RME ADI-2/4 Pro SE

[danadam]

Can you help me understand why? Is this a technical limitation of USB 2.0? If so, is it a limitation on the number of channels? Or is it sample rate?

The maximum transfers defined by USB Audio Class are:

• 1023 bytes every 1 ms for Class 1 (USB 1.0/1.1)
• 1024 bytes every 125 µs for Class 2 (USB 2.0)

That's max 65 Mbps for UAC2. 8x 352 kHz is already 90 Mbps.

From Fundamentals of USB Audio (Wayback Machine):

The isochronous endpoints run at a rate of one transfer per 125µs; or 8,000 transfers per second.
...
A single isochronous transfer can carry 1024 bytes, and can carry at most 256 samples (at 24/32 bits).

 

[catastrofe]

The maximum transfers defined by USB Audio Class are:

• 1023 bytes every 1 ms for Class 1 (USB 1.0/1.1)
• 1024 bytes every 125 µs for Class 2 (USB 2.0)

That's max 65 Mbps for UAC2. 8x 352 kHz is already 90 Mbps.

From Fundamentals of USB Audio (Wayback Machine):

Thank you. So Exasound is “lying” when they claim 384k on USB2.0 for 8 channels?

 

[danadam]

Thank you. So Exasound is “lying” when they claim 384k on USB2.0 for 8 channels?

On their page they say:

s88 features proprietary asynchronous USB streaming
...
USB Input: Proprietary ZeroJitter™ asynchronous USB interface with error correction on classic B–Type connector, USB cable included

so maybe they require a special driver to support that even if it won't work with standard UAC2 drivers.

 

[danadam]

That's max 65 Mbps for UAC2.

Hm... on the other hand, microchip.com claims that the max throughput in isochronous transfer is 3x higher. According to them there are three 1024-byte transfers allowed per frame. So ¯\_(ツ)_/¯

 

[catastrofe]

On their page they say:

so maybe they require a special driver to support that even if it won't work with standard UAC2 drivers.

Yes, I saw that verbiage and attributed it to marketing speak since their specs show USB2.0.

I do have the option of running Roon via a network connection which would render (no pun intended) the USB question moot.

Apologies for going off-topic with the Exasound discussion...I was hoping the RME would fit my needs.

 

[rcstevensonaz]

Yes, I saw that verbiage and attributed it to marketing speak since their specs show USB2.0.

I do have the option of running Roon via a network connection which would render (no pun intended) the USB question moot.

Apologies for going off-topic with the Exasound discussion...I was hoping the RME would fit my needs.

On a personal note: I find it is more useful to be a truth-seeker than a cynical challenger when trying to understand why things are the way they are. And further, to assume the other side is acting / speaking with good faith until your expertise is deep enough that you can specifically identify where the other side is in error... at least to the best of your knowledge at that time... since you could still be wrong even then.

For example: consider that USB 2.0 specification defines a wide range of Device Classes. Perhaps that ExaSound device declares itself as a different USB Class instead of UAC2 (e.g., Application Specific, Vendor Specific)? Or perhaps their proprietary driver will operate in a manner that is mutually agreed between the ExaSound device and their proprietary driver to allow them to enter into an "override" mode of operation that is uniquely written to allow the higher level streaming, while still allowing the device to operate under standard UAC2 mechanisms when it is attached to any computer that is just using the standard UAC2 drivers. Just a guess based on reading these very interesting posts.*

USB Device Classes - Developer Help

USB Device Classes

developerhelp.microchip.com

* P.s., I've found this fascinating reading because I was trying to figure out whether an RME Digiface USB could send 16 channels (8 play, 8 record) to the attached computer at full 192 kHz/24 bit.

 

[MC_RME]

USB 2 audio real-world limit is at about 70 channels (140 both ways) in single speed (48 kHz). And even that is not easy to achieve. 8 channels at 384 kHz equal 64 channels at 48 kHz. So obviously 8 channels at 768 kHz are impossible.

> 1024 bytes every 125 µs for Class 2 (USB 2.0)

There is a so called high bandwidth mode that allows to transfer 3 x 1024 bytes. Again in real world for audio we can use about 2 x 1024 - see above.

 

[MC_RME]

* P.s., I've found this fascinating reading because I was trying to figure out whether an RME Digiface USB could send 16 channels (8 play, 8 record) to the attached computer at full 192 kHz/24 bit.

The Digiface USB transmits 4 x stereo (SPDIF) or 4 x ADAT (2 channels at SMUX4) aka 8 channels, at 192 kHz, in each direction. So yes.

 

[catastrofe]

USB 2 audio real-world limit is at about 70 channels (140 both ways) in single speed (48 kHz). And even that is not easy to achieve. 8 channels at 384 kHz equal 64 channels at 48 kHz. So obviously 8 channels at 768 kHz are impossible.

> 1024 bytes every 125 µs for Class 2 (USB 2.0)

There is a so called high bandwidth mode that allows to transfer 3 x 1024 bytes. Again in real world for audio we can use about 2 x 1024 - see above.

Thank you.

 

[nomograf]

Contrary to RME, I would add, which to my knowledge and experience is flawless.

Agree here, my experience and that of the community I serve has been flawless