Aggregate audio device on Windows 10 64 bit?

[Ken P]

I'm moving back to Windows for my next machine, almost certainly coming from Scan following the recommendations on the forum.

Is it possible to use multiple audio devices simultaneously on windows these days? I use two, and sometimes 3 USB audio devices on a mac laptop at the moment by using what mac calls an aggregate audio device and it works very well. I recall struggling with this on W7 - is there a solution in W10 or with a third party application?

I'm using Ableton Live and just getting to grips with Bitwig Studio but both applications only allow me to select one audio device.

[CS70]

No, in general you can’t. Most interfaces still use ASIO which allows for one device at a time. You can have multiple devices connected but use only one at a time for audo. So if you need many inputs, you need an interface with many inputs, or create stems via a mixer at analogue level, or switch interfaces at DAW level.

Note that for non audio use, multiple interfaces work fine, for example I have an Apollo twin that I use to run UAD plugins on the audio processed by the RME

[CS70]

One more thing - there’s nothing in the hardware or OS that prevents aggregate devices, it’s just not something people particularly demand I guess.

[N i g e l]

Im using Sonar X1 & CbB on Windows 10.

I cant run 2 ASIO sound cards in my DAW but I can run my DAW & watch youtube at the same time on the same sound card, so it looks like theres some kind of sharing scheme going on.

[James Perrett]

If you can put up with a little more latency you can use a different driver model - the standard Windows drivers allow more than one interface although software often limits you to using one (or separate ones for input and output). ASIO4All will, in some cases, also support multiple drivers if your software doesn't.

[N i g e l]

Windows drivers allow more than one interface although software often limits you to using one

yes, Cakewalk only allows for selecting one driver model. Then there is another menu to select which card to use, of which you can only select one.

I have used WDM/ks sucessfully in the past [latency < 5ms] but im using ASIO now on a Behinger box.

Traditionally ASIO was one device only so i was surprised I could Youtube & DAW on the same card.

I understand that ASIO4ALL is just a bridge in series to WDM/ks drivers for devices that dont have ASIO ?

Theres probably a Martin Walker article about this somewhere !

[CS70]

Im using Sonar X1 & CbB on Windows 10.

I cant run 2 ASIO sound cards in my DAW but I can run my DAW & watch youtube at the same time on the same sound card, so it looks like theres some kind of sharing scheme going on.

That's because they are two different things.

The ASIO driver provides a direct access to the hardware, hence minimizing latency in recording. Basically an ASIO driver keeps filling a memory buffer with what's coming into the inputs (read from the hardware), and an application (for example a DAW) can decide to read that buffer and memorize the contents (i.e. record) - or fill the buffer for the interface to play back to its audio outputs (i.e. play).

The ASIO buffer exists independently and in parallel with the standard Windows audio subsystem which in many cases is not the most direct access to the hardware, and introduces unacceptable latency in recording.

When you're using the DAW with ASIO and using your browser (which has to work with any sound hardware so uses the Windows subsystem) at the same time, you're using these two independent subsystems - they both read from the same hardware so - as long you have enough computational resources to do so - you are able to play them together.

With WASAPI is possible to have a driver which talks directly to the hardware as ASIO does.. but most interface don't yet come with a suitable WASAPI driver.

[CS70]

I understand that ASIO4ALL is just a bridge in series to WDM/ks drivers for devices that dont have ASIO ?

Yes, Asio4All simply expose the kernel streaming into a buffer that conforms to ASIO specs - basically allowing them to be accessed via ASIO api calls.

It's all nice for playback, but obviously doesn't help much for recording as you will get all the latency of kernel streaming plus a little something due to the wrapping itself.

[N i g e l]

Thank you, That all seems very reasonable and makes sense.

The other issue mentioned was setting up Windows for audio. I think all I ve done so far in W10 was turn off the beeps and set "Optimise for background processes" & "optimise for performance".

[openaspace]

I'm really disappointed...

I used dante on osx to receive signals from the main workstation (32ch) to my laptop and make myself sound when playing drums... and using the integrated soundcard for headphones..

how is possible that in windows it's not possible??!!!

In the daw I can see only the dante audio interface without any possibility to listen what is arriving in the laptop

[Tony Carpenter]

Here I was about to make the choice not to buy the new Mac Pro. I can get a pro audio PC, right?, wrong!. The useless audio support in Windoze continues... sad, truly. I was really ready to save a bunch of money and live with having a special setup user for recording. Oh well, guess we'll be waiting still for Mickeysoft to realise modern hardware can actually handle talking to more than one input device in the audio world.

[James Perrett]

Oh well, guess we'll be waiting still for Mickeysoft to realise modern hardware can actually handle talking to more than one input device in the audio world.

I think Microsoft have the technology with Kernel Streaming but audio interface manufacturers don't support it as ASIO is better established. Blame Steinberg for the one device limitation as they're the people who created ASIO.

[Tony Carpenter]

Oh well, guess we'll be waiting still for Mickeysoft to realise modern hardware can actually handle talking to more than one input device in the audio world.

I think Microsoft have the technology with Kernel Streaming but audio interface manufacturers don't support it as ASIO is better established. Blame Steinberg for the one device limitation as they're the people who created ASIO.

I’ll gladly place the blame where needed. The point remains.. can’t be done!. I’m serious, sad it’s still a thing.

Tony

[The Elf]

I've been similarly annoyed about this for a loooong time. I was once told by a certain company... "watch this space". I'd be curled up in a corner gibbering by now if I was still watching that space.

