| |
spider-j
Registered: Oct 2004
Posts: 449 |
SID recordings – opinions on normalizing?
This isn't really aimed at composers but more on people who do real SID recordings:
What is your opinion on "normalizing" (or even more post processing) of SID recordings?
I usually record my music collections in one go and choose the input volume by the "loudest" song. When I later export those to individual files I'm always asking myself if I should normalize them all to -0.2 db or not. I'm unsure if it will destroy the "natural" feeling when you then again listen to the tunes in an MP3 playlist or similar.
In recording "normal" / acoustic music over the years it was never a question to normalize and even compress the heck out of everything. But just because "everyone does it" and sometimes it is even necessary to simulate the energy of a raw live rock sound when beeing in a recording / studio environment.
With SID it seems different. Even the tunes that have lower peaks already sound great. The "mixing" and in parts even "mastering" has already been done by the SID musician himself.
I would be interested in how others approach this subject. Do you normalize or even compress SID? Do you use even more post processing like EQ / reverb etc.? And if: why? And to what extend?
Cheers,
spider. |
|
| |
Krill
Registered: Apr 2002
Posts: 2855 |
One would think this is an artistic choice, depending a lot on the author's ideas about their anthology? :)
I mean, pretty sure some tunes are supposed to be lower-volume than others. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting KrillI mean, pretty sure some tunes are supposed to be lower-volume than others.
Not sure. Thinking about "volume" is usually not a part of SID composing. Afaik the huge majority of tunes uses "full volume" / $D418 lo-nib $F besides fade-in / fade-out effects.
The output volume is then a combination of how your instruments and filters are designed, which ADSR settings you use and what frequencies are put out. But as a SID musician you most likely consider those things in relation to the other instruments in your song itself not in relation to your other tunes. |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
I see.
Well, they do apply individual per-MOD gains in Amiga music disks.
So i guess it remains an artistic choice, albeit by whoever made the collection. I.e., YOU! =D |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting KrillSo i guess it remains an artistic choice, albeit on whoever made the collection. I.e., YOU! =D
Exactly. And because I want to "widen my horizon" is why I asked for opinions by other folks who record SID tunes :-) |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
Quoting spider-jQuoting KrillSo i guess it remains an artistic choice, albeit on whoever made the collection. I.e., YOU! =D
Exactly. And because I want to "widen my horizon" is why I asked for opinions by other folks who record SID tunes :-) Yes, sorry. Finally read the entire OP now, will see myself out. =) |
| |
Mixer
Registered: Apr 2008
Posts: 422 |
Some compression/limiter may be necessary to avoid clipping due to the snaps, crackles and pops of the SID, when streaming live for instance. |
| |
Flotsam
Registered: Jan 2004
Posts: 82 |
Personal choice of course, but when people upload stuff to youtube etc., I wish they'd normalize so that I wouldn't have to manually normalize by cracnking the volume slider up and down depending on the loudness of the current song.
Also, by normalizing, you ensure highest possible resolution (in digital formats) or SNR (in analog formats).
In a case of a multi-song SID, it would make sense to normalize to the loudest of the bunch, but otherwise per song. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting Mixerdue to the snaps, crackles and pops of the SID
From my recording experience there are no really "pops" in the usual "sudden high peak" sense in SIDs. Usually audible "pops" are 1.) at init (which I usually cut off – silence a litte) 2.) from "sexy hardrestart" especially with triangle waveform which imho is part of the "nature" with "modern" music routines.
I might have been just lucky with my SID / recording equipment in that regard though :-)
Quoting FlotsamAlso, by normalizing, you ensure highest possible resolution (in digital formats) or SNR (in analog formats).
In a case of a multi-song SID, it would make sense to normalize to the loudest of the bunch, but otherwise per song.
