|
Post by craigmorris74 on Jul 16, 2020 17:31:12 GMT -6
I measured the UAD 1176 rev E, and the UA reissue (compression off). The top end on the plugin looks pretty different, and looked that way no matter how hard I hit it. Attachments:
|
|
|
Post by craigmorris74 on Jul 16, 2020 17:45:57 GMT -6
OK, I was able to get closer with the plugin, still different. Attachments:
|
|
|
Post by Guitar on Jul 16, 2020 18:58:59 GMT -6
EDIT: this is wrong: I'd like to attempt to define what even and odd order harmonics, based on my simple understanding. 1K is zero in these graphs. 2K would be the first harmonic. 3K would be the second harmonic, 4K the third harmonic, and so on. The integer multiples are even or odd. So the KT 1176 is showing almost no odd order harmonics, only even. The Slate 1176 is showing a blend of both even and odd harmonics. thanks for posting those mr holmes.
Craig Morris graphs look like there's a ship-ton of aliasing going on, with both hardware and software. I'm just going to have to assume this is due to a 44.1 sample rate or something, because I don't have an intellectual broad understanding. The highest harmonics beyond the nyquist frequency are folding back down from 20.5 kHz and on. Maybe something clipped. That pulse looks really wide too going well past 1K in both directions at lower levels. Mr Holmes test signal looks more narrow in frequency, more sinusoidal. The UAD plugins from craig morris look closer to the UAD hardware, than the Slate looks to the KT.
|
|
|
Post by matt@IAA on Jul 16, 2020 19:03:38 GMT -6
Even / odd are multiples of the fundamental, not integer numbers in series.
So the first harmonic after the fundamental is 2x the fundamental, and it is even order. For a 1k tone even orders are 2,4,6,8k and odd order are 3,5,7,9k.
|
|
|
Post by Guitar on Jul 16, 2020 19:11:15 GMT -6
Even / odd are multiples of the fundamental, not integer numbers in series. So the first harmonic after the fundamental is 2x the fundamental, and it is even order. For a 1k tone even orders are 2,4,6,8k and odd order are 3,5,7,9k. OK thanks! So it looks like the KT 1176 hardware is producing only odd harmonics, no even. People like how those sound. I guess people use 1K as a test frequency to keep the counting simple, I don't know. I confused that with counting integers, my bad!
|
|
|
Post by Guitar on Jul 16, 2020 19:16:06 GMT -6
It’s not really the type of harmonics that makes gear sound the way it does. Yes there’s a difference, pleasing vs unpleasant. When I was at a guitar shop we had some early transistor amps come in, and it’s amazing how tube like they sound vs the 80s-2000s solid state amps. And the early transistor amps, while sounding a million times better than 80s versions, still were dissed by guitarists of the 60s/70s because it wasn’t as pleasing as the tube amps. Today they are very warm sounding and getting some sort of respect for that.. probably thanks to discrete big components. The reason plugins don’t sound as good is because the harmonics are not real, and don’t sound real. Check out Melda plugins.. there’s one that allows you to make your own blend of whatever harmonics you want to add. The harmonics sound like auto tune I'd like to challenge the general public to a "real" vs "fake" harmonic listening blind test but I don't care enough. The harmonics my plugins make are in fact real. Same as the harmonics that come off your favorite CD in your collection.
|
|
|
Post by matt@IAA on Jul 16, 2020 19:25:07 GMT -6
I suspect the issue is that plugins are sampling discrete instances of transfer functions (input to output at a given state and amplitude). So they’re not continuously producing different harmonic balances based on state and level.
I have seen different plugins “jump” from one state to another doing an increasing sine sweep. I imagine this could create some oddness as transients go through that level.
Also - easy way to remember orders vs number of harmonics. The integer number is the number of nodes - the first harmonic has one node. Think the 12th fret on a guitar string. But it’s order is node plus one, because each node adds an integer number of standing waves. First node creates two waves, at 2x the fundamental frequency, which is even order. Second harmonic has two nodes, and makes three waves, at 3x the fundamental, and so is odd order.
|
|
|
Post by Guitar on Jul 16, 2020 19:34:46 GMT -6
That's a great visual aid, thanks for spelling that out. Something I can easily relate too. Here's an internet picture to go with.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jul 16, 2020 21:31:10 GMT -6
That’s other stuff in the hardware. Analog can’t alias because it’s not frequency limited like digital. Many digital emulations have fairly static harmonics. They sound 2d or shouldn’t be used outside of the small range in which they behave like hardware that sounds good over a much wider range or if they don’t sound like hardware, outside or the range where they sound fairly pleasant, e.g. Kazrog True Iron is a cool but fairly static wave shaper. It doesn’t really behave like a transformer but has a lot of fun shapes. If you crank it though, it becomes yet another digital pos. Same with the Waves API 2500. It’s cool on drums, but odd order only. Deep gain reductions on a mix don’t tend to go over well; it’s way too bright and can make guitars reedy. The analog switch that should add in some even just adds noise and and audible aliasing.
