Questions & Answers

Improve efficiency of Studio One Dropout protection

+33 votes
1,840 views
asked Jun 28, 2018 in Recording by alancloughley (2,330 points)
Turning on dropout protection causes dropout issues on large projects that use CPU intensive VST plugins. Even on smaller projects Kontakt and other CPU hungry VST instruments can be a big problem and these normally run better with Dropout protection off.

I've tested the S1 dropout protection feature extensively on 4 different PCs using the Quantum, StudioLive 32 Series III, Propellerhead Balance audio interfaces as well as on-board audio on a couple of laptops (all WIndows 10).

With dropout protection on, selecting record enable on Multiple tracks (e.g. a set of drum mics) or VST instruments often causes dropouts to occur as the CPU spikes on one of the cores.

I therefore never turn this feature on as it causes more problems than it solves. Dropout protection is a fantastic concept but unfortunately it doesn't work well under load which is when it would be most useful.

An improvement to the audio engine to either run VST instruments and plugins more efficiently OR spread the CPU load better across cores would make this feature a lot more useful.

13 Answers

0 votes
answered Jul 15, 2018 by alancloughley (2,330 points)
 
Best answer

Dropout protection is hype and most users have bought into this however it doesn't actually deliver any real world benefits.

Switching it on appears to make CPU performance dramatically improve but all benefit is lost when you record enable tracks which use CPU hungry plugins or virtual instruments as it then causes dropouts. 

After extensive testing on numerous machines from low power laptops to powerful DAW optimised PCs I have found that it is very often best to leave this feature entirely switched off.

At it's best Dropout protection offers a marginal improvement on some systems but beware I have never found this to be the case as it always makes real world performance on my projects worse.

Presonus support and even experienced users have been duped by marketing into believing the hype but the extensive real world testing I've done has proven that this feature would need to be significantly more efficient to make it worth turning on in most cases.

I'd love someone to prove that I am wrong however that is yet to happen and until they do I'd consider this feature offers no significant gains at best and actually degrades performance at worst. Presonus need to VASTLY improve the feature so there is actually a real world benefit to switching it on. Another forum member is looking at a project I've sent over but I've yet to hear back, his testing has confirmed some of my findings (read thread to see his results) he has seen a small marginal performance increase on some projects but this is hardly earth shattering stuff in terms of real world gains.

A practical alternative approach

If Presonus can fix then then hooray! If not then another option might be to abandon the vastly inefficient aspirational double buffer architecture and implement a less sophisticated simpler and more pragmatic single buffer solution. 

There could be a recording mode toggle button which drops the buffers size so audio tracks and virtual instruments can be recorded in real time at a low buffer. When switched on this could disable heavyweight plugins (perhaps just the worst offenders? until the PC can cope with the load at the low buffer size, it would need to auto-compensate levels a little to prevent gain issues on tracks from EQ and compressors but this would be fairly trivial to implement. The mix wouldn't be perfect during tracking but there would be no dropout issues during tracking at low latency and when the recording mode was toggled off again the buffer would return to a high settting again allowing users to run many more plugins. This would be much simpler to implement, yes more crude and not perfect but it would deliver MASSIVE real world benefit unlike the present approach which is deeply flawed.

Maybe Multicore support will help!?

The number one feature request for Studio One is for general performance improvements for VST plugins and instruments (not sure how to add the link here just look at answers and select "Hot" from the main menu it's the top by a mile please select it while you're there!) and is a request to add Multicore support (especially for VST instruments) this might be related and therefore improve things by dramatically reducing the risk of single core overloads which I've seen due to the terrible efficiency of Dropout protection as it stands.  

0 votes
answered Jul 2, 2018 by robertgray3 (27,490 points)

Just curious, can you PM me a slimmed-down project where a few tracks run better without Dropout Protection than with it? My username on forums is robertgray3.

