Hi all,
I am curious what Audio interfaces people are using and what latency they are able to use reliably? I have an Audient id22 and an older Native Instruments Kontrol One. The id22 is my home unit and uses USB 2, has excellent AD/DA converters and as a result has very good sound. The input preamps are from the audient console channel strip design and provide great audio results for microphones and built in DI circuit make it an excellent unit for guitar and bass input. A few drawbacks are that it doesn’t perform well at sample buffers below 128 “samples” (but works really well at 256 “samples” which is what they recommend for Synthogy Ivory) and it requires a wall wart for operation. The Kontrol One; though older, has A USB 2 powered design, is more portable and runs very well at lower latency (64, 128 “samples”). It has been my workhorse on gigs for 4 years and is still working well. It doesn’t have the quality of the Audient but is very bright, stable and great for live work. I beleive that some of what prevents the id22 from achieving the lower latencies revolves around the Windows drivers for the interface. It looks like the drivers for the Native Instruments product have better written drivers in this case. I run Win 7 SP1 both laptop (i5) and PC(i7). Please share what you use and latency you achieve as well as any advice on new or upcoming interfaces you have spotted.
I’m not familiar with the Audient you mention, but have you tried using ASIO4ALL with it? For devices with poorly written asio drivers I’ve found that sometimes asio4all works better - more stable and sometimes able to get lower latency.
Also, about latency… I think many people get overly hung up on this. I’ve written about this in my mythical “Glitch Free” book, but since it’s not available yet here’s my thoughts on it:
Most people can’t distinguish sound differences less than about 10 milliseconds. ie: two sounds 10ms apart sound instantaneous.
The smaller you make the audio buffer, the more risk of a glitch and the harder the machine has to work to keep up.
So, why push for a smaller latency when you typically can’t hear it and it just increases load.
In general, I recommend sample rate 44.1Khz and buffer size of 256 - that works out just under 6 ms and gives some headroom for small stalls in the audio processing. If your machine can’t cope with that, you can generally push it out to 512, or if your machine is coping fine drop it down to 128.
There are always exceptions though… if your running from you host through other gear which also introduces latency, then you might need to squeeze down everything in the chain to make it acceptable.
Similarly I think some people also overly concerned about sample rate. Unless you’re running some very long chains of effects plugins the benefit is negligible (if anything) and just increases the processing load.
Of course this is just my opinion and not everyone agrees.
PS: The correct units for buffer sizes are “samples” not “Kb”. (unless you really mean Kb, in which case it’s no wonder you have latency issues )
Oops, better get that straightened . I agree that 256 “samples” works fine for me and I’m used to it after playing keyboards a great deal that way. I got interested in lowering it after some discussion on the S-Gear site regarding guitar to speaker response times. A number of the posters said the latency affected the way they played and that shorter delay was better. So I was just testing the edges I guess. Audient is pretty new to interface game they have id22 (needs wall wart more channels) and id14 (fully usb 2 powered), their converters are both Burr-Brown.See them here . Thanks for the feedback and advice. I think i’ll dial back the workload on the laptop by switching to larger buffer.
I’m using RME interfaces: on my PC it is an HDSP Multiface, on my laptop a Babyface does the job. CPUs are performant quad cores.
Frequncy is always 44,1kHz. I could easily use 64 samples on both but as 128 feel very well I stay there which gives me some additional cpu power if necessary.
When using audio inputs 128 samples are just the maximum I’d accept (you have twice the buffer size then). 256 samples feel like playing in chewing gum for me.
When I worked at Yamaha we did some research and testing of this, and you’re right - only something like 10% of people (from a mixed sample of musicians and non-musicians) are capable of discerning lower than 10ms, with the average point where they could discern a delay being something like 30ms. Some were as high as 50-60ms, surprisingly!
Actually I should make clear, this was a test to determine at what point people could detect latency, from hitting a key to hearing the sound, rather than just hearing two sounds. The results we saw tended to show that musicians, and in particular keyboard players, tended to have a lower tolerance for latency, presumably because they were accustomed to the feel/delay of low-latency electronic keyboards.
Perhaps the figure for detecting a delay between two sounds is more consistent amongst people, driven more by biology than experience.
I have tried 512 samples, it’s a no go because I can really feel the delay between that and 256 samples. I can actually feel the response difference between 128 samples (I use this on my laptop) and 256 samples but it is still playable. I read somewhere that the mind compensates and adjusts to very small time differences so that might be why I don’t notice as much after playing a while at the higher sample buffer (256 samples). Anyway a person who also plays a real piano could attest to the fact that the actions and response varies from piano to piano and playing one a while helps to get used to it.
I’m curious what the “latency” is on a standard piano?
I have also read that RME makes the best low latency and most reliable windows drivers, I’ll have to put the “babyface pro” on my wish list . I’ve not heard of “performant” quad cores. Great stuff guys
yes, RME interfaces are really good. F.e. multiface has extremly low latencies (down to 32 samples) and hosts can be treated to high cpu loads without crackles and the like.
