Login / Register
 
Pro Tools 10 & HDX For Post
New Reply
Subscribe
#31
5th November 2011
Old 5th November 2011
  #31
Gear maniac
 
Joined: Nov 2009
Location: Beverly Hills
Posts: 178

SDDP is offline
Quote:
Originally Posted by soundthinker View Post
I don't think PT can reallocate processing while playing without potentially interrupting the audio.

RE: Channel Strip

I've got 100 tracks with it on the first 5 inserts (500 total) and the CPU(Native) fluctuates between 70 and 74%. 8-Core 3Ghz Mac Pro, 16GB ram.

what is your buffer setting with all this?
and at 70-74%, do you get any errors or does it play super smooth all the time with audio and automation, etc?
#32
8th November 2011
Old 8th November 2011
  #32
Lives for gear
 
ggegan's Avatar
 
Joined: Mar 2008
Location: The Heart of Screenland
Posts: 2,069

ggegan is offline
Just did another test with Channel Strip and in this session 160 instances = 68%

Not sure what is going on here, but that is very different from my previous test under similar system setting.
__________________
Gary Gegan
#33
8th November 2011
Old 8th November 2011
  #33
Lives for gear
 
danijel's Avatar
 
Joined: Jan 2008
Location: Belgrade, Serbia
Posts: 2,046
My Recordings/Credits

danijel is offline
Quote:
Originally Posted by ggegan View Post
Just did another test with Channel Strip and in this session 160 instances = 68%

Not sure what is going on here, but that is very different from my previous test under similar system setting.
I know this is unlikely, but did you mistakenly use stereo tracks?
#34
9th November 2011
Old 9th November 2011
  #34
Lives for gear
 
ggegan's Avatar
 
Joined: Mar 2008
Location: The Heart of Screenland
Posts: 2,069

ggegan is offline
Quote:
Originally Posted by danijel View Post
I know this is unlikely, but did you mistakenly use stereo tracks?
No, it was all mono.
kdm
#35
9th November 2011
Old 9th November 2011
  #35
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Hi all - was just reading in, but since I had run these tests last week I thought the results might be of some interest, just for a rough comparison to an all native system with PT10 (Q9550 Quad, Win7x64, RME latency 128 samples):

Channel Strip:
165 mono instances, cpu 94%

Fabfilter Q and C with the same config - 6 bands of EQ (lowest/non-linear setting):
64 mono instances of both (i.e. 128 plugin instances, 64-EQ, 64-comp), cpu 90%.

When the AAX versions of Fabfilter plugins come out I'll run this again to see if there is any change vs. RTAS.

If I have a chance at some point I'll put PT on the new i7 since the Q9550 isn't the best/fastest reference point, but the i7 is a scoring slave at the moment.

I'm curious how Channel Strip will perform on an HDX card. Nice plugin - just wish the EQ display could be a bit bigger.
#36
9th November 2011
Old 9th November 2011
  #36
Lives for gear
 
danijel's Avatar
 
Joined: Jan 2008
Location: Belgrade, Serbia
Posts: 2,046
My Recordings/Credits

danijel is offline
These tests tell me something I wanted to know - AAX is always computing - not only when there is signal. A HUGE advantage for post is the option to suspend computing when there is no incoming signal, because we never have all tracks playing - you'd have to search hard to find a place in the project where even half the tracks are playing at once.

I just did a quick test in Nuendo - an EQ, compressor and de-esser on every track - 192 tracks load 90% of my i5 laptop CPU. When I enable 'suspend VST3 plug-in processing when no audio is received' (which I normally do), it drops to 10% during playback with no audio. This is why I never ever have to think about the CPU in Nuendo. The project I'm currently mixing has 120 tracks (24 of which are stereo) with EQ on each and comps and de-essers only on dialog tracks, two surround and one mono reverb, three stereo limiters and a couple of smaller plugs, and in my busiest scene, I hardly touch 10% CPU usage.....

