|
|
Post by Johnkenn on Feb 8, 2024 19:25:27 GMT -6
RIP Dan...now this is a subject he could have waxed poetic about...
Anyway - watching some videos - and I must say, I was rather uninformed. Is aliasing a large part of the reason we think HW sounds better than SW? I know it's not as simple as that, but...
Some of the descriptions of what aliasing "sounds" like were - "metallic and thin," or a "whistle" or "crunchy/clipping." Occasionally, I'll have a mix that I'm just not happy with...it's just sounding more thin than I want and I can't figure it out. I've always thought maybe it was just the limitation of being ITB or maybe I could have gain staged better, etc...I paid attention today and tried to low pass a lot of stuff...made sure oversampling was used on stuff that could do it and then swapped out the Ozone and IK Stealth Limiters I was using for a clipper that has oversampling. And it was interesting to hear how the first mix I did and then this one compared. There was A LOT less crunchy top and just more muscular altogether. Just reading about Ozone - and apparently you have to turn the oversampling on in the Limiter module by selecting "Prevent Inter-sample clipping." I don't think I've ever checked that...
So a couple of questions:
Is aliasing one of the main reasons we prefer HW over SW?
How can you ensure that you're preventing it?
|
|
|
|
Post by Johnkenn on Feb 8, 2024 19:30:51 GMT -6
BTW - he dogs UAD by showing how the LEGACY UAD plugs (2003ish) don't have any harmonics sampled...I think that's pretty well known. Obviously all the new stuff does.
|
|
|
|
Post by doubledog on Feb 8, 2024 19:43:07 GMT -6
Hardware does not have an infinite frequency response. Most of the legacy hardware units we know and love probably struggle to get 20Hz-20KHz (and by that I mean most have a narrower bandwidth).
|
|
|
|
Post by ironinthepath on Feb 8, 2024 20:40:35 GMT -6
I'll definitely be watching that video - at my day job aliasing is a common topic at the data converter level and it's pretty clear you need to limit the bandwidth of what's coming into your ADC (in terms of jargon this is the "Nyquist" or "Shannon" sampling theorem named after the Bell Labs researchers who figured it out from a theoretical perspective). Sampling basically creates copies of whatever is being sampled at multiples of the sample frequency, each copy can be mistaken for the original and this is why it's called an image. Aliasing occurs when the bandwidth of the incoming signal is too high for the sample rate and the images overlap and essentially can never be separated again.
It comes again on the other side (DAC) with something called the "reconstruction filter" whose job it is to pick only one of these images to make it to the final analog output (for audio we want the lowest frequency image).
For plugins though the aliasing is taking place fully in the digital domain, not at the boundaries. It's happening because if we want to model non-linearity (say like the nice distortion of a tube), by definition, we create distortion having harmonics that can have (more like will have) energy outside of the input's bandwidth and these harmonics cause the unwanted overlap in the frequency domain, comes out as non-musical hash.
I understand at a very basic level that this occurs in a DSP context, but other than using the concept of upsampling before attempting to model non-linearity and then filtering out the higher frequency stuff before downsampling back to the original sample-rate, I have no clue. I certainly couldn't tell you which plugins are doing this properly and which are not.
Looking forward to the video!
-Chris
|
|
|
|
Post by bossanova on Feb 8, 2024 21:02:21 GMT -6
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 9, 2024 1:02:45 GMT -6
So a couple of questions: Is aliasing one of the main reasons we prefer HW over SW? How can you ensure that you're preventing it? This is a bit of a loaded question so I'll try to keep it simple. Technically speaking high order harmonics can feed aliasing back into the first Nyquist band so plugins without filters will cause issues, when it comes to HW the ADC filters in a converter will remove this by default and IME there isn't a professional audio interface without filters. So, if you record with HW & real instruments then generally you're going to be fine.
Now, if some of these plugin doctor analysis tests (which includes a few major dev's plugs) are to be believed then what's going on is simply baffling. A bit of aliasing by itself isn't the end of the world but added multiple times in succession with IMD & stuff like automatic delay compensation knocking tracks out of time these software errors begin to stack up. As you mentioned it seems to be a two fold issue, either the end user doesn't know what they're doing (like the Ozone example you gave) and believe me that's not a dig at anyone. I mean are you going to read a 50 - 100 page manual on every plugin you use? Or the dev's aren't sorting the issues out.
