Forum Replies Created
-
Posts
-
LynxTwo is the fastest of them all as far as I know, and the most expensive, at over $1000 per card. It’s also probably the least reliable.
As I was saying, the card is not the limitation at these extremely low buffer sizes — the algorithms are.
The inherent delay of my low latency FM algorithm is 12ms, and that assumes the audio to already be in blocks of 512 samples. If you run my algorithm as such, the output will also be in blocks of 512 samples, but the data in those blocks will be delayed by 2304 samples due to internal algorithm filtering, so in fact it gets delayed by 12 milliseconds.
Calculating ASIO latencies is a pain in the neck. When designing the low latency algorithm, I actually ended up using scraps of paper on the table (to symbolize audio buffers) to try to wrap my head around it!
///Leif
ESI Juli@, M Audio Audiophile 192, Marian Trace 8, and LynxTWO will all yield basically the lowest possible latency, with buffer size 256.
Smaller buffer size than this could be a little bit faster, BUT, the internal buffer size of the algorithm is actually 512 samples. So, for ASIO buffer size 256, the program must buffer up two blocks (so that it has 512 samples), and then process ALL those 512 samples in the time of just ONE 256 sample block, so that it can output the first of those 256 samples during the current time period instead of having to wait for the next one. This means you need a faster cpu to run buffer size 256 than what is really necessary for the algorithm, because you need some headroom.
If we used buffer size 128, it would be slightly faster — (latency might be 15ms instead of 17ms) — but you’d need a much faster cpu, because now the CPU must process 512 samples in one 128-sample time period (and sit idle for 3 periods until it has the next 512-sample block!)… Unfortunately due to the nature of my clipper algorithm, it cannot be broken up into smaller buffer size without compromising audio quality.
Please note that this applies only to BBP ASIO.
Breakaway Live does not use the clipper back-end, and can use ANY block size down to 48 samples without uneven cpu load or extra latency. It does not have to be an even multiple of anything.
Marian Trace 8 is a nice card. 8 balanced analog inputs and outputs, DC STRAIGHT outputs (!!), and 192 kHz support. This card is ideal for an FM processor. It is quite a bit more expensive than ESI Juli@ or M-Audio AP192 — Trace 8 is €399 — but you get what you pay for. DC straight outputs means no tilt adjustment. It’s a different grade of card from pro-sumer ESI and M-Audio.
That said, ESI Juli@ and M-Audio AP192 are a great match with BBP ASIO.
Best regards,
///LeifESI Juli@ is nice! I don’t have one myself, but I’ve used it, and I can definitely recommend it.
It has 2 analog and 2 digital channels (single spdif). So, using only analog, you get FM Stereo MPX (17ms latency), and a mono ultra low latency studio output (8ms latency) to drive headphones, if 17ms is too slow for you to comfortably monitor. If 17ms is fast enough, just monitor off air. If you have a D/A converter which can handle 192k S/PDIF, you could get stereo ultra low latency output too.
You can also of course run in full quality mode, for 500ms on-air latency, and still have the 8ms low latency output.
///Leif
It processes just fine from Cyberlink PowerDVD when playing DVDs, I use it all the time.
For Blu-ray, I think we have our good friend DRM to thank. The audio is probably bypassing the pipeline for your security! 🙂
Jokes aside, that’s gotta be it. Try playing a DVD in PowerDVD. If the audio then goes through Breakaway, then there’s the answer. 🙁
///Leif
Okay, I’m at the bottom of it. 🙂
There’s one more thing to trigger the problem.. Aero has to be on.
With Aero, and a high DPI setting, Windows helpfully scales the whole window, and reports back scaled pixel coordinates for everything. When I ask Windows what resolution my screen is (1280×1024), it responds 860×563 or something like that. When I ask "where is the mouse cursor", it responds in scaled-down coordinates. However, when I tell it "please move the mouse cursor to X,Y", it moves to the REAL coordinates! So, if I ask windows "where’s the mouse", and then tell windows to send the mouse to that same place, the mouse cursor will move.. a lot. Doh!
If I didn’t know better (which I don’t), I’d say it was a bug in Windows 7 RC. When this happens, my program gets thrown for a loop, where the disparity gets interpreted as the user quickly dragging the mouse left, and the slider jumps all the way to the left.
The other bad thing about this DPI scaling is that it looks like crap! Breakaway is already a resizable app (at least the main window), so to avoid this problem (and the ugly scaling), I found out that there is an API to tell windows that "I’m DPI aware, thanks".
So, I now call that API, and the problem is gone. It’s not a great solution, because it’s now incompatible with visually impaired individuals. I will have to implement proper DPI awareness in the future, but at least now it runs.
0.90.75 is released! Check the first page of this thread.
