- This topic has 14 replies, 3 voices, and was last updated 1 year, 5 months ago by
.
-
Posts
-
Hi, I wonder what is the lowest or average latency are you able to achieve on the HD Cores with Kernel Streaming ? Is it bad to have like 90% jitter ? This value was not prexent in older Breakaway versions, we used to set the buffers to the lowest stable value, maybe we were running it with very high jitter percentage without knowing it. Also in older versions there was a Buffer size and Buffers count values but no Buffers count in BreakawayOne, does it mean that there is only one buffer ? Like if I used to make the buffers size at 128 with 2 buffers in Breakaway Live is equivalent to 256 in BreakawayOne ?
Thanks
The general aim is to minimise the jitter figure, because that number represents the number of “lost” frames where the incoming (your playout software) data doesn’t quite fill out the sound device’s buffer size.
Sometimes, it is better to not use KS, even though that is the optimum setting, but you can always experiment and possibly come up with the best KS buffer size to give you the lowest jitter (and therefore) latency/jitter.Ok, what I have found so far in experimenting with maybe 6 or 7 sound cards is that Wave input and output can’t never have smaller buffers than Kernel streaming, at best it will works the same and begins to becomes glitchy at the same buffer size. Also the automtamic buffer selection thing almost always come up with a pretty big buffer like I just tried with the Pipeline and the result is 960 with 8% jitter and Safe alternative 1024, I just tried again the automatic thing and it’s now 480, Alternative: Safe 1024, Low latency 240, but I run it since last year at 160 with 50-60% jitter and I never have drop outs, clicks, pop, etc. That’s why I asked because the difference is huge between 160 and 1024, it quickly comes to a point that the audio is not anymore in sync with the video when you’re listening TV programs or movies.
The sound cards I tried also give me results in this range, and strangely the one that is stable at the smaller buffer size is one of those cheap no name usb sound card based on the C-Media chipset CM6206, you can find them for 20$ everywhere on shopping sites, they work with a buffer of only 96, and on top of that and what I don’t understand is that the EQ and effects things that come with the driver continue to work like they were implemented in hardware, but they are not, so I don’t how they managed to tweak their drivers to achieve that. All the others the effects and DSP are bypassed like it’s supposed to be when using Kernel streaming and the Behringer UCA202 also has it’s volume control bypassed. I tested all this on 4 computers with Windows 8.1, 10 and 11, none of them are really dedicated exclusively for this task, it’s mostly for local listening.
Greeting, folks!
Following the latency questios I’ll use this thread to ask something about jitter and how BreakawayOne inputs work.
So, can a 40% jitter figure at ASIO degrade the final audio quality? In my case, I run one interface for audio input and low latency monitoring, and I run a second interface in KS mode, with a large buffer with very low figure.
I noticed that I can use the same physical input of the ASIO interface for a separate input via KS, but I can use a bigger buffer resulting a very low jitter figure.
So I would like to understand if the KS input is independent of ASIO at this point. Because if it’s independent, I wonder if it’s not interesting to use a separate HD Core in low latency mode, just for monitoring purposes, while the FM Core uses the KS input.
And is it possible to run a Full FM Core with LL processing with ASIO drivers and thr FM audio feed through KS?
If it is possible to ask Leif, I would like to know if there is no possibility to change the ‘Warning’ counters in the jitter indicators running BreakawayOne. Currently, the warning numbers here, after 6 months are scary and occupying a big display space…
You’re often left on your own with theses softwares, and many questions remain unanswered with these details. I’ve noticed that the size of the buffer is irrelevant to the number of times a device is used. For example, I can use Breakaway Pipeline 1 in 3 HD cores at the same time and still need to use the same buffer in all 3 places. And yet, and this is one of the things I didn’t understand, it’s written that you must not use the same device in several places at the same time. I’ve always done this, but often the output of an audio interface will work in only 2 places at the same time, in Kernel Streaming. I always try to keep the jitter below 100% and normally I don’t get any glitches, but sometimes it can vary a lot for no apparent reason.
Hey @milky
Any information about what quote above?Hey @skytrax
My quote about audio quality is not only about glitches, it’s also about how BA1 deals with high jitter figures. If it discards audio frames to allocate in a buffer packet, for example.
One thing I noticed when I see high jitter oscilations is music ‘losing rhythm’.