Dropout Protection should be pretty intensive when you monitor multiple tracks (assuming you have low latency monitoring for instruments enabled) but it shouldn’t perform worse than when it’s disabled.

I’d love to help figure out what’s going on and how the projects you’re talking about are pushing the limitations of Dropout Protection- might help figure out how to improve it.

I do wish Dropout Protection didn’t have as big as an affect on bounce / render accuracy as well as multi instrument playback timing, but that’s a separate issue.

0 votes
answered Jul 3, 2018 by alancloughley (2,330 points)
edited Jul 3, 2018 by alancloughley
Hi Robert,

Pretty well any S1 project that loads the CPU has this issue. It's all about load though so you need to tax your CPU. Dropout protection is only really useful under load so if there's no load issue there's no need to turn it on!

It's easiest to see the problem on an under-powered dual core laptop (where Dropout protection would be really useful!), you can simulate load of a heavy mix by dropping your buffer size to make the CPU work harder.

1. Open S1
2. Select File > New Song
3. Select the second template, Band Template
4. Go to "Audio setup" and set dropout protection to Minimum
5. Set Device block size to something that might push your PC a little(on an i5 dual core laptop 128 samples worked for me)
6. Select Record enable on the Drumset folder track which will record enable all the drum mics

Everything should be fine at this point, the performance meter should sit around 25-30% and will record the drums fine even on a modest laptop. If your PC is not hovering around 25-30% you might need to drop the buffer a little more to reproduce the issue but you should still be well within the capabilities of the machine and be able to easily record an entire drum kit with the 16 insert effects on this project at reasonable latency with a decent audio interface.

6. Deselect record enable on the Drumset folder track
7. Set dropout protection to maximum and enable the green Z
8. Select record enable on the  Drumset folder track

I find typically the CPU will spike to 100% on a low spec dual core i5.

This should demonstrate the problem which is that almost any project running under a reasonable load can easily cause spikes if anything a little power hungry runs on the secondary low latency buffer i.e. Dropout protection is switched on.

NOTE: The key point is that a modest laptop has enough power to run this project at a low buffer setting with dropout protection OFF but the act of switching it ON causes a massive CPU spike on one of the cores. On this project there are 16 Presonus stock plugins and this causes dropouts on my laptop, but obviously mileage can vary. The more you load a PC with dropout protection ON the worse it gets. If you load up a couple of VST instrument like Kontakt, Maschine and record enable a track at low latency on a heavily loaded 30-40 track session then the same pattern occurs even on powerful overclocked workstations.

NOTE 2: It also depends on the track you select as they will be distributed across the available CPU cores, on my poor old laptop I can record enable a single drum track and max out one of the cores but this will vary across machines and sessions. I suggested selecting all drum tracks to help demonstrate the issue. You can perhaps try disabling some tracks and you'll see that some will NOT be on the worst performing core which is the one the S1 performance meter displays, it can look like record enabling the tracks has no effect on the CPU. This is unlikely to be the case, they are likely running on another core which is just not showing up on the performance meter. This might cause some confusion as you will not see this issue until a project is loaded sufficiently AND a track running on the most loaded core is record enabled AND dropout protection is turned on i.e. the effect of poor dropout performance can be masked by S1's own performance meter implementation!

I could be missing a trick here though so if you've got any ideas about how to stop this happening that would be awesome!

Thanks, Alan
0 votes
answered Jul 4, 2018 by robertgray3 (27,490 points)
edited Jul 4, 2018 by robertgray3

I think when you get down to certain computer specs Dropout Protection may not help as much as you would assume. 

It sounds like you have “low latency monitoring for virtual instruments” enabled in the Audio Setup submenu. I believe Low Latency Monitoring works by loading a duplicate of the plugin instance. If your processor/CPU is underpowered enough, the simple act of managing the duplicate instance may be too much. Assuming you already disabled hyperthreading.

