Omnisphere - multiout bug?

[kdm]

Anyone here using Omnisphere had problems with a random full scale digital buzz erupting from an unused instrument's output?

I had this happen 3 times, and I think it's related to enabling multiple outputs in Omni. Omni will emit a digital full scale buzz similar to a ground lift buzz. Now imagine monitoring with your system calibrated to 85dbc. This one has me scared to use Omni at all. Love the synth, but this a nasty nasty bug.

The buzz only seems to come from an unused instrument's output (i.e. it's a "default" or "empty" slot - last time it was the 8th slot, and 8th output channel, and started as soon as I finished reassigning outputs in Omni to A-H. I'm going to call Spectrasonics tomorrow, but wondering if anyone else has seen this. Can't be a system problem as I have no problems with other VSTi's. Could however be a VST2.4/VST3 compliance issue I am guessing.

I'm sure I'll have no hope of this being addressed if there aren't other reports, and if it isn't addressed, I'm not using it anymore - can't risk it. Along with Play's muting problem, that's two major VSTi's that I'm having to sideline.... not really happy about it. I have to say, I've never had problems like this with NI VSTi's throughout the Komplete 2-5 family and upgrade cycles.

Thanks.

[D]

I'll try this out and see if I get the same bug.

[kdm]

Just be sure your monitor volume is very low!!! Seriously. The buzz comes out at 0dB Full scale.

[zvenx]

hasn't happened with me so far, and I have been using omnisphere since september.. does this happen from a particular output pair?
rsp

edit: just really read what you said, and you were saying it doesn't happen from any specific output.....I rarely if ever use output 8..and what you describe happens to my computer occasionally long before I even got omnisphere..... I always assumed it had to do with my lynx drivers but now I think of it, it only happens when I use vsti's that use a lot of memory. so maybe it is a out of memory thingamajig.


For vin, just so you know this never happens on my mac :-)
rsp

[kdm]

Richard - have you not seen it on your PC, or is Omni running on your Mac? I know there are many Omni users on VI-Control, but most seem to be Mac/Logic users, so we Nuendo/PC users may be in a vast minority there. I posted this on VI and haven't gotten any responses at all there (since Spectrasonics doesn't have their own forum).

It could be memory related in some way perhaps, except that this project only had one instance of Omni and 3 instruments - the rest of the plugins were EQ and comps for post mix, and only about 30 sparsely populated tracks of interview/VO. It was a very light project. Probably the biggest resource contender would be video - SD res DV quicktime file (10 minute clip). I doubt I had more than 1G of ram used out of 4G.

It has happened on various outputs, and even two at a time, but usually only outputs from empty instrument channels - as if someone forgot to connect the software ground on unused instrument ports. lol.

[zvenx]

Hi, My omnisphere is running on both. Yes I have heard it on the PC, but I have heard it before I even got Omnisphere. I think it is memory related for me. As it only happens when I use a lot of vsti's.

What audio card are you running?
I always assume it had to do with my lynx two c drivers.

rsp

[kdm]

RME HDSP Digiface here. If it was driver related I would think it would affect other audio - it's only Omnisphere doing this. All my other VSTi's are fine. I'm wondering if it's a VST 2.4/3.0 compatibility issue on Omni's part.

[zvenx]

ok..... well like I said from my own experience with it, it doesnt' only happen with omnisphere.. but maybe stylus rmx also causes the problem, cause I use that in almost every project I do.
rsp

[kdm]

Interesting. When you have seen/heard this, did it only come from a VSTi, or did it affect other audio/VSTi's within Nuendo? I could kill it by simply shutting down Omnishere, and all other VSTi's would be fine.

This is the kind of thing that ends relationships with clients quickly, so if it's a Nuendo issue, it's a deal breaker. If it's an Omnisphere problem, same thing. VST spec problem, still a deal breaker.

Any ideas you have on how/when it happens would be great. I either have to find a permanent solution, or look for a way to completely avoid it, whether that's ProTools, Logic, or ditching Omnisphere.

[zvenx]

Firstly I would urge you to contact Spectrasonics.

