wellllll


is it so unfathomable that a software company would reuse an oversampling module in their products?

i wasnt even on GS at the time (had an account but was inactive for years)

no, it isn't unfathomable at all - because a lot of their oversampled plugins, even late releases such as SHMC Class-A have 32samples of latency. That was my point

Just to reinforce that despite the fact that latency is not a good indicator of oversampling, it might be in the case of PA, and lack of it in SPL IRON might indicate absence of oversampling.

I'm surprise there is no more guess than one (Macaroni) regarding the blind test.
Or maybe not so surprised...

My guess is that A is OS. If I am wrong then will I blame the headset.

Edit: listened to the clips for 5 secs each just to get an overall impression. Clip A sounded more open and smoother, but did not like the treble / excited highs. Clip B sounded a little bit too hot/crunchy but I preferred the darker overall sound.

my guess is B is oversampled. Dont know how much gain reduction is there going on tho

I thought maybe A sounded slightly better so i guess that is oversampled...

I could barley tell a difference at all tho x

My vote is B is oversampled.

I believe that B is the oversampled version. More open in the higher mids it seems. Didn't download yet though but just listened on the files in browser (but through my system). I did the same test yesterday on a drumbus and I could hear a difference to the better definitely. Will be interesting, because in your files it feels like there is some more changes going on. I think they sound quite different.

I cant say that I hear any think that obviously better than the other. But my numeric analyse indicate that B is the oversampled.
And when doing a null-test it is fascinating that it then is so big difference.

i would say B is the oversampled one and the one i prefer the sound of.

Prefer the sound of B here too.

Wouldnt it be funny if we majorily preferred sound of aliasing?

Yes, I was surprised to hear a tiny difference in the compressor's action, but I used the exacts same settings for each.
Maybe the filter in the highs (Ultrasonic Filter from TDR) did make a difference in the energy received by the compressor, and so changed its action. Or maybe it's something else, but I'm sure I used the exact same settings for each track.

it's a known fact that oversampling in compressor doesn't only alleviate aliasing - it also improves timing. Not sure if this was the case here.

The null test could be a wrong method here : I've got impression that DDMF Metaplugin has some problems to report the correct latency. So it may be a tiny difference of some samples that explains the difference.
Or, like I said before, the amount of energy may be different in each case, because of the aliasing in the non upsampled version that fold back and add to the sound the compressor will treat. It may explain too the tiny difference in dynamics, like if the non upsampled version would compress a bit more...

By the way, what do you mean by "numeric analysis" ?
Is there a difference you can "see" in a frequency analyzer ?

I can only speak for my own 'numerical investigations', but after A/Bing them by ear and finding A to have this thing in the Lo Hi / Hi Mids,
I then looked at a few graphical representations: Spectrogram in and Fourier Analysis.
The slopes in Fourier analysis were identical until about 15k, then the slopes toward Nyquist were slightly different, one being less steep than the other.
On my spectrogram one was ever so slightly less busy in that same range between transients, and had cleaner transients.

I'm just learning how to mix and the infinite contributing variables involved, so I find using visual representations to confirm and/or guide my ears is helpful. I'm finding this an interesting and persistent topic and want to learn as much as I can about it.
I'm confident anyone here could make Iron work as is on a mix if they had to. It's the artist not the tools. But... in music Every atom matters. We just get to pick and choose our battles and hills to die on.

The one with where aliasing artefacts removed should have lower energy level, and with statistics it seems to be a difference. Still it is only statistics, i should probably also calculate the confidence interval too and be more confident if there are bigger changes in high frequency's then on the low. It was fun quick check.

Tonite Mastering a Live Jazz Trio Concert.

SPL IRON running inside METAPLUGIN @ 4x OS.

Firmed up the presentation in a beautiful way.

Not anymore guess regarding blind test ?

It could be cool to have more feedback from guys very active few days ago regarding aliasing in Iron...

Then I will reveal the truth...

