Stereo Tool

Could be caused by the high CPU load (have you tried turning some filters off? Especially Loudness?). Or by a real bug in the code - SAM is always way more sensitive for bugs than Winamp.

yes.. eat more cpu .. most noticable on AMD (i think) on my e8400 it's 18-22% beta3 and 20-25% defaut_v4.22. Hope that will be solved.
More important is quality... comparing v4.22 and beta3 (Normal Latency)
- hard limit is OK, no peeks.. but i swear i saw it on dual core AMD pc for our radio streaming. must try again later tonight.
- Sound level is 0.1-0.2dB higher .. tested preset "Web Radio Stereo"
- in beta3 multiband clipping is much better, beta2 too
- Avoid Distortion in Final Limiter still have crackling on right channel .. without that i can hear it is good too.
- bandpass filter is good
- Very deep bas protection OK
- ILDP - OK
question .. remove remaining peeks in PreLimiter is working or not ?
-----------
comparing v4.22 and beta3 (Low Latency) Web Radio Preset

Overall much much better is beta3

- cpu usage is 15-22%, close usage with normal latency
- i am glad that bandpass is working wery well
- Latency is very small even with loudness
- ILDP in loudness must be ON or low bass is little unstable + some little harmonics
- Stereo boost is very well and i think it's better then default 4.22 normal/low latency
- Same question for Pre Limiter, remove remaining peeks working or not ?
- Pre Limiter, maybe needs few miliseconds LookAhead.

I did't yet try MPX output. Before that, question.. is there any changes with LowLatency ?

One thing what i see many times on default 4.22 version is Meter for Stereo boost knows to freeze.

New BETA3 is online now.


Some final remarks for today:
1. The sound of the new version(s) is indeed clearly worse for certain types of sound.
2. I'm not really sure if that's also the case for "Low quality" mode. (Bojcha, if you'd like to check...)

Ad 2.: IF the difference is NOT present in Low Quality mode, then it must be possible to solve it in High quality mode also. But it won't be easy...

Bojcha:
- CPU load will be lowered (but that will cost SOME quality).
- Hard Limit should be fixed now (I *think* I know what caused it)
- Remove Remaining Peaks is working
- Still working on the quality (first goal, as I mentioned earlier, is to get the same quality in _normal_ latency mode. When that's achieved, I'll try to improve Low Latency mode, lower the CPU load etc.
- Sliders are not working very well at the moment, and some don't work at all. The framework around the processing is largely rewritten, and I haven't adjusted the slider to the new setup yet.

About Low Latency mode: It's currently set to 17 ms, but later I'll also enable selecting 34 ms - which should sound much closer to the normal latency version. And maybe 9 ms for very low (or no) Loudness values... (Bass protection will be almost impossible there).

For Low Latency Cpu usage is about good, but on my e8400 30-38% standalone version with FM out and clipper at 101% and all rest on.

This LowLatency mode in beta3 is awesome, i just try FM output, and for this latency, this is super!

edit:

FM output on beta3 (low and normal latency) .. has no constant peeks to 100% modulation ..
according to mpxtol .. audio level is lower, about 90-91% .. and with default v4.22 Audio level is 93-94%

Ah, I missed something there. In v422, I turned the final stage of the "hard limit" filter off when FM Overdrive was used, and used only the FM Overdrive limiter. So I let peaks through if they didn't collide with the stereo pilot or RDS signal.

In the BETA3, the final stage of the Hard Limit filter is not turned off, so the signal is already limited perfectly before it enters the FM Overdrive filter. For now, set the FM Overdrive volume to about 106%, I'll fix this later.

I think I've found (part of) what's causing the quality difference between v4.22 and the BETA3 in normal latency mode (but it should affect low latency mode as well).

In v4.22, I upsampled the audio, and then I removed peaks that weren't visible without upsampling.

In the BETA3, I'm using an algorithm that "estimates" the maximum peak levels and clips based on those. But this method seems to have a number of drawbacks, so I guess I'll need to go back to what I did in v4.22. The disadvantage of that is that the CPU load will go up.


I'll upload a version that should sound slightly better tomorrow morning, I'll try to create a really good solution tomorrow. (Build is running now, if it doesn't fail I'll upload it before I go to work)


--> EDIT: NEW IDEA: INSTEAD OF CLIPPING, FILTERING USING A BANDPASS FILTER MIGHT GIVE *MUCH* BETTER AND ONLY SLIGHTLY SOFTER RESULTS. (It would avoid the quality loss that "hard limit" caused in all older versions). Extra advantage: CPU load will hardly go up.

New BETA3 is online.

Sound should be a bit better. I'll try my new idea (previous post) tomorrow (if I have time then, otherwise it will be next weekend).

I've done some more tests, and the results look good:

1. I've turned off the FM hard limit filter. Because that's the biggest difference in processing between 4.22 and BETA3, except for the low latency. [Note: Part of it is ALWAYS turned on in BETA3 if FM output is enabled. You can't try this at home...].

2. I've turned off a filter that removes loud spikes in the sound, because I don't really trust it (it might work fine, will test that later).

3. Because the output of the BETA3 is louder, I've turned down the Loudness setting - for example, in 4.22 4.00 roughty equals 3.70 in BETA3. Which is still (slightly) louder if I measure it over a longer period of time (only the highs are louder, the lows are nearly identical).

The output of this is not really useful (it contains huge spikes in the FM output, and it can sound bad because of point "2"), but: I can hear almost no difference anymore between the two versions! (In normal latency mode)


The good news is that this means that ALL of the filters of the "FM Loud" preset (with Loudness set to 4.0) combined in version BETA3 can give nearly the exact same sound as v4.22. It will just require some tweaking of settings (the lower Loudness value with the same output means that I'll need to re-calibrate the filter of point "2", and the filter might need some changes anyway because this is one of the filters that is hit hardest by the low latency version), and lower the Loudness settings a bit.

Note: Those lower Loudness settings might even be a good thing, because this probably means that the dynamic range of the sound will increase slightly, with the same average RMS volume (measured with CoolEdit). On the other hand, really loud spikes (loud "S" sounds) sound a bit distorted in the new version - but that should be solved when filter "2" is working again.


Conclusion:
I'm getting much more confident that it's possible to reach a sound that's nearly identical to that of v4.22 using the BETA3 processing (68 ms latency setting). (Although I still have to see what happens when I enable the full "hard limit" processing again, and I haven't tested lower Loudness settings yet).

Hi Hans,
Can you remove second input in next beta .. it's anoying when i start standalone version, i always must select another empty input.

Curently i am buliding (again) PC for mic processing only
with beta3 in LowLatency mode i still have more latency with only used Single compressor .

edit:
i found beta1 ... and i am use it for now.