New limiter plugin - VladG Limiter6

[bezusheist]

Well, I've got pretty cool idea how to check ISP stuff. It can be done with simple impulse audio files. Let's start!

TEST#1

Wave file looks like this (where numbers are sample values, 0 is -inf dB and 1 is 0 dBFS):

... 0 0 0 0 +1 +1 0 0 0 0 ...

http://dl.dropbox.com/u/18475891/impulse_1.wav

Mathematically, maximum ISP value will be:

sin(pi/2) / (pi/2) * 2 = 4/pi = 2.098 dB

TB EBU (I downloaded latest demo from toneboosters) shows: 1.9 dB

L6 ISP detector: 1.275 (sample value) = 2.111 dB

As you can see TB EBU have 0.1 dB precision and L6 have 0.02 dB precision for this test. For L6 this value is hidden inside ISP limiting module. But I can just output it somewhere on the screen or I can move this detector out and compile simple true peak metering plugin. After that it can be interesting to create dedicated thread on this forum and analyze all known true peak meters!...

this is a quick example...(i don't want to get too OT on the thread...)
i took a measurement of ^this^ file (impulse_1.wav) using the oscilloscope in SignalScope...
routing = 44.1K file > Wave Editor ("realtime" iZotope SRC > 192K) > Soundflower > SignalScope...
peak value = 2.102 dBFS > screenshot

do you think this is a reasonable ("accurate") method to obtain "true peak" values...?

*i get a peak value of 2.094 dBFS when converting the file from 44.1K to 192K (Wave Editor/iZotope offline...)

(in this case, the average of "realtime" SRC (2.102) and "offline" SRC (2.094) = 2.098...)

[vladg]

Well, I've got pretty cool idea how to check ISP stuff. It can be done with simple impulse audio files. Let's start!

TEST#1

Wave file looks like this (where numbers are sample values, 0 is -inf dB and 1 is 0 dBFS):

... 0 0 0 0 +1 +1 0 0 0 0 ...

http://dl.dropbox.com/u/18475891/impulse_1.wav

Mathematically, maximum ISP value will be:

sin(pi/2) / (pi/2) * 2 = 4/pi = 2.098 dB

TB EBU (I downloaded latest demo from toneboosters) shows: 1.9 dB

L6 ISP detector: 1.275 (sample value) = 2.111 dB

(...)

this is a quick example...(i don't want to get too OT on the thread...)
i took a measurement of ^this^ file (impulse_1.wav) using the oscilloscope in SignalScope...
routing = 44.1K file > Wave Editor ("realtime" iZotope SRC > 192K) > Soundflower > SignalScope...
peak value = 2.102 dBFS > screenshot

do you think this is a reasonable ("accurate") method to obtain "true peak" values...?

*i get a peak value of 2.094 dBFS when converting the file from 44.1K to 192K (Wave Editor/iZotope offline...)

(in this case, the average of "realtime" SRC (2.102) and "offline" SRC (2.094) = 2.098...)

So the conslusion is that "pure mathematical" calculation of peak (4/pi = 2.098 dB) in impulse_1.wav is close enough to your results. Thanks a lot for your help!

So, the last question. Is this so much interest to create new dedicated thread of comparing inter-sample peaks meters and limiters?

If yes, we can move these offtopic results as some information to start with.

PS. And sorry for a lot of offtopic posts.

[t3toooo]

Conclusion. Soon you, guys will have true ISP meter for free

what a nice side effect.

[djeroen]

this is a quick example...(i don't want to get too OT on the thread...)
i took a measurement of ^this^ file (impulse_1.wav) using the oscilloscope in SignalScope...
routing = 44.1K file > Wave Editor ("realtime" iZotope SRC > 192K) > Soundflower > SignalScope...
peak value = 2.102 dBFS

do you think this is a reasonable ("accurate") method to obtain "true peak" values...?

You now have shown that iZotope SRC results in a true peak value of 2.102 dB. This is great, but I think the more realistic test case would be (since EBU R128 is concerned about true peak clipping in the D/A converter and analog signal paths):

- Play the signal at the original sampling frequency (44.1 kHz) on a high-quality D/A converter to generate a continuous, analog signal.
- Re-route the analog output to analog input, and capture the analog signal with a high sampling rate (192 kHz).
- Look at the peak height of the 192 kHz captured signal.

