Blue Mangoo’s take on a compressor shoot out

«1

Comments

  • I would be really interested to hear other developers’ take on this.

    (cc @Lars @woodman @Chris_Randall @sundhage)

  • Which one would you choose?
    The one that creates better graphics or the one that sounds better?

  • @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    Loaded question :wink:

  • @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    ?

  • @espiegel123 said:

    @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    ?

    In other words: Would you judge a sound, a synth, an effect by how it looks on the oscilloscope or by how it makes you feel when you listen to it?

  • @gusgranite said:

    @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    Loaded question :wink:

    It certainly is :#

  • @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    This a fair and reasonable point. The sound should not be judged by its appearance on a graph. The visual tools should only be used as tools for training your ears. Once one has learned to hear the details in sounds, the graphical tools become unnecessary. They are sometimes useful as confirmation though; when we wonder, “Am I hearing what I think I’m hearing?”

  • @rs2000 said:
    In other words: Would you judge a sound, a synth, an effect by how it looks on the oscilloscope or by how it makes you feel when you listen to it?

    While that can be a valid artistic choice in certain situations, in other situations (primarily mixing) the oscilloscope is showing you stuff that could mess up your mix, even if you don't hear it directly. Transparency is what you want on your master bus and for glue compression.

  • Also, I don't know that you ever really want aliasing on a compressor. It's impossible to eliminate aliasing entirely once you allow distortion (unless you oversample, but that brings its own problems), so to some degree it's a matter of trade offs, but as you can see in the video some developers do a better job of creating pleasing distortion, while minimizing aliasing. Also there are some compressors (the recent DDMF one is a good example) that will sound good so long as your sound signal doesn't have any frequencies above a certain level. This is useful information - particularly as if you don't care about those upper frequencies you can filter them before using the compressor (which will get you a better sound). And certainly you are not going to want to apply saturation before using that compressor.

  • @rs2000 said:

    @espiegel123 said:

    @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    ?

    In other words: Would you judge a sound, a synth, an effect by how it looks on the oscilloscope or by how it makes you feel when you listen to it?

    Obviously. And @Blue_Mangoo demonstrated that the visuals provide a glimpse as to what was behind audible artifacts. The question seemed to imply that his analysis and conclusions were based simply on visuals which wasn't the case.

    Additionally, any of us over 30 don't hear some things that scream to younger people. So, you can sometimes use spectographic analysis to see things that you can't hear but others can.

    While someone could misrepresent spectrographic analysis, I don't see evidence of that hear. He was pretty clear in the original video about at least one case where the audible artifacts were trivial but showed up in the graphics and said that the artifacts wouldn't be heard.

    If there is evidence that the conclusions weren't accurate, I am very interested.

  • edited June 12

    @espiegel123 said:

    @rs2000 said:

    @espiegel123 said:

    @rs2000 said:
    Which one would you choose?
    The one that creates better graphics or the one that sounds better?

    ?

    In other words: Would you judge a sound, a synth, an effect by how it looks on the oscilloscope or by how it makes you feel when you listen to it?

    Obviously. And @Blue_Mangoo demonstrated that the visuals provide a glimpse as to what was behind audible artifacts. The question seemed to imply that his analysis and conclusions were based simply on visuals which wasn't the case.

    Additionally, any of us over 30 don't hear some things that scream to younger people. So, you can sometimes use spectographic analysis to see things that you can't hear but others can.

    While someone could misrepresent spectrographic analysis, I don't see evidence of that hear. He was pretty clear in the original video about at least one case where the audible artifacts were trivial but showed up in the graphics and said that the artifacts wouldn't be heard.

    If there is evidence that the conclusions weren't accurate, I am very interested.

    I don't question the accuracy of theoretical, pure sinewave measurements.
    Doing this kind of measurements can help tremendously in learning how stuff works.

    What I question though is that you can do a compressor shootout using such measurements in order to find the one that works best for all kinds of real-world audio material, and I dare to believe that it's the idea behind this effort.
    Also, I highly doubt that a spectrum analyzer can make up for the ear's reduced sensitivity to high frequency content. Most music is simply too dynamic and sounds are too different in each song as to being able to have simple rules for interpreting a realtime spectrum.
    I've done scientific work on spectral analysis and I know how hard it is by today to even develop an algorithm that "simply" splits a song into its individual instruments without nasty side effects.
    A spectrum analyzer can be helpful in finding exact frequencies that you know exist in your track, in "seeing" noise floor, THD measurements, loudspeaker and EQ design and some other applications but I often see people using them in situations where they add no benefit except nice graphics to watch.