Truthfully I don't look to see which output it is coming from. I just dive for the volume switch on my monitor :-).....usually it last for about 20 seconds then goes away..... or I close that project and reopen it. I always look to the main output level in Nuendo instead of searching for which is the offending individual output causing it. Next time it happens I will look for that and report back.

Also what are your system specs? Are you using the /3GB switch?

rsp

[D]

After reading xvenx (hopefully you are not VenomFangX from YouTube) and his troubles, I extrapolated that his problems are probably due to bus contention bottlenecks.

Rather than simply memory, I would propose that they are more likely the cause of a combination of factors, which include memory, swap file and bus mastering. You may have several devices trying to master the bus, and the northbridge and/or southbrdige chipset simply gets bogged down and cannot process the requests fast enough.

I've seen this happen outside of DAW work, and it's usually related to that.

[zvenx]

have no idea who venomfang is, and I now I am afraid to find out :-)
I am zvenx... a play on sven......but almost everyone calls me richard.
You are probably right what is causing the problem.
rsp

[TAFKAT]

This burst issue is quite interesting..

I have a client having a random issue with Soundforge as well, odd occurrences of white noise, almost like the card is loosing sync, now heres the kicker, its across 2 cards, and M-Audio Audiophile and an RME Fireface 400..

Still trying to pinpoint it..

I'll let you know if I dig anything up that can give so clues in this instance..

@Richard,

Re it not happening on Mac, cool, I hear a lot of things don't happen on Mac, like reliable low latency performance, or decent multiprocessing scalability, but thats another thread.. :wink:

[zvenx]

yes Vin, please do......

I actually thought it was a normal part of XP...lol I kid I kid....... two words for you......SNOW leopard :-)
rsp

[kdm]

Re it not happening on Mac, cool, I hear a lot of things don't happen on Mac, like reliable low latency performance, or decent multiprocessing scalability, but thats another thread.. :wink:
:rotfl:

I talked to Spectrasonics and am putting together some task manager reports to send over. When I first tested it this morning, and while on the phone with Spectrasonics, it happened every single time I went to Nuendo's VSTi rack and enabled "All outputs", and would quit as soon as I deactivated all but the first output - but only at lower latencies up to 512 samples. At 1024 it wouldn't happen at all.

With multiple outs enabled, I reset ASIO latency back to 256 and Voila! it was back again.

I opened up task manager to watch this, and Fireworks to drop screen shots into, and it couldn't not recreate it again no matter what I did. I shut everything down, and reopened Nuendo, plugged in Omnisphere - still couldn't recreate it. Not sure what the catalyst is now, and nothing of value to send to Spectrasonics either.... sigh... still chasing this one.

Probably a grounding problem with Omnisphere. I'll put it after a power conditioner and ground lift plugin and see if that helps. :eusa_eh:

[zvenx]

ok let me try it a few times with all outputs enabled.... I usually use 64 samples buffer size so I should know quickly if that is the trigger on mine.
rsp

[TAFKAT]

... two words for you......SNOW leopard..

.. one word for you..., Garfield.. :D

[TAFKAT]

I opened up task manager to watch this, and Fireworks to drop screen shots into, and it couldn't not recreate it again no matter what I did. I shut everything down, and reopened Nuendo, plugged in Omnisphere - still couldn't recreate it. Not sure what the catalyst is now, and nothing of value to send to Spectrasonics either.... sigh... still chasing this one.

Heres an idea, run up the DPC latency check and see if it corresponds to a spike.. ?

[zvenx]

nope that isn't triggering it on mine......sorry.
rsp

[zvenx]

Heres an idea, run up the DPC latency check and see if it corresponds to a spike.. ?

I am so glad that is directed at Dedric.
You keep mentioning this DPC and I have idea what it means (yes I know it is some application that gives you some graphical output of something or the other :-)) and way too lazy to google it..

so yes Dedric, please see if the DPC thingamajig leads us anywhere :-)
rsp

[zvenx]

The Guru, I do use emulated midi sports and system timestamp and I do use midisports.

Dedric, are you using UAD's?

Now I th ink of it, I have some UAD's permanently attached to my outputs as a mastering chain. They are in bypass mode when I am recording midi parts and audio but when I am editing and mixing etc they are on. It is quite possible it only happens when they are on in my chain.
rsp

[kdm]

No UAD-1s, and never use emulated midi ports. Spectrasonics thinks it might be related to virtual memory. I'm thinking that is possible, though ASIO block/buffer boundaries seems a more likely specific cause given how it would only happen with latency under 1024, and that's exactly what tech support suggested trying... i.e., it sounds like they've seen this before.