Dude, if I am wrong they will have to change the saying to "All headset, no cattle" to describe us audio cowboys.

i thought that was only me

But isn’t doubling SR does not resolves aliasing problems? There’re video on Fabfilter channel, where it all described... and i did some research by myself, basic 7kHz sine frequency modulated by 0.5Hz LFO reveals aliasing “gremlins” very nice. Doubling SR does not erases them, and sometimes they’re not going any quieter.

That fabfilter video uncovered a programming error of not properly low pass filtering a saturation effect.

Iron sounds really good at 96k and will run in the DSP of an HDX system. I can't imagine SPL didn't play with oversampling the program in addition to the sidechain and decided it wasn't worth the extra computational requirements and latency when its sound was compared to the hardware.

I simply don't buy this idea that they are ignorant.

Personnaly, I don't hear any problem while using SPL Iron "normally" (few dB of GR and a source around -18dBfs), but testing the plugin with sweeps and stuff like that, it doesn't compare very well to other PA compressors : it aliases a lot more.
You can test it yourself, or take a listen to my blind test (above).

Personnaly I don't think there is any big problem with 3 instances of Iron in serie compressing (or better said squashing) the same drum loop (my test shows it, and like always in a blind test many people don't want to take the risk to guess the oversampled loop), but it's a fact iron could sound a bit better with a proper oversampling...

That video by Dan Worrell demonstrates that just working at higher SR does not eliminate aliasing. Anyone can test it by ear just using high pitch modulated tone or a sine sweep.

The Iron do have some aliasing at highest attack and release settings (germanium rectifiers) while having 6dB of compression... but that’s almost limiter mode, i wont ever use it that way. It has some anomaly when the threshold knob is around 8...12dB - i’d call it a bug, as aliasing here gets pretty wild and not depending on speed.

BUT: even cleanest DC8C has some amount of aliasing in such mode - fastest peak compression (smash mode) with 0.2ms attack and 20ms release...until you turn on 4x oversampling, even 2x is not enough.

Working at higher SR we can get lower aliasing levels, but it will not almost eliminate it as 4...8 times oversamling does, i guess because of filtering that good oversampling uses...we need to ask somebody like Fabien or Vladg here.

So, in normal circumstances SPL Iron is a great tool. Every compressor is a very complex thing, we need to know how our tools behave.

This is literally the most complex compressor I've ever used, I haven't seen anyone even try the different rectifiers while trying to beat the aliasing drum. I agree it's particularly aggressive at the shortest attack/release times and I don't use those on a master bus anyway.

@ RJHollins were there any complaints of bleeding ears or nyquist?

Hi dirtROBOT.

No ... no the world did not end [last time I checked .... been in the Studio]

IRON is one of my 'simpler' compression to setup and use. IMO.

With all the talk of it, I decided to pull it in on this Jazz Trio Mastering session.
It really complimented the track, and with 1 maybe 2dB compression st most, the stage had a firmness and control. Really sounded good.

Now to be very clear ... I was using METAPLUGIN, with its oversampling options.

Running normal [44.1] sound good. Meta at 2x sounded better ... 4x was the best. I did NOT like running at the 8x oversample option ... there was an edginess to the clarity. [something like that]. 4x OS was clearly the best setting.

Maybe that is what the 'aliasing' watchdogs are referring to? can't be sure. My tech test for aliasing didn't show any issue like that in SPAN ... but maybe I didn't push far enough. Again, don't know.

But also ... I'm doing Mastering ... I've no need to push one processor to the extreme.

I hadn't test IRON against my other comps ... and I have plenty, from SH class A, TDR, UNISUM, to all the ACQUAS/Nebulas, and more.

IRON is just so simple to set ... and had the sound and action that works for the track.

It is to note .... running IRON [and many other Comps and such] inside METAPLUGIN with oversampling options is the way I work.

The difference is NOT subtle.

That's some great real world application advice, cheers!
I was already looking at DDMF/metaplugin just to unify my effect chains in a way that is DAW agnostic but the oversampling seems like icing on the cake.

Oversampling can be a choice, so you can have the best of both worlds, but not at the same time.

B is the oversampled.. ( and I listened with a Sony MDR-ZX110, that's is not that good...)