Simple Buffer Question

[a100man]

All the talk recently about latency has got me thinking. I haven't found a clear answer to this, so I'd like to pose a straightforward question: Is there an unavoidable latency that in incurred when recording due to the H/W buffer size (on an 002)?

For example, if I set the buffer to some high value like 1024, and I record while playing along with recorded tracks, should I expect my recorded performance to be late (as compared to existing tracks) by the amount of the buffer, or does PT time align new tracks for the buffer latency on its own?

Thanks!

[original_min]

Hi.

As far as I understand it PT does compensate for the buffer on recorded material.
In other words it shifts it back relative to the buffer setting you have.
Where it doesn't do that is on Audio or Aux tracks monitoring outboard gear live,
which makes it a little tricky integrating things like samplers and effects units.

Here's a good pdf on it.
http://akmedia.digidesign.com/suppor...tems_33000.pdf

I'd love to know how you're going to play along with material at 1024 buffer though?

Cheers

[mano111]

Its Best to record at 64-128 buffer. Try using Low Latency Monitoring option for listening accuracy.
If you plug a mic in talk and listen back on your headphones out, you will notice the amout of latency is rediculously unusable for vocals when set to 256 or higher (you will seem to be talking slower). If you are playing keys, or guitar it will throw your timing way off.
LLM is a must when recording IMO.

[a100man]

Lol, I was only suggesting playing along with material at 1024 for the sake of illustration-- I agree it would be quite difficult!

I did some of my own tests last night and came up with some very interesting results:

On PT7.3 and DIGI002

Loopback from AO7 to AI7, (not configured as an insert, just a playing and recording track) I got that audio was recorded EARLY by 1 sample REGARDLESS of the buffer setting. So this would imply that if I track with LLM, there are no ill effects to recording with a huge buffer size, right?

Interestingly, the same test on PT6.4 and DIGI001 yielded 51 samples of latency at all buffer settings!

[mano111]

your test is intersting, as it means that PT over compensated for the recroding latency by one sample. how did you measure this?

If I send a track to outboard gear like EQ or compressor (rarely) I allways make sure to manually align it with the origilnal track by zooming in all the way and aligning the waveforms.

I've had more accurate recrdings with LE + LLM than with HD (whitout ADC enabled), same with mixing. Really minor stuff and I don't feel in made any diffrence in the quality of music produced. bottom line is its up to the musicians and producer to make it work in the studio. mixing is much more affected by room accoustics and speakers than internal time misalignment (its very low in LE). If you see a certain plugin is causing high latencey problems (eg. amplitube) just record it to a track and align it.

I did however fill out the request form on the digi site and asked for PDC as it seems like a nobrainer. you should too...

[a100man]

I actually reproduced a test that was described on another thread. I took a few seconds of the click plugin (all default settings) and recorded it to a track. I then set the output of this track to analog 7, and made a new track with a source of Analog in 7, which I had looped back with a cable (I also tested ADAT and SPDIF in, which had -4, and -8 latency respectively, although the exact gear I was using probably factored in here too) and recorded several tracks at different buffer settings. Although the buffer setting didn't change the latency, they were all 1 sample early, which I could see but zooming in to the sample level. This particular audio signal worked very good for this, because it is almost like a step function (if you try it you'll see what I mean).

The thing that strikes me as odd, though, is that PT is obviously compensating for the buffer size, but it's off by 1 sample. Why on earth wouldn't they just tweak the compensation until it is right? 1 sample is pretty insignificant, but it's very strange that they would bother compensating, but not get 100% there!

[The Weed]

Okay, let's see now. At 44.1 that makes a sample 1/44,100 of a second or 0.0000226 of a second or an error of 0.00226% meaning it's 99.99774% accurate. I don't know if it's possible to make it any more accurate than that!

Cheers,