avssamp/avshws does not appear as a system audio device

NTDEV
1

I try to write an usb audio capture device based on AVStream. So I create a filter with 2 Pins. 1 Bridge Pin and 1 CapturePin, (see below). With Graphedit, or KsStudio all works fine, but my device is not visible/selectable as a system audio device in System->Sound->Audio (mmsys.cpl). Is this a problem of the filter (Pin descriptor, etc.)? Or is this a problen in my Inf -file, because i only add an capture interface

AddInterface=%KSCATEGORY_CAPTURE%,%KSSTRING_FILTER%,DrvInstall.A.Interface

and no KSCATEGORY_AUDIO Interface?

What should I do that the device is an ancessible system audio device? Becaus lot of audio application captures the audio only from the system audio devices, not from KSfilters.

thanx for your help

Can you send me a version of avssamp/avshws wich appears as a system audio device

My Pins:
const KSPIN_DESCRIPTOR_EX CaptureFilterPinDescriptors [CAPTURE_FILTER_PIN_COUNT] =
{
//
// Capture Pin
//
{
&CapturePinDispatch,
NULL,
// KSPIN_DESCRIPTOR
{
NULL, 0, // Interfaces (NULL, 0 == default)
NULL, 0, // Mediums (NULL, 0 == default)
SIZEOF_ARRAY (CapturePinDataRanges), // Range Count
CapturePinDataRanges, // Ranges
KSPIN_DATAFLOW_OUT, // Dataflow
KSPIN_COMMUNICATION_BOTH, // Communication
&KSCATEGORY_AUDIO, // Category
&g_PINNAME_AUDIO_CAPTURE, // Name
0 // Reserved
},

//KSPIN_FLAG_GENERATE_MAPPINGS | // Pin Flags
KSPIN_FLAG_DISPATCH_LEVEL_PROCESSING |
KSPIN_FLAG_PROCESS_IN_RUN_STATE_ONLY,
1, // Instances Possible
1, // Instances Necessary
&CapturePinAllocatorFraming, // Allocator Framing
reinterpret_cast
(CCapturePin::IntersectHandler)
},

// bridge pin
{
NULL,
NULL,
{
0, NULL, // SIZEOF_ARRAY(PinInterfaces), PinInterfaces,
0, NULL, // SIZEOF_ARRAY(PinMediums), PinMediums,
0, NULL, //Ranges…
KSPIN_DATAFLOW_IN,
KSPIN_COMMUNICATION_NONE,
&KSNODETYPE_LINE_CONNECTOR, // Category
&KSNODETYPE_LINE_CONNECTOR, // Name
0
},
KSPIN_FLAG_PROCESS_IN_RUN_STATE_ONLY | // Flags
KSPIN_FLAG_FIXED_FORMAT,
1, //InstancesPossible
0, //InstancesNecessary
NULL,
NULL
}
};

2

xxxxx@gmx.de wrote:

I try to write an usb audio capture device based on AVStream.

May I ask why you are not using usbaudio.sys? I cannot imagine anyone
in the last 5 years who would have designed a USB-based audio device
that was not USB Audio Class compliant. The spec is easy to meet, and
doing so allows you to avoid all of the hassles of interfacing with the
audio subsystem on your own.

So I create a filter with 2 Pins. 1 Bridge Pin and 1 CapturePin, (see below). With Graphedit, or KsStudio all works fine, but my device is not visible/selectable as a system audio device in System->Sound->Audio (mmsys.cpl). Is this a problem of the filter (Pin descriptor, etc.)? Or is this a problen in my Inf -file, because i only add an capture interface

AddInterface=%KSCATEGORY_CAPTURE%,%KSSTRING_FILTER%,DrvInstall.A.Interface

and no KSCATEGORY_AUDIO Interface?

Wouldn’t it have been much quicker just to try adding the
KSCATEGORY_AUDIO interface, rather than submitting a question and
waiting for an answer that most folks are unlikely to have off the top
of their heads?

–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

3

Yes, of course you can.

My device has 32 channels at 96khz with 32 Bit/sample. This are 32*96000*4=
12Mbyte/s !!! The only way to transfer this is USB2.0, but the USB Audio
Device specification is only for USB1.1 devices, even so USBAudioclass
driver from microsoft is only for USB1.1 device.

Carsten

4

xxxxx@gmx.de wrote:

Yes, of course you can.

