Flat Dither Question Mbit+

[Rossd25]

And will using Mbit flat be as effective as a standard TPDF dither?

[Alexey Lukin]

When noise shaping is none, the applied dither is flat (non-shaped) TPDF.

[Rossd25]

Are you talking about Megabit +? I know type 2 flat is TPDF, but I was wondering about Megabit + flat.

[stinkyfingers]

mbit+ (with no shaping) is flat

[Alexey Lukin]

Yes, you can check it with a spectrum analyzer.

[Rossd25]

Great, thank you.

Which is the better dither - type 2 or megabit?

[Rossd25]

And should I keep dither amount at normal without shaping?

[Alexey Lukin]

To get pure TPDF dither, keep it at High.

[Rossd25]

Will normal be sufficient? High seems like it is adding too much noise.

[Rossd25]

Will normal be sufficient? High seems like it would be adding too much noise.

[Alexey Lukin]

When noise shaping is none, normal and high dithering amounts are almost the same: they differ by less than 1 dB.

[Rossd25]

Hm, okay. I am definitely a bit confused now. If dither amount is at normal will that work as well as at high? High is TPDF, but I thought TPDF was type 2. I have no idea. I think I am just going to leave it at normal with no shaping. It seems to work well enough with shaping, so why not without it. Anything with a high amount of dither seems like adding more than is necessary.

And also, is Waves IDR without shaping flat too? Or is it still a shaped dither without noise shaping?

[grrrayson]

Why not simply try them all and see which sounds best to you?

[Rossd25]

What am I listening for? Not to hear the dither or for it to be pleasing? I know when noise shaping is enabled I hear much less of the dither, but it depends on the track. Some tracks I don't hear it at all flat... others it is really apparent.

Dither is the most confusing thing to me. Everyone tells me that I need it, but... really what does it do? I know it covers up quantization error, but is that it? Will my tracks sound more high res with a certain dither over another?

Maybe I should just forget it all and export with pow r on and call it a day.

[dcollins]

Originally Posted by Rossd25

What am I listening for? Not to hear the dither or for it to be pleasing? I know when noise shaping is enabled I hear much less of the dither, but it depends on the track. Some tracks I don't hear it at all flat... others it is really apparent.

In mastering more energy has been expended for what is arguably the least audible thing in the entire chain. Why this is, I'm not sure.

Yes, you need it. Assuming it's competently implemented, they all will work. Highly noise-shaped versions are "fragile" as they cannot withstand any additional processing.

Using flat d*ther guarantees big sales, Grammy™ awards, and the respect of your peers throughout the Internets, so I would personally I would just stick with that and worry about the sounds above -90.


DC

[Bob Olhsson]

Originally Posted by Rossd25

... I know it covers up quantization error...

No, dither prevents it.

You have a choice between 3 dB. of additional noise or 6 dB of edgy distortion and a loss of fine detail from not dithering. Everything ought to simply have flat dither built in and be done with it.

[Rossd25]

Well, I have Finalplug 5 which has a flat TPDF dither. I hate the limiter itself, but I could just disable it and use the dither and be done with it. I wanted to ask why most engineers don't like noise shaped dither? Is it because you can't process the audio again once it is noise shaped?

[Alexey Lukin]

Originally Posted by Rossd25

If dither amount is at normal will that work as well as at high?

Yes, Normal is the recommended setting by iZotope. However if you want what is called "standard 2 LSB peak-to-peak TPDF dither", set it to High.

Originally Posted by Rossd25

High is TPDF, but I thought TPDF was type 2.

Both of them: TPDF is a baseline for both MBIT+ and older dithers coming from Ozone 2.

Originally Posted by Rossd25

And also, is Waves IDR without shaping flat too? Or is it still a shaped dither without noise shaping?

I don't remember exactly, but I thought that Waves IDR is flat when noise shaping is disabled, both with Type 1 and Type 2 dither.

[Bob Olhsson]

My experience has been that the theoretical standard 2 LSB peak-to-peak TPDF dither is the best at preserving tonality, width, depth and room tone provided the audio has never been truncated. In the early days of digital most of the audio I got for mastering had been truncated several times and shaped or reduced level dither often sounded more transparent.

