Comment Re:Using real world audio waveforms? (Score 1) 50
Oh! And 'linear' was completely wrong. I don't know how I braino-ed that in there, the [at very least perceived] distortion/noise tradeoff is not linear.
Oh! And 'linear' was completely wrong. I don't know how I braino-ed that in there, the [at very least perceived] distortion/noise tradeoff is not linear.
Right, and this is why dither is only applied to 'last-mile' audio intended to be consumed. Dither 'screws' you in other ways if you intend to use that audio in production, such as losing all the property of removing the distortion, yet still having the additive noise. But we're still talking about changes 100+dB down.
>Counter nitpick: Monty, as a professional motion picture sound designer, I cannot tell you how distracting it is to hear your voice constantly changing its pan across the stereo field
The audio was recorded with a stereo pair. It wasn't panned artificially
I think it's more "I want people to know why I do the stupid things I do." Latent fear of being committed.
As a nitpick, you get dithering losses _or_ quantization distortion, or a linear tradeoff between the two. You don't get the worst case of both on top of each other unless you screw up.
Without dither, worst case, all your 16 bit quant distortion products will be under -100dB regardless of input amplitude. I actually display the worst case in the video to make it easy to see. Quantization distortion aliases, and I chose an integer sample period so the aliased distortion would always land in the same bins after folding. If I hadn't, it would have spread out more and been even lower. If I had chosen a relatively prime frequency, the quantization distortion would have spread out across all bins equally.
> Stop using cdparanoia - it isn't very good, at all. It tests poorly, we're sad to say.
Really! As the author, I'd love to hear hard specifics. or maybe a bug report.
> You want to use Secure Mode with NO C2, accurate stream, disable cache.
You can't disable the cache on a SATA/PATA ATAPI drive. The whole point of cdparanoia's extensive cache analysis is to figure out a way to defeat the cache because it can't be turned off. There is no FUA bit for optical drives in ATA or MMC.
The 'accurate stream' bit is similarly useless (every manufacturer interprets it differently) and C2 information is similarly untrustworthy.
Plextors are not recommended for error free or fast ripping. They try to implement their own paranoia-like retry algorithm in firmware and do a rather bad job about it. They also lie about error correcting information (you do not get raw data, you get what the drive thinks it has successfully reconstructed). Plextors often look OK on pristine disks, but if you hit a bit error (like on just about any burned disk), you don't know what it's going to do. Plextors are, overall, among the more troublesome drives _unless_ you're using a ripper that does no retry checking (ie, NOT cdparanoia and NOT EAC). If you use iTunes, you want a Plextor. Otherwise, avoid them.
That was true ~15 years ago. Since then, Plextor's firmware gets along very badly with the rippers that try to be frame accurate, because Plextor tries to implement a much lighter-weight more error prone version of the same algo on the drive. The drive still doesn't do a realiable job, and it seriously mucks up the ripper.
Regarding hte first point, that 120dB broadband noise figure is giving you at least 140dB of SFDR, probably much better, and the depth of any critical band is going to be even better yet. Even 16 bit data with a decent noise shaper is going to be 120dB deep in the 2-4kHz critical bands. (all of which doesn't disagree with anything you said of course)
Oh! I remember this one
>I guess my point is that it's too easy to make an error when seeing an "interesting idea and no data" and dismissing it.
I agree with you completely. Interesting ideas should be published; no paper is born in the state of being independently verified. I object to those who take these papers as evidence to support a position when no such validation has taken place. Thinking aloud is useful, but thinking aloud != hard data.
I'm actually giggling that 20 bits of real dept is 'disappointing' in any sense of the word. It's bloody fantastic!
(I hadn't meant to trash the ESS chip in any sense. I've not used it, but it looks impressive)
Well, for the record, I've not been rejected, but I've only published within AES once.
It's not an attack, it's more a statement of truth. The AES publishes all sorts of things. Papers with interesting ideas and no data (eg, the J. Dunn 'equiripple filters cause preecho' paper, which presents a fascinating insight, even if it doesn't work out in practice), papers with data that are effectively WTFLOL (the famous Oohashi MRI paper) and papers that are more careful controlled studies. It runs the whole gamut on both sides, just as I said.
Do you deny that a substantial portion of the membership, including many elders of the group, are not 'bigger numbers are always audibly better' audiophiles? It was Andy Moorer himself who, with no hard data, kicked off the insane sampling rate race that now has some hardcore audiophiles wondering if 192kHz is enough.. they're holding out for 384kHz!
Is the AES a worthless cesspool? Oh heck no. Never said that. But treating its publications as more than a good industry rag (where it's sometimes hard to tell the research from the advertisements).. or perhaps an advanced debating club... is probably not a very good idea. Treating any one AES paper as gospel is just insane.
Oh, what do you know. In that case, go forth and unsubscribe
As for me, it was 'membership'. Peer review means little when a substantial portion of the peers don't believe in the scientific method.
Dynamic range (not SNR) is 22.5 bits ("How quiet the chip can get when it's doing nothing/off"). SNR is 20 bits at best given the 120dB THD+N spec.
Surely you mean membership...?
ESS's own specs list it at THD+N of 20 bits, and SNR can't be higher than that.
This is also an 'ultimate' DAC, supposedly the Terminator of the current generation (in ESS's own marketing anyway)
Parent post is talking about broadband SNR, not narrowband depth or SFDR. Very few ADCs/DACs reach 20 bits SNR. Almost all exceed 20 bits SFDR. Noise shaping has nothing to do with it; it actually penalizes SNR even though it is generally a [great] benefit to perceived fidelity.
"I am, therefore I am." -- Akira