Sample rates.....what's up with 88.2?

A place to talk about whatever Scope music/gear related stuff you want.

Moderators: valis, garyb

Neil
Posts: 47
Joined: Sat Dec 09, 2006 5:33 pm
Location: The Vast Western Desert

Post by Neil »

astroman wrote: This is surely not my cup of math, but the load increase you report does in no way corelate to a 'slightly higher demand of resources'.
That's true, but I even tried two different methods of conversion, and - on my PC at least - it STILL puts the system over the top, resouce-wise, to try & run a particular project at that samplerate... however Mr. Weiss was talking about the conversion process ALONE, and at the time of tha post I was referring to not the conversion process itself at all, but the resources required to run the project & it's plugins, etc at 96k vs. 88.2k.

It does seem strange that there's that much of a resouce hit, I'll admit; but I've done it twice now using different methods of conversion, just to see if I f*ked something up the first time, and in both instances the project goes from running fine at around 60% of my CPU meter, to pretty much full lockup & not being able to run at all at 96k... it's not the disk streaming aspect, either - that meter in Cubase stays fairly low, but the CPU usage maxes out at 100%.

Neil
areptiledysfunction
Posts: 137
Joined: Tue Oct 17, 2006 4:00 pm
Location: Durango, CO

Post by areptiledysfunction »

astroman wrote:
Btw increased transparancy isn't always an advantage - I once bought the Curtis Mayfield 'Superfly' soundtrack on CD (one of those nice price things).
Remastered crisp as hell - someone had mixed the funk out of it entirely.
On the original topic I'm as shocked as GaryB that the system doesn't sync at 88.2 with wordclock. The original manual says 'any rate', they may have modified it...

cheers, Tom
the same thing happened with the remastered version of Hotel California (Eagles) Sounded terribly neutered to my ears. As far as being shocked that the system doesn't work at 88.2, that is very strange, but having been working with a system for 9 years (Paris) that was also proprietary/hardware DSP and had a few quirks of it's own and which was discontiued 5 years ago, he and I are used to strange and bizarre workarounds in order to get a system to perform unnatural acts. I think Neil needs to purchase a 24 channel 88k to 96k outboard SRC, buy one of my RME HDSP 9652's and stream the files to the Weiss converter input, SRC them in real time while streaming them at 96k from the Weiss into a Pulsar system on a new computer that he will need to build, process them there, then send the stems out of Pulsar analog on a A16U to an SSL console for final mixing/summing.........at least that's what I would do. :wink: We need to get as many DAWs and external hardware boxes involved in this as possible, cause the more gear you know how to throw at a mix, the better the possiblity the mix may suck, or not..........sorta' like an extreme sport.
:lol:
User avatar
astroman
Posts: 8446
Joined: Fri Feb 08, 2002 4:00 pm
Location: Germany

Post by astroman »

Neil wrote:...and in both instances the project goes from running fine at around 60% of my CPU meter, to pretty much full lockup & not being able to run at all at 96k... it's not the disk streaming aspect, either - that meter in Cubase stays fairly low, but the CPU usage maxes out at 100%....
I hardly dare to mention it, as it's known for several years, but could it have something to do with the infamous 'denormalization' feature of P4 ? those 'underflows' when dealing with (very) small numeric values ?
Current software should be adapted to it, but you might have something active which is from an older date and was 'overlooked' on a regular base.

cheers, Tom
Neil
Posts: 47
Joined: Sat Dec 09, 2006 5:33 pm
Location: The Vast Western Desert

Post by Neil »

astroman wrote:
Neil wrote:...and in both instances the project goes from running fine at around 60% of my CPU meter, to pretty much full lockup & not being able to run at all at 96k... it's not the disk streaming aspect, either - that meter in Cubase stays fairly low, but the CPU usage maxes out at 100%....
I hardly dare to mention it, as it's known for several years, but could it have something to do with the infamous 'denormalization' feature of P4 ? those 'underflows' when dealing with (very) small numeric values ?
Current software should be adapted to it, but you might have something active which is from an older date and was 'overlooked' on a regular base.

cheers, Tom
I don't even know what that is which you're referring to... that's a Pulsar (P4?) bug, you mean?

