…at a level we can’t perceive…
What I said was that we don’t know exactly which spot will add and which spot it does not - you really twist my words.
In some spots it will add other might cancel something - from the same file.
So instead we just use better SRC from somewhere else.
You don’t want to fiddle with a lot of workarounds - and gate these levels out and stuff.
I know for sure - next time I pay for Cubase update - it has better SRC - up to Steinberg to show how professional they are and when that will be. 
If I deliver a bunch of tracks to a client or collaborator I want to know it measure up to standards.
Rather embarrassing to explain if anybody see this:
- yes, you see, Cubase does this, you know, not me
I am responsible for that either way.
If that is blown up to speakers in a cinema theater you may hear this crap as some harsch haze.
Compressors will lift this as well as any other noise - as well - all up to any audible level commonly used.
And since Cubase IMD is 20 dB above noise floor you will hear that first.
And about dither - for bitreduction to 16 bit this is not a problem.
Well outside range of 90 dB - I talk about import where it is processed further.
My only point is that you’re making a big deal out of something “we know even less” about…
You can’t gate it out. The distortion is surely dependent on the input signal. No input signal = no noise.
I’ve mixed for theater and used plenty of samples from a 44.1kHz library, all SRCd to 48kHz. I’m pretty sure that if you go into the SRC and look at raw data you’ll find artifacts there too. Did I hear it? No. Did anybody else? No.
If Cubase is good enough for Hans Zimmer it’s probably good enough for the rest of us.
Speaking of theatrical releases though: What do you think bothers people the most, this SRC “issue” that isn’t even proven to be audible, or bad ADR? Do I as a producer of audio for picture gain more by getting better ADR tools or a different SRC algorithm?
There are like a million things that go into audio-for-post in general, and this is pretty far down on the list I’d say.
Ok, I just looked again and that image you posted said - if I’m reading it correctly - -110dBFS, correct? So, you boost your signal so that 0dBFS is equal to 120dBSPL, which causes discomfort and pain in peoples’ ears. Now subtract 110. Ok, we’re at 10dBSPL for that distortion. Average home is about 50dB.
Who hears distortion 40dB below the room tone in which they’re listening, while actual content is blasting at 120dBSPL???
Bueller?.. Bueller?..
Ok, fair enough, you compress. Again; how much? How loud is the signal you’re compressing? How much are you raising all levels?
+1,0000, with the caveat of my edit added in red, above. Addresses djw’s point also - it’s probably very hard to do what he mentioned, but it’s been demonstrated as not very hard to make a better SRC by virtually every other DAW maker, and Steinberg as well (Wavelab’s SRC is much better quality than Steinberg’s).
I thought you were curious about DAC’s.
What my system do is not the point - it’s well below -120 dB in this analog loopback test, and Cubase IMD is at -110 dB - meaning +20 dB above that.
I’ll add, then likely bow out: If the SRC replacement waa a simple “Pop out the old, pop in the new” kind of thing, there really would be no excuse to not do so.
Odds are it’s more complicated than that, and I do have to admit there are other things I’d personally like to see fixed first (see my feature suggestions, as well as all the ones I’ve "+1"d to). But I do empathize with the guy who had to answer a client about why he’s using an SRC that he’s heard (and even seen the graph of!) is clearly inferior to those found in other DAWs. And the fact that the chief technical editor at SOS spares little criticism of Cubase’s SRC puts those here with similar criticisms in company of very good standing indeed.
Re:
…If Cubase is good enough for Hans Zimmer it’s probably good enough for the rest of us …
- I wouldn’t be surprised at all if he used a third party SRC … as some senior, very experienced posters on this forum have said they do … but who knows!
Thanks!
Well put - why is everybody else putting more attention to this than Steinberg?
So are we a riot of three now then? 
There is some love in the feature requests also - and maybe me added some graphs will make it more obvious it would be good to attend to it.
Since it’s been annual request - don’t know for how long.
Testing in Samplitude, I found that just drag n drop files into workspace - used the stretching stuff which is much worse than best SRC’s.
But you could say no to immediate conversion and go into a dialog and choose from a range of algos - whatever you wanted to use to convert it.
At least something like that would be nice - something built in. Or just do as Sonar and most, put something really good there - it’s just a call to another piece of code. At infinitewave you could see that many daws like Reaper has improved it over time - even if Reaper does not convert anything on import, it’s all in realtime.
