USB Audio Interfaces

Hi, Long time supporter and follower.
Ardour became unusable for me on the laptop I used to use, latency was atrocious - not ardours fault, more the crappy lappy I was using :slight_smile:
Getting more serious now, with a more serious machine getting set up almost ready for a live recording session.
I’ll be using my AudioBox 1818VSL again, my only concern is avoiding latency …

How accurate/current is the statement about USB audio here :

Do not plan to use USB audio interfaces if low latency is important to you

This statement still causes me concern. Is it still the case? or will it still be the case in 6

Thanks for a great daw!!


USB audio uses isochronous data transfer. It sends one packet every 1ms. That means USB always adds 2ms of latency. This is the same on USB 2, USB 3, USB 3.1 if that’s what it’s called – all of them are the same. This is why there are no USB 3 soundcards – USB 3 doesn’t reduce latency.

2ms isn’t perceptible but if you want low latency it’s something that could be done without.

If you notice latency then the problem is not your interface – it’s likely to be the configuration.

Live recording isn’t an application that needs low latency – avoiding Xruns is more important.

I can name two off the top of my head: Zoom UAC-2 and Steinberg UR816C :wink: I don’t disagree with your latency statement though and any extra bandwidth benefits seem pointless too for the typical channel counts of these interfaces.

OK, well I’ll take a chance and say they’re USB 2 with a ‘C’ connector, then look it up, which isn’t the recommended order, but let’s see. :grin:

I was thinking the same thing :slight_smile: There are plenty of devices with USB-C connectors (I think like the SD MixPre series) but running 2.0 under the hood. I would hope the ones I mentioned are different but marketing can be deceptive. I’m officially worried that you might be right…

EDIT (from Zoom website):

The USB 3.0 SuperSpeed transfer protocol supported by the UAC-2 is ten times faster than USB 2.0 and six times faster than FireWire 800. The result? Low-latency (2.2 ms roundtrip @ 96 kHz / 32 samples) for improved performance from your DAW software, and a stable data stream that is unaffected by computer jitter (slight variations in timing), so there’s no need to connect an external master clock source. What’s more, the extra bus voltage carried by USB 3.0 (900 mA, as opposed to the 500 mA carried by USB 2.0) allows for the use of higher-end electronic components such as the advanced preamps and converters in the UAC-2.

Lets look at that a little more carefully. (I was just looking at this myself)

The USB 3.0 SuperSpeed transfer protocol supported by the UAC-2 is ten times faster than USB 2.0 and six times faster than FireWire 800.

So what… this is a two channel device and there are already many 32 channel USB2.0 devices around. Speed is about throughput so no advantage in this case.

The result? Low-latency (2.2 ms roundtrip @ 96 kHz / 32 samples)

Actually, this statement is inaccurate. As discussed above, 2ms of that are just because USB.

for improved performance from your DAW software,

Ah no, actually both 2.2 ms latency and 96k SR will make the DAW work harder to remain stable and still be able to use many plugins. Lets call this misinformed rather than Lie though.

and a stable data stream that is unaffected by computer jitter

Nothing to do with USB 3 and everything to do with the internal clock in the audio device itself.

What’s more, the extra bus voltage carried by USB 3.0 (900 mA, as opposed to the 500 mA carried by USB 2.0) allows for the use of higher-end electronic components such as the advanced preamps and converters in the UAC-2.

a couple of things here, many USB 2.0 connectors allow as much as a 2 amp draw (2000mA) already. Also 900mA is not much when it comes to high quality preamp circuits. And 5 volts is already not enough for many. A good presamp has it’s own (Known quality) power supply. Computer power supplies are about as dirty as can be imagined… Did you notice the use of “higher voltage” when talking about current? can any of the rest of this statement be trusted?

Realistically the only advantage to of USB 3 over USB 2 is that “10 times” as many channels can be transferred at the same latency as USN2.0. If you hold your mouth right…


I’m pretty sure that a USB 3 device looks more impressive too. You never know what will get you the gig :wink: That said, you are right on most if not all of this, it seems. Marketing, smarketing. But what about the RME UFX+ and other RME USB 3 devices? According to an SOS review, the UFX+ cannot record or play back over MADI without the USB3 given the lack of bandwidth. There’s also a great article here describing a classical recording session with up to 64 channels routed via a single 50m MADI cable but here we are dealing with a USB 2 device? Hard to figure out therefore why the UFX+ needs USB 3 for the same MADI stuff…

I do not know what the upper limit is for USB2.0 in terms of channel count, but 64 has got to be close when all the overhead is added in. RME has a good reputation for stable operation (and no linux support) and may have chosen to make 64 channels on USB2.0 not work because it would not be stable. RME has never had a problem with telling the user they should upgrade their HW and the addition of a PCIe USB3 card is not expensive anyway (compared to RME anything). Why mess with making 64ch USB 2.0 work on whatever computer the user may have when USB3 is now mature and USB3 cards are cheap. I suspect RME would not make the mistake of calling 900mA a voltage as Zoom did in your quote above.