Neil
User avatar
astroman
Posts: 8446
Joined: Fri Feb 08, 2002 4:00 pm
Location: Germany

Post by astroman »

no, it's a design flaw in Intel's Pentium 4 processor - when that CPU was released a couple of years ago it was pretty widespread - meanwhile most software should be patched (or modified) according to it.
Never experienced it myself, but it's said to lock up machines with sudden high CPU loads very similiar to your symptoms. It's happening with extremely small numbers, and as such found an excellent playground in audio processsing.
One never knows, but I wouldn't expect Cubase to still suffer from it today, but you may have a favourite plugin that's always engaged, so you don't even think about it. Or you may have an Athlon CPU and laugh about it all... ;)

cheers, Tom
Neil
Posts: 47
Joined: Sat Dec 09, 2006 5:33 pm
Location: The Vast Western Desert

Post by Neil »

astroman wrote:no, it's a design flaw in Intel's Pentium 4 processor - when that CPU was released a couple of years ago it was pretty widespread - meanwhile most software should be patched (or modified) according to it.
Never experienced it myself, but it's said to lock up machines with sudden high CPU loads very similiar to your symptoms. It's happening with extremely small numbers, and as such found an excellent playground in audio processsing.
One never knows, but I wouldn't expect Cubase to still suffer from it today, but you may have a favourite plugin that's always engaged, so you don't even think about it. Or you may have an Athlon CPU and laugh about it all... ;)

cheers, Tom

AHHH! PENTIUM 4... no, in my case that can't be it - I have Athlons in all my PC's.
User avatar
valis
Posts: 7650
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Post by valis »

Athlons can denormal too, in fact any digital calculation that recedes to infinitesimally smaller values without checking for a threshold could cause this (in theory). P4 is just affected more due to having a deep pipeline that gets 'flushed' by the denormalling thus causing a loss of many cpu cycles/increase in cpu usage by any offending plugin. Its important to understand that the same plugin STILL wastes cycles on a cpu with a shorter pipeiine (athlon, core2 etc) its just less apparent to the end user.
hubird

Post by hubird »

another adventage of OSX... :-)
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

i'm sure osx is also prone to this.
User avatar
astroman
Posts: 8446
Joined: Fri Feb 08, 2002 4:00 pm
Location: Germany

Post by astroman »

...when it runs on a Intel CPU... (?) ;)
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

is there any other way these days?
i'll bet motorollas can do the trick as well with poor coding as a catalyst....
User avatar
astroman
Posts: 8446
Joined: Fri Feb 08, 2002 4:00 pm
Location: Germany

Post by astroman »

Gary, the old Motorola coding (pre-PPC) was pretty damn smart... but they retired the entire product line, possibly it was a bit too smart to fit with the upgrade spiral :D
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

you bet.
User avatar
valis
Posts: 7650
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Post by valis »

hubird wrote:another adventage of OSX... :-)
OSX or the operating system in general has nothing to do with denormalling. I'ts purely a math thing.
Neil
Posts: 47
Joined: Sat Dec 09, 2006 5:33 pm
Location: The Vast Western Desert

Wonder of Wonders...

Post by Neil »

Well, I find this very interesting: I got my Project Card to sync to 88.2k yesterday - what I did was to pull it out of my Cubase PC & put it in my Paris rig, which happens to be running WinMe, loaded the drivers, the software (which "hung" a few times & thusly scared me a bit, based on a previous posters' experience with that phenomenon - but finally I got it to fully launch), coupla reboots & lo & behold I have it syncing perfectly to both S-Mux/lightpipe and/or word clock via BNC.

Now, the Pulsar samplerate display does NOT read 88.2k, but rather 96k, but I get not one, but TWO little red lock-lights (I assume signifying that both S-Mux inputs are receiving synced data?).

Why this would not lock properly on the Cubase Machine, I have no clue - perhaps due to there ebing two ASIO drivers present (my RMEs & the Pulsar), even though I had disabled the RME's? Dunno.

Very strange, esp considering that GaryB had said that CWA had definitively stated that it should, in fact, NOT work at this samplerate under any circumstances.

Neil
User avatar
valis
Posts: 7650
Joined: Sun Sep 23, 2001 4:00 pm
Location: West Coast USA
Contact:

Post by valis »

For what it's worth a reinstall fixed my issues. Since the Scope installer wouldn't even work I don't think my problems were related to 88.2, but rather to one of the MS updates from that same week (I rarely reboot unless I need to).
User avatar
astroman
Posts: 8446
Joined: Fri Feb 08, 2002 4:00 pm
Location: Germany

Post by astroman »

a very special merry xmas to both of you, congrats :D
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

Neil i think you misunderstood me. :wink:

but it's true. cwa is making no official 88.2k claims. i knew that your card was locking to 88.2k, the problem is that it reported that samplerate as 96k. that made cubase play back at the wrong speed. i'm really glad that you found a way to make it work.
Neil
Posts: 47
Joined: Sat Dec 09, 2006 5:33 pm
Location: The Vast Western Desert

Post by Neil »

garyb wrote:Neil i think you misunderstood me. :wink:

but it's true. cwa is making no official 88.2k claims. i knew that your card was locking to 88.2k, the problem is that it reported that samplerate as 96k. that made cubase play back at the wrong speed.
So, if I take a Sharpie & draw "88.2k" on the screen over the reported 96k, then I could fool Cubase into playing back at the right speed? :D
User avatar
garyb
Moderator
Posts: 23364
Joined: Sun Apr 15, 2001 4:00 pm
Location: ghetto by the sea

Post by garyb »

:)
Post Reply