low latency monitoring doesn't work on 002r latest drivers

[TOM@METRO]

We have been recording paying clients for years at a 64-sample buffer. Typically at 44.1. With many commercial releases over the years, we have never, I repeat never had a artist complain about latency or phase issues or comb filtering in the cans with this setting unless there was something else causing it.

Are you sure there are no plug-ins in the signal path at all, bypassed or not? Have you double-checked all routing in the headphone mix? There must be something else going on here. Seriously.

I do agree though, that LLM should be working for those of us who feel we need it, for whatever reason.

[daeron80]

By "round trip" I just mean the hardware buffer delays the signal a little at A/D and then a little more at D/A, and the software delays it some more in between. That's my current understanding, anyway, subject to reproof and correction as always. In any event, firewire and USB interfaces generally create a recording monitoring latency about 2.5 to 5 times the size of the hardware buffer setting, and it's not difficult to see why if you work through the whole signal path in your head.

I worked with one artist who was bothered by latency on a TDM system with no plug-ins! She also needed huge amounts of herself in the cans, to the point that we were always riding the edge of feedback. Highly unusual. She was highly sensitive to any change in the sound, and extraordinarily particular about what she was getting in the cans. It took her a long time to get used to recording in PT, and she always preferred tracking analog when possible. Of the hundreds of vocalists I've recorded (including myself), she's the only one who had difficulty adjusting to latencies under 10 or 15 ms.

[SKI]

This thread is actually interesting.
Ever since PT LE gave the ability to do 32 buffers the whole pretense of LLM became questionable.
If your system was strong enough to handle running at 32 buffers you didn;t have to bother using LLM since the latency was almost the same.

I wonder if the reason your LLM isn't responding is related to the delay compensation?

[daeron80]

Originally Posted by SKI

Ever since PT LE gave the ability to do 32 buffers the whole pretense of LLM became questionable.
If your system was strong enough to handle running at 32 buffers you didn;t have to bother using LLM since the latency was almost the same.

What interfaces allow 32 buffer on Windows? I can only get down to 64.

[mykhal c]

Originally Posted by daeron80

By "round trip" I just mean the hardware buffer delays the signal a little at A/D and then a little more at D/A, and the software delays it some more in between. That's my current understanding, anyway, subject to reproof and correction as always. In any event, firewire and USB interfaces generally create a recording monitoring latency about 2.5 to 5 times the size of the hardware buffer setting, and it's not difficult to see why if you work through the whole signal path in your head.

yeah...gotta disagree with you here on the inner workings of PT. and this is nothin' other than my 'opinion' comin' from a system programming background. but 'hardware buffer' is really a software only thing. the conversion inside the interface is not dependent on this setting at all...IMHO. the conversion takes place at a static value/rate dependent on sample rate and the firmware/bios of that unit...002 in this case. the hardware buffer setting tells the software at what rate your rig wishes to handle the data. IOW at 64 sample buffer says, grab this amount in one get which will lead to more 'gets' by your cpu thus causin' a bigger cpu load. a 512 buffer says i got more data per get therefore decreasin' the load on the cpu but increasin' the latency. so in this example it takes 8x the number of gets to process the same amount of data at the lower hardware buffer. but the lower buffer places the process more in real-time than the higher buffer...and real-time DSP is what PT code was based on from the beginnin'. anyway, my 2cents...and i could be waaaayyyy off. but then again...

[daeron80]

That sounds like how it should work, mykhal. Thanks for that explanation. Maybe the "get" cycles on input become equivalent "give" cycles on output, yielding a latency that is always more than twice the buffer size plus a static amount.

[JMS40]

Originally Posted by daeron80

What interfaces allow 32 buffer on Windows? I can only get down to 64.

32 is Mac only.

[mykhal c]

and PT9HD Native interfaces...OMNI (Shane shared some screen-shots with me showin' the 32 sample buffer option in W7)

[JMS40]

Cool.

[jntracks]

i checked every permutation of delay comp on or off and LLM on or off.

i've had delay comp enabled for the whole session but, still in the tracking phase, of the few plugins instantiated none are reporting delay anyway.
and none are on the track i'm recording vocals to.

no sends on that track. just straight input. i just gathered all the gear in my control room to check every link in the chain. even when i take the headphone amp and monitor controller out of the chain, plug the headphones directly into the headphone out on the 002r i can hear the slight phasing effect on my voice. if i record and play it back it sounds normal.

and it's only the person on the mike that hears it. when i'm recording another vocalist it sounds normal to me when they are recording, and normal to them when i play it back.
i can't think of anything else i can eliminate to isolate this any more. just microphone>preamp>analog input>audio track>analog output.

even though i've pretty much always used LLM when recording i've always heard people use 64 samples and can't tell the difference. i'm sure they're not having this problem. i think it's a bug. LLM isn't working and maybe this is a related bug. more latency than it should have at 64 samples. enough to make the voice in your head comb filter with the voice coming back to the headphones.