And if I know that the KS input with high buffer sizes and low jitter reduces these effects, I guess is more interesting to run a BA1 LL instance for monitoring purposes and other for quality audio processing.
Ok I see, but what is an acceptable jitter figure ? I saw that it will tell that it’s low with 1-2 % but it will often jump to several hundreds or even thousands % without obvious dropouts. That’s why I said in the beginning that since the old versions didn’t have jitter indicators and we used to adjust the buffer size by ear maybe we were running it with like 200% jitter all the time without knowing it. It looks scary the first time you realize that your is at 250% and it still play without being completely choppy.
Just for curiosity purposes: can you describe the system you’re running (audio interfaces, CPU, OS…)?
And theres no dropouts in any ocasion with the 100-200% jitter figures?
I got used to see 5% jitter in my Delta 44 buffered at 128 and now I’m running an UMC404 with the sammer buffersize but with continuous 45% jitter (both ASIO). For me this is something alarming and may you comfort my soul. 🙂
I used Breakaway Live for several years and switched to BreakawayOne like 2 years ago. Iwanted the lowest possible latency because I was using DJ CD Players and now mostly DJ Controllers and was annoyed by the delay when you press Play, it needs to be mostly instant. This was in the first time on a Core i5 3570K a Core i5 6500, a laptop Core i3 4030U and since a couple of month a Core i5 13400F, all with Windows 8.1 upgraded to Windows 11 last year. Like I said earlier for the interfaces I’ve been trough a long list with almost all consumers grade Sound Blaster Audigy SE and 2ZS that have seen the most use and that were great for the latency and an X-Fi surround 5.1 pro USB and a Behringer UAC202 USB that needed bigger buffers plus some cheap no name USB sound card based on a C-media chipset that were strangely really good for the playback latency but unusable for the line input, they clip like crazy and have way too much noise. Now I use 2 Asus Xonar DX that work great.
It begins to dropout between 150% and 250% depending on how it feels but BreakawayOne is more stable and can achieve lower latency when running like a service, some periods when running not like a service I always had the “I/O Thread” (with a percentage) flashing yellow, without actually dropping, it’s another thing that we don’t know what it means exactly and how it works. Sometimes it will flashes yellow at 12% I/O Thread and sometimes it will not flash at 25-30% I/O Thread.
I have tried to maintain the jitter in the 5-6% range but that means buffers most of the tim in the 480-1920 range, and it’s too much for me. Most of the time if it doesn’t work with the buffer at 256 ar lower I simply stop using it. If it’s for listen to the TV or background music I can use 480 but not much more because i’m annoyed by the image starting to go out of the sync with audio.
Also for the PC’s CPU and OS I have not seen a significant difference between any of them. I had the same latency on the old Core i3 laptop that for the i5 13400 that is 10 times more powertful. No difference either between Windows 8.1, 11 or 7 and 10 for the short time I used them, but all the PC’s have always had the same Windows configuration and the same set of software, almost like if they were cloned.
As stated elsewhere, jitter is simply a percentage of discarded frames because the incoming samples “didn’t fit”. This is why usually (but not always), the buffer count should be a multiple of the sample rate. For instance, if the sample rate is 44.1, theoretically buffers of 441 or 882 would “fit” all samples, with very few frames rejected. However, it takes longer to fill 882 buffer than 441, so this introduces latency, because nothing is passed to the output until the input buffer is filled with samples.
A low jitter rate would probably not produce audible artefacts, simply because there are enough useable complete frames to mask the dropouts. The higher the jitter percentage, the more rejected frames, and therefore sooner or later, you will hear artefacts. I am amazed to hear reports of no dropouts with 100% jitter. The processor must be running on the edge to keep processing the frames and passing them to the output.
Latency simply comes from processor capability. The audio has to come into the input buffers, and then feed through the audio processing algorithm and preset parameters, and then pass to the output buffers. It stands to reason that the quicker it can do this, the less latency is going to be experienced. Sure it is possible to run BA1 on a very low powered CPU, but the latency will increase. Similarly, if you run a top-end i7, but have other CPU-cycle-hogging processes running on the same computer, BA1 will not get enough of the CPU time slice, and latency will be the result.
Just as a finely tuned race car delivers better speed and therefore more wins, a high speed CPU with very few other tasks running will deliver better latency.
I made a quick video of it playing a song in foobar 2000, it’s on Google photo.
I’m not sure if the first link works, I made another one
- You must be logged in to reply to this topic.