I’ll post that and the DMG 1176 in bit along with some atrocious plugins
|
|
|
Post by drbill on Jul 16, 2020 21:37:16 GMT -6
along with some atrocious plugins please name names!!!
|
|
|
Post by ragan on Jul 16, 2020 23:19:49 GMT -6
That’s other stuff in the hardware. Analog can’t alias because it’s not frequency limited like digital. Digital is frequency limited (filtered) because of aliasing, not the other way around. Analog can't alias because...well, it's analog (continuous). There's nothing to reconstruct from sampled data points and therefore no guesswork re: frequency spectra.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jul 17, 2020 0:11:19 GMT -6
along with some atrocious plugins please name names!!! Attached are the DMG 1176 A and D all buttons in with Aliasing all below -150 dbfs and the Overloud TapeDesk running in Neve mode driven into the red to saturate a bit of the top end off. That's aliased to hell even not driven into the red. The Trident is even worse. Without the console emulation on, your air suffers from aliasing. Stack the tape and the mix becomes 2d and suffers from Waves syndrome. Stack the console emulation, get Indiana Jones the Crystal Skull digital monkey vomit.
He didn't enable oversampling in many of the compressors though. Not that Waves has oversampling.
|
|
|
Post by mrholmes on Jul 17, 2020 2:56:18 GMT -6
This therad turned intresting and made me to do more measurements....
I dont have time to post shots right now.
A quick summary in no special order:
Most of the 1176 simulations fail.
They alias like crazy and thats not what I can see in my hardware 1176s.
Very intresting is the simulation of the PA SSL 4 K.
I did this in the past with some SSL gear and the PA SSL has no aliasing going on at all and looks very similar to what I remember by the hardware.
The Slate SSL 4k has aliasing like crazy but the Slate API has nearly no aliasing.
All plug in compressors which try to sumulate tube gear alias like crazy. Even though PA SPL IRON sounds fantastic to my ears.
PA ACME aliasing like crazy in the upper range of the spectrum, but sounds gorgious on base guitars, very usable for the low end.
A BIG surprise was PA Lindell 80 as long as you gain stage it nice it is very similar to my NEVE Clone preamp by Vintage Design. Dont rasie the THD button!!!
Antoher BIG Surprise was Slate VTM no aliasing at all as long I dont go overbias....
I am bit disapointed by the Slate they cash every month and deliver some big compromises? They may correct this asap? In this unfair to say?
|
|
|
Post by Guitar on Jul 17, 2020 7:44:59 GMT -6
I believe if you over sample either within the plugin (2X, High Quality, 4X, 8X etc), or within the entire DAW project, to a higher sample rate, your aliasing will be significantly reduced. Of course this comes at the cost of realtime audio computer processing power. There was a cool Dan Worral video on this.
The problem is you'd have to watch every track on a graph to see what's going on because you likely won't hear it all the time.
Too many questions, here's the video on the topic:
|
|
|
Post by Guitar on Jul 17, 2020 9:49:18 GMT -6
One tidbit I got from re-watching that video is amplitude modulation (compression, limiting) behaves more and more like clipping the faster the attack and release times are. So maybe that graph from a few posts back of the UAD and the UA was a fast attack and release time resulting in something like clipping, which in my imagination makes side bands and intermodulation distortion, a much more dense bunch of harmonics that acts more like noise.
Don't know if that's accuarate or not. It's also interesting that ring modulators (effect) seem to have the same description as intermodulation distortion, sum and difference of two frequencies being added to the signal. I wonder if a ring modulator pedal is maybe an intermodulation distortion pedal.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Jul 17, 2020 13:22:19 GMT -6
This therad turned intresting and made me to do more measurements....
I dont have time to post shots right now. A quick summary in no special order:
Most of the 1176 simulations fail.
They alias like crazy and thats not what I can see in my hardware 1176s. Very intresting is the simulation of the PA SSL 4 K.
I did this in the past with some SSL gear and the PA SSL has no aliasing going on at all and looks very similar to what I remember by the hardware.
