms       60 us   Isoch Transfer                  in      01:01:83        85057030h       USBPDO-5        usbhub  94D72B00h       Unsuccessful (Internal HC Error)
URB     0192-0189       8:49:33.364     87.790 ms       68 us   Isoch Transfer                  in      01:01:83        867F9A80h               ACPI    94D72B00h       Unsuccessful (Internal HC Error)
URB     0193-0188       8:49:33.364     87.793 ms       78 us   Isoch Transfer                  in      01:01:83        8699BC68h       00000068        usbccgp 94D72B00h       Unsuccessful (Internal HC Error)
URB     0329-0328       8:49:33.424     143.782 ms      57 us   Isoch Transfer                  in      01:01:83        85057030h       USBPDO-5        usbhub  94D72B00h       Unsuccessful (Internal HC Error)

and I am getting some 0 packet transfers as well, but I think this is because of the new mods you made to usbuser.c

URB     0091-0084       9:25:52.942     2.192497 s      7.821 ms        Isoch Transfer  0 bytes data            in      01:01:83        84306030h       USBPDO-5        usbhub  842D0670h       Success (Success)
URB     0092-0083       9:25:52.942     2.192501 s      7.831 ms        Isoch Transfer  0 bytes data            in      01:01:83        84396020h               ACPI    842D0670h       Success (Success)
URB     0093-0082       9:25:52.942     2.192504 s      7.839 ms        Isoch Transfer  0 bytes data            in      01:01:83        82124990h       00000071        usbccgp 842D0670h       Success (Succes

I made the changes for keils's mask method

- I touched to the mods for Keils's "mask" method, in this post
http://www.keil.com/forum/19557/

you wrote

WBVAL(66) (wMaxPacketSize of isoc OUT EP)

The original example sets this value to WBVAL(64), ie. 32 samples for 32kHz monaural 16bits.
But MacOSX doesn't put any feedback IN transaction for this value.
To receive feedback IN transactions, at least WBVAL(66) (32 + 1 samples) is required.

does the audio.h need to be modded

/* Audio Definitions */
#define DATA_FREQ 32000                 /* Audio Data Frequency */
#define P_S       32                    /* Packet Size */  <-------------32  + 1 samples
#if USB_DMA
#define P_C       4                     /* Packet Count */
#else
#define P_C       1                     /* Packet Count */
#endif
#define B_S       (8*P_C*P_S)           /* Buffer Size */