Today on my own recordings, ones I know haven't ever been truncated, standard 2 LSB peak-to-peak TPDF dither clearly preserves a more convincing image and sounds less edgy. My conclusion is that the theoreticians were actually right and that the other dither flavors were just concealing problems.

A further argument against noise shaping and reduced level dither is that today most of the audio we produce is going to be subjected to plenty of additional digital signal processing after it leaves our hands.

Waves/Gerzon type 1 IDR is flat and type 2 is a reduced level smart dither that mutes when there is no signal.

[Alexey Lukin]

I think Type 2 is 1 LSB peak-to-peak RPDF — that's why it mutes when the signal is at exact quants.

[Bob Olhsson]

Michael Gerzon told me it was reduced level TPDF that mutes on black. This was to compete with a feature of the Sonic Solutions workstation which also offered Sony noise-shaping. Samplitude also has this feature in addition to the ability to set the number of bits.

Michael's L1 plug-in also had inter-sample peak protection that he called "analog mode" and was the very first look-ahead limiter in addition to being the very first limiter plug-in.

[Alexey Lukin]

Ah, ok, then it must be 1 LSB peak-to-peak TPDF.

Yes, I have a good deal of respect for L1+. It was a model for how intersample peaks are detected in Ozone's maximizer too. I'm wondering why Waves has dropped this "Analog" mode from other maximizers.

[The Listener]

Now this thread desperately needs that guy, what's his name, that says dither doesn't matter...

[dcollins]

Originally Posted by The Listener

Now this thread desperately needs that guy, what's his name, that says dither doesn't matter...

He must be some kind of Audio Expert!


DC

[Rossd25]

Okay, just to clarify a few things. I am using Ozone 4, so is this the same as Ozone 5 (the most recent version)? So, if I use megabit + with high dither amount I get TPDF. I always use Type 2 with dither amount set to 1. Is that enough dither for it to be TPDF, too?

Waves IDR Type 1 with shaping to none is flat?

[Rossd25]

Originally Posted by Alexey Lukin

Ah, ok, then it must be 1 LSB peak-to-peak TPDF.

Yes, I have a good deal of respect for L1+. It was a model for how intersample peaks are detected in Ozone's maximizer too. I'm wondering why Waves has dropped this "Analog" mode from other maximizers.

I do not like the sound of Ozone with the intersample peak detection turned on. I usually drop the ceiling by default to -.3 and am wondering if that is enough to prevent intersample overs. Quite frankly I do not think that I have ever had a problem with them and most commercial CDs are mastered full scale anyway.

[HB Stone]

Originally Posted by Rossd25

I usually drop the ceiling by default to -.3 and am wondering if that is enough to prevent intersample overs.

Depends on the material; intersample overs could be up to 3 dB over. But as many folks will tell you, do what works. If you've never had anyone come back to you and say the CD's defective, keep doin' it.

[Rossd25]

Originally Posted by HB Stone

Depends on the material; intersample overs could be up to 3 dB over. But as many folks will tell you, do what works. If you've never had anyone come back to you and say the CD's defective, keep doin' it.

They should bring back 50% of their CDs then. I have seen them all over commercial releases... at least according to the meters that I have used to read them. Never seen them in my own material.

[Rossd25]

Alexey, just one more question.

Should normal dither amount Mbit + flat work well enough to eliminate quantization distortion? I do not like it with Mbit set to high and normal seems sufficient to my ears. It doesn't matter if it is TPDF, right?

I ask because I have auditioned the dither at 8 bits on and off and normal seems to eliminate quantization distortion. Low doesn't seem to be sufficient or enough to eliminate it, but I don't hear it with normal. High just seems to add a lot of hiss which is really distracting flat.

Does flat dither have to be pure TPDF or does it matter as long as it does its job? Mbit normal sounds the best to me flat. It eliminates distortion and seems pleasant to the ears as a background hiss. That's just my opinion, though. I do agree that normal truncation sounds much worse than dither, especially at the end of the audio where it is most noticeable.