Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Best Dither Settings In Cool Edit (Read 5684 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Best Dither Settings In Cool Edit

Source Audio: 44.1 Khz 24 bit.
Destination:  44.1 Khz 16 bit (Audio CD)

I want to use Cool Edit to dither the above source
to the above destination and I wonder what are the best settings to use.

David Robinson recommends:
--------------------------
48kHz Noise shaping curve
Triangular p.d.f.
0.5 bits dither depth

"I use 0.5 bits of triangular dither, with the '48kHz Noise shaping'
If you don't like the idea of using the 48kHz noise shaping curve at 44.1kHz, then use the 44.1kHz curve - but to my ears, the one labelled "48kHz" sounds slightly better. Also, I believe the correct amount of dither would be 1-bit, but again to my ears (and with all the tests I can throw at it) 0.5 seems to behave correctly, and gives a lower noise floor."

From:
http://pub41.ezboard.com/fr3mixfrm6.showMe...opicID=13.topic

Alexey Lukin recommends:
44.1 Noise shapping curve
Triangular p.d.f.
1.0 bit dither depth (2 LSB peak-to-peak).

He says this setting is better than Sony's Super Bitmapping technology and
WaveLab's (Type3 curve).  He also comments that:

"[The] Cool Edit 44.1 KHz curve is a more adequate approximation of the human hearing threshold. Nevertheless the curve is far from optimal (too low filter order again?). It could be much closer to optimal if the spectral pit at 21 kHz – 22 kHz is filled with frequencies from around 12 kHz."

From:
http://audio.rightmark.org/lukin/dither/dither.htm

What do you guys think are the best settings to use?
Anyone fooled around with this stuff in cool edit?

RD

Best Dither Settings In Cool Edit

Reply #1 –
Triangular pdf is considered the best.

As to bit depth, it's better to use as lower bit depth as possible, because the lower the depth, the lower the added dither noise level.

With triangular pdf and no noise shaping, you need to use at least 0.7 bits of dither to properly dither a signal.

Using noise shaping, you can use lower bit depths and properly dither the signal. Using highly noise shaped dither (44.1 KHz and 48 KHz curves for example), just 0.3 bits are enough to properly dither a signal.

The noise shaping curves only differ in the audibility of the added dither noise. With no noise shaping, the dither noise has same level at all frequencies. With proper noise shaping, you move the "higher" levels of the noise up in the frequency spectrum, where the human ear is less sensitive, leaving very low noise levels where the ear is more sensitive.

In my opinion, even with the most basic dither, which is dither with no noise shaping at all, it's impossible to hear the dither added noise, under real world listening conditions.

So, practically, there's not much difference between 44.1 or 48 KHz noise shaping, or no noise shaping at all, as long as the miminum bit depth required to properly avoid digital distortion is used.


Edit:

After playing further with very low-level signals (-90 dB) , it seems that to totally avoid digital distortion with this type of signals, you need at least 0.9 bits with no noise shaping dither, 0.3 bits with 44.1 KHz noise shaping profile, and 0.6 bits with the 48 KHz noise shaping profile. So, it seems that the 44.1 Khz profile with 0.3 bits dither would be the less objectively/subjectively "noisy" of all them, taking into account that what I said at the previous paragraphs about real world audibility still applies.

2nd Edit:

As 2Bdecided says, 0.3 dB is not enough with the 44.1 KHz profile, because you get noise modulation (similar to noise 'pumping'). You need at least 0.5 dB of dither to avoid this noise modulation effect with this profile. The other values remain OK.

Best Dither Settings In Cool Edit

Reply #2 –
Dear KikeG,

Thanks, so in sum you recommend trying:
trinagular pdf
0.3 bits
44.1 khz curve

Is that correct?

Anybody else have some ideas?

Best Dither Settings In Cool Edit

Reply #3 –
Test it yourself. I'm at work and don't have CE here but I've done this at home. Record a low-level log amplitude decay  in 32-bit. I did this in 1.2a and I seem to remember that the max amplitude boost was 80dB. (Haven't tried it in CE 2.) Fade a sequence of a  1KHz sine wave, a 8KHz sine wave and maybe a heavily compressed music sample (so there is little dynamic range variation - modern pop music will conveniently provide a sample which may not require further compression) from say -80dB to -120db, if it will let you, or from -0dB to -40dB, and then reduce the amplitude by -80db. Save this as your test fadeout file.