I think this is where Dan might have gotten a lot of the information from (Plugin doc analysis) and it's interesting to say the least. TDR scored well especially in insane mode and seems to correct these issues (although precise seems more than adequate), Voxengo, DMG, Cytonic also seem to do well.
Now, back to your question. If this therotical information is correct and they can't fix something basic like aliasing then I really don't have much confidence in the rest of their plugin. All this makes me think is what else have they screwed up? I know this might sound a bit tin foil but I actually think this merits further analysis. So yeah, you probably are better off with HW unless you know what to go for.
Oh, in terms of preventing it (well at least minimising)..
> Low pass everything but that has side effects > Up your sample rate > Use decent plugins with oversampling built in
|
|
|
|
Post by christopher on Feb 9, 2024 10:03:58 GMT -6
My ears always told me the top end on of plugins wasn’t as gorgeous as analog. My ears have lost some extention, plugins have gotten better, but I’m never truly in love with it. And yes, my trick is low pass so much I forget that I do it. As soon as I hear a top end I don’t love, I low pass (or subtract high shelf 5+dB) . Typical 12k to 8k, gentle slope, but really it’s just whatever I like the sound of. Later I choose which tracks to be the top that crests up and over the pack.
When I worked live on digital boards, I hated it so much. I’m blasting at 105dB and always harsh and sandpaper, hard to make the crowd get into it. Mackie never hurt like that. Then one day I said screw it and started the mix with a low pass on everything, and used smaller nodes to bring back highs. Much better. Crowd clearly enjoyed it more, lots of compliments. And I guess that’s why I kept doing it and never looked back. All this said, your mix style seems fine to me..
|
|
|
|
Post by seawell on Feb 9, 2024 10:14:26 GMT -6
I thought Paul did a good job in his video. I know some people that use rx and plugin doctor a lot think "use your ears" is a cop out but after going down this rabbit hole a bit that's pretty much where I've landed. In general to reduce aliasing, I use hardware as much as possible and when using plugins, turn on oversampling. Having said that, there are some plugins that are supposed to suck that I actually like and some that are supposed to be awesome that I think are pretty boring 🤷🏻♂️
Also, here's a response to Paul's video that I thought did a good job of bringing some balance to it:
|
|
|
|
Post by Johnkenn on Feb 9, 2024 12:08:18 GMT -6
The plugin blacklist was a bad idea…but maybe we can compile a list of plugs that have aliasing…
|
|
|
|
Post by notneeson on Feb 9, 2024 12:09:30 GMT -6
^^ “Having said that, there are some plugins that are supposed to suck that I actually like and some that are supposed to be awesome that I think are pretty boring 🤷🏻♂️”
This. 100%
Also, several “vetted” compressors (regardless of their lack of distortion) can really suck the life out of a mix and flatten things in an icky way.
|
|
|
|
Post by robo on Feb 9, 2024 13:06:50 GMT -6
Not to start an annoying side debate, but upping your sample rate can have some pretty yucky side effects too (jitter and IMD). Also, if Dan were here, he would remind us that not all oversampling is equal. Some plugins have weird phase smearing when enabled, so you really do have to trust your ears.
Either way, aside from good oversampling, I’ve found that cutting excess high end before compression/saturation seems to really help keep the hash at bay. You can’t just crank a high shelf like on an analog eq.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 9, 2024 13:12:30 GMT -6
I give up.. LOL..
|
|
|
|
Post by notneeson on Feb 9, 2024 14:39:25 GMT -6
|
|
|
|
Post by Johnkenn on Feb 9, 2024 14:55:42 GMT -6
Who the hell is that? Maybe it’s Dan. We love you Dan. or maybe we’re just losing people left and right. |
|
|
|
Post by skav on Feb 9, 2024 15:00:12 GMT -6
|
|
|
|
Post by seawell on Feb 9, 2024 15:00:35 GMT -6
Who the hell is that? Maybe it’s Dan. We love you Dan. or maybe we’re just losing people left and right. It was ShadowK. As far as people leaving, I can't see a single thing in this thread that would cause someone else to leave. The dramatic exits are getting a little ridiculous in my opinion. |
|
|
|
Post by notneeson on Feb 9, 2024 15:07:38 GMT -6
Who the hell is that? Maybe it’s Dan. We love you Dan. or maybe we’re just losing people left and right. It was ShadowK. As far as people leaving, I can't see a single thing in this thread that would cause someone to leave. The dramatic exits are getting a little ridiculous in my opinion. Wow. I'm very surprised. |
|
|
|
Post by Johnkenn on Feb 9, 2024 15:54:08 GMT -6
This is a bitch-free zone, so if you’re a bitch, feel free to delete your account.
