The No.1 Website for Pro Audio
Lets do it: The Ultimate Plugin Analysis Thread
Old 1st July 2020
  #3571
Lives for gear
 

I have a question pertaining to all this Aliasing stuff. I have an older Art Tube Levelar starved plate compressor. I have only used it with my pod xt before my the amp I use to mic my guitar.

It is a good sounding comp for the most part but it is starved plate. But it has a good rep and mine is in great condition.

This Aliasing subject now has me wondering if I should use this after my mic preamp and compress going in to my daw in order to maybe around using channel compression on my daw tracks, thus maybe cutting down my daw plug compression to just the mix bus and masterbus limiting. I could trade a digital pass of compression for an analog pass.

So there's the question, is this worth doing?. Would it be worth investing in an rnc comp or something better simply for benefit of trading at least one digital track comp in your daw for an analog input comp on tracking?
Old 1st July 2020
  #3572
Quote:
Originally Posted by Jeffguitar View Post
I have a question pertaining to all this Aliasing stuff. I have an older Art Tube Levelar starved plate compressor. I have only used it with my pod xt before my the amp I use to mic my guitar.

It is a good sounding comp for the most part but it is starved plate. But it has a good rep and mine is in great condition.

This Aliasing subject now has me wondering if I should use this after my mic preamp and compress going in to my daw in order to maybe around using channel compression on my daw tracks, thus maybe cutting down my daw plug compression to just the mix bus and masterbus limiting. I could trade a digital pass of compression for an analog pass.

So there's the question, is this worth doing?. Would it be worth investing in an rnc comp or something better simply for benefit of trading at least one digital track comp in your daw for an analog input comp on tracking?
IMHO the only reasons to do that is if you think your analog compressor has a better sound character or is easier to handle than the ones in your DAW.

If you are worried about aliasing there are lots of options where this is no problem at all. TDR Kotelnikov and Limiter 6 come to mind; they both have very good free versions.

Plus:
I bet that if someone checked the aliasing through your Pod XT the results would be very much less than stellar. So if you enjoy that sound then you're fine and dandy anyway. Just go on making music and let worries about aliasing slip away
Old 1st July 2020
  #3573
Gear Maniac
 

People sometimes go a little overboard worrying about aliasing. I'm not saying you should just forget about it (unless you want to.) I measure any potential purchase and if it doesn't look good enough (which isn't the same as "shows no aliasing at all" but rather "if it has any it is low level, below -100dB, preferably lower than -120dB") I am not going to buy it. Because I have plugins which do the job well already and show extremely low levels of aliasing.

I also have some I bought before I started measuring which show higher levels but where I like the sound and because I'm now picky I don't worry about because I'm not stacking them with lots of other aliasing processors (or with serial multiples of themselves.) Their aliasing is most likely completely hidden in the mix.

If you're after the specific sound of a processor and it has aliasing, you can still use it and make good sounding music. You only need to care if you're after the cleanest possible sound (which is a sort of funny thing to say when considering saturation and distortion processors but aliasing is unwanted coloration) as a priority. Which I am and so are some others here. This is easy for me because I don't care if I get the compressor sound of an emulation of a particular vintage device. I care that I get good sounding compression and Kotelnikov (and a couple others) give that to me without noticeable aliasing. Ditto saturation. As long as I get a processor I like the sound of which has low/no aliasing then that's the sound I'm after.
Old 2nd July 2020
  #3574
Lives for gear
 
Laurend's Avatar
 

There's no much place in the "The Ultimate Plugin Analysis Thread" for unanalysed "I like/dislike" this sound. If you like a particular sound, it's perfect. But there isn't any analysis which could improve our knowledge. Maybe a new thread about "good sound analysis" is required ?
Old 2nd July 2020
  #3575
Gear Maniac
 

This thread ran its course as an analysis thread a long time ago. I'd argue the last couple of posts were more useful and informative (especially as they both pertain to talking about plugins that analyse well vs ones that don't vs level of personal tolerance for aliasing in context of appreciation of the overall character of an effect) than a post complaining that we're not posting analysis graphs or whatever.

Something I forgot to say to @ Jeffguitar : Recording with compression means you can never adjust it after recording. So if you aren't quite happy you're going to have to record again. Another tradeoff.

