yr wrote:I like the thought behind switching the env type to peak- it makes perfect sense. That said, having tested it I'm not sure that it works well in practice. I've tried it with several preamp presets and the end result was always inferior to my ears when compared to the default RMS17 settings.
What I did to test it was choose relatively clean presets (with low thd) and process a track that was recorded at very moderate levels multiple times. I then compared both versions (peak/RMS17). The Peak setting seems to generate more artifacts regardless of the sampled HW.
have you tried increasing the smoothing mode some? the times i've switched to peak i haven't heard artifacts, but maybe you're onto something. what kind of artifacts are you hearing? and when you say 'generates more artifacts' are you saying that you were getting some artifacts before switching?
haven't tried smoothing, but I have a feeling that the playback engine handling of transients is problematic in peak mode. The overall sound (in peak mode) is more edgy/strident and unnatural to my ears. I've also compared the RMS/peak modes by inverting the phase on one of the tracks and ended up with relatively high levels of garbled audio with occasional strange peaks, so there is something else going on beyond thd/freq response etc.
Another way of thinking/testing it is by using a very clean preset. What I've noticed that using multiple passes through the plugin, there are some changes to the sound that are not related in my opinion to the HW. I can run the same audio material more times through the HW with smaller changes to the sound as compared to the Nebula version. At least that was my conclusion with the clean preamps I have (I've used the line inputs on the ULN-2 & TC).
When I tested the peak mode using clean presets, I had a feeling that the changes to the original material were rather dramatic (and unnatural) relative to what I can expect from the sampled HW. I hope Giancarlo can help with a more technical explanation.
maybe try backing off on the input level a bit when using peak? also i think you should try adding a little smoothing. i haven't noticed any of these issues with the things that i've switched to peak, but i always have at least cubic smoothing because i think artifacts are a pretty big likelihood with no smoothing at all. nobody has complained of artifacts with any of the stuff i've released using peak, and plenty of the people using that stuff are studio types who i would expect to notice...
smoothing is, from my experience, the biggest thing that has direct impact on artifacts.
I can't argue with what someone else hears, I can offer my impressions and testing. Below are a couple of files: 3x ULN-2 preset, RMS vs Peak modes with cubic smoothing. Max thd of the preset (as shown by the vst analyzer) is -80.81dB, so you are not hearing more natural saturation of the preamp with the peak mode. The HW stays very clean and loyal to the source in those levels, whereas (I feel) Nebula using Peak doesn't.
personally i don't really feel that vst analyser is perfect for measuring this kind of stuff because it's sending a constant tone. so there are no transients involved. why would there be 'more natural saturation of the preamp' when you are talking about a constant tone? i was saying that transients would be affected differently with peak mode, not constant tones.
also i don't quite understand what you are saying here. first you say "Max thd of the preset (as shown by the vst analyzer) is -80.81dB, so you are not hearing more natural saturation of the preamp with the peak mode." which i take to mean that the 'saturation' levels are lower with peak compared to the actual preamp. but then you say "The HW stays very clean and loyal to the source in those levels, whereas (I feel) Nebula using Peak doesn't." which implies that nebula with peak has higher saturation levels (because you are saying the HW stays clean). ? but mainly i don't think using constant tones is going to show you anything about how something reacts to transients.
the analyzer can show you the max thd levels using a constant test tone. The reason I quoted the very low figures (-80.81dB) was to make it clear that the preset used for the samples generates ridiculously low harmonic distortion even under extreme conditions.
I don't believe that the difference that you hear between the posted samples can be attributed to levels of harmonic distortion. The difference in sound (peak/rms) is simply too big under the given test conditions. Listening to the summed difference of both versions (one being phase inverted) and also comparing them to the original track, I tend to think the "peak" mode actually adds some non-harmonic distortion which contributes to the (subjective) impression of edginess.
yr wrote:the analyzer can show you the max thd levels using a constant test tone. The reason I quoted the very low figures (-80.81dB) was to make it clear that the preset used for the samples generates ridiculously low harmonic distortion even under extreme conditions.
ah, i gotcha, you were just saying that it's a program with minimal harmonics so the difference must be caused by something else. makes sense.
yr wrote:I don't believe that the difference that you hear between the posted samples can be attributed to levels of harmonic distortion. The difference in sound (peak/rms) is simply too big under the given test conditions. Listening to the summed difference of both versions (one being phase inverted) and also comparing them to the original track, I tend to think the "peak" mode actually adds some non-harmonic distortion which contributes to the (subjective) impression of edginess.
i don't hear enough of a difference to make this a clear thing one way or the other. and i don't think that i agree that the result you get from nulling the RMS/peak versions is showing you stuff that the peak version is adding. you couldn't know because it's the difference between the two so you can't know which of the two the bulk of that difference is coming from. if it were just a simple matter of it being material that peak version is adding, then i would expect that you could copy it, and add it to the rms version and then its average rms level would exactly match that of the peak version, but it doesn't. i think it's not a result of either one having something the other one doesn't, so much as it could be the result of some kind of maybe subtle phase differences caused by onset of the attack envelope happening at slightly at different times, between the two versions. also i selected identical (time-wise) segments of the rms/peak versions and did a freq spectrum analysis on both and the results were identical. i'd expect to be able to see some tiny difference if one had some kind of distortion added that wasn't in the other.
i did the same test with a preamp program, and when i lowered the attack and release to 0 before rendering the peak/rms versions, then inverted one and mixed the two, the stuff left over became much lower in level (compared to what i got with attack/release at 15/5ms), and some of it disappeared. i think that supports my theory.
Last edited by Cupwise on Tue Mar 19, 2013 2:03 am, edited 1 time in total.
i just did another where i used only rms mode but at different attack/release settings, and you get the same thing. null between attack/release at 15/5ms, and 0/0ms. then between 15/5 and 30/10ms. you get stuff left over just like doing between peak and rms. the difference left between the 15/5 and 30/10ms nulls is more from the body of the drums so has some lower freq content, but the 15/5 and 0/0 is more from the transients and so has more high freq content. this makes me think its from some slight phase differences caused by the attack stage.
Since we can compare directly to the source (unprocessed file) + have knowledge about how the HW behaves (ULN-2: Harmonic Distortion @1kHz +9dBu in @ 6 dB Gain = 0.0005%), it's safe to say that we can know which "mode" represents the HW better in this case.
Just did another test using a clean (no thd), flat freq preset and I'm pretty much convinced that "peak" mode's handling of transients is worse then the rms mode (be it phase distortion or any other artifact). I'm attaching the original sound file + preset so that you could repeat my test. I've also attached both "summed difference" files (raw+rms / raw+peak). I've added 18db gain to the sum files so that you don't need to crank your speakers.
You may find the difference small or unimportant, but the question at hand is which "mode" brings us closer to the HW. I'm pretty sure that if you bought a new "clean" preamp or A/D converter and realized that it has a built-in random transient exciter (as "peak" mode does) you would send it back to the manufacturer...