The other aspect is service. As far as I remember I purchased multiface in 2001 - latest driver and app updates were released some weeks ago without any fees or the like. They keep drivers up to date for new OS without rumors and upgrade applications even after years.
All this combined with stable drivers and solid audio hardware makes their products reliable workhorses.
Concerning “performant quadcores”: sorry my mistake: performant is a german expression meaning “with good performance”. What I simply wanted to express was the cpus were really some of the faster ones when I purchased them a year ago.
Wow, that’s some high praise for RME… and I’ll back it up with this: in all the years of supporting Cantabile, I don’t think I’ve ever had a crash report mentioning RME drivers, nor a single user having trouble with one of their devices.
RME Babyface here as well - rock solid and minimal latencies. Also, I like the internal mixer - convenient for some applications (although not needed at all for what I’m doing with Cantabile). The only thing not to like about it is the fact that its audio and midi ports need an external breakout cable, which is just one more thing to possible go wrong live…
As a backup solution, I’ve just ordered a Presonus AudioBox - specs and price look good (and its ports are all directly on the box!). I’ll report back once I’ve tested it!
Essentially, my setup is pretty simple: MIDI In for main master keyboard, Audio In for guitar, MIDI out to VoiceLive, Audio Out to mixer. So I need a solid but simple box with 2/2 audio and MIDI In/Out.
BTW: Once I’ve finalized setting everything up, I’ll give a breakdown of my (by now pretty complex) C3 setup in the Testimonials section - may be useful for others
MOTU UltraLite Mk3 here, giving me 8 analogue inputs and 8 analogue outputs - works great for me running with a 192 sample buffer (it also runs fine at 128, but I prefer not to push it for gigs). No problems, clear sound, works well!
Here’s an update on the Presonus experiment: Just received and tested the AudioBox USB - nice basic mobile interface with a good feature set: two audio ins / outs, MIDI pair, good gain/volume handling with physical knobs (not just a software dialog window) and good sound quality.
Unfortunately, the latency performance is not quite as I need it to be - I get crackles and noise at 128 samples, even with a pretty powerful laptop (i7 quad core). At 256 samples, performance is good, but I don’t really want to go to 256 for live, especially playing guitar through an amp plugin with the added input latency.
So I pulled out my old DJ interface (Native Instruments Komplete Audio 6) and was pleased to find that its perfomance is perfectly solid at 128 samples! So now the Komplete Audio is my backup interface, with the Babyface as primary. And the Presonus box goes back to big T…
I simply use a Peavey USB P at 512 samples running VB3, Hybrid 3, NI Stuff (FM8 etc), and The Grand 3. Plus Sylenth and I run it all through L1 Limiter from Waves as the final out. I chose the USB P for live because it is cheap and has an XLR output so I have zero noise. My studio has much more serious hardware for audio. I am primarily a B3 player and latency is not an issue for me. Like Neil said, even up to 40 to 60 ms is okay for some. I am at 12ms. My only issue is CPU load. ASIO4ALL helps with that. Brad explained very well his measurement of CPU load. Makes be respect the software even more.
As an amusing side-comment, one person taking part in the latency acceptability tests we did was quite happy with something like a quarter of a second latency!! It turned out he also played organ in a cathedral, so was used to there being a substantial time delay between hitting a key, and hearing sound from pipes a reasonable distance away
That is funny! I am not much of a piano player. I would think in that case latency would be an issue. Synth pads and sounds like that are also forgiving. I plan to buy an audio interface that can handle more load, though. I am using 3rd Gen i7 with 8gb ram, and I hit 80 percent very often. Mostly on pads.
Surprising - I’m using a 4th Gen i7 with 16Gb RAM, and hardly ever get above 30% with Cantabile 3, with a 192 sample buffer, including with a bunch of separate instances of Omnisphere, Ivory II etc. That goes out of the window if I have WiFi or Bluetooth enabled though - the average load goes up and I start to get nasty load spikes.
I’m using the Focusrite Scarlett 2i2. It doesn’t offer the bells and whistles or number of outputs of some of the others, but is good enough for my purposes.
I run it at 44100 Hz and 256 samples. (I would like to use 48000, but one of my plugins has a bug that makes it out of tune at anything other than 44100.) I could probably go lower than 256 samples, but the 5-6ms latency is fine with me. I think I can start to notice the delay above about 10-15ms, although I haven’t done a formal experiment.
Hi,
I like the look of the Peavey USB P for a pure output interface. I looked at it online but they didn’t give the operating modes: i.e. Available sample rates and buffer settings. Do you know these specs? I had honestly never seen one before. Thanks for posting about this!
)
. I agree that 256 “samples” works fine for me and I’m used to it after playing keyboards a great deal that way. I got interested in lowering it after some discussion on the S-Gear site regarding guitar to speaker response times. A number of the posters said the latency affected the way they played and that shorter delay was better. So I was just testing the edges I guess. Audient is pretty new to interface game they have id22 (needs wall wart more channels) and id14 (fully usb 2 powered), their converters are both Burr-Brown.See them 
