Part 1:
The VLevel latency mystery
The low latency using VLevel with a buffer length of 2.0 s (default setting) inside of foobar2000 is really mysterious.
Why is there almost no delay (latency) when you click and step from one track to another up and down the foobar2000 playlist ?
You would expect that (using the default 2.0 s buffer length) VLevel should add this 2 seconds to the overall latency.
That is just the way it works with StereoTool 3.40 using Hans' settings:
Quote:
1. Press RESET
2. Set Pre Amp higher (as high as you like, VLevel recommends 10-20)
3. ONLY turn on Final Limiter
4. Set the time (right slider) to maximum (1.60 seconds)
5. Keep everything else turned off.
This adds 1.60 seconds to the latency (click a new track you have to wait 1.60 seconds before you can hear it).
But not with VLevel.
Even when you set the buffer length to 5.0 seconds.
There is no extra delay (latency).
It does not make any difference if you use VLevel or not for the overall latency of foobars audio output.
So how could this lookahead (default 2.0 seconds) work without extra playback latency ?
Any idea ?
Part 2:
VLevel - the dark side of the moon
I have done some serious listening tests with VLevel.
It does compress (with default settings) !
Technically: the only way to avoid any compression would recommend a lookahead as long as the processed track - 2 or 5 seconds lookahead automatically leads to compression of the dynamic range of the whole track.
Though this is not a compression in the traditional meaning of audio engineering (using a compressor), because it should work as lookahead AGC only -> so without any change in the (temporary) sound signature.
I thought it should work this way with VLevel.
But it does not.
Using the default settings for VLevel there must be some kind of "real" compression applied, too.
Try some older tracks with wide dynamics, long fade ins and fade outs and little noise.
VLevel amplifies this noise in a very ugly and annoying way.
I have compared it with Stereo Tool and simple Pre Amp settings to get the same
loudness as with VLevel.
Using Stereo Tool the noise is a little bit louder (dependant on the Pre Amp Gain)
than the unprocessed original, but not that annoying as with VLevel default processing on similar percepted loudness (VLevel vs. Stereo Tool 5).
The only way to avoid this extraordinary noise gain is to change the default settings for VLevel.
Instead of (default):
strenght: 0.80
max multiplier: 25 (+28.0 dB)
buffer length: 2.0 s
Use this:
strenght: 1.00
max multiplier: 3 (+9.5 dB)
buffer length: 0.1 s
This should be to 99 % very close to the default characteristic, but without that extreme noise gain.
Ironically, though this works with a very low buffer length (compared to the default 2.0 s setting) it leads to more consistent results.
I had some older tracks with originally quiet segments - using the default VLevel settings - that became louder than the originally loud segments (I think Bojcha knows this side effect, too ?).
Based on this experience with VLevel that the lookahead buffer can be relative low I have tried to use lower lookahead buffer lengths with ST 3.40, too.
Instead of 1.60 seconds I tried 0.80, 0.40 and 0.20.
0.20 s was too low, but 0.40 s works.
But ST 3.40 standard (low) latency plus 0.40 s is still too much overall latency.
But this gave me an idea that there might be another way (without big lookahead
buffers) to emulate VLevel characteristic.
This is were
Stereo Tool 5 VLevel Emulation Preset
story starts:
viewtopic.php?p=5107#p5107