If it was uncomplicated it would have been done for sure.
Well, that dude at SOS called it “fairly poor”, which to me is a lot softer than “spares little criticism”. And secondly, even a guy with a nice title and an audience can evaluate the significance of something incorrectly.
Incidentally it’d be interesting seeing some more testing on this.
As for the “client”, that was as far as I can see not something that happened, but just a hypothetical.
He’s been more descriptive in the forum itself, though to be honest I really can’t say how much criticism he spared or didn’t, can I? Sorry!
He doesn’t just have a nice title - he worked for years (decades?) as a BBC engineer. He definitely has the chops.
Why even add things you don’t need or want?
The point - in not so many words is - everybody else do a better job in SRC - obviously caring about this.
Steinberg can just copy 100 lines of code from Wavelab and done.
I looked at Secret Rabbit Code - and it’s not like it is a new application or anything.
Steinberg may be ignorant about this - so we point it out - EOS.
We argue because we care!
And I do it because I’m worth it. 
Because you add pure signal too.
- 1000 x 1 unit of pure signal is 1000 units
- 1000 x 0.000002 (0.0002%) units of distortion is 0.002 units of distortion
0.002 units of distortion in 1000 units of pure sound is still 0.0002%.
EDIT: Oh cr*p. Mattias already explained this.
That is measured as THD 1 KHz sine - and nothing else. Considered low, like on a hifi amp or something.
It tells nothing about full spectrum which in some cases sound shite.
When measured distortion on a single frequency at a time - in a sweep - and you see in a graph what it does to a full spectra which is closer to what music is. This is the additive component. Then another track and another track and so on…depending how much you import.
Why deliberately add this if you don’t need to?
Some part are maybe lower signal level and you have a limiter with threshold at -20 dB or so, usually creating a makeup gain of 20 dB.
And suddenly you in the risk this crap is directly audible. And maybe you do what some do multiple compressors in series - and similar things reduce dynamic range. And suddenly you are in the CD range if doing bitreduction.
blah blah blah
Compressors/limiters/gates/etc have no effect on Cubase SRC’s distortion levels because they are fixed amount below signal level, NOT below FS.
It won’t be, it will be a lot higher, typical real life analog noisefloor is around -75dB to -80dB
Now, where are you going to find that additional 60dB (remember logarithmic) in order to be able to playback that distortion.
You system is NOT capable of playing it back.
The 2nd talkingpoint would be that -80dB thermal noise would mask / outplay ANY distortion at the levels you measured.
Well, you better learn what compression does then - it lift lower noise floor with make up gain to start with.
If working below threshold - actually compressing, gain reduce - and above threshold make up gain is there.
So the artifacts, if being at -120 dB or something are raising up in level if compressing.
You can try Waves L2 and lower threshold, and you hear how noisefloor go up. It’s automatic make up gain working.
I assume that if you are THAT anal about sound quality, you will prepare and process your samples upfront in order to achieve the best SQ.
You will not boost them with 60dB or have them realtime SRC’ed. But again, your analog output will be limited to a -80dB floor, still not an issue.
Best noise floor doing mike recordings was about as you say 75 dB or so. I had to shut a door, not to hear fridge in kitchen sounding to get that. This Focusrite preamp at 60 dB gain and SM7B - so preamp involved to full extent.
Where I lived at the time was more noise and traffic outside as well - here probably better values(if hens on the farm are quiet).
Analog loopback is much lower than that - just dac output to adc input at +4 dB. I even did through Mackie mixer compared to pure line level looped back by cable - it values cruised along the original pattern at - 120 dB. I did this so I could compare my own dac also put into Mackie to raise level to normal +4 dB from the 0.3V thingy normal in hifi stuff.
But what I hear - and my system does is not the point. What leave my studio and delivered as files should be as top notch as it can be.
And using even free tools they do a better job than Steinberg flagship daw - which makes me think it’s just overlooked and haven’t gotten their attention.This is my goal to remedy that.
The RMAA tests are completely ITB - RMAA test pattern saved to disk, imported into Cubase project and then rendered back to disk and taken in for analysis in RMAA. Graphs shown here - and comparison doing the same thing in r8brain, old Sonar 8.5, and original testpattern by RMAA and Cubase in green standing out like a christmas tree. No analog involved.