This is a huge advantage for small post shops like me, without external DSP.
I'd love to use PT more, but I can currently use it only for mid-category projects, which don't require much CPU, and also don't require faster then realtime render for ridiculous turnaround schedules....
__________________
www.danijelmilosevic.com
kdm
#37
9th November 2011
Old 9th November 2011
  #37
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Quote:
Originally Posted by danijel View Post
I just did a quick test in Nuendo - an EQ, compressor and de-esser on every track - 192 tracks load 90% of my i5 laptop CPU. When I enable 'suspend VST3 plug-in processing when no audio is received' (which I normally do), it drops to 10% during playback with no audio.
How about reverb tails when you stop playback? Don't you have to set a process length for post-stop? Logic had this (dynamic processing) a while back. It was fine, other than tails, and since it didn't have a disable switch, you had to constantly watch where the high cpu points were to figure out if you had processing available for another plugin.

At least with Nuendo you can disable it, but it's still another step to go through everytime you want to add more plugins and think (or even don't really know) if you will max out during a section, or small clip - that's the worst case for dynamic processing - short clips you don't really see in a large project that spike the cpu over the limit when playback reaches that point.

It's a cool feature, but not without some workflow limitations, imho. Ideally this would be available on a track by track basis so it could be more easily managed without dropping in and out globally (which is what many of us requested a few years back on the Nuendo forum).
#38
9th November 2011
Old 9th November 2011
  #38
Lives for gear
 
danijel's Avatar
 
Joined: Jan 2008
Location: Belgrade, Serbia
Posts: 2,046
My Recordings/Credits

danijel is offline
Quote:
Originally Posted by kdm View Post
How about reverb tails when you stop playback? Don't you have to set a process length for post-stop?
It all works for me, I don't know how, but it does - maybe the trick is that I use almost only stock VST3 plugins that come with Nuendo (including reverbs), and VST3 format was pre-conceived with this idea, so maybe the plugins tell the DAW when they think they've had enough of tail. Or maybe some plugins, like reverbs, tell the DAW to process them irregardless of the input.
To tell you the truth, I didn't read much about this feature on the forums, and there could possibly be problems with third-party plugins, maybe someone who knows will chime in.... I only use this since Nuendo version 5, and the way I use it, it works.
#39
10th November 2011
Old 10th November 2011
  #39
Lives for gear
 
jahtao's Avatar
 
Joined: Dec 2005
Location: London, UK

jahtao is offline
Quote:
Originally Posted by danijel View Post
'suspend VST3 plug-in processing when no audio is received'
That's smart!
#40
11th November 2011
Old 11th November 2011
  #40
Lives for gear
 
ggegan's Avatar
 
Joined: Mar 2008
Location: The Heart of Screenland
Posts: 2,069

ggegan is offline
Okay, just did another test with Channel Strip on 192 mono channels. This time The CPU usage is 25%!!!. This should not change from day to day.
#41
12th November 2011
Old 12th November 2011
  #41
Lives for gear
 
Airon's Avatar
 
Joined: Aug 2005
Posts: 552

Airon is offline
Some bugs to sort out then.

On the DUC I was a little surprised to see bugs in relation to the clip gain and painting out stuff with the pencil tool. Might only be a workflow problem, but it seems to hurt the workflow of dialogue editors, which is what I do a lot. Still no news on the trial either it seems.
kdm
#42
13th November 2011
Old 13th November 2011
  #42
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Quote:
Originally Posted by Airon View Post
Some bugs to sort out then.

On the DUC I was a little surprised to see bugs in relation to the clip gain and painting out stuff with the pencil tool. Might only be a workflow problem, but it seems to hurt the workflow of dialogue editors, which is what I do a lot. Still no news on the trial either it seems.
The pencil tool bug appears if clip gain is active (i.e. a change made) - drawing creates what looks like a random waveform instead of what you draw (or a large amplitude, sine-modulated version of the original).