I run a version of Studio One v3.5.6 on a single core laptop and I think I can only use up to a certain level of Dropout Protection... I think “Medium”. It smooths out my CPU usage but I seem to recall overall usage going up as well. And if I turn it up further I run up against my CPU’s limits.

To be clear, Dropout Protection works as expected on my four-core system. Turn it off, CPU goes up, turn it on, CPU goes down.

In theory, yeah, I get the expectation but Dropout Protection, like ASIO Guard and other schemes, is a bit of a “trick” that I’m going to guess requires a certain amount of processing power to pull of. As much as I’d love it to for my single core POS netbook :) I don’t get the impression that Dropout Protection was designed to make Studio One run really well on underpowered systems. A system like REAPER’s anticipative effects processing is more geared to that style of performance. 

So if S1 runs OK in a low buffer setting for you, maybe just keep it that way.

+1 vote
answered Jul 4, 2018 by alancloughley (2,330 points)
edited Jul 4, 2018 by alancloughley
Thanks Robert,

I use a 6 core overclocked audio PC, unfortunately exactly same thing, dropout protection causes crazy dropout issues so is always set OFF. It's workable but kills stone dead my plan for that dream tracking and writing template with everything setup and ready to use though :(

Yes switching dropout protection on causes the performance meter to drop for me too, I assumed things would therefore run better but it has crippled every PC I've ever tried it on in terms of CPU headroom for record enabled tracks on my typical sessions.

Are you using any VST instruments like Kontakt and do you ever need to record enable more than one track in a session,? Also do you track with any VST inserts enabled?

I've had such a bad experience with this I'd advise only ever turning it on if you must, absolutely NEVER turn it on until you absolutely run out of CPU..Turning it on at this point might help in which case great, but I've found it always makes things worse for me if I need to record something.

Yes I use low latency monitoring for instruments, I can't see how S1 can run two instances simultaneously of VST instruments though as that would really confuse Maschine and Console 1, as they talk to their own software and everything would fall apart if it worked this way. My guess is the problem comes from a loss of efficiency running two sets of tracks on two different audio buffers and having to sync them back up. I can see why it's a challenging feature to get right but as it's so bad at the moment there's huge scope for squeezing out more CPU headroom whatever the cause is..

I've tried it before but will give the medium setting another go this weekend to see if there's any difference on the latest version of S1, thanks for that suggestion, fingers crossed!

I've have had a quick shot at Reaper it seems to be a fair bit better at running VSTs..S1 is a lot slicker though, I love the arranger.and workflow add in the integration with SL32 and I can't really contemplate moving to another DAW right now! This is the only feature I hate about S1 it really sucks badly for my typical projects!

Cheers, Alan
0 votes
answered Jul 6, 2018 by robertgray3 (27,490 points)
edited Jul 6, 2018 by robertgray3

Hey Alan,

You know what, it's funny, now that I've done some more testing I noticed the following things:

On a minimal or less complex song Dropout Protection on "High" brings the idling / playback CPU usage as indicated in the Activity Meter down a lot. But when I enable monitoring a track, the Activity meter shows more usage between the tracks playing back and the monitoring track than when I turn off Dropout Protection entirely. This particular situation fits your description of Dropout Protection as adding to the CPU as indicated in the Activity Meter

NOTE: the activity meter can be a bit deceiving. It shows a sort of compiled number of estimated % of total capacity and is skewed towards the most loaded core.

 When I take that same song and create a duplicate of every track to simulate a more complex song, Dropout Protection on "High" has the Activity Meter shows an average of 42% with no monitoring and 57% with monitoring on and me playing tons of voices on one instrument. No audible dropouts. With Dropout Protection disabled (Minimum) the Activity Meter shows an average of 64% to 72% when I'm playing that monitored track. Still no audible dropouts, but this situation fits the intention of Dropout Protection as allowing you to manage more complex sessions while allowing some room for one low latency track to do an overdub etc.