  • @rs2000 said:

    >

    What I question though is that you can do a compressor shootout using such measurements in order to find the one that works best for all kinds of real-world audio material, and I dare to believe that it's the idea behind this effort.
    Also, I highly doubt that a spectrum analyzer can make up for the ear's reduced sensitivity to high frequency content. Most music is simply too dynamic and sounds are too different in each song as to being able to have simple rules for interpreting a realtime spectrum.
    I've done scientific work on spectral analysis and I know how hard it is by today to even develop an algorithm that "simply" splits a song into its individual instruments without nasty side effects.
    A spectrum analyzer can be helpful in finding exact frequencies that you know exist in your track, in "seeing" noise floor, THD measurements, loudspeaker and EQ design and some other applications

    but I often see people using them in situations where they add no benefit except nice graphics to watch.

    Ok. I acknowledged that people sometimes do that. But that doesn't seem to be the case here. If you think the analysis is flawed in this actual case, how about demonstrating that to us.

  • What I question though is that you can do a compressor shootout using such measurements in order to find the one that works best for all kinds of real-world audio material, and I dare to believe that it's the idea behind this effort.

    You uncover a lot of flaws with a compressor using this kind of analysis, which is why audio engineers do them. There is nothing in these analyses that wouldn't also apply to real world signals. You can literally prove this mathematically.

    There are other things for which this analysis would not be appropriate and you would use different tests (envelope response for example). Though these things are more subjective up to a point.

    Also, I highly doubt that a spectrum analyzer can make up for the ear's reduced sensitivity to high frequency content.

    Not sure what the point you're trying to make here is, but aliasing will be heard a lower frequencies that you absolutely can hear as 'harsh' inharmonic sounds. Typically people describe it as digital noise, or harshness. It certainly does not sound 'analog', which matters if a plugin is marketed as having an analog sound.

    How much aliasing matters depends upon the input signal, and how much of it there is. If you can keep aliasing at around the noise floor, most people won't worry about it. If aliasing only affects higher frequencies then your compress may sound great for some instruments, while sounding like total garbage for instruments with higher frequencies (something you could literally hear with the DDMF compressor).

    I though the video actually did a great job of clarifying all these issues and was very fair.

    I've done scientific work on spectral analysis and I know how hard it is by today to even develop an algorithm that "simply" splits a song into its individual instruments without nasty side effects.

    Building a spectrum analyzer accurate enough for the kind of work that Blue Mangoo was doing is pretty trivial. Splitting a song into individual instruments is in a completely different category. I know how to build a spectrum analyzer (and have done so - because I have strange hobbies). No idea how to even begin with one which splits a song into instruments.

    A spectrum analyzer can be helpful in finding exact frequencies that you know exist in your track, in "seeing" noise floor, THD measurements, loudspeaker and EQ design and some other applications

    The use of the spectrum analyzer in the video seemed totally appropriate.

  • A very interesting video.

    Got me interested.

    But again, just not able to find nothing.
    Cannot remember the right name.

    Any posting promoting an app needs an app store link, because of this:

  • tjatja
    edited June 13

    Many thanks!

    I was totally confused to see this as an App from Ngo Minh Ngoc, who I know.

    But the video clearly shows this compressor and also "Blue Mango Software".
    Makes me wonder, why this name is not also used for the developer name in the App Store.
    Or the other way around. One company or developer, one name!
    @Blue_Mangoo

    We have a similar confusion with another developer with "veek" in his name, cannot remember now.

    It really makes things complicated and easily confuses people who are as easily confused as me :-D

    Anyway, bought the compressor!

  • @Blue_Mangoo : I am getting around to seeing if Audulus is generating a clean enough sine and sine sweep for the type of analysis you did and whether SpectrumView is up to the task. Is there a free AU plugin that you can think of that would introduce bad aliasing that I can use to check out ut my set up?

    Thanks.

  • God I love this forum

  • tjatja
    edited 5:17PM

    Pure Audulus Sine Wave Sweep:

  • tjatja
    edited 5:20PM

    Still not ideal, and start and end where not cut at a clear transient.

    What I used:

    The volume LFO is disabled.