Disabling or clearing clip gain allows pencil draws to work as expected.
The workaround would be to make all drawn edits before doing any clip gain adjustments for a track/clip.
#43
13th November 2011
Old 13th November 2011
  #43
Lives for gear
 
kk@jamsync.com's Avatar
 
Joined: May 2005
Posts: 928

kk@jamsync.com is offline
Quote:
Originally Posted by kdm View Post
The pencil tool bug appears if clip gain is active (i.e. a change made) - drawing creates what looks like a random waveform instead of what you draw (or a large amplitude, sine-modulated version of the original).

Disabling or clearing clip gain allows pencil draws to work as expected.
The workaround would be to make all drawn edits before doing any clip gain adjustments for a track/clip.
Problem comes when you have a 6-hour track with terrible s/n ratio, so you start laying in clip lines just so you can focus on those clicks. You wind up with many more clips than you'd like. I work with Izotope RX and after, declick, NR, and Spectral, I get to focus on the left over stuff. Very badly recorded audio for a court case, so there are many thumps, clicks, and scratches from hidden mikes. The pencil tool bug really can negate the advantage of clip gain with some material. Frustrating.
__________________
___________________
K. K. Proffitt
President, JamSync®, Nashville
www.jamsync.com
http://jamsyncnashville.blogspot.com
(615) 320-5050
kdm
#44
14th November 2011
Old 14th November 2011
  #44
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Quote:
Originally Posted by kk@jamsync.com View Post
Problem comes when you have a 6-hour track with terrible s/n ratio, so you start laying in clip lines just so you can focus on those clicks. You wind up with many more clips than you'd like. I work with Izotope RX and after, declick, NR, and Spectral, I get to focus on the left over stuff. Very badly recorded audio for a court case, so there are many thumps, clicks, and scratches from hidden mikes. The pencil tool bug really can negate the advantage of clip gain with some material. Frustrating.
Absolutely. I have to try to get by with a pass with the pencil tool before doing any clip gain additions, but that only works some of the time.

I hope it's fixed soon.
#45
14th November 2011
Old 14th November 2011
  #45
Lives for gear
 
ggegan's Avatar
 
Joined: Mar 2008
Location: The Heart of Screenland
Posts: 2,069

ggegan is offline
Quote:
Originally Posted by kdm View Post
Absolutely. I have to try to get by with a pass with the pencil tool before doing any clip gain additions, but that only works some of the time.

I hope it's fixed soon.
Have you tried writing regular volume automation and then converting to clip automation? You can use any combination of fader, pencil and click/drag to write the automation, then select the clip and convert or coallesce to clip automation. It's a lot more versatile and efficient than trying to work within the clip.
kdm
#46
14th November 2011
Old 14th November 2011
  #46
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Quote:
Originally Posted by ggegan View Post
Have you tried writing regular volume automation and then converting to clip automation? You can use any combination of fader, pencil and click/drag to write the automation, then select the clip and convert or coallesce to clip automation. It's a lot more versatile and efficient than trying to work within the clip.
True, but for some clicks/pops, I prefer to edit it out of the waveform itself very small but prominent clicks that automation would be too broad or might not be accurate enough to grab, esp. where I might need to crossfade between edits/clips - if there is time of course and depending on necessity. Usually the exception rather than the norm.

Your suggestion is more efficient in most cases though.
#47
14th November 2011
Old 14th November 2011
  #47
Lives for gear
 
tamasdragon's Avatar
 
Joined: Dec 2005
Location: Hungary
Posts: 1,557

tamasdragon is offline
Quote:
Originally Posted by ggegan View Post
Okay, just did another test with Channel Strip on 192 mono channels. This time The CPU usage is 25%!!!. This should not change from day to day.
Now you get me curious, I'll try some tests tomorrow.
#48
19th November 2011
Old 19th November 2011
  #48
Gear maniac
 
bizzle's Avatar
 
Joined: Apr 2005
Location: Los Angeles
Posts: 197

bizzle is offline
#49
20th November 2011
Old 20th November 2011
  #49
Lives for gear
 
kk@jamsync.com's Avatar
 
Joined: May 2005
Posts: 928

kk@jamsync.com is offline
Quote:
Originally Posted by ggegan View Post
Have you tried writing regular volume automation and then converting to clip automation? You can use any combination of fader, pencil and click/drag to write the automation, then select the clip and convert or coallesce to clip automation. It's a lot more versatile and efficient than trying to work within the clip.