So it seems like depending on the situation it could be a balancing act... I remember testing this a lot when I first moved to S1 with a big session I was working on and Dropout Protection was useful in that situation but I'm really just monitoring one or two tracks at a time. Unless it's a multi-instrument, then it places it all on one core due to the core per track layout. That can get pretty hairy.

My instrumental sessions aren't terribly complex. Usually a combination of audio tracks and virtual instruments totaling up to 25 - 45 tracks. When I'm at the higher end I usually end up freezing two or three tracks, usually my complex synth leads or something like that. 3 to 4 instances of Kontakt. Usually 1 of Serum, Mai Tai, Vacuum Pro, and Strike respectively. Some instances of Presence and Structure 2. Impact and Sample One are always in there but they're not very CPU hungry. 2.5 ghz i7 macbook pro Quad Core. so its not really that great a machine. Also I'm typically only record enabling one to two tracks at a time though.

And yes, Low Latency monitoring 100% causes some issues in a few plugins. Theres a list on the forum about it. It's a short list though but I think Softube Console One is on it.

I hope you can figure this out and not have to switch! :D REAPER's great, bulletproof, and it performs well on computers you can literally find in the bottom of a dumpster. I used it for years, but time is money and S1 for me is miles more efficient as a working environment. I just resigned to the fact that I may have to pick and choose systems based on Studio One performance and it not having any weird gremlins that would tank that.

0 votes
answered Jul 6, 2018 by alancloughley (2,330 points)

Thanks SO much for taking the time to test that Robert very much appreciated :) 

Sounds like you have exactly the same issue but as low latency monitoring doesn't work you've turned it off. Low latency monitoring when tracking Maschine drum parts is pretty important to me unfortunately :( 

I bought the Presonus StudioLive 32 (awesome and recommended btw! - 32 channels of USB audio at very low latency) When tracking a live band I set the buffer a little higher and monitor through the desk so again dropout protection is switched off otherwise I get dropouts when record enabling multiple drum tracks.

This is all still a little mind blowing though when you consider what's going on under the hood. To provide context I find on my typical projects: 

A. Dropout protection OFF:  
ALL tracks and plugins, VST instruments (often 30+ tracks tons of CPU intensive plugins + 3-4 Kontakt instances!) 
ALL tracks running at a VERY low latency and low buffer of 64 samples 
Result: Works fine on my 6 core :)

B. Dropout protection ON:
ALL tracks in the session are now running at a massively increased buffer size of 2000+ samples taking a massive load of the CPU 
Record enable a SINGLE track so just one track now runs under low latency monitoring at 64 samples(remember this works fine with ALL tracks running at 64 samples not just one!) 
Result: Performance meter hits 100% and dropouts occur :(

There should theoretically be a lot more headroom with this feature on to run a single VST instrument on one of the cores. 

Presonus hopefully can find a way make this run more efficiently with low latency monitoring of instruments ON. Of course it's a challenging feature to get right but I'm not looking for miracles, my feature request is simply to improve efficiency a bit to make it more workable on a heavily loaded session for two these two main scenarios:

-Drum tracking (ability to record enable a reasonable number of tracks without dropouts)
-VST instruments like Kontakt/Maschine (ability to record at least one instrument with LLM without dropouts)

I'll try disabling Console 1 plugins too that might help but it's not a very helpful workaround lol!

I'm with you though S1 is too good a DAW in every other way so I'll just work round this by using less plugins, enabling/disabling plugins where required and bouncing tracks to audio. 

I'm hoping they improve this feature so it works with LLM as it would be a complete game changer for me in terms of S1 workflow!

Thanks again for taking the time to test this, the useful suggestions and the additional info! 

Cheers, Alan

+1 vote
answered Jul 6, 2018 by alancloughley (2,330 points)
Sorry Robert, having read your post again it sounds like you are using low latency monitoring for instruments. I think you were referring to some VST insert FX having issues when this is switched on is that the case?