Yeah, that's kind of what I also thought so far. When "normalizing only" the RMS stays untouched so I think it still represents the original recording. There are a few exceptions though, which i.e. are heavily lo-pass filtered songs vs. songs that don't use much filter at all. When normalizing everything the latter ones sound much "louder" in comparison to the former than if you just playback them after one another on the real machine.
Thanks for your input so far! |
| |
vincenzo
Registered: Sep 2003
Posts: 79 |
All of the SIDs I own have pretty low output volume level and they don't clip - unless you crank up the incoming gain level on your input channel. Personally, I use ADSR to "mix" my tracks and because of this, their output is usually a tad quieter than others. Example of opposite side where the output is pretty crazy louder than usual is Jammer's tracks :) I'm not sure what's the trick but they are definitely louder than the average SIDs and they still don't clip the input of my soundcard's.
If I would do multiple SID recordings in one row, I do the same as you Spidey: set the input level to the loudest SID, record the bunch then normalize all. They probably will be still different because of the "dense-ness" of the composition, instruments, filter usage, etc. but at least there wouldn't be too much difference after normalization.
Using 48kHz/24bit for recording should be enough for normalization - the output of the SID can be pretty noisy anyway, depends on the motherboard and on the actual SID, etc. For this reason I use noise reduction too (record the pure output noise only, then use it as a noise pattern in eg. Izotope RX or similar plugin).
I believe the average listener doesn't really care and can't even tell the various SID revisions or replacements apart so it's up to you what you do with your recorded music. I hate to touch the volume knob whenever I listen to (any kind of) music, I prefer them on a steady and equal level so I appreciate normalization. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting vincenzoFor this reason I use noise reduction too (record the pure output noise only, then use it as a noise pattern in eg. Izotope RX or similar plugin).
Oh, that is interesting. Coming from acoustic music I completely avoid noise "reduction" (I only used "noise gate" on things like bass drum / snare i.e.) – always fearing that noise reduction will destroy the "natural" sound. Especially with those noise footprint functions you mention.
Do you have any (example) SID recordings where that technique was used? Would really love to hear those and compare to raw SID ouput. |
| |
ws
Registered: Apr 2012
Posts: 230 |
firstly, please don't normalize to -0.2 but -0.3 db. i don't find the explanatory article on that right now, but if you want to test it, open a -0.2 db limited mp3 as wav and you can actually see the effect.
normalizing shouldn't do anything bad to the overall artistical value for a SID recording, in the contrary: if you do not normalize, you actually waste amplitude power and leave unnecessary headroom.
compression: you would rather want brickwall limiting or multiband limiting. compression would change the dynamics of the song while limiting makes your track "louder" while also cutting off some dynamic peaks and introducing some kind of distortion, but for a 3-voice signal, it works pretty well.
if i want to do things like that quickly (and 99% of the time sloppy as it is a trade mark of WS), i'd recommend izotope Ozone. Also Izotope RX has some very nice spectral denoising, if that is ever desirable because you will at some point always get artifacts.
:-)
ps.: if you recently won the lottery, i recommend these monitors, they are like an electron microscope for the ears: https://www.abacus-electronics.de/c-box4.html |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting wsfirstly, please don't normalize to -0.2 but -0.3 db. i don't find the explanatory article on that right now, but if you want to test it, open a -0.2 db limited mp3 as wav and you can actually see the effect.
Okay, I didn't know that. If you stumble over that article again post here or PM. I'm very interested. The -0.2 db normalization peak was what my audio engineer friends taught me in the 90ies XD
Quoting wscompression: you would rather want brickwall limiting or multiband limiting.
Yeah, brickwall limiting seems to be the best option of you want to go for compression/limiter on SIDs for me too.
Quoting wsi'd recommend izotope Ozone. Also Izotope RX has some very nice spectral denoising, if that is ever desirable because you will at some point always get artifacts.