I think technology is set to overtake this requirement now. If network audio does properly take off (and it is getting some traction, though it's still far too expensive as yet), then adding an audio device to a network is trivially simple - and requires no extra drivers. If I were to place money on it I'd say this is the way we are going - and ASIO et al will be a distant memory.

[resistorman]

Don’t let the door hit you on the way out, ASIO....

[Hugh Robjohns]

I agree. AoIP is quickly becoming the de facto technology in new professional installations in broadcast, live sound, theatre and education. The more common it becomes, the lower the costs will move. It's very noticeable how much kit is crossing my desk now with Dante, Ravenna or AES67 connectivity.

And as the Elf says, multiple I/O boxes are no problem at all with AoIP systems, so I think USB and ASIO interfaces will probably be quite rare within decade, at least in new-build systems.

[The Elf]

Don’t let the door hit you on the way out, ASIO....

I'm not so cruel! ASIO may not be perfect, but it's brought us a long, long way. Think of it as a Pony Express horse - it was never mean to take us the whole journey...

[resistorman]

Don’t let the door hit you on the way out, ASIO....

I'm not so cruel! ASIO may not be perfect, but it's brought us a long, long way. Think of it as a Pony Express horse - it was never mean to take us the whole journey...

True, but it may also be responsible for 33 1/3 of my white hairs

[Pete Kaine]

True, but it may also be responsible for 33 1/3 of my white hairs

I'll see your white hairs and raise you one bald head.

[Johnsy]

I agree. AoIP is quickly becoming the de facto technology in new professional installations in broadcast, live sound, theatre and education. The more common it becomes, the lower the costs will move. It's very noticeable how much kit is crossing my desk now with Dante, Ravenna or AES67 connectivity.

And as the Elf says, multiple I/O boxes are no problem at all with AoIP systems, so I think USB and ASIO interfaces will probably be quite rare within decade, at least in new-build systems.

Has absolutely nothing to do with drivers though. The Dante PCIe card is an ASIO device. The network paradigm solves the the multi-device issue for ASIO, rather than supplanting it.

Re aggregate devices under macOS: unless those devices are synchronised in hardware (i.e., wordclock or digital stream daisychain), then you're getting sample rate-converted (rather than bit-perfect) audio (and SRC done in software, to boot). Doesn't matter to some, does to others.

[The Elf]

Has absolutely nothing to do with drivers though. The Dante PCIe card is an ASIO device. The network paradigm solves the the multi-device issue for ASIO, rather than supplanting it.

Yep, true enough. I had forgotten the link between Dante and DAW! Though it wouldn't surprise me to see a new standard to supplant ASIO and bring Dante channels into a DAW more directly. I would love to think of the likes of Audinate and Steinberg beginning to have some serious talks along these lines - and let's not forget that Yamaha are very much in the Dante vanguard...

Don't forget that you don't need a Dante 'card' - the software only virtual 'card' works really well IME. You just need at least one piece of Dante-enabled hardware to begin to make use (and sense!) of it.

[Johnsy]

Don't forget that you don't need a Dante 'card' - the software only virtual 'card' works really well IME. You just need at least one piece of Dante-enabled hardware to begin to make use (and sense!) of it.

Sure - but the VSC's an ASIO device too (albeit a virtual one).

And audio I/O really doesn't get much more 'direct' than ASIO.

[ZapZap]

ASIO4ALL allows recording from multiple devices simultaneously, but I'm not sure if it does sample rate conversion to correct for their clock drift:

“Support multiple sample rates concurrently - for as long as they are physically derived from a common master clock, are integer multiples of each other (except 44.1/48kHz special case), etc...
Support a number of sample rates on the ASIO side that the audio device does not physically support. Basically, this means an extension to the on-the-fly rate conversion capabilities.”

“ASIO4ALL now does sample rate conversion whenever required in order to support 44.1 kHz.”

“Support for single common sample rate that is *not* a multiple of 8 or 11.025”

[The Elf]

ASIO4ALL allows recording from multiple devices simultaneously

I've seen this claimed many times, but I've never met anyone who has got it to work?

[CS70]

ASIO4ALL allows recording from multiple devices simultaneously, but I'm not sure if it does sample rate conversion to correct for their clock drift:

“Support multiple sample rates concurrently - for as long as they are physically derived from a common master clock, are integer multiples of each other (except 44.1/48kHz special case), etc...
Support a number of sample rates on the ASIO side that the audio device does not physically support. Basically, this means an extension to the on-the-fly rate conversion capabilities.”

“ASIO4ALL now does sample rate conversion whenever required in order to support 44.1 kHz.”

“Support for single common sample rate that is *not* a multiple of 8 or 11.025”

If you understand what ASIO4All is (if not, read my blog post at https://www.theaudioblog.org/post/should-i-use-asio4all), you understand why in theory it can aggregate devices.. because Windows WDM can aggregate devices, by mixing the sound.

But in practice, these devices will have no reference clock because there are no provisions for that in the WDM architecture, so they will either be out of whack at start or soon get out of whack. There are techniques to try to synchronize them (for example using the USB frequency) and I am sure that Asio4All attempts this, but while it works fine for loose applications, for strict audio samples ordering, anything less than an accurate timestamp on each sample buffer, synched to a common clock will generally not do, at last not for enough time.

Note that an ASIO driver theoretically could support concurrent devices - ASIO2.3 added CanTimeInfo and CanTimeCode queries to the driver and also allowed for internal buffering so it'd be possible to write an aggregate wrapper. But this would requires manufacturers to do a little work (implementing timecode info in their drivers) and someone to write the aggregate, and there's so little demand (and need, really, with inexpensive multi-input interfaces and audio over internet for larger things) that nobody has great interest in doing it.