The Slate SSL 4k has aliasing like crazy but the Slate API has nearly no aliasing. All plug in compressors which try to sumulate tube gear alias like crazy. Even though PA SPL IRON sounds fantastic to my ears. PA ACME aliasing like crazy in the upper range of the spectrum, but sounds gorgious on base guitars, very usable for the low end. A BIG surprise was PA Lindell 80 as long as you gain stage it nice it is very similar to my NEVE Clone preamp by Vintage Design. Dont rasie the THD button!!!
Antoher BIG Surprise was Slate VTM no aliasing at all as long I dont go overbias.... I am bit disapointed by the Slate they cash every month and deliver some big compromises? They may correct this asap? In this unfair to say?
Yeah every 1176 plug but DMG Trackcomp aliased hard or had no even major order harmonics like the Behringer clone of the Warm clone. The DMG is SPICE modeled circuit emulation but obviously is choosy with what to model because it just doesn’t do what the DBX, SSL, and API stuff does perfectly right at the end of the chain. Klanghelm SDRR and MJUC don’t alias in HQ mode unless you drive them into the red. The Magic Death Eye plugs also don’t audibly alias either but they’re modeled after modern cleanish tube gear. It also might have more static type of distortion than MJUC but I haven’t tried it yet. I don’t think either MJUC or MDE are fully SPICEd but they behave better than many of the DMG comps. Plugin Alliance console THD function is just a shitty, non-oversampled soft clipper. Slate has always done that. It’s a meme at this point. Even the hardware Paul Wolff helped design is still using buyers as a beta test. The interface looks killer if you drank the Slate kool aid with both PCI-E (in 2020! Awesome!) and thunderbolt but who knows how good it sounds or how the drivers are. It’s still a Chinese interface with chinese condensers for 2k that’s obviously meant to take a bite out of UAD and doesn’t seem to be flying off the shelves. It could be awesome but it sits used online for 1200. The Steinberg thunderbolt thing used initial buyers as a beta test too but Yamaha is Yamaha and came through. The money shots of the AXR4 board look sick meanwhile there’s nothing for the Slate; only hype, just like the plugs.
|
|
|
Post by mrholmes on Jul 17, 2020 14:51:58 GMT -6
This therad turned intresting and made me to do more measurements....
I dont have time to post shots right now. A quick summary in no special order:
Most of the 1176 simulations fail.
They alias like crazy and thats not what I can see in my hardware 1176s. Very intresting is the simulation of the PA SSL 4 K.
I did this in the past with some SSL gear and the PA SSL has no aliasing going on at all and looks very similar to what I remember by the hardware.
The Slate SSL 4k has aliasing like crazy but the Slate API has nearly no aliasing. All plug in compressors which try to sumulate tube gear alias like crazy. Even though PA SPL IRON sounds fantastic to my ears. PA ACME aliasing like crazy in the upper range of the spectrum, but sounds gorgious on base guitars, very usable for the low end. A BIG surprise was PA Lindell 80 as long as you gain stage it nice it is very similar to my NEVE Clone preamp by Vintage Design. Dont rasie the THD button!!!
Antoher BIG Surprise was Slate VTM no aliasing at all as long I dont go overbias.... I am bit disapointed by the Slate they cash every month and deliver some big compromises? They may correct this asap? In this unfair to say?
Yeah every 1176 plug but DMG Trackcomp aliased hard or had no even major order harmonics like the Behringer clone of the Warm clone. The DMG is SPICE modeled circuit emulation but obviously is choosy with what to model because it just doesn’t do what the DBX, SSL, and API stuff does perfectly right at the end of the chain. Klanghelm SDRR and MJUC don’t alias in HQ mode unless you drive them into the red. The Magic Death Eye plugs also don’t audibly alias either but they’re modeled after modern cleanish tube gear. It also might have more static type of distortion than MJUC but I haven’t tried it yet. I don’t think either MJUC or MDE are fully SPICEd but they behave better than many of the DMG comps. Plugin Alliance console THD function is just a shitty, non-oversampled soft clipper. Slate has always done that. It’s a meme at this point. Even the hardware Paul Wolff helped design is still using buyers as a beta test. The interface looks killer if you drank the Slate kool aid with both PCI-E (in 2020! Awesome!) and thunderbolt but who knows how good it sounds or how the drivers are. It’s still a Chinese interface with chinese condensers for 2k that’s obviously meant to take a bite out of UAD and doesn’t seem to be flying off the shelves. It could be awesome but it sits used online for 1200. The Steinberg thunderbolt thing used initial buyers as a beta test too but Yamaha is Yamaha and came through. The money shots of the AXR4 board look sick meanwhile there’s nothing for the Slate; only hype, just like the plugs.
