Are double floats more accurate than single floats? Yes.
Are doubles sometimes necessary in a plugin to combat noise at the expense of speed? Yes.
Is summing two or more tracks together to make a song using single floats going to sound any different? No.
Let's talk about summing
-
- KVRAF
- 14658 posts since 19 Oct, 2003 from Berlin, Germany
Denormal issues usually happen in plugins, not necessarily in the summing engine.HobbyCore wrote:This is actually quite an interesting discussion because these sort of errors are an issue in some algorithms. It may not be a major issue for summing, but that certainly doesn't mean it's useless in the realm of digital audio. Perhaps it's worth mentioning that this is related to the use of denormals in floating-point streams. If you've been in the digital audio world at last ~10 years ago, that should be somewhat familiar to you.
But your arguments turn into that. Same with the ones by the OP.HobbyCore wrote:It is also of interest to see which company's programmers have paid attention to designing their engine to deal with, alleviate, eliminate or circumvent this problem. It does speak highly of their computer science prowess and forethought. Historically it would also be interesting to see which engines show the least susceptibility to underflows like this.
Not every comparison has to be a contest.
These posts imply that x-host is inferior since it's "still using 32bit float" as mixing engine. But I tell you what...
How long are certain hosts existing by now?
Cubase, Logic, ProTools... I don't remember them offering 64bit float mix engines. ProTools is still rumored to be locked at 34-36bit (int) in terms of a mix engine as well.
How many tracks are produced on these host?
And how many of them have "issues" on the long run if we look at the plain summing engine only?!
Not to mention, how many plugins on the market are still internally 32bit float in terms of signal processing?
Of course, if you don't mix properly, and constantly have your sh*t in the red zones (read - peaking around 0dBFS digital), it's no wonder that your material will result in a crap mixdown. But if you worked carefully otherwise, sticked to certain work levels and never exceeded 0dBFS on either channel OR the summing bus... then you should be absolutely fine.
I don't say that this is not an interesting topic, but it clearly implies the undertone "your host is sh*t in terms of the mix engine". Which in turn is another "my host sounds better than your host" or "I can hear bits flop" thread in disguise.
I've said that how many times by now?!
