Sorry to go slightly off-topic, but if anyone can help it will be someone that frequents this thread ...
In light of the clear benefits of letting plugins process at 88.1 or 96 kHz, can someone tell if this is 'ok'...
Basically I recorded at 48 kHz so that's the sample rate of my source wav files. When I first created my project file (in Reaper), it was set at 48 kHz sample rate, so I believe the files were imported, untouched.
Now, I know I can change the project setting to 96 kHz. I believe this means Reaper up-samples my 48 kHz files on the fly - so presumably the effects processing is done at 96 kHz?
Is this ok? Are there any issues with letting Reaper do the up-sampling? (except presumably the real-time CPU demands...
Or... should I re-sample my source files to 96 kHz first (e.g. with iZotope SRC) then import them to a 96 kHz project?
If Reaper has high quality resampling filters and you have oodles of cpu to burn then you can just switch the sample rate in the project settings, otherwise it's best to use a decent resampler offline and import the tracks.
If Reaper has high quality resampling filters and you have oodles of cpu to burn then you can just switch the sample rate in the project settings, otherwise it's best to use a decent resampler offline and import the tracks.
Thanks Andy.
Yep - with brain engaged I realised it is probabably doing what I expected!
Checking the bible on SRC performance (SRC Comparisons), the Reaper SRC isn't so great, so I'll resample offline with the izotope SRC.
Hmmm . . the linear plot does make it even easier to identify aliasing . . maybe Andy thought that seeing the aliasing on a linear plot would then help people identify it on a logarithmic plot? Or maybe it's just an Australian thing? :P
It would be cool to define a standard for the plots and perhaps even start a new thread with those standards being followed from the get-go . . . the plots already posted in this thread are all over the place as you've pointed out! Personally I've seen enough plots to last a lifetime but I think it'd be helpful for people new to this topic!
I really enjoyed this thread . . these days if I'm using very fast attack and/or release times I might run a sine wave through the plug-in and see what it looks like! I don't think I've yet changed something based on this though . . I certainly haven't stopped using any plug-ins . . . . yet. :]
Quick question: Does anyone on this thread know how to change the sine wave frequency in VST Plug-in Analyser?
i was wondering why you only posted a plot with the linear scale and not the logarithmic scale like Span shows and everyone is viewing...?
I made those other plots in Mathematica which by default plots things "Linear" , I just forgot to make them "Log" plots like most spectrum analysers display things. As long as I can see the aliasing I don't really mind if the plot is linear or log.
edit: ah nevermind...answered my own question...i couldn't figure out why i was getting 2nd order harmonics from the Glue when there are none in the early graphs from the thread but then i read on and found out they were added later on...
edit again: i see your post below...sorry about the confusion, i was talking about comparing the new version of the Glue to the old screenshots of it in this thread...
__________________
"it's a long way to the top if you wanna rock -n- roll" -Jesus
edit: ah nevermind...answered my own question...i couldn't figure out why i was getting 2nd order harmonics from the Glue when there are none in the early graphs from the thread but then i read on and found out they were added later on...
If you read the post #60 you referenced it says the plots were generated by a "trivial digital rectifier" which is the function y = abs (x). This trivial function is not used in The Glue. The point of the graphs was to show that basic maths like this causes issues in compressors since it requires a huge amount of oversampling to sound decent, and that other methods are needed apart from just brute force oversampling to reduce aliasing.
There are a whole lot of people out there that would hunt you down if they knew you were distracting Andy from The Drop for even one second stinkyfingers! :]
^the what now...? didn't the beastie boys do that...?
anyway, did anyone test the Vertigo VSC-2 yet...
i did...
lots of aliasing...
(also a shitload of DC, not pictured...actual level @ DC = -62dB...level @ 1,986Hz = -58.5dB...)
(*it's 32bpf, so ignore the red light on the tone generator, it's clean...)
Yes, I did the same test long ago on Vertigo, too much aliasing. I hope they will release update with oversampling option, this compressors deserves it.