Great sound is available ITB. With minimal aliasing if you're picky about what you use.
Old 2nd July 2020
  #3576
Lives for gear
 

Quote:
Originally Posted by strange loop View Post
This thread ran its course as an analysis thread a long time ago. I'd argue the last couple of posts were more useful and informative (especially as they both pertain to talking about plugins that analyse well vs ones that don't vs level of personal tolerance for aliasing in context of appreciation of the overall character of an effect) than a post complaining that we're not posting analysis graphs or whatever.

Something I forgot to say to @ Jeffguitar : Recording with compression means you can never adjust it after recording. So if you aren't quite happy you're going to have to record again. Another tradeoff.

Great sound is available ITB. With minimal aliasing if you're picky about what you use.
Well, I was just mentioning the pod xt to say that I was only using my Levelar with that and not with other mic stuff like vocals.

I COULD get rid of 4 or 5 digital compressor plugs and use my analog Levelar going in instead. So my question was intending to ask would it be worth it to do some rerouting (which I did and will try) and maybe be able to get rid of 4 or 5 digital compression plugs on individual tracks using my Levelar on tracking and just use one compressor and a limited on the mix bus afterwards.

I have the GE of Kotelnokov and limiter 6 but really I mainly use either Magic deatheye, or My on the mix or sometimes Touwnhouse if the song is more upbeat and calls for more punchy ssl style comps.

I only use the pod xt for convenience as I have my amp surrounded by rigid fiberglass panels and it would be miserable to have to open it up and feel around for the eq knobs and such.

Soon I will get an eq pedal, use my overdrive pedal and can then stay all analog on my guitar amp signal as well.
Old 13th July 2020
  #3577
Gear Maniac
 

Could anyone spare some kind of idiot proof rule of thumb? Most of my plugins (UAD, DDMF, Kush) Hover around -150dB as far as the noise in the harmonic analysis section of Plugin doctor. Is this good, bad, average? Should I back off having 30 instances of a plugin above lets say -125dB for example? Just some margins to have some perspective would be really useful.
Old 13th July 2020
  #3578
Cytomic
 
andy-cytomic's Avatar
 

Quote:
Originally Posted by kev1n28 View Post
Could anyone spare some kind of idiot proof rule of thumb? Most of my plugins (UAD, DDMF, Kush) Hover around -150dB as far as the noise in the harmonic analysis section of Plugin doctor. Is this good, bad, average? Should I back off having 30 instances of a plugin above lets say -125dB for example? Just some margins to have some perspective would be really useful.
You can safely ignore stuff below -120 dB as it will never be audible unless you do deliberately bad gain staging.
Old 17th July 2020
  #3579
Gear Addict
 

Have a look at these screenshots of the T-Racks Pultec EQ... at 48k there's this substantial "rippling" of above ~1khz, which is totally gone at 96k. What exactly is it? Is it a side effect of the oversampling?

48khz


96khz
Old 18th July 2020
  #3580
Gear Nut
 

Most of t-racks stuff and old waves emulation are trash wouldn't recommend anyone these
Old 18th July 2020
  #3581
Gear Maniac
Quote:
Originally Posted by mustgroove View Post
Have a look at these screenshots of the T-Racks Pultec EQ... at 48k there's this substantial "rippling" of above ~1khz, which is totally gone at 96k. What exactly is it? Is it a side effect of the oversampling?
It’s probably oversampling filter ripple. Plot the spectrum in linear frequency scale and if the ripples are at regular intervals, you’ve confirmed the source to be that.
Old 18th July 2020
  #3582
Gear Nut
 

you mean downsampling
Old 18th July 2020
  #3583
Gear Maniac
Quote:
Originally Posted by jams3223 View Post
you mean downsampling
As it’s a plugin, the input and output rates must be the same and thus if there’s downsampling, there’s also upsampling. Aka oversampling.
Old 18th July 2020
  #3584
Gear Addict
 

I've never seen such a pronounced rippling in an oversampled plugin, why would they have chosen to do it this way? Even other T-Racks plugins don't ripple this much, if at all
Old 18th July 2020
  #3585
Lives for gear
 
EvgenyStudio's Avatar
Quote:
Originally Posted by mustgroove View Post
I've never seen such a pronounced rippling in an oversampled plugin, why would they have chosen to do it this way? Even other T-Racks plugins don't ripple this much, if at all
Even Apollo Twin has this kinda quality in its oversampling filter in the converter