You can test this by yourself - anybody can. RMAA is free version also. Only thing to remember, that I struggled with a bit, was to set levels with a pattern first so they become green - then do the analysis. Just a tip to shorten time getting results.
And to emulate as close as InfiniteWave site - I importen in a 96k project in Cubase and worked with 24 bit 44k testpattern.
This artifact stuff - not needed or wanted - is added by Cubase inferior SRC.
My reasoning all the way.
If it’s in your face as imported is not the point. It may show up as something you later struggle with in mix and mastering.
If you use an exciter or something you know add harmonics - this also work on everything on that track - including possible SRC artifacts.
And simple things like a compressor also raise this floor of noise and SRC artifacts.
I think we all strive for getting as good quality recordings we can - and remove any tendenzy to harsch components. That’s why we don’t make levels so peaks overs and all the rest. Many things are limited by the system and equipment we have and we can’t do anything about it.
But SRC can be fixed - rather easy too. so why not do what the rest of main daws out there did?
If devs at Steinberg spent less hours attending to this - than we spent discussing in this thread - they would have fixed it already.
Pick some code from Wavelab, or a developer library used in Wavelab and change the call that do SRC at import - done.
Wavelab is not top notch in comparison - but way better than current code in Cubase.
If Steinberg want to give us named options for SRC algorithms - like Reaper and Samplitude - fine.
It could be in import dialog and export dialog.
If I had Wavelab I would have done the same thing in RMAA - but have to trust Inifinitewave place since other results presented there clearly showed up in RMAA tests as well. And had some trouble getting RMAA recognize Reaper output wave - so left that out for the moment.
Noise floor on my Nordlead 2X was about -96 dB last time I looked, fed right into line inputs of RME card.
So IMD artifacts from Cubase at - 114 dB(in graph earlier) would clearly be there if that was imported as we talk about now.
Rember this test - two distinct frequencies present - do that - what happends to compiled music material with every frequency just about, we can only guess where levels are then - it’s not better I can tell you. This test work as a comparison between equipment - not absolute for music material.
Then a compressor 10-20 dB and you are already in the noise floor with miked stuff - assuming still just two frequencies in the test.
As of now yes, I do SRC externally. Some vocals that I hired people and do 44k at the time will be reused in new context. Some tracks with synths I had at the time, and no longer have, will also be imported.
Maybe one day in Cubase, in a dream, I can just drag n drop files onto track without event hinking about format of files and just trust Cubase to do the best with it.
All things matter, believe it or not.
Why use a tool that do worse job than all competitors for this task - I know I won’t.
- let’s see, yes that track was imported with Cubase SRC involved, I have to be careful with that.
- put gate first maybe, just not use compressor as first insert…
I was astonished by how much cables do to alter sound. On electric guitar changing to GeorgeL cables - brightened sound quite a bit - since so much lower capacitance together with pickups internal resistance, being less of LP filter. As much as changing pickups to a model brighter sounding - just changing cables. All the years spent on guitar forums and what people do, me too.
I will not spend $2000 on a single preamp like big studios does. So I try to make the very most of what I have.
So again - attending to details - if that is anal - then be it.
Then SoundOnSound guys are anal as well as the rest I read on the matter on the web.
I’m sure a good bunch of people do projects at 96k - and why is that then?
I mean everybody know only babies hear up to 20k, so 22k present for 44k - why do that.
So much extra resources used for nothing - ooooooooooooor?
For now I only do synths at 96k for aliasing purposes being so obvious when listening.
Did some tests recording acoustic guitar at 96k, and it was something about it compared to 48k.
But not sure yet if imagination - so will wait a bit.
As with many tiny things - they may not be obvious alone, but together for a whole mix you might reconsider.
All things matter…
I know very well what compression does and how it works, thank you. You on the other
hand should learn about floating-point digital audio and how rounding errors in it behave.
-120 dB compared to what?
It kind of irks me that Cubase has the worst SRC in the industry, on the other hand it is not something worth loosing sleep over.
Afaik wavelab elements uses SOXR, and that is already far better.
https://sourceforge.net/p/soxr/wiki/Home/
I would not mind for something like that being implemented , even if it only were for marketing purposes
Cubase now with better SRC, buy it while it’s hot.