The main algorithm that people use for EQ is called a "direct form 1 biquad". This has terrible performance on dc signals, as well as low frequency cutoffs. DC blocking is a combination of both of these, so it really shows which EQs are using decent algorithms. Partially the terrible performance of direct form 1 biquads can be mitigated by the developer using double precision numbers to compute the filter, but this doesn't help all issues. Cytomic never uses direct form 1 biquads and never will. I have come up with a much better alternative that can also handle smooth automation and I've published a technical paper on it here: Technical Papers | Cytomic
A good test signal is a 20 kHz frequency sine wave of ampitude 0 to 1 (+- 0.5 amplitude sine wave with a dc offset of 0.5), then stick that through a high pass filter at the lowest frequency possible, say 1 hz if you can, but not all EQs go down this low, most only go to 20 hz. Have a look at the results in a spectrum analyser, you should only see only a single sine wave at 20 kHz at the output.
A good test signal is a 20 kHz frequency sine wave of ampitude 0 to 1 (+- 0.5 amplitude sine wave with a dc offset of 0.5)...
pardon my ignorance, but does this mean that the 20Hz sine and DC are each at -6dBFS (or 50% each)...?
i think so but i'm not good with math and technical terms...
No a 20 kHz sine wave. It should remain untouched. Many eqs will introduce what is called quantization error, which will be shown as noise, most likely at the low end of the spectrum. Some high pass filters don't actually reduce low frequency rumble but generate it. And yes they are both at an amplitude of 0.5, where 20 * log(10, 0.5) = -6.0206 dB. The decibel scale is an approximation to exact doubling and halving of amplitudes.
i tried to do a few different IMD tests on some of these compressors, but using "traditional" test methods, things got too "dirty" to make any sense of it...
it's pretty scary...the last thing you want to do is run more than one sine wave through these things...it's a mess!
anyway, i came up with something i thought was interesting...
pic 1: source > left = 11,025Hz sine @ 0dBFS / right = 12,000Hz sine @ 0dBFS
when you compress each tone individually, you can see the 11,025Hz sine (pic 2) has no aliasing (aliases on itself...), but the 12kHz sine does show aliasing (pic 3)...
pic 4: when you compress them both together (panned hard L & R) you get artifacts on both channels, the 11,025Hz sine now shows the same aliasing "pattern" as the 12kHz sine....
now, try the oversampling...it may work on the 12kHz tone by itself, but it is useless in this situation...
*edit: seems like the oversampling issue was a bug running in Reaper as VST...can't reproduce now...
hmmm...
When you have two frequencies both combining to do amplitude modulation on the other then you should expect "skirt" harmonics on either side of each source sine of the sum and difference of the two frequencies.
For example for 11025 hz and 12000 hz then 11025 - 12000 = 975 hz, so you should add and subtract this from each sin and that is where you should expect harmonics. This is not aliasing, this is amplitude modulation:
{9075, 10050, 11025, 12000, 12975, 13950, 14925}
{8100, 9075, 10050, 11025, 12000, 12975, 13950}
In the case of a single sine wave the modulation from itself only both add and subtract and bounce back around dc but always land on multiples (harmonics) of the source sine.
I suggest you try the plots again with and without oversampling and have a look at the difference between them. You can use x64 offline oversampling and record the result and play it back through the spectrum analyser, otherwise you are flying blind as it is too hard to tell what is and what isn't aliasing. Either way the results aren't harmonic and this will happen in an analog compressor as well.
in pic 3, the aliasing frequencies for the 12kHz sine are spaced 300Hz apart...
in pic 4, when 12k is combined with the 11,025Hz sine, the spacing is now 75Hz apart for each one...(12k maintains original 300Hz spacing but is now further divided by 4)
it seems to be both harmonics from amplitude modulation and from the 12k aliasing...
Nothing sounds better on the master bus, and nothing even comes close to how well this plugin enhances transients.
give it a shot, well worth it.
Anytime anyone uses such absolutes about plug ins the best thing to do is discount it. There are almost no periods in music! Certainly no absolutes, period.