1 Like

I was lazy and just looked it up instead of doing the math…


At the theoretical maximum USB 2 bandwidth, you’d be able to record just over 40 tracks of 24-bit, 96kHz audio, while halving the sample rate to 48kHz would give you 80 tracks. Staying at 24-bit/48kHz, consider a more realistic real-world USB 2 bandwidth of 240Mbps (a slightly conservative figure, giving us plenty of overhead to allow for the connection limitations discussed earlier): you’d still have the ability to work with up to 40 channels of broadcast-quality audio simultaneously!

So given that RME often supports 96k it makes sense they limit their interfaces like this. So above in that 32-40 tracks range it the most you are likely to do with USB2 at least at 96k. 48k is a bit more open but the conservative numbers listed here are more likely to be real world due to overhead etc.

So if you are recording that many tracks at one time then yes USB3 may make sense. I would wager a guess most people don’t fall into that category. I am not sure what the driver situation looks like for USB3 audio honestly though. At that point I am typically tracking through Dante anyways (Sadly just not in Linux at this time).

1 Like

Ya, in this case this is a MADI interface. I am assuming a single MADI port, so 64 channels at 48k and 32 channels at 96K. That is the same band width for either 48k or 96k. However, I am realizing that the setup for recording 64 channels is being compared to the RME card which is an I/O card. That is 128 channels at 48K and 64 channels at 96k. So I think it is a case of record only vs. record and playback at the same time.

Technically, today’s USB interfaces (>= 2.0) have a penalty of about 500 microseconds compared to a PCIe/Thunderbolt interface. Any additional latency is due to driver design.

A two channel USB 3 interface is pure marketing. USB 1.1 can do two channels!

MADI has a bandwidth of 100Mbit/s – theoretically USB 2 could do four channels of MADI, or two using the more conservative 240Mbit/s figure. It’s not to do with bandwidth then --it’s the maximum size of a packet that can be sent in isochronous mode. If I’ve got that right it means isochronous transfer mode ‘wastes’ a lot of bandwidth.

The RME MADIface does something more sophisticated which implies a proprietary driver which implies it won’t work with Linux, and indeed the MADIface doesn’t apper in the ALSA soundcard matrix.

MADI has a bandwidth of 100M, yes, but most of the single MADI interfaces actually have two connectors on them, one for each direction. So 100M transmit and 100M receive. The USB path has to deal with 200M so in most people’s mind a 64 channel interface is 200m bandwidth of signal at 48k/24bits. 64 channel i/o at 96k/24bits would be 400M bandwidth (four MADI connectors). I’m not really sure which actual MADI interface we are talking about here :slight_smile:
(are any of us?)

@bachstudies mentioned a USB 2 interface that can do MADI above – I think it’s the RME MADIface USB :

That’s more channels than the SOS article suggests USB 2 can handle, so that means the MADIface isn’t using isochronous transfer.

Getting back to the point I was making…there are USB 3 interfaces that actually use USB 3 technology and not just some USB-C connector running at USB 2.0 :wink: Whether that is pointless (i.e. marketing) I’m not arguing about.

As per the point @lenovens made, RME seems to have limited certain high MADI channel counts to USB 3.0 at this point, despite USB 2.0’s ability to handle them in RME’s prior devices. Odd, but true. I’m still drawn in by Zoom’s claim that 900mA allows for higher quality electronic components. I was under the impression that this particular interface used the same preamps as the Zoom F8…

How would that work? The higher current is for charging phones quickly I would think :slight_smile:

1 Like

There is some difference from the web page to the manual. The web page does say 128 channels right off and the device does have two MADI in and two MADI out. The web page does say 96k and 192k but does not mention that the channel count is reduced to do so. However, the USB setup only deals with one input and one output (they do not have to be the same set) at a time. The manual says that at double speed (96k) there are 32 i/o and at quad speed there are 16i/o. It is possible to set up one pair of MADI i/o connectors as a pass-through… or pass-through the internal mixer. That is the device will handle a total of 256 channels (128i/o) at 48k but will only send and accept up to 64 i/o on the USB at 48k (32i/o at 96k and 32 at 192k). From the manual:

"The MADIface allows the use of sample rates up to 192 kHz via the MADI interface. For this to work single-channel data is spread to two or four channels using the Sample Multiplexing technique. Therefore the number of channels is reduced to 32 or 16 respectively. "

So to answer our earlier question, the USB link needs to deal with 200mbit bandwidth plus overhead. I do not know if the audio is sent over USB as 24bit int or 32 bit floats or 32bitAES3 (which has other data piggy backed on top). I suspect that in any case the extra data is not lost and so AES3 would be my guess.


To get back on topic and help out the OP the ‘System Requirements’ page does look kind of old – there are references to kernel 2.6.

With USB 2 @Drumfix has clarified that the latency is 0.5ms which is not noticable, but not as good as a PCIe card.

USB 3 will not improve latency. The legitimate use of USB 3 is for high channel counts and/or high sample rates.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.