Thanks for the recommendations. I must admit I'm completely out of the "audio engineer" business for a long time now. Last album I made was in 2012 and that was mixed and mastered by a friend. Never bothered with audio stuff in the last >10 years because my 100% switch to linux. But a lot of VST(i) stuff seems to work nowadays. Have to look into it. Just bought a REAPER license one or two years ago and have only used that to capture raw SID so far. |
| |
ws
Registered: Apr 2012
Posts: 230 |
and before the wiseguys pop up and quote "apples and oranges" or "outdated", yes, we're talking LUFS these days, when it comes to limiting/mastering for the digital realm, i know.
https://bobbyowsinskiblog.com/lufs-standards/
[but i nevertheless personally insist on -0.3 db headroom wherever i go ;-) ] |
| |
vincenzo
Registered: Sep 2003
Posts: 79 |
Quoting spider-jQuoting vincenzoFor this reason I use noise reduction too (record the pure output noise only, then use it as a noise pattern in eg. Izotope RX or similar plugin).
Oh, that is interesting. Coming from acoustic music I completely avoid noise "reduction" (I only used "noise gate" on things like bass drum / snare i.e.) – always fearing that noise reduction will destroy the "natural" sound. Especially with those noise footprint functions you mention.
Do you have any (example) SID recordings where that technique was used? Would really love to hear those and compare to raw SID ouput.
Thing is, the lower the volume level of a music, the noisier it gets when you normalize it after recording. Obviously, the noise will be "normalized" too.
A gate could help but it's pretty tricky to set it up properly because a fast gate/treshold might cause stutter, a slow gate might not close quick enough.
Whenever there's no silence or gap but music is playing, the SID's and motherboard's noise is practically inaudible, but as mentioned, where the music is fairly quiet or has a gap, the noise is more audible.
Check out my C64 releases here, they are all denoised with Izotope RX: https://strayboom.bandcamp.com
(Funktastic SID, Clockwork Factory, ByteMorphoSID) |
| |
Flotsam
Registered: Jan 2004
Posts: 82 |
Vincenzo said: "Thing is, the lower the volume level of a music, the noisier it gets when you normalize it after recording. Obviously, the noise will be "normalized" too."
Sure, the volume of the noise will be higher when normalized, but it does so in 1:1 ratio to signal. So, the SNR will remain the same and that's really all that matters. If you think of the source (C64), the target (human ear) and everything in between (most likely lossy formats), the earlier you normalize, the better. That way you'll have the best possible resolution for the following stages to work with. There's no way around it.
But I agree, using a gate seems wrong, because inevitably it will alter the sound itself too, unless you set the limit so low that it basically does nothing. Side-chaining the effect to react to just the upper spectrum won't help either because a tune might have a part with just noise playing at a low level. Using more sophisticated algos like the one in RX might work slightly better, but then again... we're dealing with an analog device, why should all noise be killed? I think noise is part of the sound, it smooths things out a bit and creates atmosphere just like it does in photos. |
| |
vincenzo
Registered: Sep 2003
Posts: 79 |
Quoting FlotsamVincenzo said: "Thing is, the lower the volume level of a music, the noisier it gets when you normalize it after recording. Obviously, the noise will be "normalized" too."
Sure, the volume of the noise will be higher when normalized, but it does so in 1:1 ratio to signal. So, the SNR will remain the same and that's really all that matters. If you think of the source (C64), the target (human ear) and everything in between (most likely lossy formats), the earlier you normalize, the better. That way you'll have the best possible resolution for the following stages to work with. There's no way around it.
But I agree, using a gate seems wrong, because inevitably it will alter the sound itself too, unless you set the limit so low that it basically does nothing. Side-chaining the effect to react to just the upper spectrum won't help either because a tune might have a part with just noise playing at a low level. Using more sophisticated algos like the one in RX might work slightly better, but then again... we're dealing with an analog device, why should all noise be killed? I think noise is part of the sound, it smooths things out a bit and creates atmosphere just like it does in photos.
