Soundcard Latency measurements- Firestudio, Soundblaster, GX

[Pickasso]

I've been through several sound cards, trying to achieve the best recordings. (buying/selling, etc)
The 2 main criteria which seem to affect D.I. recording quality are preamp sound quality, & audio latency when record monitoring using ampsims/cabs. If you can't record & have it playback TIGHT, it's never going to sound pro. The lower the glitchfree latency, the more accurate you can play in time, & the more likely the DAW will play it back in time. One problem is the driver settings for different units often use different terms for their settings, some are samples, some are milliseconds, so how do you compare? Wish I had tested all the units I've owned, but just found this free latency tester utility:
http://centrance.com/downloads/ltu/

The method to test is to run a patch cable from the interfaces output to the input, the utility plays a short signal, and measures the roundtrip audio latency delay.

Test results using the CEntrance Latency tester:
I Set the soundcards drivers to the lowest glitch free record monitoring settings. Example- a Reaper session running 2 basic guitar tracks w/fx (one set for record monitoring), EZ drummer, & no other fx in the session anywhere to slow it down. Guitar tracks were noise gate/loner or legion/boogex, no other fx on. Then ran the latency tester, using that same buffer size/latency setting.

Presonus Firestudio Mobile, '128 samples buffer size':
32000 hz sample rate = 431 samples/13.47 ms
44100= 455/10.32 ms
48000= 463/9.65 ms
96000= 590/6.15 ms


Creative Sound Blaster X-Fi pci-e card, '5 ms ASIO Buffer Latency':
32000= N/A
44100= 899/20.39 ms
48000= 923/19.23 ms
96000= 1863/19.41 ms


Line 6 GX, 'extra small/128 Samples buffer':
32000= N/A
44100= 852/19.32 ms
48000= 869/18.10 ms
96000= N/A


At least on my system, the Firewire device outperforms PCI-e & USB. In case you want to compare computer specs:
W7 64 bit PC, AMD 2.9 ghz, 6 gb ram

Hope this info helps someone.
Anyone care to test their cards?

[michalkaszczyszyn]

Hi,

I'm also interested in latency measurements.

http://www.michalkaszczyszyn.com/en/tutorials.html - here are reviews of more than 10 audio interfaces with latency measurement results

Centrance Latency Test Utility is accurate only within 0.5ms. Latency isn't always stable - in some interfaces it drifts (latency jitter). For example - M-AUDIO Firewire 410 has a latency jitter of 0.23ms (10 samples @ 44.1kHz).

There is a more accurate method of measuring latency, that is well described in Cockos Wiki:
http://wiki.cockos.com/wiki/index.php/H ... ce_Latency

Using your DAW gives better results as you can signal levels properly and measure with an accuracy of a single sample.

Line 6 POD Studio - on a PC the smallest actual latency is around 15.4ms (WIndows 7 32-bit), on a Mac it's around 7ms... BUT these interfaces can do much more - when you're playing through POD Farm or GearBox standalone you get only about 7ms of latency (300 samples @ 44.1kHz). You can measure this by using another audio interface with known round trip latency - launch the POD Farm standalone app, set up a tone with not too much gain and then measure that other interface latency in with the POD Studio between it's input and output.

Devices such as Line 6 POD XT (2.65ms), POD X3 (1.5ms) and Fractal Audio Axe FX (1ms, tone processing times) are superior to typical audio interfaces as they're equipped with their own processors, memory and they do the job instantly for monitoring while working at a large buffer size with DAW.

I agree that the lower the latency, the better and tighter the groove is. There's another cool thing that you can do - use DI-BOX/splitter (Waves/PRS GTR interface, LittleLabs RedEye, Behringer GIto feed your guitar signal to both recording interface and an analog amp. It's something that I'm experimenting nowadays.

It's important to have a total latency of less than 12ms, but you also have to take into account that sound waves travel through air at the speed of 340 meters per second, so every meter of distance between your ears and your speakers adds 2.941ms of latency.