__________________
Please take a listen to My Tunes
iMac i5 quad core 2.8GHz
12 gigs RAM
Lion 10.7
27 inch screen
Pro Tools 10.3
i7 quad core 3.48GHz
Windows 7 Ultimate
12 gigs RAM
dual 24" monitors
Studio One V2 2.5
Samplitude Pro X Suites
Yep those graphs look great to me too! Looks like you are in an area with 60 Hz mains electricity but the hum shown :-) Any noise at around -120 dB is not an issue, even at -100 dB it's still completely inaudible on a CD!
The right kind of harmonic distortion on a monophonic signal to me sounds great, and is what these graphs are showing me. You can see 20 harmonics of the 1k sin wave. Generating this much harmonic drive in digital as cleanly as is shown here (ie without aliasing) is not easy.
Digital is great at recording and mixing as long as you don't do anything non-linear that generates harmonics. I would guess that the drive heavy parts of most great sounding modern music are generated in the analog domain, and if plugins are used that generate drive they are either running at a high sample rate or used sparingly.
revisiting this topic two years later and I still agree with your statement 100%
This thread has some really interesting information in it - thanks everyone for your contributions.
After reading/skimming all the way through, I had a hypothetical question and I wondered what people thought.
Let's say you own both a hardware comp and it's aliasing software version - for sake of argument the Vertigo VSC-2 - and you wanted to put one of them on your mix buss in a 44.1k session.
Which would ultimately be more "damaging" to your mix. The plugin with loads of aliasing or the round trip of conversion involved with going out and back in through your interface (keeping in mind you are re-sampling a 44.1k track at 44.1 again)
I certainly don't own the hardware so I can't test but I'd love to hear your thoughts.
__________________ - "You only have a certain amount of headroom with Pro Tools... if you start pushing it a little bit too hard it's starts squawking like a chicken, if you go too low, it starts squawkin', you have to work within a certain realm otherwise you get zapped either way!" - Tad Donley (2006) -
This thread has some really interesting information in it - thanks everyone for your contributions.
After reading/skimming all the way through, I had a hypothetical question and I wondered what people thought.
Let's say you own both a hardware comp and it's aliasing software version - for sake of argument the Vertigo VSC-2 - and you wanted to put one of them on your mix buss in a 44.1k session.
Which would ultimately be more "damaging" to your mix. The plugin with loads of aliasing or the round trip of conversion involved with going out and back in through your interface (keeping in mind you are re-sampling a 44.1k track at 44.1 again)
I certainly don't own the hardware so I can't test but I'd love to hear your thoughts.
That would depend on the quality of your conversion, cabling, the closeness of the emulation to the specific piece of hardware, and probably a bunch of other factors. If you owned a 3k hardware compressor, you are probably just going to use it anyway!
The best signal to put through plugins is not 1khz it's best to use an odd frequency like 1.1255khz this seems to show aliasing better.
If you really wanna trip out, run a sine sweep and watch the Aliasing dance around like being on a trampoline!
This way shows up much more and don't forget this is just ONE SIMPLE WAVEFORM that causes these artifacts. Not a full complex wave form as everything in the real world of production would be!
192k is the only time when these artifacts completely disappear.
Major eye opening cause some crunchy edm will need lots of plugin processing and i have always found my self against the wall with doing it plugin style and never understood why they get muddy even when im doing everything as supposed to. Anyhow im not letting go the plugs i now and like so for the first time im convinced to go upper than 44.1khz. Can some good person upload and pm me (i dont want to make this thread even longer for lil o me)the 96khz needed 996 hz test tone as im unfortunately cubase user. Advanced- thank u very much!
Major eye opening cause some crunchy edm will need lots of plugin processing and i have always found my self against the wall with doing it plugin style and never understood why they get muddy even when im doing everything as supposed to. Anyhow im not letting go the plugs i now and like so for the first time im convinced to go upper than 44.1khz. Can some good person upload and pm me (i dont want to make this thread even longer for lil o me)the 96khz needed 996 hz test tone as im unfortunately cubase user. Advanced- thank u very much!
New Reply 
Facebook
Twitter
LinkedIn
...