To be fair, I don't mind having noise in the signal. What I dislike is the buzz of the modulator and/or other motherboard components. Probably it can be killed with some modding, but I have no further knowledge about it and never experiminted with changing the components. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Dang. I should have asked earlier. Due to some misinformation our release is already tonight at 00:01 and not tomorrow as I thought.
What I did, encouraged by this thread was:
* normalizing (at -0.2 db, sorry WS, but I will release FLAC not MP3 only)
* "manual" normalizing / cutting out clicks / heavy pops especially at init
* a tiny – almost unnoticable bit of reverb
* a tiny bit of compressor
* brick wall limiting
Personally I'm quite happy with the result.
But I already learned a lot for the future. -0.3 db it will be and I guess I'll start playing around with noise reduction by patterns ... Although I still don't feel it is neccessary.
@Vicenzo: thanks for the link, will check those recordings out. |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
Where does normalising to -0.2 or -0.3 dB(FS?) come from?
Some considerations wrt the corresponding real-world peaks being closer to 0 dB, i.e. physical speakers overshooting the digital curves due to their inertia and momentum? (In other words, normalising to just under 0 dB would be a recipe for audible clipping.) |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting KrillWhere does normalising to -0.2 or -0.3 dB(FS?) come from?
In my case: mouth to mouth propaganda. Don't know if it's really important. WS sent me some links about what happens if you convert to MP3 then which I have to check / try out. |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
MP3 is still a thing? :-O |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting KrillMP3 is still a thing? :-O
It's quite common still. And I must admit as someone who tries to keep everything as FLAC I had some streaming problem because of filesize when I was in holiday. I stream from my own private nextcloud running in a Ubuntu VM and have "only" a "classic" 100MBit / ~30-40MBit upload, because this is the maximum I can get here.
FLAC files had problems, MP3 of course didn't. And VBR highest quality is for the most cases quite close to FLAC/WAV. |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
I meant, it's still a thing while AAC and other more modern psycho-acoustic encodings allowing for better perceived quality at equal bitrates exist? =) |
| |
vincenzo
Registered: Sep 2003
Posts: 79 |
Quoting KrillI meant, it's still a thing while AAC and other more modern psycho-acoustic encodings allowing for better perceived quality at equal bitrates exist? =)
Well, what do you expect? People listen to music on everything that has a speaker. It's not really a common knowledge to "how to listen to" music and quality doesn't make a difference, musicality and easy/memorable melodies do. Most people don't listen to music because of quality but because of background music. Not to mention that in video gaming, especially on mobile games everyone plays with muted audio :) So without getting deeper into desperate thoughts about audio as an audio person :) I'm just saying that MP3 will be still a thing and nobody will care if it's lossy.
Percentage of people who care about audio quality is pretty low. We do care, we are the minority.
On a sidenote:
Personally, I don't understand the existence of FLAC. I don't see a reason for storing audio in FLAC that is similar sized to WAV, and it also requires more CPU for decompression (which is basically 0 spike on nowadays CPU's, but still).
Highest bitrate in any lossy compression will do the job for the average listener and well, to be honest I don't hear above 13k anyway and still make a living of working in game audio :) Shame on me, I know. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting vincenzoPersonally, I don't understand the existence of FLAC. I don't see a reason for storing audio in FLAC that is similar sized to WAV, and it also requires more CPU for decompression (which is basically 0 spike on nowadays CPU's, but still).
Highest bitrate in any lossy compression will do the job for the average listener and well, to be honest I don't hear above 13k anyway and still make a living of working in game audio :) Shame on me, I know.