It would be of great value to see what the result is of such an experiment (while taking good care of the input level calibration obviously),and how this would differ from the theoretical true peak values.

A related question: in the mathematically exact true peak values, you make assumptions about the phase relationship of the anti-aliasing filter (e.g. that is it linear phase). This is not always the case; what would be the true-peak value for a minimum-phase filter for example?

[Alexey Lukin]

Min-phase filters may result in higher reconstructed peak values, but they are uncommon in real-world D/A converters. The EBU standard for "true peak" measurement requests a linear-phase filter and 4x oversampling. In practice, reconstructed peaks may be slightly higher than this measurement due to higher oversampling ratios and different filter designs.

[vladg]

Conclusion. Soon you, guys will have true ISP meter for free

what a nice side effect.

It is here (if anyone skipped):

http://vladgsound.wordpress.com/2012/07 ... few-tests/

Okay, we move to separate thread with all this inter-sample peaks stuff. I'm collecting data for new thread to start...

[Trakstar]

Is this based on the Neve 33609 by any chance, it seems like it?

[maxxxter]

DELETED

[Trakstar]

it seems like it?

wtf? in what way??

that Neve is a standard stereo tracking/buss compressor, while this, OTOH, is a mastering limiter with a completely different set of controls and wholly different purpose.

design wise, I should have been more specific. In molot the two algorithms are based on a fairchild type and neve 33609 I think Ive read, and I was comparing just the compressor and limiterfunctions alone because they are functionally simalr to the waves Vcomp. So off you silly

[maxxxter]

DELETED

[Trakstar]

it seems like it?

wtf? in what way??

that Neve is a standard stereo tracking/buss compressor, while this, OTOH, is a mastering limiter with a completely different set of controls and wholly different purpose.

design wise, I should have been more specific. In molot the two algorithms are based on a fairchild type and neve 33609 I think Ive read, and I was comparing just the compressor and limiterfunctions alone because they are functionally simalr to the waves Vcomp. So off you silly

This is the LIMITER6 thread, not MOLOT!!

Im referring to THE LIMITER 6. The first stage is compressor with just a release, and then the limiter stage, like normad factorys british bundle, and waves Vcomp. So naturally I was wondering If Vlad had made the dynamics processing such as the way it reacts to transients similar in design to the algorithms in MOLOT.

[vladg]

design wise, I should have been more specific. In molot the two algorithms are based on a fairchild type and neve 33609 I think Ive read, and I was comparing just the compressor and limiterfunctions alone because they are functionally simalr to the waves Vcomp. So off you silly

No, it was not _based_. It (by accidential) sounds a bit like these compressors. BTW, I never heard neither original Fairchild nor Neve 33609 but if respectable guys say that Molot sounds like them, I don't want to deny it

For L6 slow RMS compressor I don't know what does it sound like. I didn't have any known compressor in mind when I did L6 compressor section (and the same was with Molot).

So I think it is not.

EDIT: There's only one similarity between L6 compressor section and Molot - sidechain HP filter. There was no more shared code between them both.

[jam92189]

[quote="Trakstar]design wise, I should have been more specific. In molot the two algorithms are based on a fairchild type and neve 33609 I think Ive read, and I was comparing just the compressor and limiterfunctions alone because they are functionally simalr to the waves Vcomp. So off you silly

No, it was not _based_. It (by accidential) sounds a bit like these compressors. BTW, I never heard neither original Fairchild nor Neve 33609 but if respectable guys say that Molot sounds like them, I don't want to deny it

For L6 slow RMS compressor I don't know what does it sound like. I didn't have any known compressor in mind when I did compressor section (and the same was with Molot).

So I think it is not. [/quote][/quote]
I have tried bother the comp and the lmiter. I must admit they kick ass some of the best out there

[vladg]

Okay, we move to separate thread with all this inter-sample peaks stuff. I'm collecting data for new thread to start...

For inter-sample peaks stuff we move here:

http://www.kvraudio.com/forum/viewtopic.php?t=354995

[risome]

Use this on our latest remix and it is awesome cant thank you enough Vladg