[D]

I first tested it this morning, and while on the phone with Spectrasonics, it happened every single time I went to Nuendo's VSTi rack and enabled "All outputs", and would quit as soon as I deactivated all but the first output - but only at lower latencies up to 512 samples. At 1024 it wouldn't happen at all.

This tells me that my analysis is spot on. Some combination of what Omnisphere is requesting and what the rest of the system needs is causing a bottleneck on the bus. Perhaps Spectrasonics need to do a bit of assembler language optimization to clear up this bottleneck.

[D]

I think there are different issues here. Are people aware sync can be lost with some devices if Emulated MIDI ports aren't used with some devices? The MIDSports are devices that require the use of such ports in Cubendo.

System timestamp may also be needed in other situations and with other kit as the clocks associated with different drivers can vary widely and cause loss of sync. I've seen it affect audio as well as MIDI. Don't ask me why, but it does.

This is not a sync issue. This is an issue of bus arbitration or overload.

[kdm]

D - any suggestions on what to test or try to isolate? I'm going to put dpclat on this system and run it later per Vin's suggestion.

[D]

Well, I too suspect it's some sort of disc activity, very possibly the virtual memory settings. Depending upon how your system is set up, you might be able to turn off virtual memory altogether.

If that is the problem, you might be able to minimize it by setting your virtual memory to a fixed size, rather than using the default variable size. If you've never done this before, it's not as straightforward as it might seem to do it right. Let me know.

[D]

Also, are you running any RAIDs? The nice thing about RAID is the redundancy or the extra speed, but for every disc in a RAID, you have to double the bus activity related to disc activity. I would even extrapolate that there is quite possibly some overhead processing in the RAID operation that needs to use the bus, as well, which would contribute to the congestion.

Also, if you are using an ATI video card, their latest drivers are throttling back the GPU under "normal" operating conditions, and only upping the speed in full-screen 3D applications. This kind of sucks, because I own a number of 3D applications, and the windowed applications get no speed boost. When working in D3D or OpenGL, the system sometimes gets a bus backlog which occasionally causes a blurpbleepbuzzfizzle in the audio system.

I am also extrapolating that these things are being mitigated by the handling of priorities within the OS itself, combined with the way the hardware handles it through bus arbitration and contention. For instance, the OS would put virtual memory access at the highest priority level (and this is indeed the case with Windows, and I've seen it many times during particularly geometry overladen 3D renders) and if you've never been into the netherworld of overloaded virtual memory calls, you've probably still got all that hair you would have torn out otherwise. Once you hit that wall, the hard drives go into constant spin mode, and keyboard and mouse input are almost completely ignored.

:rotfl:

Anyway, what might be happening here is that some very small version of this virtual memory queue overload syndrome, and once you hit that wall, the system pretty much is out of your control (or any other control), hence the noise.

[kdm]

Definitely not the wall of VM - it's only Omnisphere that outputs noise - system response is fine, VM doesn't change, plenty of ram, ASIO unaffected, other VSTi's still playing fine, etc. If resetting ASIO to 1024 can stop it, and resetting back to 256 restarts it, there is something else happening within Omnisphere.

No raid on this system currently - too many drives for samples, video, audio, etc to go raid just yet, but I'm migrating all of my farms systems and the host to raids at some point.

Nvidia based video card, and Decklink - Decklink not in use when this has happened, and it's only with Omnisphere - Kontakt and all others are fine.

I'm migrating to a new system today and hoping it just won't be an issue, but not holding my breath since Omnisphere has to be addressing ASIO/buffers differently from Kontakt and other VSTis to cause this, and only affect Omnisphere in the first place.

[D]

Well, good luck with this problem. Obviously there are a lot of variables involved, and the way Omnisphere is dealing with either the ASIO stream or virtual memory could be at the root of it. Hopefully Spectrasonics will deal with it in code.

[kdm]

I hope so too. Thanks for your input and advice D and everyone else - much appreciated.