  • @tja said:
    Still not ideal, and start and end where not cut at a clear transient.

    Can you explain the image a bit. I see white edges and a red core to the graph lines. Does this imply
    a range of values for a lot of cycles or a comparison of a pure input to an output? Is this a desktop tool
    for spectrum analysis like the BlueMangoo guy used? If the details are in the prior thread comment I'm sorry.

  • tjatja
    edited 5:33PM

    @McD said:

    @tja said:
    Still not ideal, and start and end where not cut at a clear transient.

    Can you explain the image a bit. I see white edges and a red core to the graph lines. Does this imply
    a range of values for a lot of cycles or a comparison of a pure input to an output? Is this a desktop tool
    for spectrum analysis like the BlueMangoo guy used? If the details are in the prior thread comment I'm sorry.

    I cannot explain the details of the graphic.
    You noticed the colored db-bar at the right?

    But this was just from one single sweep of Audulus, as shown above, not several runs.
    Not sure if this is "really clean"...

    The App is this:

    SpectrumView Plus by Oxford Wave Research Ltd.
    https://apps.apple.com/us/app/spectrumview-plus/id571455198?l=en

    They also offer a free version.

  • tjatja
    edited 5:34PM

    Interesting is, that if you enable the LFO to modulate the volume, you get bands of aliases (from a different run) and also harmonics:

    Always around the highest volume.

  • @tja said:

    @McD said:

    @tja said:
    Still not ideal, and start and end where not cut at a clear transient.

    Can you explain the image a bit. I see white edges and a red core to the graph lines. Does this imply
    a range of values for a lot of cycles or a comparison of a pure input to an output? Is this a desktop tool
    for spectrum analysis like the BlueMangoo guy used? If the details are in the prior thread comment I'm sorry.

    I cannot explain the details of the graphic.
    You noticed the colored db-bar at the right?

    But this was just from one single sweep of Audulus, as shown above, not several runs.
    Not sure if this is "really clean"...

    The App is this:

    SpectrumView Plus by Oxford Wave Research Ltd.
    https://apps.apple.com/us/app/spectrumview-plus/id571455198?l=en

    They also offer a free version.

    Nice. Are you feeding it audio files or do you have a way to feed it a live signal from Audulus?
    If files, how do you record them? Again, too lazy to read the whole thread.

  • @tja said:
    Pure Audulus Sine Wave Sweep:

    Yes. My tests are also showing a clean sine ... cleaner than I expected. Hence, I want to see if that means the sine is clean or if SpectrumView is not sensitive enough (or something to do with settings).

    Didn't you have an earlier test in which there were artifacts in the Audulus signal or am I mis-remembering?

  • tjatja
    edited 5:52PM

    @McD Recording in AUM, sharing from AudioShare directly to this App

    Sadly, I cannot find a live Spectrogram App.

    @espiegel123 Yes, I added it above. Here I also modulated the volume, faster than the frequency.

  • @tja said:
    @McD Recording in AUM, sharing from AudioShare directly to this App

    Sadly, I cannot find a live Spectrogram App.

    @espiegel123 Yes, I added it above. Here I also modulated the volume, faster than the frequency.

    What is different about how you generated the sounds that aliased and those that didn't?

  • @espiegel123 said:

    @tja said:
    @McD Recording in AUM, sharing from AudioShare directly to this App

    Sadly, I cannot find a live Spectrogram App.

    @espiegel123 Yes, I added it above. Here I also modulated the volume, faster than the frequency.

    What is different about how you generated the sounds that aliased and those that didn't?

    I wrote that already, I just modulated the volume (Amp) with the the second LFO, see the picture of the Audulus patch.

  • @tja: btw, it seems to be that there is something screwy about spectrumview's audio playback. there is distortion and aliasing that I hear when it plays the test sounds (including a test sweep I downloaded) that aren't there when played with other apps. Do you experience that also?

  • @espiegel123 said:
    @tja: btw, it seems to be that there is something screwy about spectrumview's audio playback. there is distortion and aliasing that I hear when it plays the test sounds (including a test sweep I downloaded) that aren't there when played with other apps. Do you experience that also?

    I did not notice that.
    But then, my ears are especially bad... :D :)

    If we had a hostable Spectrogram App, we could compare before and after visually.
    Or maybe someone with an audio interface?!?

Sign In or Register to comment.