Last edited by EvgenyStudio; 18th July 2020 at 08:08 PM..
Old 18th July 2020
  #3586
Gear Addict
 

Quote:
Originally Posted by EvgenyStudio View Post
Even filter in Apollo Twin has this kinda quality in its filter
Which filter are you talking about?
Old 18th July 2020
  #3587
Gear Maniac
Quote:
Originally Posted by mustgroove View Post
I've never seen such a pronounced rippling in an oversampled plugin, why would they have chosen to do it this way? Even other T-Racks plugins don't ripple this much, if at all
That ripple seems to be around +- 0.03 dB which is a normal amount in oversampling ADCs / DACs for example. It won't be audible and only affects the analysis result.

As for the root cause, that's probably an equiripple filter (or two of them) which minimizes absolute error until the very end of the passband in exchange for greater ripple at mid frequencies. I personally favor least squares (or some windowed sinc) filters since nobody is going to hear if the response is down by 0.3 dB at 20 kHz.
Old 18th July 2020
  #3588
Gear Addict
 

Quote:
Originally Posted by Antti H View Post
That ripple seems to be around +- 0.03 dB which is a normal amount in oversampling ADCs / DACs for example. It won't be audible and only affects the analysis result.

As for the root cause, that's probably an equiripple filter (or two of them) which minimizes absolute error until the very end of the passband in exchange for greater ripple at mid frequencies. I personally favor least squares (or some windowed sinc) filters since nobody is going to hear if the response is down by 0.3 dB at 20 kHz.
Thanks for that.

I'm sure there's no issue with it, it's just interesting that after looking at a fair few oversampled plugins in Plugindoctor, this is the only one I've seen which has this rippling. Apart from some Acustica Acqua plugins, but I figured that was related to the sampling of the gear in the first place, but perhaps it's an oversampling thing there too?
Old 10th August 2020
  #3589
Quote:
Originally Posted by mustgroove View Post
I've never seen such a pronounced rippling in an oversampled plugin, why would they have chosen to do it this way? Even other T-Racks plugins don't ripple this much, if at all
I think it's the type of filter being used.

Look at the filters in Equilibrium and you'll see that there are options for filters with and without ripples.

The steeper the filter, which you'd want to avoid rolling off the top end, the more ripples.
Old 11th August 2020
  #3590
Here for the gear
 

has anyone compared real amps vs amp sims with plugin doctor?
Old 11th August 2020
  #3591
Lives for gear
 
EvgenyStudio's Avatar
Quote:
Originally Posted by Ne555 View Post
has anyone compared real amps vs amp sims with plugin doctor?
I did. It’s pretty complicated. I wouldn’t post publicly a video of me testing my Vox AC15 tube head
Overall in ampsim world, NeuralDSP/Helix give you the closest sound to real amp sim in technical way
Old 3 weeks ago
  #3592
Gear Head
 

So how much aliasing is to be determined ok, and not?

To test, I set Voxengo Span to show all the way to -180dB, and run sine tones through compressors. I've tested probably 100 compressors and found maybe 3-4 compressors that has no visible aliasing at all (on span) at 24k without oversampling. But when you look as far down as -180dB, even plugins with 4- or 8x oversampling shows a little visible aliasing. Almost no compressors shows no aliasing from a pure sine at 24k or above.

To 'approve' plugins or not, for a time I listened to the sine tones on headphones, turned up as loud as I could take it, or dared to listen at, and listened for the whistling sound of the aliasing artefacts. Even if I could see aliasing on Span, I approved the plugin if I couldn't discern any of the aliasing in the headphones. With lots of compressor plugins, you could hear it easily.

But something tells me there's more to it than this. I don't buy into the idea that all aliasing below -144.5dB will get cut out at render. Even low level aliasing should be able to affect- or multiply with plugins later in the signal chain, etc etc.

