"Pono, the Neil Young-endorsed Kickstarter project, is drawing more and more pledges. Now past the $2 million mark (with an expected goal of $800K), this project aims to create a audiophile friendly FLAC player along with its ecosystem (and by that they mean their own music store and syncing application).
The device itself features 2 audio outputs, one 'specially designed for headphones' and the other 'specifically designed for listening on your home audio system'. The player is controlled by an LCD touchscreen, and its triangular 'Toblerone' shape makes it easy to hold it upright with one hand or to lay it flat on surfaces. The player, which has 64GB of internal memory, comes together with a 64GB microSD card.
The board and its components, as well as a 'pre-prototype' model, are pictured in the project's Kickstarter page.
I read that article back in 2012. It certainly makes some very good points, but skips over a few facts, such as:
1. Yes, a sampling rate of 192 kHz is way higher than what's necessary. No, it's not pointless in itself, particularly for devices capable of recording (from an analog source).
Should an analog signal source contain components (or noise or whatever) in the ultrasonic range, a low sampling rate will actually push that noise down into the audible range (yes, really; think about it). The solution is to record at a much higher sample rate, remove the ultrasonic components with a (steep) digital filter, and then downsample because... why? To save space?
2. Yes, a CD/DVD player playing back a high-frequency sine wave recorded using a sampling frequency of 44.1 kHz will produce something that looks remarkably like a sine wave. No, that doesn't mean the Nyquist theorem has anything to do with sound quality, or that 44.1 kHz is good enough for high-quality sound reproduction.
A 44.1 kHz sampling frequency will result in roughly 6 samples per cycle for an 8 kHz sine wave signal. That's not a lot, and as the article says, some people visualize a staircase effect and expect to hear distortion. Yet the output from a CD player will indeed look (and sound) pretty smooth.
Well, guess what, a sampled 8 kHz triangle wave will look like a smooth sine wave too. Try it with an oscilloscope and see, or just listen to what's coming from the CD player and tell me it sounds anything like the analog source.
This should come as no surprise to anyone. The actual shape of the waveform doesn't get magically embedded in the samples; the player just creates a (possibly quite inaccurate) approximation using aggressive oversampling. This works reasonably well, great even, for lower frequencies. For frequencies above 6-8 kHz, not so much. Beyond 12 kHz it's just terrible, but fortunately most music is mostly sine waves at those frequencies anyway, and our ears are rather forgiving at such high frequencies. (My 'scope has problems locking on to a reproduced 18-20 kHz signal, though, as there's massive phase distortion. Not sure how our ears could possibly ignore that.)
3. Yes, a dynamic range of 16 bit does provide (barely) enough headroom to cover everything from silence to the pain threshold of the human ear. No, it's not really good enough, and again, it doesn't work at all for recording from an analog source.
To avoid clipping when recording, you have to turn down input sensitivity to a point where you can be reasonably cartain anything you record will be well below the 0db point. This usually means sacrificing a lot of headroom. With a 16 bit dynamic range there's not really a lot to sacrifice without affecting sound quality.
So we can either record and mix in 24 bit and downmix to 16 bit for the final master, or we could just, you know, use 24 bit all the way since storage space really isn't an issue anymore.
As for 16 bits being "enough", that's only true if you overlook an obvious characteristic of the human ear: Its sensitivity is not linear, while the samples are. This could be a problem if the source material contains both very loud and very quiet (and noise-free) segments. In very quiet segments of some classical pieces, the quantization noise can be quite noticeable. It's not a huge problem by any means, but one that can be easily eliminated.
This is where the article gets a bit silly, as Monty proceeds to create a wav file containing a 1 kHz sine wave at -105 db, and then uses a spectral analyzer to "prove" that the signal is there. Well, a quick look at the diagram shows considerable distortion in the 500 Hz - 2 kHz range, and a massive noise component appearing out of nowhere at around 6-8 kHz. Yes, something is certainly there, including stuff that shouldn't be there at all.
Fortunately, he provides the source file for download. Load the .wav file into Audacity and amplify the signal and you'll see why the analysis looks weird: The signal looks [i]nothing[/i] like a 1 Khz sine wave, or any wave at all for that matter. The quantization noise is just massive. The amplified signal sounds like a 1 kHz tone being played over a poor cell phone connection, complete with artificial background noise.
So yes, 16 bit at 44.1 kHz isn't horrible by any means, but there's plenty of room for improvement. 24 bit/96 kHz would be great, and would probably push both sampling inaccuracies and quantization noise well out of the audible range, but if the master recording will be using 192 kHz to handle ultrasonic noise, why bother with downsampling?
But otherwise I certainly do agree with Monty: Before you start worrying about sample sizes and frequencies, buy better headphones and speakers, and replace those MP3s with lossless files.