Stereo Tool

No, because pre-emphasis makes highs much louder which causes huge peaks.

The following filters need pre-emphasized audio to function correctly because otherwise peaks would be introduced afterwards:
- Singleband compressor
- Final Limiter
- Loudness
- Hard Limit
The Multiband filter also needs to work on pre-emphasized audio if the right top slider is set below 100%.

Only Noise Gate, AGC and Stereo can be run without pre-emphasis. (And Multiband if the top right slider is set to 100%).

About trumpet filter:
Sound much clearer, highs and mid-highs most.
So, if you Hans think this filter still need to bi in processing,can i ask for option On and Off (On by default)?

I have an idea about 512, 1024, and 2048 samples, but i must try first to be shure.

Update to Bojcha:

I've just checked what happens if I re-introduce the crash bug in the Trumpet filter. And I see now that the audio that comes out is nearly identical to the audio without a trumpet filter. The version with repaired Trumpet filter is very different.

So it's VERY likely that removing the Trumpet filter improves the sound.

LOL I waited too long before pressing Submit

I also noticed that the Trumpet filter apparently doesn't remove the type of distortion it's supposed to remove. So it's not working properly at all.

I'm now comparing v4.22 with the latest version to see why the trumpet filter works so badly. Before it seemed to improve the sound...

in 512 samples 40Hz EQ slider not working!
i want to set it lower (for test) because there is much more this freq. then in 4096.

There is differences between Latency modes 512, 1024, 2048 and 4096 in quality but and in freq. responce.

I am comparing 512 and 4096 ..
512 much thinner sound .. so i turned on EQ and start to set it to sound just as 4096 mode, and i get almost same sound.
Better way will be if there is a "output level" settings for each band in multiband .. so this tune can be done much closer.
Maybe it can be done internaly in ST - Output level settings from each band for evrey sample ?

In v4.22 the Trumpet filter works much better. Trumpets sound fuller and far less distorted. So I guess my performance optimization of this filter ruined it. And that may also explain the "vibrating" female voices problem.


Edit: It's not the trumpet filter. Apparently there is a difference between v4.22 and the current version that - in case of certain sounds - pushes the lows and mids down while increasing the highest frequencies, which causes a rather annoying sound. I fear I'll have to move step-by-step through the processing chain to find out from where the difference originates. But I do think that it needs to be fixed - for these trumpet sounds v4.22 sounds clearly better (and it's louder too).


Edit #2: Surprisingly, the first difference seems to be in the Multiband filter. The last "soft limit" slider must be set at about 3/4th of the value in v4.22 to get the same sound... But that's not all. Will continue tomorrow.

Edit #3: The other difference is in Loudness. Hard Limit doesn't cause any changes. If I turn on Loudness the frequency response is very different, and the sound is pretty bad.

I'm adding 'fixing Multiband and Loudness for trumpet sounds' to my todo list.


The bad news here is that I have more work to do than I expected.
The good news is that these differences might explain the few sounds for which the quality got worse since v4.22 - if I can fix those, everything is better... (female voices with vibration etc.)


Planned steps:
1: Fix frequency response
2: Add new Trumpet filter if needed (I think it will be needed)

The problem here is that in lower latency modes the channel separation is not very good. So I cannot even reliably measure the amount of audio at - say - 40 Hz. (Well I could if I would use some history, that would work if it's just for displaying it...)

At Latency 512:
- The first 4 Multiband bands cannot differ in volume much or you'll get horrible artifacts. Basically this means that if the volume of one band drops, I'm taking the others down as well. Maybe I should average them instead...
- Improved Loudness Distortion Protection (including it's Deep Bass Boost) and Deep Bass Protection don't work AT ALL (they are turned off, they would only generate artifacts)


I suppose to get similar sound for 512 as for 4096, you had to increase the equalizer levels of the first few bands? (If so, than averaging might indeed work).

I'm confused now. I've done a lot of new measurements v4.22 without Trumpet filter vs. the latest version without Trumpet filter, and now the audio that I'm getting is nearly identical (and graphs even show that the sound of the latest version is cleaner). So it seems that I did something wrong yesterday (but yesterday I've also done multiple measurements, which consistently showed big differences.)