Although I do use FLAC I must admit: It's mostly only "in my head". I use FLAC especially for archiving my own recordings and for backups of CDs that I bought. For some reason I have this feeling: well I paid for lossless back then and I want to keep it lossless. But nowadays when I buy music on bandcamp I just download the MP3 :-)
Oh: and AAC is also lossy. I guess Krill meant that it is a "better" codec (?!?) – don't know for sure. I think MP3 is still a "thing" and will be, because people are used to it since $forever and don't see the need to switch because of the reasons you mentioned. |
| |
ws
Registered: Apr 2012
Posts: 230 |
@krill
mp3 at 320kbps is as far as i am in contact with commercial audio (for example bandcamp) still very much a thing. The clipping problem concerns all lossy compressions - which can even just pop up unbeknownst to you in your listening chain as a wireless transmission codec.
"The 0.3 dB headroom is left at the top, because some oversampling DA converters do not perform very well at full-scale levels. They sometimes make distortion from arithmetic overflow."
In conversation with a few fellow audio engineers and when it comes to mastering (for a physical release for example) -0.3 db is -to my knowledge- the defacto standard upper amplitude limit for "zero clipping problems".
While yes, ofcourse there are the bold and the brave who go for -0.2 and even -0.1 db -- nobody forbids this.
And if you go PCM release, sure, why not 0 db. Most probably you won't even hear the clipping, or notice the clipping indicator on your gear. But if you want to hand out high quality audio in a form that is fit to be converted to almost any other format and is safe for any DAC, for me, -0.3 is the value where i will have no complaints or difficulties in any forseeable future use.
On the other hand, if you prepare audio for youtube releases, for instance, it makes no difference anyways, since they (and also others) heavily process the audio in an undisclosed way. |
| |
ws
Registered: Apr 2012
Posts: 230 |
And a word on FLAC, i also like it very much, it really helps for example reducing massive file size for multi-track production when it comes to dialogue editing.
There is an oddity, though: it seems that for example the FLAC decoder in Winamp (which still exists, yes) doesnt decode the audio very well. There are quite some "rounding errors", as i have noticed some months ago. I know that i get 1:1 decoding in my DAW or my Editing Software, but there seem to be some suboptimal decoding libraries floating around. |
| |
Krill
Registered: Apr 2002
Posts: 2855 |
Quoting spider-jI think MP3 is still a "thing" and will be, because people are used to it since $forever and don't see the need to switch The choice is mostly up to the content providers, no?
Most consumer devices should eat AAC just as well as MP3.
Quoting ws"The 0.3 dB headroom is left at the top, because some oversampling DA converters do not perform very well at full-scale levels. They sometimes make distortion from arithmetic overflow."
Yes, so pretty much the thing i suspected.
It's just how oversampling works (not a problem of the DA not performing well) - and you can oversample yourself (to 96 KHz, e.g.), then find out the headroom you need to leave for your specific recording.
But... this would be a bit of a cult in general, i guess. =) |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting KrillThe choice is mostly up to the content providers, no?
Yes. I don't even know if anything besides Amazon and Bandcamp exists where you can still "buy" music. And on both default to MP3. Bandcamp offers also other formats, but this site in general is the digital equivalent to that record store where only musicians and die hard fans hang out.
Every "normie" only uses Spotify anyway and I must admit I don't even know what they are streaming. May even be AAC *shrug* |
| |
vincenzo
Registered: Sep 2003
Posts: 79 |
Quoting spider-jQuoting KrillThe choice is mostly up to the content providers, no?
Yes. I don't even know if anything besides Amazon and Bandcamp exists where you can still "buy" music. And on both default to MP3. Bandcamp offers also other formats, but this site in general is the digital equivalent to that record store where only musicians and die hard fans hang out.
Every "normie" only uses Spotify anyway and I must admit I don't even know what they are streaming. May even be AAC *shrug*
I believe https://www.discogs.com still sells physical copies and there's a renaissance of vinyl (again). People sometimes send a query about my music being available on vinyl, which, to be honest is a nice collector's thing but doesn't really worth the investment from my point of view.
