Satin - tech question about differences when rendering / bouncing

[Mr D]

Hi Urs,

Loving Satin!

I have a technical question:

With most plugins, if i render to a wav file, the wav file will be identical each time. But not with Satin.

Clear as mud, right? Let me explain:

Imagine i have a track in Samplitude with Satin loaded on it. Now i export to a wav, naming that wav "Export_A.wav"

Once the export is complete, i export again, naming the second wav: "Export_B.wav"

So now i have two identical wav files, right? - WRONG!!!

If i compare wav A & wav B, they're not the same, not even remotely, then don't even come close to nulling if i flip the phase on one.

Why is this important?

Well, it's very important:

In mastering, i might often master an album with Satin on various songs. Say the artist asks me to change only one song. So, I'll open up the session, change that one song, and then re-render the whole album. But now (as we know from the experiment above) that all songs on the album will sound slightly different from the first version. The difference is tiny and not audible when you A/B in a normal way. However, sometimes situations occur when artists or producers are going to check if the manufactured CD has the same sound as the reference files they received from me (via a null test or some other method). So this is going to be a bit of a problem. Obviously they compare to the first set of refs they approved. I can't send them all wavs from the album again to check, when they only asked me to change one song! Or otherwise i'll have to explain: "please note that all the other songs, even though they weren't changed, are slightly different, due to a technical issue, so please check them again" - well, that sounds a bit weird!

Obviously i could only re-render the one song the artist wanted changed, but for work-flow reasons this isn't ideal and makes thing complicated.

I should just mention that my other tape sim Kramer Tape doesn't exhibit this behaviour, it will re-render an identical wav file each time.

Could you let me know your thoughts on this?

Thanks! ........ Mr D

[whyterabbyt]

Obviously i could only re-render the one song the artist wanted changed, but for work-flow reasons this isn't ideal and makes thing complicated.

Suggestion : If its a significant issue, freeze the tracks with Satin on them, or bounce a post-processed copy to a spare track.

FWIW, I wouldnt expect any recording of anythng through analog equipment to ever completely null with a second pass through the same gear. Anything 'emulation' that does might be considered not to be emulating quite as well as it could.

[Burillo]

so Satin is one of the first analog emulations that actually behaves like hardware then?

Suggestion : If its a significant issue, freeze the tracks with Satin on them, or bounce a post-processed copy to a spare track.

that wouldn't help if Satin is on the master, would it?

[Mr D]

I can see that this behaviour (different sound each time) might be pleasing philosophically but i'm afraid in a mastering context it's a bit of a PITA!

[whyterabbyt]

so Satin is one of the first analog emulations that actually behaves like hardware then?

Suggestion : If its a significant issue, freeze the tracks with Satin on them, or bounce a post-processed copy to a spare track.

that wouldn't help if Satin is on the master, would it?

Its ambiguous to me if that's the case, but what's going to be on the master bus is never going to null between instances if anything at all is changed in between time. And if nothing is changed, why redo it?

[whyterabbyt]

I can see that this behaviour (different sound each time) might be pleasing philosophically but i'm afraid in a mastering context it's a bit of a PITA!

For the record, and going by what you said here

The difference is tiny and not audible when you A/B in a normal way. However, sometimes situations occur when artists or producers are going to check if the manufactured CD has the same sound as the reference files they received from me (via a null test or some other method)

Im assuming you're talking about a situation including dithering down to 16-bit for CD? If that's the case, are you sure that your dithering algorithm (which is usually shaped noise) doesnt also vary from one instance to the next.

[Mr D]

I did the test without dither, although in my program (Samplitude) if i had done the test with Kramer Tape and dither ON, it would null 100%, so it's Satin that's doing it i'm afraid!

[whyterabbyt]

I did the test without dither, although in my program (Samplitude) if i had done the test with Kramer Tape and dither ON, it would null 100%, so it's Satin that's doing it i'm afraid!

Oh, Im not doubting it, for the reason I mentioned. For some Im actually slightly surprised that dither isnt truly random though.

[Mr D]

Well, dither is random distribution of noise, but of course a good dither algorhythm will generate the same dither each time, there's no reason not to (and there are benefits to it not being different each time, as outlined in my first post). If there was a reason, the coder would have had to have built in some kind of randomness generator as the start point, but i guess they don't as there's no reason to in a dither algorhythm!

[whyterabbyt]

Well, dither is random distribution of noise, but of course a good dither algorhythm will generate the same dither each time, there's no reason not to (and there are benefits to it not being different each time, as outlined in my first post). If there was a reason, the coder would have had to have built in some kind of randomness generator as the start point, but i guess they don't as there's no reason to in a dither algorhythm!

I'd expect there to be a seeded pseudo-random number generator in there; the difference will be in whether the seed is consistent or not.

[Mr D]

I guess it boils down to this:

When you open a session in your DAW, do you want the sound to be 100% identical to when you last closed the session?

From me, the answer is a resounding: "YES"!

[Ch00rD]

I guess it boils down to this:

When you open a session in your DAW, do you want the sound to be 100% identical to when you last closed the session?

From me, the answer is a resounding: "YES"!

But, as mentioned before, then you can simply bounce to audio, and any halfway decent DAW is able to give you 100% the same result whenever you play the same audio file(s) with the same settings.

Btw, for me, the question is a resounding "NO!" whenever I'm using analog gear, or (good) emulations in software. Much like I'd expect e.g. the oscillators of a virtual synth to be free-running, resulting in non-nulling audio files if I'd record the same sequence multiple times, whereas I would expect bit-idential output when using a sampler.

[sascha]

We have a lot of random stuff going on in Satin, since we model physics and electronics, so it's no wonder that stuff doesn't null perfectly. Have to look at our framework (out of office today), but I'm guessing our noise and random generator functions don't start with the same values every time.

Mr D, to quantify the PITA you're having, can you please do the following:

- in Samplitude, bounce the track to a 32bits float wav (NO integer format!). You could as well do a track freeze, as this automatically does the same.
- load this file in on a seperate track, flip the phase. Perhaps you need to align this to the one containing the realtime instance of Satin, although our latency compensation should take care of this.
- the 2 might not 'null' perfectly. What's the RMS of the delta you're auditioning?
- as this realtime check can be faulty at times, consider a 32-bit float export of this test. You may provide a short snippet here. Thanks.

[Mr D]

But, as mentioned before, then you can simply bounce to audio, and any halfway decent DAW is able to give you 100% the same result whenever you play the same audio file(s) with the same settings.

Seems like you didn't read or understand my first post. When bouncing using exactly the same settings, Satin doesn't give the same result twice. I outlined the reason why this might be a problem in a mastering scenario also in my first post.

Btw, for me, the question is a resounding "NO!" whenever I'm using emulations of analog gear, btw. Much like I'd expect e.g. the oscillators of a virtual synth to be free-running, resulting in non-nulling audio files if I'd record the same sequence multiple times, whereas I would expect bit-idential output when using a sampler.

I understand, but again see my first post for reasons against. It would be a great optional extra, a setting you could turn on if you like, but personally i want a seesion to be the same each time. I undestand about analog imperfections / emulations etxc, but i don't notice any DAW makers advertising the fact that each time you open a session it'll be subtly different! Actually very few peole want that, unless it's built into a plugin, like an analog synth or tape delay or whatever. And Satin is a plugin that has one very big eye on the mastering market, therefore it should (in my opinion) have features that make it appeal to mastering dudes like my self.

[Burillo]

well, the session would be exactly the same - just the render will be different every time