Now experiment with dithering it down to 16-bit and either turn your volume way up and listen through headphones to the fadeout or increase the amplitude by +80dB.

You can also do things like look at the freq analysis graph of of a low-level -80db or so 1Khz sine wave thats been downsampled.  I think I viewed this dynamically and looked for harmonic peaks i.e. distortion if it's not dithered sufficiently.

Best Dither Settings In Cool Edit

Reply #4 –
I edited my previous post, please take a look at it.

Quote
Thanks, so in sum you recommend trying:
trinagular pdf
0.3 bits
44.1 khz curve

Is that correct?


Yes, that seems to be the best noise shaping profile in CEP. EDIT: better use 0.5 bits of amplitude, see edit of my previous post.

You can do your own tests (that's what I did), as dosdan suggests. Generate a 5 sec. 32 bit 1 KHz sine signal with any amplitude, normalize it to -90 dB, and convert it to 16 bits using dither. Analyze the result using the Frequency Analysis tool, using 66535 FFT points and averaging (Scan button) the whole signal, and see the results. If some isolated peaks different from the 1 KHz signal appear at the spectrum, that means that the signal is insufficiently dithered (digital distortion) . Also, you can listen at the result, boosting up the volume of your card/amp, and see if you hear any difference.

Best Dither Settings In Cool Edit

Reply #5 –
Thanks to everyone for their help.

Of course I should do my own tests... but its nice to have some recommended values to test.  Thanks for the input....

Best Dither Settings In Cool Edit

Reply #6 –
32-bit, 44.1kHz. Generate a 0.5Hz sine wave (2 seconds duration) at -80dB. Reduce it by 8dB. Now use the settings suggested above: 44.1kHz noise shaping, 0.3 bits tri dither. Look at the result. The dither cuts out for certain signal amplitudes. (Well, it does on my PC - YMMV!)

I think this is why I suggested 0.5bits originally - 0.3 doesn't seem to be enough.

Let me know how you get on - I'm using CEP 1.2a.

Cheers,
David.
http://www.David.Robinson.org/

P.S. - FWIW I typically use 1-bit Tri no noise shaping - as KikeG said, it's difficult (impossible?) to hear the difference at reasonable listening levels in most listening environments, and noise shaped dither could cause problems if the signal was ever sampled and pitch shifted down. Also, it doesn't help psychoacoustic (e.g. mp3) encoding.

Best Dither Settings In Cool Edit

Reply #7 –
Quote
Let me know how you get on - I'm using CEP 1.2a.

In my computer, using CEP 1.2a, seems to be little or no difference compared to using 0.3 bits. However, practically there's almost no difference in noise level using 0.3 bits or 0.5 bits, so doesn't matter much.

Quote
and noise shaped dither ...  doesn't help psychoacoustic (e.g. mp3) encoding.


For test signals I prefer to use no noise shaping. For other things, I instinctively prefer to use no noise shaping too, but the reasons you give are in fact good ones not to use it on a regular basis.

Best Dither Settings In Cool Edit

Reply #8 –
Quote
In my computer, using CEP 1.2a, seems to be little or no difference compared to using 0.3 bits.

2Bdecided, I've done some more testing, and you're right, with just 0.3 bits of dither you get noise modulation ('pumping'). As you said, you need at least 0.5 bits of dither with this profile to avoid this noise modulation.

I've added this to my first post.

Best Dither Settings In Cool Edit

Reply #9 –
David,

I'm so glad you found this thread... that is what
I was hoping for...
When I get a chance (I'm at work now) I have to generate
some cool edit sine waves and test, etc...

So far based on your testing and Alexey Lukin's this seems to
be both the safest and best tradeoff:

trinagular pdf
0.5 bits
44.1 khz curve

I value KikeG's suggestions too, but have to do some tests at 0.3 bits...

Well I guess i should stop dithering around (no pun intended).

RD