Anyway here's a brief list of my results (44.1kHz):

Apple Mac mini 2011 built-in audio: 1.904ms (OS X Lion)
ASUS K50IJ built-in audio (VT1708S): 3.7ms (Windows 7)
ESI Maya44 XTe PCIe: 9.932ms (Windows 7)
ESI UGM96: 5.9ms (OS X Lion), 8.7ms (Windows 7)
E-MU 0202 USB: 10.4ms (Windows 7)
E-MU 0204 USB: 9.455ms (Windows 7)
E-MU 0404 PCIe: 6.2ms (Windows 7)
E-MU 1616M PCIe: 6.077ms (Windows 7)
Focusrite Saffire PRO 14, 3.741ms (Windows 7), 5.011ms (OS X Lion), MixControl 3.0
Focusrite Saffire PRO 24, 4.399ms (Windows 7), 5.714ms (OS X Lion), MixControl 3.0
Line 6 POD Studio: 15.4ms for DAW, 6.8ms for monitoring through the POD Farm/GearBox standalone
Line 6 POD XT: 2.65ms hardware monitoring through XT, more than 12ms for DAW
Line 6 POD X3: 1.5ms for hardware monitoring through X3, more than 12ms for DAW
M-Audio Firewire 410: 6.077ms (14 samples, Mac OS X Lion)
Peresonus Firestudio Mobile: 4.739ms (Windows 7), 6.077ms (14 samples, Mac OS X Lion), Universal Control 1.6
TC Electronic Konnekt 6: 4.693ms (Windows 7), 6.031ms (Mac OS X Lion), TC Near 3.0 f22

Not all of these values are easy to achieve and stable, read more on my website:
http://www.michalkaszczyszyn.com/en/tut ... l#database

Regards

[Pickasso]

Michal,

Thanks- That's some great info, as well as on your website- great stuff. I'll have to take some time to digest this & try your tip from the Reaper wiki!

[Pickasso]

I agree that the lower the latency, the better and tighter the groove is. There's another cool thing that you can do - use DI-BOX/splitter (Waves/PRS GTR interface, LittleLabs RedEye, Behringer GIto feed your guitar signal to both recording interface and an analog amp. It's something that I'm experimenting nowadays.

Brilliant! That's it- that plus the latency adjustment is the solution. Noise is not an option for me here, so I'll attempt the split signal with a distortion pedal, instead of an amp. Thanks for the tips!

[michalkaszczyszyn]

Thanks, I'm glad that you find these ideas interesting.

I happen to have the WAVES/PRS GTR, it does the job with very little noise.

I'd like to try out the Redeye splitter and other Di-Boxes/splitters.

Here's a link to a cool Ultimate Metal Forum thread with dry guitar tracks and a backing track of a Killswitch song:
http://www.ultimatemetal.com/forum/rate ... cking.html

User JeffTD recorded four rhytm tracks using RME Fireface 800, the Redeye and a POD for monitoring. It sounds nice and tight.

If you discover something interesting, please let me know.

Regards!

[michalkaszczyszyn]

Apple Mac mini's built-in audio round trip latency measurement results. Aggregate Device was created using Built-in input and Built-in output. Choosing seperate input and output devices increases the latency by 0.5 to few miliseconds depending on the buffer size.

@128:
7.233ms (319 samples)

@64:
4.308ms (190 samples)

@32:
2.811ms (124 samples), very usable on an i5-2410M (2.3GHz) with several tracks loaded with Overloud TH2

@14:
1.904ms (84 samples), some occasional pops and cracks in sound

This is the lowest actual round trip latency that I've heard of. It sounds great even with Behringer ULTRA-DI DI100.

[michalkaszczyszyn]

I just got a Presonus Firestudio Mobile, under Mac OS X Lion it works great with RTL of 6.077ms (14 samples buffer, Low Latency Mode, 44.1kHz). I updated my RTL database :
http://www.michalkaszczyszyn.com/en/tut ... l#database

[michalkaszczyszyn]

Focusrite Saffire PRO 14, Mac OS X Lion, 5.011ms RTL, good performance, 14 samples buffer, Firewire Driver Latency set to Short mode, 44.1kHz. Updated RTL database :
http://www.michalkaszczyszyn.com/en/tut ... l#database

[michalkaszczyszyn]

Finally I got a chance to test some firewire interfaces under Windows OS. Here are the RTL results:

Focusrite Saffire PRO 14, 3.741ms (Windows 7, 32 samples), 5.011ms (OS X Lion, 14 samples), MixControl 3.0
Focusrite Saffire PRO 24, 4.399ms (Windows 7, 32 samples), 5.714ms (OS X Lion, 14 samples), MixControl 3.0
Presonus Firestudio Mobile: 4.739ms (Windows 7, 32 samples), 6.077ms (Mac OS X Lion, 14 samples), Universal Control 1.6
TC Electronic Konnekt 6: 4.693ms (Windows 7, 32 samples), 6.031ms (Mac OS X Lion, 14 samples), TC Near 3.0 f22

It appears that devices that use TC Electronic Dice chip offer better latency under Windows OS!

Regards