Forum Replies Created
-
Posts
-
quote :Skype Limited announced that SILK can use sampling frequency 8, 12, 16 or 24 kHz and bit rate from 6 to 40 kbit/s.
That seems pretty low quality to me, as is SVOPC which is where SILK came from. Not to mention it’s in mono.
It is half-decent for casual voice transmission, but certainly not broadcast quality because of the sampling rate limitations for starters.
-A- have you confirmed that the "Breakaway Pipeline" device shows up in your Device Manager under the "Sounds, video and game controllers" category?
-B- Did you restart SAM after you confirmed that "Breakaway Pipeline" has been properly installed? It’s probable that SAM doesn’t rescan for new soundcards every time you select the dropdown to pick an input in the encoder. 😉
I recommend starting with EVERYTHING in the middle to begin with. Speed 100 is VERY drastic. Even Speed 60 or Speed 40 can make a big difference.
Power adjusts the ratios of the internal dynamics processing. More power will have a higher ratio which will "push down on" the audio more when there is gain reduction. Some of those presets, such as New York, are already running at Infinite:1 ratio, so turning power above 50 will do nothing.
This is a good time to point out that deciding any of your adjustments without extensive long-term listening (like a day or two) and comparing good & bad with any previous adjustments… is not a smart thing to do, and you will not end up with the best sound you can get. It actually helps you finish sooner in the long run & get closer to your goals as well, if you make smaller adjustments then just listen for a day or two.
If you try to rush this process, or just guess settings without listening, then it will sound like that’s what you did. Not good.
in the 0 to 48kbps area, I would go with the latest Coding Technologies PS-AAC codec.
In the 48-96 area it’s pretty much a tie between CT’s latest HE-AAC, and the current aoTuV tuning of Vorbis (v5.7)… Vorbis has the advantage the higher the bitrate goes (by around 64kbps), and sits very pretty between the weak point of the transition from HE-AAC to LC-AAC, in the 80 to 160 kbps area.
What I would do is go with Vorbis at 128-ish kbps if you have the STABILITY at that bitrate. Otherwise go with PS-AAC at 48 kbps.
Breakaway blows away Sound Solution v2 which blows away Sound Solution v1.3 (which is what you meant)… I can attest to that having made great presets for all of them and pushing their limits.
The advantage Breakaway has over Sound Solution v2 even… is the ability to have much more consistency without having to sound like you’re slamming the bejesus out of the audio. The best I could get with SS2 was indeed pretty good, but no comparison to the maturity and clarity of Breakaway.
As far as your LPFM at home, you might check out the consumer (or do we call it pro-sumer?) Breakaway Audio Enhancer, because it has an "FM Mode" that’s especially designed for LPFM transmitters that only have un-emphasized L/R inputs. I forget where Leif put it (maybe he can enlighten us) but there’s somewhere a description of how to calibrate the "FM Mode" in BAE to sort of pre-compensate for the effect your LPFM’s pre-emph/limit/stereo-gen/etc will have on the signal, and you can end up getting a decently competitive signal out of it on air. Nothing major market, but certainly better than just running un-pre-compensated into the L/R inputs.
Yeah the netcode in EdCast is just whack. But really you can’t blame it on Ed, he’s a darn busy dude as one of the technology VPs and one of the lead coders here:
http://www.andomedia.com/
I know first hand having consulted for them before.LeifCast has great netcode, very stable in the presence of networking whackness. 😛 And I coded one myself for a private webcasting software for a station that is as stable, and has something I’ve never seen done before… a meter of the lag between data waiting to go out and data actually sent. Currently I have it set with a 30 seconds of audio sized buffer, and as long as the meter/buffer doesn’t fill up, you can eventually catch up to the server and nobody will be any the wiser.
EdCast on the other hand will glitch like crazy if it’s not able to send every single packet before the next one is encoded, so if your Tx buffers on your networking devices get filled up, bam… skipping like a lil kid on its way to church.
Not sure on the buffer size in LeifCast, but I’ve never had any skipping or problems while using it.
And for New York preset, while you’re at it, set the pre-emphasis to 50uS. It sounds better there when you have the option to do so.
Soft clipping just means that the peaks start to distort before they reach the threshold of pure straight-edged clipping. Think of it like a compressor’s soft-knee function. 😉
Distortion-canceled clipping is more of a generic term and can vary from somewhat complicated and not very effective (like an Orban 8100A) to really really freaking complicated and totally effective (like Breakaway).