To my surprise the tube Preamp Collection by Slate is not aliasing at all. I dont understand .... if they can do it with a few ones .... why cant they do it with all plug ins?
|
|
|
Post by ragan on Jul 17, 2020 17:51:04 GMT -6
Let's not fall into the trap of seeing aliasing as a pure dichotomy of good/bad. Looking at plugins with analyzers is interesting and fun but it's a different pursuit than using DSP to make music sound good. There are all kinds of reasons devs choose to expend system resources combating aliasing (or not) in a given scenario. Outside of the microcosm of internet forum plugin dogma, it's a nuanced subject with case-specific design considerations.
|
|
|
Post by EmRR on Jul 17, 2020 18:16:33 GMT -6
Let's not fall into the trap of seeing aliasing as a pure dichotomy of good/bad. It will make software output differ from original hardware, if that's the only question. Without the hardware in hand, you can't really do a comparison. One of the big arguments against tube gear was all the intermodulation distortion; a similar effect.......there's also aliasing intermodulation distortion, a doubling of evils.
|
|
ericn
Temp
Balance Engineer
Posts: 14,937
|
Post by ericn on Jul 17, 2020 19:29:56 GMT -6
Even / odd are multiples of the fundamental, not integer numbers in series. So the first harmonic after the fundamental is 2x the fundamental, and it is even order. For a 1k tone even orders are 2,4,6,8k and odd order are 3,5,7,9k. OK thanks! So it looks like the KT 1176 hardware is producing only odd harmonics, no even. People like how those sound. I guess people use 1K as a test frequency to keep the counting simple, I don't know. I confused that with counting integers, my bad! Unless you love the Aphex Exciter or BBE, don’t get hung up on even/ odd harmonic distortion. Because that’s all those boxes are I can never remember which is which. You have to understand any deveation from the original waveform is distortion. Some of it is very very subtle, you can hear it but you can’t measure it. And if you compare on a digital scope depending on the resolution you will completely miss it.
|
|
|
Post by ragan on Jul 17, 2020 20:16:02 GMT -6
Let's not fall into the trap of seeing aliasing as a pure dichotomy of good/bad. It will make software output differ from original hardware, if that's the only question. Without the hardware in hand, you can't really do a comparison. One of the big arguments against tube gear was all the intermodulation distortion; a similar effect.......there's also aliasing intermodulation distortion, a doubling of evils. Right and the question for DSP devs is what matters audibly and what resource costs are associated with what methods. As for the audibility question, well, that gets into some pretty murky (and often religiously passionate) territory.
|
|
|
Post by Guitar on Jul 17, 2020 20:22:27 GMT -6
OK thanks! So it looks like the KT 1176 hardware is producing only odd harmonics, no even. People like how those sound. I guess people use 1K as a test frequency to keep the counting simple, I don't know. I confused that with counting integers, my bad! Unless you love the Aphex Exciter or BBE, don’t get hung up on even/ odd harmonic distortion. Because that’s all those boxes are I can never remember which is which. You have to understand any deveation from the original waveform is distortion. Some of it is very very subtle, you can hear it but you can’t measure it. And if you compare on a digital scope depending on the resolution you will completely miss it. To be quite blunt about it, I have tested even and odd distortion modes in some guitar overdrives, and I like them both. Same with the Andrew Scheps Omni Channel plugin. I almost have no preference for one, the other, or a mix of the two. So it's a very good point you make.
|
|
|
Post by the other mark williams on Jul 18, 2020 6:44:26 GMT -6
Yeah, if “no aliasing” was the most important goal, I’d be using Waves RVox all over the place. It doesn’t alias at all IIRC. Aliasing is not the only consideration, just as self-noise isn’t the only consideration in a microphone.
|
|
|
Post by Ward on Jul 18, 2020 6:51:36 GMT -6
If you replace the Field Effect Transisters with Operational Amplifiers, in effect, are you not turning it into a pseudo Voltage Controlled Amplifier as you are no longer using FETs?
|
|
|
Post by Guitar on Jul 18, 2020 7:31:12 GMT -6
If you replace the Field Effect Transisters with Operational Amplifiers, in effect, are you not turning it into a pseudo Voltage Controlled Amplifier as you are no longer using FETs? Most of the transistors in the 1176 are BJT type transistors, bipolar junction transistors. There is one FET in the "compression element," and a second FET for the visual meter. There are a lot of audio amplifiers built out of discrete components, and the BJTs. You can replace a lot of those "blocks" with op amp circuits which simplifies the layout a great deal. here's a schematic to look at from Gyraf.
|
|