My device has 32 channels at 96khz with 32 Bit/sample. This are
32*96000*4= 12Mbyte/s !!! The only way to transfer this is USB2.0, but
the USB Audio Device specification is only for USB1.1 devices,

This is not true. MANY of today’s web cameras are USB 2.0 high-speed
devices that include a USB Video Class interface and a USB Audio Class
interface. The audio class spec makes no mention of this kind of
limitation.

even so USBAudioclass driver from microsoft is only for USB1.1 device.

Have you tried it? I know for a fact that usbaudio.sys works just fine
with USB 2.0 high-speed devices. I have 4 or 5 right here in my office.

I don’t know whether it will, in fact, compute the isochronous buffer
sizes correctly in a case like yours, but I would be surprised if it did
not.

–
Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

5

I tried (some days now) to stream data over usbaudio.sys.
I use a Cypress FX2 demo board. The device enuerates under Windos XP and is
selectable and working (system->osound, GraphEdit), but when I try to play
the graph, Graphedit hangs an no data is streamed. The Isochronous Endpoint
(EP2, in my case) is not polled by host. This is, why I thought, that
usbaudio.sys is not capable of handling USB2.0 high speed devices as audio
devices. (my audiodevice is based on USB Audio Class 1.0, because Microsoft
definitly says, that USB Audio Class2.0 is NOT supported). Maybe you can
send me a log of your USB2.0 Audio device descriptors(possible sampled with
usbview.exe). Are you shure, that they are working in high-speed mode, not
in full speed mode and that they use the class driver? But we’ll see in the
log…

Carsten

“Tim Roberts” schrieb im Newsbeitrag news:xxxxx@ntdev…
> xxxxx@gmx.de wrote:
>> Yes, of course you can.
>>
>> My device has 32 channels at 96khz with 32 Bit/sample. This are
>> 32960004= 12Mbyte/s !!! The only way to transfer this is USB2.0, but
>> the USB Audio Device specification is only for USB1.1 devices,
>
> This is not true. MANY of today’s web cameras are USB 2.0 high-speed
> devices that include a USB Video Class interface and a USB Audio Class
> interface. The audio class spec makes no mention of this kind of
> limitation.
>
>
>> even so USBAudioclass driver from microsoft is only for USB1.1 device.
>
> Have you tried it? I know for a fact that usbaudio.sys works just fine
> with USB 2.0 high-speed devices. I have 4 or 5 right here in my office.
>
> I don’t know whether it will, in fact, compute the isochronous buffer
> sizes correctly in a case like yours, but I would be surprised if it did
> not.
>
> –
> Tim Roberts, xxxxx@probo.com
> Providenza & Boekelheide, Inc.
>
>

6

Carsten Karpowitz wrote:

I use a Cypress FX2 demo board. The device enuerates under Windos
XP and is selectable and working (system->osound, GraphEdit), but
when I try to play the graph, Graphedit hangs an no data is
streamed. The Isochronous Endpoint (EP2, in my case) is not
polled by host. This is, why I thought, that usbaudio.sys is not
capable of handling USB2.0 high speed devices as audio devices.

I’m dumbfounded by this analysis. At the very minimum, if I observed this, I would have dropped a USB 1.1 hub in the mix to see if it started working.

Otherwise, you’re basically saying, “I tried one device, once, and it didn’t work. Without messing around with it at all, including anything having to do with full versus high speed, I declared that usbaudio.sys does not support 2.0 high speed devices.”

7

Many thanks for the descriptors Tim,

now I know the trick. I have had a look at the audio streaming endpoint
descriptor. Here is bIntervall=4, wich means the polling intervall is 2
^(bIntervall-1) = 2^(4-1)=2^3=8. This means, that the iso-endpoint is
polled only every 8th microframe and indeed there are 8 microframes in a
frame, so the endpoint is polled once per millisecond. This is compatible
to USB1.1 but this is a very big performance leak! You can only transfer the
same datarate as you can with USB1.1. Ok, there is more free space on the
2.0 bus for other devices, and of course we can use this free bandwith with
our device with a second endpoint, but there is no chance to trasmit over a
single iso streaming endpoint with more bandwith than USB1.1. So this is not
really an advantage and we still have to wait for the USB audio class 2.0
implementation from Microsoft and up to then we must implement our own
streaming driver wich can handle higher data transfer rates. Back to the
roots…at this point I return to my first question: How can I make my
kernel streaming filter capture device visible as a system audio device? Or
to simplify the question, what must be done to make avssamp or avshws
visible as a system audio device?

carsten

PS: The endpoint descriptor layout doesn’t matter at all, USB1.1 endpoints
with length=9 can be used, USB2.0 endpoints with length=7 can also be used,
but the polling intervall (bIntervall) had to be 4, else usbaudio.sys
doesn’t work.