I've now done 5 different tests with different settings, and the difference between v4.22 and the latest version is marginal. The latest version does have slightly more distortion in the highs caused by loud bass sounds, but I really have to switch back and forth multiple times to hear the difference. The difference is probably caused by some settings that I had to re-calibrate to lower the latency.

Much more important: v4.22 with Trumpet filter DOES sound clearly better than without. So I still need to add a new Trumpet filter to the latest version. (The one that was present in the latest versions didn't function very well, I've now confirmed that by doing some comparisons; I hear no difference in the type of distortion that it's supposed to solve).


I may have an idea for a Trumpet filter that doesn't cause distortion - but I need some time to think first... If I can find a frequency-independent solution it might solve the vibrating female voices issue also.

Confusion here too ..
i said that without trumpet filter sound is clearer, but today i again hear something similar but not same.
it's like i pushed high and mids to high and loudness or final limiter can't handle it.

I installed this no-trumpet version .. and org.v4.22 .. and testing it side by side. (fm out)

With v4.22 i recall the bad bass followed by strange early-tail sound (multibang and loudness bug) .. and something i hear first time, is on some deeeep-bass sound there is something like "sand" sound .. in last beta there is nothing like that ..except there is still small early-tail sound in bass caused by loudness or Finall limiter (i think), but it's really small.

v4.22 have bit softer mid-highs and highs and i think more natural last beta have stronger highs but seems loudnes can't handle that.

edit:
4.22 vs beta2
My conclusion is that it's not problem in mid-highs and highs in latest beta. Problem is in mid and mid-low which are missing a bit. Like comparing 4096 samples mode and 512 samples mode, but smaller difference, of course.

And the 'similar' sound is not present in v4.22?
So, the latest beta is better in all these aspects?
I suppose then that you mean that it sounds less full/warm? Or is there really distortion?


Note: It may be difficult to compare v4.22 against the latest beta's due to the big difference between the original 'pre limiter' and the new 'agc' filter. If you're using a dualband AGC, it might help to lower the

Note #2: Please see my earlier remark about the Multiband filter: For some reason, the last 'soft limit' slider must be set at about 77.5% of the value in v4.22 to get the same sound. I really don't have a clue why. If you don't do this, you'll sometimes get really loud (+ 2dB) very high frequencies, which sound really bad and also push the other frequencies down (especially when using pre-emphasis, as the highs are already loud then!)


I've just performed a measurement to compare the frequency response of v4.22 against the latest (BETA2) beta.
Results, with AGC turned OFF and Multiband 'soft limit' slider 10 set to 1.56 in v4.22 and 1.13 in BETA2:
- Frequency response is roughly similar.
- But around 150 Hz-250 Hz the sound in BETA2 is a bit softer (on average about 0.3 dB).
- At frequencies above 7 kHz the highs in BETA2 start to get a bit louder, despite the much lower 'soft limit' setting (it does help a bit though). At 14 kHz the difference is about 0.5 dB. (I still need to look into Multiband to find out what's wrong).
- At low frequencies the BETA2 is softer (-0.3 dB at 40 Hz, upto -4 dB at 0 Hz, with a highpass filter frequency of 30 Hz). So that should be an improvement.
- The difference in RMS volume is 0.03 dB - negligible.

Except for the slightly louder highs, I would expect the difference to be hardly noticeable. Also, based on the output graphs, I would expect the mid freqs (400 - 7000 Hz) to sound better - they look cleaner in the output graph. I don't know if that's also the case for higher frequencies (difficult to compare).


All of this is completely negligible compared to the huge impact that the AGC will have on the sound, compared to the Pre Limiter filter. The only reliable way to compare the quality of the other filters in the two is to turn AGC/Pre Limiter off in both versions.


Edit #2: I just did a quick comparison between Pre Limiter and AGC. Result: AGC usually results in a softer sound (which means that Multiband works less hard, so you might need to increase the AGC soft limit slider a bit, say 10%). Also, at least for the tracks that I tested, I see that the highs are relatively lower in volume in AGC than in Pre Limiter. Which unfortunately contradicts what you're describing... Although I've seen one track where the high are (relatively) getting louder.


Edit #3: Note that you're comparing v4.22 WITH Trumpet filter against BETA2 WITHOUT Trumpet filter. I'm now building a version of v4.22 without this filter, to enable a more 'honest' comparison. Will post it tomorrow morning (if the build succeeds).