///Leif
Hi Guillou!
As always, there is no single easy answer!
What I/O do you need? How many channels?
For example, EMU Tracker Pre is a nice cheap external USB interface, and it handles 192 kHz well enough for MPX output, but it’s only 2 channels! So, you have FM Stereo MPX out, and MONO studio ultra low latency out, and that’s it.
I have not personally used the EMU Tracker Pre.
I have an EMU 0202 USB, and audio performance is good, but it usually fails to initialize when starting the computer, so I would not use it for unattended applications. Also it’s still only 2 channels.
I have an EMU 0404 USB, and it works well, but it’s also just 2 channels.
I have an Edirol FA-101 and Edirol UA-101, and they both work well. These are 8-channel cards, so you have enough I/O for everything.
However.. NONE of the above interfaces will give you exactly buffer size 256 or 512! They give you many other sizes, but not those too. And that’s too bad, because those are the sizes used internally by the algorithm. It will run fine on a different buffer size, but CPU usage will be more uneven, so you may need a faster CPU to avoid glitches. Also latency will be a tiny bit higher, because it must still buffer to meet the internal buffer size.
Also, don’t forget that external interfaces are more expensive than PCI cards for the same audio performance. If you’re building a dedicated PC (which you should be, they’re so cheap), I recommend you buy an internal PCI card, and get the lowest possible price AND lowest possible latency.
Best regards,
///LeifI see it! Breakaway currently runs extremely poorly in Windows 7 x64 with font size 150%. Not only does the sliders pull all the way to one side with no hope of adjusting (WTF?) but the child process also completely fails to start, so I have to uncheck Audio Realtime Priority for it to run.
That’s not a very good user experience. I’m going to have to get to the bottom of this.
///Leif
Hi Torsten,
Glad you like it! Awesome.
1. Plug-ins get loaded before the Input AGC, so it is indeed performed on the raw mic input.
2. Phase tornado uses a total of 48 poles, in three groups at three different frequencies from very low to pretty high. I actually tuned the coefficients by ear, so I’m not sure of the exact frequency. The others, I need to keep secret. Part of the hard earned know-how that helps me to make a living 🙂.
3. There are no multiband clippers in any version of Breakaway, and there is no final clipper in Breakaway Live. However, the multiband limiters have impeccable peak control on their own, and only half a millisecond attack and release time. The output of the multiband limiters first go through an optional wideband compressor used in certain presets. When in use you can see it on the AGC meter, as three separate shades on the bar. It’s somewhat counter-intuitive to show it on the same meter as the AGC, but there was sensible place to put it, and in some presets this compressor IS situated after the input agc, before the multiband.
Best regards,
///Leifquote :. But the 0.5 sec delay on MPX don’t border me, as I listen to the studio output,I rather have “Full Quality” than lower quality and low latency on MPX output.I am so happy to hear this! I completely agree. Forget what I said about Low Latency, full quality indeed sounds better. 🙂
quote :So my request is that BBP ASIO could support two ASIO cards, one for studio output and one card for MPX.I’m afraid it’s not that simple. ASIO achieves low latency because Input and Output are hard synchronized, sample-for-sample, and are handled on the same callback. To use two ASIO devices at the same time, I have to put an asynchronous sample rate converter and buffers between them, and this will add at least 10ms latency — enough to push it over the edge for realtime monitoring not to be possible.
So, the short answer is, if you want low latency studio output through a 44/48k asio card, and also want mpx output in the same machine, there is a good solution, and you own it already. Breakaway Live + Breakaway Broadcast Processor 🙂.
Best,
///LeifThat’s an important clue! I never would have found that.
But.. *How* could this possibly be?? I can think of no rational explanation yet — it definitely shouldn’t have an effect.
I wonder if the same thing happens in XP with large fonts. Time to try it 🙂.
///Leif
Downloading Windows 7 RC x64. Will get to the bottom of it.
Damn I wish I had this kind of bandwidth at home!
///Leif
quote :I hade to set latency on my ESI-MAYA44 to 1024 (outhervise I got poping/cracking sound) , but stil studio out is good for live performance (I can’t here any delay on the mic). MPX is about 0.5 sec deleyed.Try running with or without realtime priority, it may make a difference. Also, if MPX is 0.5 seconds delayed, you’re running in Full Quality mode. Try low latency mode 🙂. There’s a switch in the signal settings window.
///Leif
It used to pump audibly (intentionally) and be really top-heavy and have excessive bass clipping. All those aspects should be significantly improved. Basically I tweaked it to sound GOOD, whereas the old version was tweaked to sound like 3FM.
///Leif
Hi Martijn!
I haven’t worked this out with Keith Edwards yet, but will very shortly. Do you own both BBP and Live already?
///Leif
