What are the differences of these ditherings do you guys think?
Any guys have ever tried to compare them?

--------------------
Cubase 5.1 + Foobar2000 -> ASIO + MMCSS + Windows7 -> SIIG TI chips 1394 PCIe -> Echo AudioFire 4 -> HH MX250 + Wharf Delta70
A native USA style contemporary R&B artist in HongKong

They are all variations on the theme of pyscho-acoustic manipulation. Essentially noise energy is removed from parts of the spectrum where the ear is more sensitive (nominally the 2-4kHz region) and moved to areas where it is less sensitive (usually into the area above 15kHz).

The different dithering schemes you list -- and many others such as Prism Sound's four SNS dither variations -- each have a slightly different solution to the same problem, shaping the dither noise spectrum in subtly different ways. The UV22 system is the least 'psycho-acoustic' of these systems because it simply places most of the noise in a wacking great narrowband lump at just below half the sample rate.

In most cases, with reasonably loud and compressed material, I doubt you'd hear any appreciable difference between the different algorithms, although the required headroom margins may vary a little. With more dynamic, low noise source recordings the differences can be more obvious, but are always subtle. The psycho-acoustic shaping then comes into play with some algorithms or algorithm variations sounding more natural or pleasant, and others less natural and unpleasant, depending on the source material, the listener, and the listening volume.

If you have the option of different dithering algorithms it's worth experimenting with the different choices to see if one works significantly better on a specific piece of music. Most of the time, you won't hear a difference, but sometimes you might...

Personally, I quite like POW-R, but generally use Prism Sound's SNS 2 dither these days. I have used UV22 in the past, but occasionally had problems with it when further processing UV22 encoded tracks.

Hugh

--------------------
Technical Editor, Sound On Sound

Thx Hugh.

Then how about the Waves IDR and the Izotope MBit+ (MegabitMax) ?
Some people said Waves IDR Type 1 sounds like POW-R 2

--------------------
Cubase 5.1 + Foobar2000 -> ASIO + MMCSS + Windows7 -> SIIG TI chips 1394 PCIe -> Echo AudioFire 4 -> HH MX250 + Wharf Delta70
A native USA style contemporary R&B artist in HongKong

Quote Parco:

---

Then how about the Waves IDR and the Izotope MBit+ (MegabitMax) ?
Some people said Waves IDR Type 1 sounds like POW-R 2

---

They may be right... although the noise-shaped dither shouldn't really have a sound -- the whole point of the psycho-acoustic shaping is that you shouldn't hear the dither noise energy! I've not compared these dithers directly, so can't offer any qualitative opinion.

However, different noise-shaped dither algorthms do put energy in different places, and it's really about finding a dither algorithm that suits the material you're working with. To illustrate the kinds of variations on offer, here is a plot of the four Prism Sound SNS dither options:



As I said, I mostly use SNS2, because it offers a lower perceived noise floor than SNS 1 or flat TPDF dither, but is relatively modest in the amount of noise shaping applied -- particularly at the high end. In contrast, the highly shaped SNS3 and SNS4 offerings, while being subjectively quieter still, can sometimes sound 'strained' and less natural to my ears... and they wack in 30dB of HF noise which can rob the headroom in very bright recordings, like close-miked brass bands!

In 99% of recordings the material's own noise floor will completely swamp any dithering subtleties and you really won't notice an audible difference between any of the noise-shaped dither algorithms on the market. The main thing is that you do apply dither when reducing the wordlength. The kind of dither is really just the flavourings of the icing on the cake!

Hugh

--------------------
Technical Editor, Sound On Sound