Ideally… a distortion-canceled clipper will either only clip the audio when the distortion won’t be audible, or it won’t clip the audio. That’s what Breakaway does. All other previous clipping designs are just using methods of reducing the audible distortion caused by the clipping. Not to actually prevent the clipping if it were to be audible.
For this reason, Breakaway should probably be considered to be the first true distortion-masking clipper. If the distortion can’t be masked, it will not clip the audio. Other supposed "masked" clippers are, for lack of a better word… clearly… not living up to the definition.
It won’t be that way forever though. I know of at least one forthcoming box from one of the large broadcast processor companies that has a clipper that sounds pretty darn clean loud. Will it be available in a $199 software version? Dream on. 😆
lol, Leif already knows I can code. we bounce stuff off each other from time to time. 😉
as far as your account, ehh… if i figure that out, i’m not telling ANYONE. lol
I would say the first big problem is you’re comparing their net-radio to your FM. Two different beasts ESPECIALLY if you’re trying to be loud at all on the FM dial, especially at 75uS.
First off what you should do is start with a preset that is the closest to the sound you want to get, changing ONLY the final drive to see how it changes your sound on air. Take at least an hour listening to the presets you think get pretty close, and adjusting the final drive slowly till you are getting close as they will get.
Then pick the preset you think is closest of all, and then sit on that for a few hours, and really *THINK* about how it doesn’t operate the way you want to hear, and how increasing or decreasing any of the other sliders will make those aspects better and what other aspects they could make worse. THEN start to very very slowly adjust the settings.
Audio processing is a series of compromises, and you try to make the compromises that interfere the least with everything else that you like about the sound. It takes a while to figure out if any of the decisions you are making were good or not in the long-term average.
Also "final drive" doesn’t go to 100. It goes from -6db to +6db. I can’t think of any preset that sounds good at +6db ESPECIALLY with power and speed also cranked to the max. That’s probably a huge part of the problem. (you didn’t say what preset you were using, which is the largest factor in trying to figure out why you sound bad, but the cranked drive & speed is pretty easy to figure out as being a bad idea)
The other thing I can think of would be to make sure you calibrated the tilt correctly at least using a scope, if not a decent tuner into a scope or a calibrated (or DC straight) input and using MPXTool. If you don’t have the tilt and frequency response perfect, you’ll be compromising your sound no matter what presets and settings you’re using.
[quote author=”Leif”]That should do it![/quote]
nope. that would also only increase the value of the pointer. not the value that the pointer is pointing to. for instance in C++, if "MyAccountNumber" was a 32bit unsigned int…
code :int *MyAccountNumber = *(int*)[some_pointer], a;
while (a < 100000)
{
// add 1000 euro to my account
MyAccountNumber = MyAccountNumber + 1000;
a++;
}or if MyAccountNumber was already defined only as an offset and not a full fledged locally mapped pointer…
code :int *cashToSet, a;
while (a < 100000)
{
// add 1000 euro to my account
cashToSet = &MyAccountNumber + 1000;
memcpy(&MyAccountNumber, &cashToSet, sizeof(int));
a++;
}profit?
extra credit bonus:
code :for a = 1 to 100000
rem add 1000 euro to my account
&MyAccountNumber += 1000
next a8)
(someone someone got the puns, lol)
There used to be Volume Logic by Leif for iTunes, but obviously not system-wide. And that was sold to Plantronics and shortly/unfortunately the product was dropped.
There’s a possibility that Breakaway could find its way to OSX as a system-wide processor. But that would be a question that only Leif could answer as far as probability.
Yet it "looks" ok. Proving you can sometimes look at a track to tell if it is awful, but you can’t just look at a track to tell if it sounds good. SeeDeClip actually does try on Say It Right, and it certainly comes out better, but… nothing like the Because Of Your or other "legitimately" digitally clipped recordings.
Still… better than anything else I’ve used to try to "fix" stuff like this, including software over a grand, and the SADiE with CEDAR I used to own. A lot of declipping tries to artificially shape the clipped peaks. SeeDeClip is actually somewhat simpler at face value but a much more ingenious way of "seeding" the information for the process, by taking the other channel into account.
This actually DOES sound less distorted. Not remarkably, but noticeably for sure. (sure enough for me not to bother blind ABXing it)
Check it:
