[LAU] Affordable well-working USB interface with low latency at 48kHz?
dak at gnu.org
Thu Dec 17 15:08:54 CET 2020
David Kastrup <dak at gnu.org> writes:
> Sam Kuper <sampablokuper at posteo.net> writes:
>> Apparently they are USB class-compliant, with low latency & reasonable
>> build quality. Audio quality reportedly close enough to Focusrite for
>> most uses. Here are some of my bookmarks:
> That does not match my prejudices :)
Back to my prejudices: just got an UMC22. Its codec is a Burr-Brown
(TI) PCM2902. Looking up its specs, it's USB1.1 full speed and sports
• Stereo ADC
– Analog Performance at V BUS = 5 V
– THD+N = 0.01%
– SNR = 89 dB
– Dynamic Range = 89 dB
– Decimation Digital Filter
– Pass-Band Ripple = ±0.05 dB
– Stop-Band Attenuation = 65 dB
– Single-Ended Voltage Input
– Antialiasing Filter Included
– Digital LCF Included
"Not recommended for new designs" (currently the card is being sold new
everywhere), an USB1.1 (and USB is half-duplex, so this is certainly
responsible for part of the latency) 48kHz design with a 16-bit ADC
specified for slightly above 14bit of precision.
And that's the naked codec, without being impacted by any circuitry
It's more compact for sure than my current solution at one ensemble
member (mic input into a Mackie Onyx 400F with dead digital circuitry,
out via the plugin jack into the mic input of a laptop).
For -p64 (without --sync) I get
464.320 frames 9.673 ms total roundtrip latency
extra loopback latency: 272 frames
use 136 for the backend arguments -I and -O
(the -O and -I values are independent of -pxx I think).
For comparison: with my Alesis iO|14 (Firewire, a lot more inputs and
outputs) I get 81 for the backend arguments regarding additional
loopback latency. With a Hammerfall DSP (PCMCIA card via expresscard
adapter, so routed through a PCI to PCIx converter) it's 31 if I
So even in the latency department, I am not really impressed though
given the tests I've seen, this does not seem too uncommon for USB
cards. Though there are plenty by now that are at least USB2.0 and can
consequently, for example, do 96kHz at full-duplex (full-duplex audio,
the USB is still half-duplex).
> For my particular use case, sound quality and to some degree noise
> floor are secondary considerations compared to standard compliance,
> reliability, latency. While I am personally more of an audio quality
> buff with a focus on recording, if the U-Phoria supposedly checks the
> boxes I need, that would be the pricing on the cake.
It's most definitely not what I'd use for recording (and with a single
mic input, it's not overly suited for that either). For jamming, it
might fit the bill if you need a phantom power input. Not much
justification for it beyond that I'd say over basically anything cheap
with a line input to be had these days.
> I need to ask the exact model of Behringer that my one orchestra
> colleague got. That one's performance (regarding latency/dropouts)
> did not really appear to improve upon my stone-age Tascam US-122L
> brick (which you can throw basically in any bag without worrying about
> anything breaking off). Maybe something related to drivers.
I think it might be that model but I haven't asked back yet.
More information about the Linux-audio-user