So here you're largely talking about flexibility which the simple limiters don't give you. But the 670 clones don't give you much here either, just some tailoring of the time constant. So what are some examples of limiters that are state of the art in this category? Granted, this is the kind of tool where it's more about user skill than technology and, off the cuff, it's not clear to me that this kind of thing couldn't be implemented in a cheap limiter, it sounds like it's something that there's not as much market for in the low end?Gamma-UT wrote:The circuit in the Fairchild 660 is pretty simple but effective. People are pretty insensitive to short-term distortion – if you clip a sharp peak, I doubt anyone could hear the difference. It's why hard clipping can work really well and up to a point is practically unnoticeable. With a limiter you can turn the audio down and back up again real fast as long as the short-term peak is short-lived. The release circuit in the Fairchild does this, but if the audio power is strong after the peak, the release takes longer to kick in to avoid pumping.ghettosynth wrote:Ok, can you elaborate?
Every maximising limiter since then has built on that approach - look at what the audio does post-peak and work out how best to manipulate it to minimise apparent distortion. From what I understand, you can play with release shapes and settings and match them to the music or speech to avoid apparent distortion - which is where you get a lot of variability in limiter performance.
So this is more interesting, thanks for elaborating. Do you know which limiters implement either Izotope's patent or the older patent by Bob Orban.Also, with fast limiting, you are doing amplitude modulation. That tends to lead to sidebands forming around strong frequency peaks and you get lots of intermodulation distortion. You want to hide those effects if you can.
Over time, the design principle behind limiters seems to have headed further and further into psychoacoustics. Alexey Lukin's patent for Izotope (https://www.google.com/patents/US9225310) centres on the same kind of ideas that drive MP3 and AAC compression: psychoacoustic masking. It basically does a lot of parallel processing then picks the one with the lowest subjective distortion. There's a quite old patent by Bob Orban that knocks out the lower sidebands (as these tend to be more audible) and take advantage of psychoacoustic masking of the higher sidebands.
So the upshot is that more sophisticated limiters are pushing the envelope using psychoacoustic principles to mask intermodulation products. So this comes down to being able to hear these products at the edge of performance of lower cost limiters that don't employ these principles.The result, in principle, is you get high subjective loudness that sounds 'transparent' in that the listener can't spot obvious differences in sound at equivalent volume. And I guess there's a lot going on in between playing with release profiles and psychoacoustics - only a select few companies have filed patents on this stuff, so I doubt many are going to come forward with details.
I'll take a look at the patents and some of the source for the JS limiters, but I'm genuinely interested in what precisely the performance difference is when these principles are employed.
It strikes me that the extent to which this is an issue is also a function of material. A two-tone test, for example, is a more stringent test of a system than what one typically encounters with real material. That said, I don't really have any intuition about what kinds of material might leverage this more than others.
So, again, I want to know which limiters are state of the art. Obviously this rules out those tools that are expensive just because they model certain hardware or are being sold for character, just those whose goal is to be as effective as possible at raising the average loudness transparently.