In my case, I have to change level and then de-noise the clip to get at radically low-level stuff, sometimes several times for one clip. One guy has a loud voice and the others practically whisper. It's a "wire" so there's clothing noise and the environment has heavy air conditioning. When I bring up the level of the clip sometimes I de-noise, find a click, then back out of the de-noising if I think it's too weird. If I had relatively clean audio that was fairly even, I'd just automate it and clean later, but I'm literally denoising, decrackling, dehumming and declipping with some of the material. One clip was raised by 24 dB! I'm working around by zooming in deep and making the click a clip and just reducing it that way. Probably wouldn't work for music, but in a pinch...
#50
21st November 2011
Old 21st November 2011
  #50
Lives for gear
 
ggegan's Avatar
 
Joined: Mar 2008
Location: The Heart of Screenland
Posts: 2,069

ggegan is offline
Quote:
Originally Posted by kk@jamsync.com View Post
In my case, I have to change level and then de-noise the clip to get at radically low-level stuff, sometimes several times for one clip. One guy has a loud voice and the others practically whisper. It's a "wire" so there's clothing noise and the environment has heavy air conditioning. When I bring up the level of the clip sometimes I de-noise, find a click, then back out of the de-noising if I think it's too weird. If I had relatively clean audio that was fairly, I'd just automate it and clean later, but I'm literally denoising, decrackling, dehumming and declipping with some of the material. One clip was raised by 24 dB! I'm working around by zooming in deep and making the click a clip and just reducing it that way. Probably wouldn't work for music, but in a pinch...
Creative use of clip gain! Thanks for the ideas!
#51
21st November 2011
Old 21st November 2011
  #51
Gear addict
 
Joined: Aug 2007

ThisIsSka is offline
One thing I've done when de-clicking (without using RX) is to highlight a short period before the click and then duplicating it over the click.

Speaking of clicks, though OT - my Neumann M149 tends to click whenever someone turns a light on in the next room. I can't do any rewiring so could someone recommend an affordable power-conditioner (if it would help at all) that may reduce this problem? No issues with other mics, just this particular tube mic.
#52
21st November 2011
Old 21st November 2011
  #52
Lives for gear
 
kk@jamsync.com's Avatar
 
Joined: May 2005
Posts: 928

kk@jamsync.com is offline
Quote:
Originally Posted by ThisIsSka View Post
One thing I've done when de-clicking (without using RX) is to highlight a short period before the click and then duplicating it over the click.

Speaking of clicks, though OT - my Neumann M149 tends to click whenever someone turns a light on in the next room. I can't do any rewiring so could someone recommend an affordable power-conditioner (if it would help at all) that may reduce this problem? No issues with other mics, just this particular tube mic.
Yep that works well especially for repetitive wave forms (and it's also one of the ways a declicker works).

As for a power conditioner, I like the ones by Furman. Even though I had all my stuff on Square D isolation transformers in one of my home studios, a lightning blast to my outside ground rod fried the one keyboard that wasn't on a Furman. A couple of resistors toasted, absolutely black. Since then, even though we have isolated power, everything is routed through a Furman.
#53
27th November 2011
Old 27th November 2011
  #53
Gear interested
 
Joined: Jul 2010
Posts: 17

cidmonkey is offline
Quote:
Originally Posted by danijel View Post
These tests tell me something I wanted to know - AAX is always computing - not only when there is signal. A HUGE advantage for post is the option to suspend computing when there is no incoming signal, because we never have all tracks playing - you'd have to search hard to find a place in the project where even half the tracks are playing at once.

