Low Latency Monitoring with HD Native

[David Sandwisch]

Today I was using HD Native with an Omni I/O for the first time on a Foley stage and I was having problems with latency in the monitoring chain. I'm a very experienced PT HD user who owns the old DSP cards and have never had issues with latency on my system. I turned the low latency monitoring option on and off but it didn't seem to make a difference.

Does HD Native have more latency than HDX or the old HD core/process card systems? It was annoying but workable. I was running a session of off a USB 2 drive which was not my preference but the only way to work on this system at the time.

It's a mac tower (not sure exact model but it's only a couple months old and running the latest OS.)

There's nothing after the I/O in the B-chain to induce latency, it's all analog and very simple.

Thoughts? How do you get HD Native to have 0 or really close to 0 latency so you can record with it without driving the talent nuts? I did not try adjusting any playback engine settings but they were set to the defaults for everything. It's a very powerful compute so I doubt the CPU is even taxed at all.

Thanks,
David

[Craig F]

http://www.avid.com/US/products/Pro-Tools-HD-Native
Then Compare Latency link
Native is higher then DSP but less than FW or USB

[chrisdee]

Originally Posted by David Sandwisch

How do you get HD Native to have 0 or really close to 0 latency so you can record with it without driving the talent nuts?

Have you tried using direct monitoring?

[BobbyDazzler]

Were the outputs assigned to the correct Low Latency Output pair?
If you get it right, LLM on HDNative is extremely quick from input to output.

[Brandonx1]

What version of pt where you using? What was the hardware buffer set to? Any plug ins in the recording chain? When was the foley artist where headphones? The only time I've every had a foley artist where headphones is the cloth pass.

The way to get hd native to me really low latency, you need to set the playback hardware buffer really low. 128 or 64 samples.

[David Sandwisch]

That's how this artist likes to work so that's what we do. It's protools hd native 10.3.6 on a Mac running the latest os. No plugins in the session. The hardware buffer was set to the default setting which was 512k I believe.

I've never heard of a low latency pair of outputs I need to assign things to. Aren't all the outputs the same? This is very different from my pt hd 2 dsp based system. How do I assign the low latency pair? Or know which pair it is?

Thanks!

[P_J]

Originally Posted by Brandonx1

The way to get hd native to me really low latency, you need to set the playback hardware buffer really low. 128 or 64 samples.

This, I would try first. Mac you have should run 64 buffer without problems.

You can find OMNI manual online here: http://akmedia.digidesign.com/suppor..._v81_67825.pdf

[Jack Ruston UK]

OK, so there are two approaches to this...You can monitor through the PT mixer in the same way as you used to do with your old HD cards, by setting the buffer size very low. If you have a powerful machine and you're not running virtual instruments you should find that you can easily do that. A 96k session with a 64 sample buffer has very low latency. But you can also use low latency monitoring whereby you assign an output pair in the IO setup, turn LLM on and then feed the talent their cue mix via that output pair. That bypasses the PT mixer allowing you to get extremely low latency but with the caveat that inserts and sends applied in the mixer will not be available in the LLM path. With LLM you can leave the buffer set at whatever you like. Make sure the compensate input and output delays are ticked in the IO as with the old system.

PT11 offers a new and very elegant way around this by using a very low record buffer while keeping the playback buffer long. ie the track you put into record has a different (and minimal) buffer, while the rest of the session comfortably purrs along at higher buffer. You'd still be running through the PT mixer though, so there will be SOME latency. The figures are available on the website...I think it's 1.7ms for HD Native with an Omni at 96k (to put this in perspective the older HD cards with a 192 at that rate would have been about 1.2 I think). You can still use LLM if you want to get even lower. It depends on the singer really. Some people can be affected by even tiny amounts whereas others are seemingly immune.

J

[propower]

Totally agree with post above ---

I mistakenly posted all of this in the Windows forum but I have just measured all the Latencies of my HDN thunderbolt system a couple of days ago. The Low Latency path at 96kHz is incredibly good and I am surprised AVID does not spend some serious time touting it .... 0.47ms is almost HDX fast!

With the ease of creating a second input or aux channel to feed a reverb (many descriptions how to do this - like create an aux"1" path whose input is the channel/channels you are recording and set its output to the reverb aux - use fader on "aux1" to set reverb level) I now have a cue mix for recording Vocals or Fingerstyle acoustic with less than 0.5ms latency while leaving the main buffer anywhere I like (512 is a nice low system stress number). Working great so far!

From my other post
---------------------------------------------------
I broke out the oscilloscope and function generator and answered my own question

Using an input on an Avid 16x16 and an output on an OMNI and measured time delay from Analog in to Analog out.

96kHz I/O RT
64 samples ---- 1.80ms
128 ------------- 3.12ms
256 ------------- 5.8ms
512 ------------- 11.12ms
LLM ------------- 0.47ms (regardless of Buffer)

44.1kHz I/O RT
32 samples ----- 3.35ms
64 --------------- 4.80ms
128 -------------- 7.68ms
256 -------------- 13.52ms
LLM -------------- 1.9ms (regardless of Buffer)

[Jack Ruston UK]

Awesome! Very useful measurements. To add further context, Hd accel with a 192 is 2.4ms at 44.1kHz.