Toronto meet at philodox'

[Hopstretch]

120dB peak?

[HiWire]

Theoretically. Since the decibel system is logarithmic, SACD's 120 dB is a lot more headroom than CD's 96 dB.

If you take a look at Stereophile's review of the dCS Scarlatti system, you can see some real measurements from a state of the art SACD player. I'd like to see measurements from an EMM Labs TSD1 + DAC2 combo for comparison.

[grawk]

The point is that 96db is more than you'll ever need at the listening end of things. More headroom is helpful in recording, but if you have a greater than 96db swing in your finished product, you're gonna blow things up. Including eardrums.

[HiWire]

Agreed. Nobody really needs that kind of peak volume in their room. According to the traditional reference scale, a jet plane taking off is about 140 dB – nobody wants that, even Spinal Tap.

If you read Fremer's review of the Scarlatti, he states that there is an audible difference between the various upsampling methods and SACD. I'm simply waiting for this kind of SACD performance to become affordable – for now, I've invested in an upsampling Redbook player and I'm collecting hybrid SACDs to hedge my bets.

The price to get "perfect" digital is still too high for me, and I don't plan to invest in vinyl. Also, I'm not currently paying for any music downloads, although I'm slowly ripping my CDs to iTunes' AAC 320 kbps VBR – a compromised format, but only for convenient listening on my computers.

[Dusty Chalk]

The point is that 96db is more than you'll ever need at the listening end of things. More headroom is helpful in recording, but if you have a greater than 96db swing in your finished product, you're gonna blow things up. Including eardrums.

Well, if you're going to repeat yourself, I'm going to repeat myself, too -- you're talking about dynamic range, there's also resolution, which you're not addressing at all.

[grawk]

and as I've said before, more bits doesn't equal more resolution in the first 96 db, it just adds it to the end. The sample rate does make a difference, of course.

[HiWire]

The funny thing is that we ended the meet by listening to Jay's vinyl rig on his vintage Marantz amp. I think hi-fi is a fun (and expensive) hobby, but it should never get in the way of our enjoyment of music.

[Dusty Chalk]

and as I've said before, more bits doesn't equal more resolution in the first 96 db, it just adds it to the end. The sample rate does make a difference, of course.

I didn't understand that statement then, and I don't understand it now. It (the additional 8 bits of resolution) adds values between every pair of 16-bit values, therefore adding resolution at every point in the scale. "0db" and "96db" &c. is the overall amplitude of the envelope following the signal, not the actual amplitude of a signal. A full strength (even a simple sine wave) signal can utilize every value between 0 and 2^16.

[grawk]

because a -96db sine wave is represented with exactly the same value on a 96/24 and 96/16 representation. The same is true with a -80db sine wave and a -24db sine wave. A -97db sine wave isn't able to be represented by a 96/16 representation, but is on a 96/24.

The image you're showing represents sample rate. But nothing reproduces music with the stairsteps anyway, the sine wave is then interpolated and generated as an actual sine wave, as well.

[cetoole]

I didn't understand that statement then, and I don't understand it now. It (the additional 8 bits of resolution) adds values between every pair of 16-bit values, therefore adding resolution at every point in the scale. "0db" and "96db" &c. is the overall amplitude of the envelope following the signal, not the actual amplitude of a signal. A full strength (even a simple sine wave) signal can utilize every value between 0 and 2^16.

http://streaming.wisconsin.edu/images/audio_editing_SF_images/sf00sbit.gif

http://www.jiscdigitalmedia.ac.uk/images/ITDA-05-bitdepth.jpg

And what do you get after a proper reconstruction filter?

[Dusty Chalk]

I was using positive numbers, so my 96db == your 0db.

The first image shows the same sample rate with 2 different bit depths (differing by 1 bit). Look at the scales.

The second image shows a bit depth of 1.5 bits (it's a bad example, because they're using 1 bit for sign, which is wasted when the other value is zero) vs. a bit depth of ~5 (32=2^5).

It's the best I could come up with, with 5 seconds of googling.

because a -96db sine wave is represented with exactly the same value on a 96/24 and 96/16 representation.

Incorrect in two ways -- it's not represented with a single value, it's represented with multiple values; and #2 if you sampled at 24 bits, you'll get different more accurate values. Look, for example, at the second image again. It's exaggerated for effect. He didn't change the sample rate, it's just that you only have one value between the time slices, so it's not changing. It looks coarse because that's what you get at lower resolutions. The first time slice where the amplitude of the signal goes down below +0.5 is when the sampled version changes to zero.

I understand about the stairsteps, but I think the second image is a good example of how starting with more accurate data will produce more accurate results. At least, it does in my mind's eye.

[Dusty Chalk]

And what do you get after a proper reconstruction filter?

"Proper" -- good luck with that.

Yeah, you get two things that look a lot like the original signal -- especially with a sine wave -- but music is much more complex than that. You can hear the difference much more than you can see the difference.

[cetoole]

"Proper" -- good luck with that.

Yeah, you get two things that look a lot like the original signal -- especially with a sine wave -- but music is much more complex than that. You can hear the difference much more than you can see the difference.

Proper means fully removing the harmonics of the square wave. Take a 44.1khz input single, apply 8x OS, and then sufficient LPF to get the response down by 384khz, where the first image appears. You will get the same output as the original, in that 20-22.5khz 0-96dB range. Outside of this sure, high rez is unquestionably superior.

Oh, and dont think you can get away without a reconstruction filter on high rez either.

[swt61]

Are those cinnamon rolls I smell cooking?

[Dusty Chalk]

*set to the tune of the William Tell Overture* d-b-t, d-b-t, d-b-t-t-t, d-b-t, d-b-t, d-b-t-t

I don't know how much I can set up by Sunday, but I'll try so that Colin can at least hear what I'm talking about in a controlled environment. Don't know whether or not we'll feel up to a DBT -- I'd rather set it up with Grawk at the controls of downsampling the files from 24/96, so I'll see what I can do in terms of getting him the files. I can do it myself, but we're limited to whatever the Masterlink does, which I suspect is simple truncation.

Although the theory -- grawk's, my understanding -- should go that if the bits are below the level of human hearing, then dither shouldn't matter, but I'll save that card for later.

[grawk]

I don't have time to do much of anything in the near future, unfortunately. That truncation works suggests tho that there's not bits in the middle that dissappear when you go from 24 to 16