I just did a quick test in Nuendo - an EQ, compressor and de-esser on every track - 192 tracks load 90% of my i5 laptop CPU. When I enable 'suspend VST3 plug-in processing when no audio is received' (which I normally do), it drops to 10% during playback with no audio. This is why I never ever have to think about the CPU in Nuendo. The project I'm currently mixing has 120 tracks (24 of which are stereo) with EQ on each and comps and de-essers only on dialog tracks, two surround and one mono reverb, three stereo limiters and a couple of smaller plugs, and in my busiest scene, I hardly touch 10% CPU usage.....

This is a huge advantage for small post shops like me, without external DSP.
I'd love to use PT more, but I can currently use it only for mid-category projects, which don't require much CPU, and also don't require faster then realtime render for ridiculous turnaround schedules....

I haven't tried this but it makes sense in my head. If 'mute frees assigned voice' makes the track inactive, then the processing on the channels would be inactive and presumably free up DSP/CPU? So then you'd have to automate the mute when there's nothing on the track, but it could work well if there's a long stretch without any regions on the track?
Anyone tried this? Maybe theres something I'm overlooking but I havn't had a chance to try it out
#54
29th November 2011
Old 29th November 2011
  #54
Lives for gear
 
danijel's Avatar
 
Joined: Jan 2008
Location: Belgrade, Serbia
Posts: 2,046
My Recordings/Credits

danijel is offline
Quote:
Originally Posted by cidmonkey View Post
'mute frees assigned voice'
Is that an option in PT10?
#55
29th November 2011
Old 29th November 2011
  #55
Gear Head
 
ReZoUnD's Avatar
 
Joined: Aug 2011
Location: Houston
Posts: 65

ReZoUnD is offline
ive been a PT user sence 7.3, and i dont dont have HD so i dont feel the upgrade. it feels like it should be 9.6 cuz being native its not worth much...well what i do like is the 32bit float and gain but thats not something to jump out your seat for. ive been loyal so im taking it as bittersweet bitter: cuz its not huge like going from 7 to 8 truly felt the diffrence there, Sweet: cuz of the headroom and gain
#56
29th November 2011
Old 29th November 2011
  #56
Gear interested
 
Joined: Jul 2010
Posts: 17

cidmonkey is offline
Quote:
Originally Posted by danijel View Post
Is that an option in PT10?
It's not new, I can't find it in my native system so it must be HD only. I'm not sure how long it's been around but here's a DUC thread from 08 mentioning it so it's been around since at least then

Mute Frees Assigned Voice - question - Avid Audio Forums
#57
29th November 2011
Old 29th November 2011
  #57
Gear addict
 
iluvcapra's Avatar
 
Joined: Feb 2010
Location: Los Angeles, CA

iluvcapra is offline
Quote:
Originally Posted by kdm View Post
How about reverb tails when you stop playback? Don't you have to set a process length for post-stop? Logic had this (dynamic processing) a while back.
VST plugins (and for that matter Audio Units plugins) can advertise to the host their tail and priming requirements -- it's up to the host to do the "right thing" in the given circumstance but its not impossible.

Some plugins have an infinite impulse response and never "tail out," though, many have a tail that varies dynamically with the control envelope, and many aren't hip to the tricks the host plays on them in order to make them efficient. REAPER is another DAW that dynamically runs VSTs for clip-based effects, and in order to make the playback more reliable it seems to give them several seconds of leadtime to render out samples, which makes their UI do very strange things -- you see their meters move seconds before you hear things, for example.

You're also right, if you switch processing on and off throughout the session, the ability for the session to play becomes a lot less deterministic. Best practice is for something to fail as soon and as obviously as possible.

Quote:
Originally Posted by cidmonkey View Post
It's not new, I can't find it in my native system so it must be HD only. I'm not sure how long it's been around but here's a DUC thread from 08 mentioning it so it's been around since at least then
[/URL]
"Mute frees assigned voice" only mutes voices that are assigned, which is to say, not dynamically assigned. You can't manually assign a voice to a track in LE, thus, no Mute frees assigned voice. It sortof a dangerous feature of indeterminate usefulness in out modern era: if you mute a track long enough, and then unmute it while playing, it won't actually start playing audio again until its buffer gets filled, which means you could in theory automate a mix to sound one way, with mute automation, but wether or not it plays the same way later will depend on the memory setting of the system you open it on.