If you do have LLM on, it sounds you might be getting slightly better performance with Dropout protection on with the mac 4 core than I am on my overclocked 6 core desktop, although I'm perhaps using different plugins and instruments which is why my machine might be getting loaded differently. I can get VST instrument dropouts at a lower load that that but it depends on the instrument (Abbey Roads drums can be problematic!) As mentioned though it can be tricky to see as some record enabled instruments will be assigned to a less loaded core, track selection can be a lottery with Dropout protection on so it's worth selecting a few different instruments to see if things are better across all the tracks if you haven't already tried that i.e. the test instrument you've picked to hit hard might be running on a good core.

Although loaded heavily it sounds like you still don't actually need to turn dropout protection on (kind of impressive for a 4 core laptop!) and can run ALL your tracks at very low latency at this point across the board. If you have the time to take it to the point where you absolutely do need to turn dropout protection on (say another 10-20 tracks) it would be interesting to see if it's still possible to record VST instrument tracks with dropout protection on and off. It does sound like you are seeing a slight improvement with dropout protection ON which I've never seen with VST instruments and projects under that load, I'm perhaps using different plugins and this feature might work better on Mac than PC? It's not a huge difference but it might be useful if does work as you increase load further to the point where you would have no option but to turn the feature on.

Thanks, Alan
0 votes
answered Jul 6, 2018 by robertgray3 (27,490 points)
edited Jul 6, 2018 by robertgray3
Hey Alan, thanks for re-reading my post.

Shot you a PM on the Presonus forums so we can test further

Yes, Dropout Protection absolutely allows me to playback and monitor one to two tracks in more complex songs on my system that aren’t quite doable Iin “Minimum” mode. When I said I could observe your issue of DP increasing the cpu usage over leaving it disabled it was only on minimal songs where none of this really mattered anyway.
 

maybe we can compare a test song and isolate why my lesser computer is getting better performance on some sessions. Even audio computers have little wacky things in the BIOS or in the settings that make a marginal improvement on heavy projects, where every little bit counts.

If it helps clarify some specific situations where DP might be able to be improved that’s cool too, but I feel like there still may be room to improve within your setup from the situation you describe.

Cheers!
+1 vote
answered Aug 17, 2018 by tarsonis (2,880 points)
Same Experience. Totally useless for me running a i7@4Ghz and RME HDSP ITB. For which this "stunning" feature was advertised. Seems to work for some folks. I always got huge projects with a track count >50 tracks and sometimes a ridiculous amount (sounddesign) of processing. In those scenarios this feature would come really handy as its useless for a project when I start out with few tracks (the latency is already low with the PCI version) but it is really a performance hog.  Destroy the groove and timing of the project. And not forget huge amount of dropouts and glitches. Its useless or in beta phase and buggy. Ideally it work out if one had the chance to disable all plugins temporary. As this seems to not going to work with my workflow with a lot of latency processing, busses and pre master. What is the sense if one can play nearly without latency but the hole project goes nuts playing random timeless noise.
+1 vote
answered Jan 8 by franckmariani (1,730 points)
Same thing here. Super annoying to say the least. My I7 4790k seems like an old pentium on SOv4... Presonus please this is your top priority.
0 votes
answered Apr 18 by robertgray3 (27,490 points)
Thanks Alan for sending me that song! That particular song does in fact handle overdubs smoother with Dropout Protection off. It's funny since I have some songs quite a bit larger than that and they seem to run better with DP on. I'll research this a bit when I can to try and find a pattern.
0 votes
answered Dec 1 by robertgray3 (27,490 points)
Ran into it again- VPS Avenger uses MORE than twice the CPU when the two (Bc of dual buffer) instances are loaded in low latency monitoring mode.

so a patch using 35% of the CPU at each note-on with Dropout Protection disables, when I put it in LLM mode, somehow is spiking to 100% of my CPU on every note-on

I think it has to do with core synchronization but I’m not sure, at least that’s what the developer of Acústica said when I asked them about why their plugins instantly **** out in LLM mode.
...