Re. lossy compressions, I prefer using OGG over MP3! :) |
| |
acrouzet
Registered: May 2020
Posts: 80 |
I don't add any effects other than normalizing and adding fade-ins and fade-outs for my videos of SID tunes. Normalizing doesn't effect the volume or frequency relationships of the signal, so the resulting information that enters your ears is the exact same as if you listened to the raw SID output and increased the volume on the amp the signal's passing through.
I tend not to like any other post-processing on hardware chiptunes in general, since in my head it provides a "dishonest" representation of the sound hardware, and if you're doing it to a tune you didn't compose, it may go against how the composer intended the tune to sound like. |
| |
spider-j
Registered: Oct 2004
Posts: 449 |
Quoting acrouzetand if you're doing it to a tune you didn't compose, it may go against how the composer intended the tune to sound like.
That is an interesting take. As I come from a plain acoustic music background there it is quite "normal" that audio engineers do those decisions and *not* the composers there. Mixing and mastering is an art form itself in that domain and the results are of course discussed in a good relationship between artists and engineers but I think there is a huge tendency to "trust" your enigneers choices as an artist*.
SID music is a bit different. There is no such thing as "Mixing" that an audio engineer would do. Mixing is completely in the hand of the artist – it's part of the process to create a tune. And only "Mastering" will eventually be applied by someone else if that SID is recorded. And yes: that could be perceived as more "intrusive" than in the acoustic domain, because one is just not "used to" someone else fiddeling with their sounds.
That is one of the key points why I started this discussion here and I thank you very much for sharing your thoughts on that matter.
EDIT: * in fact the one best sounding album of my whole "career" as a musician was the one where a friend of mine didn't take any hints / infos / ideas from me at all and just did the complete mixing and mastering process alone and kept me at a distance. At first I was shocked because he did things I would not have approved in the first place. But as time went on I noticed that engineers listening to me as an artist was probably the cause for every thing before not sounding really "good" in the long run. In retrospective I have to admit: that is the best sounding album I ever released and I (as the composer and main performer) had nothing to do with how it was mixed and mastered. |
| |
acrouzet
Registered: May 2020
Posts: 80 |
Quote: Quoting acrouzetand if you're doing it to a tune you didn't compose, it may go against how the composer intended the tune to sound like.
That is an interesting take. As I come from a plain acoustic music background there it is quite "normal" that audio engineers do those decisions and *not* the composers there. Mixing and mastering is an art form itself in that domain and the results are of course discussed in a good relationship between artists and engineers but I think there is a huge tendency to "trust" your enigneers choices as an artist*.
SID music is a bit different. There is no such thing as "Mixing" that an audio engineer would do. Mixing is completely in the hand of the artist – it's part of the process to create a tune. And only "Mastering" will eventually be applied by someone else if that SID is recorded. And yes: that could be perceived as more "intrusive" than in the acoustic domain, because one is just not "used to" someone else fiddeling with their sounds.
That is one of the key points why I started this discussion here and I thank you very much for sharing your thoughts on that matter.
EDIT: * in fact the one best sounding album of my whole "career" as a musician was the one where a friend of mine didn't take any hints / infos / ideas from me at all and just did the complete mixing and mastering process alone and kept me at a distance. At first I was shocked because he did things I would not have approved in the first place. But as time went on I noticed that engineers listening to me as an artist was probably the cause for every thing before not sounding really "good" in the long run. In retrospective I have to admit: that is the best sounding album I ever released and I (as the composer and main performer) had nothing to do with how it was mixed and mastered.
Ah yeah, it really depends on what/who you're composing for. This varies even within chiptune itself. Generally in "western" chiptune, it's expected that the composer do everything themself. However, what's interesting is that in Japanese game studios it was common to separate composers and programmers. The composer wouldn't do any work with the computer/console. It was up to the programmer to port the composer's composition to the sound hardware, basically doing everything from instrument creation to mixing, and possibly even deciding what parts to omit to fit the composition into limited channels. In other words, the programmers were basically making covers. |