I don't think I've used it since the late 1990s; I use "Make Active" instead, because it also deallocates any plugin resources on the track, which "Mute Frees Voice" doesn't do.
__________________
Jamie

Last edited by iluvcapra; 29th November 2011 at 08:17 PM.. Reason: combine replies
#58
30th November 2011
Old 30th November 2011
  #58
Gear interested
 
Joined: Jul 2010
Posts: 17

cidmonkey is offline
Quote:
Originally Posted by iluvcapra View Post
I don't think I've used it since the late 1990s; I use "Make Active" instead, because it also deallocates any plugin resources on the track, which "Mute Frees Voice" doesn't do.
Oh cool, well if removing the voice like that doesn't deallocate the DSP then it wouldn't work for what I was thinking anyway.
kdm
#59
6th December 2011
Old 6th December 2011
  #59
kdm
Gear addict
 
Joined: May 2008
Posts: 439

kdm is online now
Quote:
Originally Posted by iluvcapra View Post
VST plugins (and for that matter Audio Units plugins) can advertise to the host their tail and priming requirements -- it's up to the host to do the "right thing" in the given circumstance but its not impossible.
Hence my question of how VST3 suspend worked in his case. I knew VST advertised reverb tails, but it sounds like suspend factors that in as well - makes sense that it should.

Dynamic processing has it's limitations though - I was always running into a spike in the middle of a mix in Logic when a plugin went active for just a short clip. I and others lobbied for this when I was a Nuendo user - the ability to make each track either static or dynamic depending on what was being processed. I've since reconsidered whether I would really want another variable to potentially undermine a stable playback config (for native systems at least).
#60
6th December 2011
Old 6th December 2011
  #60
Gear addict
 
iluvcapra's Avatar
 
Joined: Feb 2010
Location: Los Angeles, CA

iluvcapra is offline
Quote:
Originally Posted by kdm View Post
Dynamic processing has it's limitations though - I was always running into a spike in the middle of a mix in Logic when a plugin went active for just a short clip.
Note that it's not supposed to go from active to inactive based on signal level, but on presence of a playing file at all--the host signals the plugin out-of-band if its voice was working or not.

VST3 supports a "silence flag" that the host can set when sending a signal buffer to a plugin, thus, if the host knows there's no audio on a buss, because it's not in a play state, or the disk voice feeding that buss isn't in a play state, the host can just send the silence signal to the plugin once every block, instead of the plugin having to manually test every sample on the bus for silence. The VST plumbing around the plugin can also look for the silence flags and decide not to call the process methods on the plugin at all if all its input busses are flagged quiescent.

All this saves you in the end are cycles on your CPU, though, which for many of us might not be as big of a problem as memory. Automatically suspending and resuming processing can't save memory, unfortunately, because memory allocation is time-intensive and a plugin can't efficiently free and reclaim memory at the drop of a hat -- if a plugin tried to allocate itself memory only when samples needed it, it'd never run on time. Even when memory is retained, you can still take a hit in dynamic situations if a plugin gets called to "wake up," but it's been moved off to virtual memory due to lack of use. This causes a page fault and a very slow process wherein the page(s) of swap your plugin lives in are lifted off the drive and paged-in. Keeping the processing going constantly is a way of keeping all the plugin resources in physical memory.

If anyone here is an RTAS developer please weigh in. (Oh yeah, that's right, you're all under an NDA.)
New Reply Submit Thread to Facebook Facebook  Submit Thread to Twitter Twitter  Submit Thread to LinkedIn LinkedIn  Submit Thread to Google+ Google+ 
 
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump

SEO by vBSEO ©2011, Crawlability, Inc.