I am on Spectralayers 11 Pro.
I tried unmix some live tracks, using High Quailty option, to separate Vocal from others (I am working on live recording session).
After extract the result, i figure out a strange frequency is always appear on both tracks, vocal and others.
The exact freqquency I see is about 14800Hz (as Izotopne Neutron 4 shows in the photo I attatched).
The other quality options work well without issue.
Don’t know if developers know it or not. This issue happens to all of my live tracks, on all projects.
Where and how was the original recording made? What equipment was used and what was the environment like? What other equipment was present in the recording environment? Is the spike visible in the original audio before unmixing?
That spike in the signal looks suspiciously like electrical interference from lighting or other equipment at the time of recording. Do you see the same signal on other audio material which was not recorded in the same way when unmixing in high quality mode?
It could be that the other quality options somehow mask the spike which is present in the original audio whereas the high quality option exacerbates the issue. Just guessing.
Yes, there appears a spike at around 14.700 Hz, reproducable.
This seems just like another algorithmic glitch in SL.
After using SL for over a year now, I do not trust SL in any scientific way. It’s a nice toy, but in no way realiable for usecases, which require exact precise operations beyond artistic purpose.
I love the concept of SL, but the actual implementation leaves a lot of questions, at least to me.
Maybe there are some kind of frequency cuts in the quality settings of the algorithms, which seems plausible.
But(!) this should be documented somewhere. The documentation lags lots of information of what is going on under the hood. Especially all the parameters around FFT are a mystery. This might be, because a lot of people have some kind of fear about mathematics, but not all of them.
It should be clear, that somehow sometimes, there are people, who check the program rigorously and discover glitches and bugs, that for many are either not noticable or worth to follow - understandable, of course. But unsatisfactory on the other side.
Please, I checked for months, before posting this. The original audio is good and there is no hiss at 14800 as the photo I attached.
We are a professional studio. The audio was recorded in the studio.
Thank you.
I am very careful in audio. To be professional, we have to notice every small detail. This is just one of many problems I have to solve daily.
I work in live recording, where tools like SL (or RX advanced which I also own) are very useful.
I choose Extreme Quality as a work-around but it cost much more time than High Quality, which I feel good enough in most cases. It’s the reason why I think it’s better report this issue, hope to be fixed.
It is really great you follow and have a check, to make clear the issue.
Thank you.
Thx a lot. It might seem like nitpicking or nerdy stuff, but SL just invites to be curious.
I do learn quit a lot while using SL - and love it.
See this for instance:
Nitpicking? Maybe. But there are cases, when it matters.
When working with spectral audio, it might be difficult to define exactly, what a tool has to do and what kind of artifacts are tolerable. The documentaion of all this might require another full-time job.
So, we are fine here, but we are still free to mention the glitches we find in our daily work.
Very often I double check Wavelab and Spectral Layers with Audacity… which should not be necessary, imho. I do expect software like Wavelab or Spectral Layers to work exactly as the menu says as could be expected - on the numbers, on the bits. but they don’t. This is disappointing to some extend.
Please, do not post here assuming that we know what you did before or after posting or with regard to your recording environment, status or professionalism. We are not mind readers. My comments were not offered as an attack but as a starting point… as I said I was ‘just guessing’. Thank you for your understanding.
By the way, I did some quick tests on some live recordings here and no spike was apparent.
My apologise, you can see another comment who also figured out the problem.
So I am not the only one face this.
By the way, if I listen carefully, this noise can be heard.
By the way, I don’t think any word of my comments is breaking the forum’s rule. So are you. Peace.
→ " wanabeme308" has a point here.
And yes again: I do consider SL not reliable for exact scientic work, as I do.
SL is a phantastic tool to work in the spectrum. But the code needs a lot of quality checking, imho, because there are so many bugs and glitches all over the place. That’s a fact.
No time right now to provide ‘proof’. You can be confident I processed and checked in the exact same way you recommended and analysed with the spectrometer in Wavelab 12 (which I was already doing, by the way). No tiny spike is apparent in the Other layer.
You may be right. As to the explanation that might be another matter. I’ll test again when I get more time but probably best to get Robin’s opinion on this @Robin_Lobel.
Also, so far, we have all been testing with different audio material. If you want to be sure of a result in scientific terms we’d all have to be testing with the same file (same format) - otherwise the results will be non-conclusive. But since we are all getting different results with different audio material I’d still suggest the results are highly dependent upon the source audio / bit depth / sampling rate etc.
It’s fine man, we’re all have no time and some issues I figured out before couldn’t be seen by some other users, but they were confirm as bugs by developers. Take it easy.
Well, on this part (14700/800 Hz) I am not sure, if it’s a bug or not. This might even be an algorithmic artifcat, that even developer was not aware of - until some noise customer discovered it lol
And also, I try to keep this in mind: Without SL, we would not even have thought about this, right?
14700/800 Hz… year, that’s nerdy stuff.
Just imagine a world, in which this would be the highest ranked problem. Phantastic!
Well yes, it’s very low dB, but a complete zero could be expected, right?
This seems to happen only with the “High” setting. “Fast” and “Extreme” do not show those added frequencies. … another interesting quirk of Spectral Layers.
LOL
Addendum
It’s actually an interesting spectrum. It has some rythm in it and can be used for ambient sounds.
SL would be so much more fun, if the VST-interface would work flawless realtime as in other programs, like Wavelab, Reaper or Cubase…
Seriously?
It seems plausible, that this noisefloor depends on the inputsignal.
No! SL is not reliable as it is today. I do not trust the program and must double check the results very often with other programs. … just to find more and more glitches.
Once a program becomes as popular as SL, the developer(s) must anticipate, that there are some nitpickers (like me), who take the promise of precise work in the spectrum seriously and check every data, that the program produces to the bit.
I think, it is the responsibility of Steinberg to establish a rigorous quality testing here.
Why? Because with the spectrum, we work at the heart of music, of sound in general. SL is supposed to be like a very sharp scalpel, not some blunty sword.
They should check every mathematical procedure and the datahandling in generell for
a) numerical efficiency (to actual state of the art algorithms)
b) correctness
Well, I know, that’s not something to shout hurrah for. But considering the actual state of SL, it seems reasonable, imho.