Perhaps this has been discussed in the thread already? It's a loooong-ass thread to browse.
How do you guys decide what amounts of aliasing is ok and not ok? (plz mention your work freq of choice too)
Old 3 weeks ago
  #3593
Gear Maniac
 

Advice from people who know what they're talking about (i.e., developers who write quality plugins and understand digital audio and plugin operation in detail) is if it's below -120dB you don't need to to care, and below -100dB is generally also fine. I have span set to -130dB. I have on occasion accepted a plugin which aliases at as much as -100dB or even a little higher. If it's above -80dB it's a no. Though I have a couple I bought before I cared about this which may go higher but I still use because a little aliasing is unlikely damaging the sound in any significant way any more than intermodulation distortion - a thing that exists even in the analog world. But as long as I'm not piling on the aliasing plugins I figure I'm okay (I'm not usually using more than one or two instances of these and not in serial.) I don't bother with listening tests for new plugins because since I have options that don't alias at noticeable levels it seems pointless. I just go with the processors I like that show low levels and skip the rest. Helps with the temptation to keep buying the new and shiny .

If it's below -144.5dB it's almost certainly being masked by noise when dithering. So maybe not when you render if you're sending off higher bitrate undithered, but for the final product (unless you're one of those people who buys into the idea that there are benefits to releasing in 24bit) and if not by dithering it's probably then lost in circuitry noise or environmental noise on playback.) Yes, it will possibly get worse as you stack up aliasing plugins (so try not to do that) and maybe it will be turned up by certain plugins such as saturators. which is why you test at as close to 0dB as you can get because that shows what the level is likely to be *after* being turned up by anything, since you don't want your final channel level above 0dB. What this means is that aliasing you've measured at -144.5dB is more likely to be a little lower in volume in any real mix (remember if you're pushing a signal at -10dB through the processor the aliasing is probably at around -154.5dB.) But ultimately it's so quiet the likelihood of anything you do making it audible is slim to none.
Old 3 weeks ago
  #3594
The most problematic issues with aliasing appear in control signals, directly affecting the intended function (overshoot protection, compression, gating, limiting, saturating, whatever).

"Primitive" aliasing, i.e. having the audio signal itself alias can be tolerated here and there, e.g. in case of low bandwidth signals and very simple and weak non-linearities.

But a sidechain modulating a signal will result in bandwidth equal the sum of the modulator's and carrier's bandwidths. That is, it will ALWAYS turn up to be far greater, typically twice and large as the input signal's bandwidth. That's why all proper digital dynamics processors must operate at least at twice the audible bandwidth, or inevitably introduce aliasing.

But this is just the beginning. A side chain applying a hard knee will greatly extend the bandwidth of the signal it is trying to control. If the term "hard knee" is really taken seriously, the bandwidth, inside the sidechain, must be able to multiply infinitely! A dynamics processors consists of several more or less aggressive nonlinearities in chain: the RMS or peak detection at least doubles the bandwidth. The Attack/Release switching will easily triple the bandwidth. The thresholding and "kneeing" also at least 2 times in the very best case.

If not handled well, all this aliasing will break the device's original function, its original promises. Even if you can't spot a single partial mirroring at Nyquist, it simply doesn't fulfil what it was intended to do.

This is particularly challenging in a limiter trying to catch the true waveform's max peak, but only having a brutally aliased detected signal to follow.

Last edited by FabienTDR; 3 weeks ago at 11:15 PM..
Old 3 weeks ago
  #3595
Gear Maniac
 

Quote:
Originally Posted by FabienTDR View Post
The most problematic issues with aliasing appear in control signals, directly affecting the intended function (overshoot protection, compression, gating, limiting, saturating, whatever).

"Primitive" aliasing, i.e. having the audio signal itself alias can be tolerated here and there, e.g. in case of low bandwidth signals and very simple and weak non-linearities.

But a sidechain modulating a signal will result in bandwidth equal the sum of the modulator's and carrier's bandwidths. That is, it will ALWAYS turn up to be far greater, typically twice and large as the input signal's bandwidth. That's why all proper digital dynamics processors must operate at least at twice the audible bandwidth, or inevitably introduce aliasing.

But this is just the beginning. A side chain applying a hard knee will greatly extend the bandwidth of the signal it is trying to control. If the term "hard knee" is really taken seriously, the bandwidth, inside the sidechain, must be able to multiply infinitely! A dynamics processors consists of several more or less aggressive nonlinearities in chain: the RMS or peak detection at least doubles the bandwidth. The Attack/Release switching will easily triple the bandwidth. The thresholding and "kneeing" also at least 2 times in the very best case.

If not handled well, all this aliasing will break the device's original function, it's original promises. Even if you can't spot a single partial mirroring at Nyquist, it simply doesn't fulfil what it was intended to do.

