Glad you liked the blog, thanks. I'd have to go dig out my notes (it was the 2008 AES conference in San Francisco) as I honestly don't remember. I also don't remember if it was just a demonstration or an actual presentation of a paper. If there's a paper, I should be able to dig up the numbers in the AES library. I might also be able to dig up the presenter's slides.
But the short answer is a vague "pretty high" to where it was really obvious to most of us. This isn't uncommon, but for me at least, once I finally heard it, I much more knew what to listen for and it became easier to detect and more obvious. But until the level of jitter crossed that first threshold, the clips played all sounded the same to me. But that's just how the brain often works.
If you don't know what you're listening for, you're scanning for all sorts of possible things and probably not doing a very good job of detecting any one thing. It's like the naked guy who runs across a football field while a key play is being executed and hardly anyone notices.
After he played the highest level, the presenter mentioned it reminded him of being in school as a kid and a teacher playing piano albums on a school-issue well-abused turntable that used an idler wheel drive and apparently had large amounts of audible wow and flutter. The "sound" of jitter is a really hard thing to describe with words (for me at least). The music just gradually loses its "structure" as the jitter goes higher but it's hard to put your finger on exactly what's wrong with the sound. In this case, piano notes sort of got harder and more brittle sounding.
The other problem is it's hard to quantify jitter with a single number as it can have multiple frequency components, with some mattering more than others. For many years the Miller Jitter Analyzer was considered the "standard" but, from what I understand, it's been somewhat discredited in the last several years. What we really need is a well run blind study that comes up with a way to weight (or group) various measurable components into a reasonably correlated single number. But that might be wishful thinking on many levels.