I mean - what the actual fuck? Did I hurt his fee fees from something I said? Seriously, though. If anyone is so freaking thin skinned, I’m glad they’re leaving.
|
|
|
|
Post by Johnkenn on Feb 9, 2024 15:56:42 GMT -6
I really, really wish I still had that IP, because I would ban it.
|
|
|
|
Post by notneeson on Feb 9, 2024 16:01:38 GMT -6
The “under the hood” stuff is interesting to a point, but there’s basically no correlation between knowing a lot about how plugins work and making great records. And by far, the trend is that plugins are getting better and better. So WTF? It’s like an obsession.
|
|
|
|
Post by Johnkenn on Feb 9, 2024 16:06:30 GMT -6
The “under the hood” stuff is interesting to a point, but there’s basically no correlation between knowing a lot about how plugins work and making great records. And by far, the trend is that plugins are getting better and better. So WTF? It’s like an obsession. Seriously - do we know what made him delete his account? I’d venture to guess if we were to hear the work of several of the people that have exited, the insecurity it takes to make that much of a dramatic exit would make a looooot more sense. Anyway, I don't see a single thing that would be offensive to shadowk…so maybe he just flat out thought he was smarter than the rest of us and picked up his oscilloscope and took it home. |
|
|
|
Post by crillemannen on Feb 9, 2024 16:10:41 GMT -6
From what I understood as long you implement oversampling aliasing shouldn't be audioable. Not at X2 and definitely not at X4. Our developer who is also an engineer has tried to educate me on the matter and there are quite a few misconceptions and internet mispreads because People that are using plugin doctor and other measuring software can't read and interpret the data properly.
|
|
|
|
Post by Johnkenn on Feb 9, 2024 16:20:30 GMT -6
A list of plugins that alias would be helpful, honestly. Like the guy showed with the Helios example (which is supposedly upsampled)...it still aliases after cranking the pre over 40db. IMO, the Helios is the best sounding software emu UA does...and I've never used it over 40db I don't think.
Wonder what Drew would say about that? I'm a little scared of getting yelled at lol.
|
|
|
|
Post by ragan on Feb 9, 2024 16:26:35 GMT -6
🤷♂️
|
|
|
|
Post by viciousbliss on Feb 9, 2024 16:41:19 GMT -6
Over the years I spent a decent amount of time studying this over on the plugin analysis thread with guys like Andy from Cytomic. And I've tried to commit a lot of what Dan said to memory while also saving a lot of discussions he was involved in. He said minimum phase filters for oversampling were bad if I recall. And that this was something in Softube plugins aside from the FX. He also said Tapedesk was an "aliasing nightmare" or something like that. CLA76 was also one he hated for various flaws and I'd say the UAD 1176 is tons better. But Dan was using settings I never used. These insanely fast attack times. For all I know he had completely different taste and goals from me. But I always appreciated his deep technical analysis. Over the years I've kept sort of a mental list of plugins with aliasing issues and other objective flaws. Compressors not oversampling the sidechain was another Dan pet peeve. Often he was the only one who dissected them enough to find this stuff.
But what it really boils down to is your ears. If something has aliasing issues, you probably won't like the way it sounds. High-end hardware is classy. It's full and smooth, but clear too. Plugins that are well-engineered like Satin, have this sort of quality about them. But it also runs at 384k. Overloud Gem stuff can give you a ton of instances and there's apparently a price to pay for that. But you could still get a good sound out of it, you just have to find out what settings create problems. So, you experiment and find the strengths and weaknesses of each plugin. Dan seemed to get frustrated when he wanted to use a plugin a certain way and then just couldn't without aliasing, cramping, or some other issue popping up. Plugin compressors routinely seemed to drive him up a wall aside from the TDR stuff, Fuse, and the Weiss. There was probably a couple others in there.
I'd have to go back through old notes I've made and threads I've read in the past to contribute to a list of problematic plugins. It's been a while since I checked out the plugin analysis thread.
|
|