This is particularly challenging in a limiter trying to catch the true waveform's max peak, but only having a brutally aliased detected signal to follow.
Great explanation. This is why there are so many defective 1176 type plugins that can’t take the peaks off.
Old 2 weeks ago
  #3596
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by Carrero View Post
But something tells me there's more to it than this. I don't buy into the idea that all aliasing below -144.5dB will get cut out at render. Even low level aliasing should be able to affect- or multiply with plugins later in the signal chain, etc etc.
Well, what you need to take into account is that the total of all the aliasing on all frequencies should be below (in this case) -144dB. The individual aliases may be below, but is the sum total of all of them also still below -144dB?
Also, when the signal is dithered then you can hear 'through' the bottom of the signal, so you can actually hear things below -144dB (when at 24 bits).

In any case, low enough is low enough. There are lots of great recordings with a noise floor of only -60dB. Why obsess about signals that are a whole CD worth of dynamic range lower than that?

Then there is also masking. You were testing with sine waves, but most normal sounds have a spread out spectrum and such low level signals like your aliasing will be masked.
Old 2 weeks ago
  #3597
Gear Maniac
 

Quote:
Originally Posted by monomer View Post
Well, what you need to take into account is that the total of all the aliasing on all frequencies should be below (in this case) -144dB. The individual aliases may be below, but is the sum total of all of them also still below -144dB?
Also, when the signal is dithered then you can hear 'through' the bottom of the signal, so you can actually hear things below -144dB (when at 24 bits).

In any case, low enough is low enough. There are lots of great recordings with a noise floor of only -60dB. Why obsess about signals that are a whole CD worth of dynamic range lower than that?

Then there is also masking. You were testing with sine waves, but most normal sounds have a spread out spectrum and such low level signals like your aliasing will be masked.
Right, the question in my mind is where does a plugin need to be noise wise to not have any audible aliasing within a session with lets say 40-60+ instances, perhaps most plugins are already quiet enough for that to be the case, perhaps not?
Old 2 weeks ago
  #3598
Lives for gear
 
monomer's Avatar
 

Quote:
Originally Posted by kev1n28 View Post
Right, the question in my mind is where does a plugin need to be noise wise to not have any audible aliasing within a session with lets say 40-60+ instances, perhaps most plugins are already quiet enough for that to be the case, perhaps not?
Well, assume that to fit 40~60 tracks into your output dynamic range you will have to reduce the levels of those tracks. So things kindof balance out and typical aliasing under, say, 100dB, is fine for a compressor plugin on a track IMO.

It kindof depends on different factors. If you tend to limit the crap out of your music then -100dB for an individual track may not be enough.
If you have a lot of sine-like sounds with no harmonics you will hear it more as well.
More complex sounds mask the aliasing and make it sound more like noise, not separate unrelated frequencies.

So its kidof difficult to state some sort of magical value of 'enough'. It will depend on the situation, on the music, on the kind of mix you make.
Old 2 weeks ago
  #3599
Lives for gear
 
Laurend's Avatar
 

Because each track has a different audio content, the aliasing errors aren't correlated, thus their summing is always lower than what could be produced using the same signal on every tracks.
Old 2 weeks ago
  #3600
Gear Maniac
 

Issues with cumulative aliasing aren't only about summing - serial aliasing devices can alias the aliasing. I've seen this demonstrated. In all cases, however, overall volume of the aliasing is the real problem. This is why I tend to stick to the -100dB max, preferring under -120dB. Low level aliasing that re-aliases or combines is still low level.

One difference between analog noise and aliasing is aliasing is likely to be less pleasing to the ear. Also it's (I would think) correlated to the signal. So again, I'd say avoid it as much as possible. While not being overly paranoid. If your devices are aliasing at below -140dB you probably don't need to care even if you're using many such devices.
📝 Reply

Similar Threads

Thread / Thread Starter Replies / Views Last Post
replies: 272 views: 69865
Avatar for Roberto Alvarez
Roberto Alvarez 4th May 2017
replies: 80 views: 29259
Avatar for projektk
projektk 30th January 2012
replies: 530 views: 71467
Avatar for joe_04_04
joe_04_04 12th August 2015
replies: 441 views: 67331
Avatar for M2E
M2E 16th June 2019
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
🖨️ Show Printable Version
✉️ Email this Page
🔍 Search thread
🎙️ View mentioned gear
Forum Jump
Forum Jump