Help with best Recording Practices, is AK4524 really a 24-Bit 96kHz MD Audio codec?

[louciphre]

Long time lurker here.  I'm a DJ and I use MDs to spin with because I like them, I don't get finger prints on their surface, and they're cool. No one ever knows what they are at my gigs. But MDs do take some effort.

I often find all my MD answers by reviewing posts. So this is my first posted question. I have sifted through much conflicting info here and at other places, I would like to know your opinion given all the current tech and years of knowledge with all the available MD decks.... (excluding Hi-MD), what procedure would render the best sound quality onto a Minidisc? What is the current best practices, techniques, software, and gear to get the upper most sound/bit rate out of MD?

My current practice is to use the highest quality sound possible (96/24 or higher) to analog-in on my JA555ES using SP. I am aware codecs will down sample HD sound to 44/24 or lower depending on the method and deck. I considered using a digital source using software like sound forge but haven't tried that yet.

 

I did some research and found the MDS-E10 has the AK4524 which is a 24-Bit 96kHz Audio codec. Reference info link is here....
https://people.freebsd.org/~lofi/4524.pdf

The information listed mentioned the chip and codec processes 96/24 A/D and D/A. 44/24 was what I believed was the upper recording limit from a digital source from select MD decks. Perhaps it down-samples?

Anyway thank you for your insights.

Get some!!!

[sfbp]

All of the decks (just about) use the AK4524 or AK4584 (so you don't need the E10 or JA555ES, a lowly JE440 uses one).  But a :codec: is not a piece of h/w.

 

I don't believe it's running at 96Khz, but the real "chippy" guys here can confirm that. There seem to be two reasons higher rate sampling components might be used:

1. Upsampling can clean up waveforms. Someone should chime in about dithering which is the same process that must happen on downsampling, which may be relevant.
2. High speed copy sometimes depends on doing everything at double the normal frequency. The MDS-W1 actually doesn't use AK4524 but some other chip, though, and in any event its copy takes place in digital mode, no A/D required.

I'm no waveform expert, but there are waveforms shown in every MD service manual.

 

Bottom line - nearly all the Sony gear (I don't have knowledge about other manufacturers but Sony did invent MD) uses very good digital circuitry, and the high-end machines are sometimes more about marketing something, that is perceived to be better, by putting a higher price tag on it, than any real differences. Better capacitors, sure.

 

The 24-bit thingy is part of ATRAC's definition - it's a logarithmic encoding of the digital signal with a 24-bit mantissa, so you can get better than :pure: 16 bits in a lot less data bytes. A CD supposedly contains only 16 bits per PCM data byte - but that has been oversampled and dithered (smoothed, in effect) before writing to the physical CD. Quite how CD playback gets oversampled data (they always claim 20 bits) played back to give better fidelity is a mystery to me, but again someone else lurking can probably give us a tutorial. Exactly how CDs get ripped is a related question - there are choices to be made at rip time.

 

And ATRAC is a winning encoding. It lost out to MP3 almost 25 years ago due to (IMHO) a flawed testing regime that didn't actually use the data conversion paths that happen in real MD (and later) equipment. Yes, SP is probably better than straight PCM, but the ATRAC3 and ATRAC3+ codecs are probably more accurate than SP. But we are always using ATRAC3 (and ATRAC3+) at lower data rates than SP, so it's like comparing apples to oranges.

 

There, I've shown you a simplistic argument the real techies could drive a truck through, but differences between implementation of digital codecs are relatively unimportant, from my limited understanding and my limited experience. The crucial thing is that they are digital. That's how people (eg the police) can recover details from bad images, for example, so-called "computer enhancement".

[kgallen]

I think @sfbp has pretty much nailed it above.

 

I'll take a look at some schematics and datasheets to see what I can learn... In the E10, the AK4524 is clocked at 45.158MHz. The DFS value is set by firmware so I've no idea what sampling rate (44.1/48/96) is configured. However 45.158MHz/2^10=44.1kHz and that is supported by Table 3 in the AK4524 spec. 48kHz and 96kHz would not be achievable with that input crystal frequency.

 

The common view seems to be that with decks (particularly those with "better spec" analogue sections) and the best source, then you aren't going to go far wrong recording via that route.

 

Using a digital input should in theory give a "lower noise" source, but from CD you will be "limited" to 16-bit samples, not that should matter. Are there other digital sources that can give 20 bit samples on SPDIF/TOSLINK? (Those standards will support up to 24-bit).

 

96kHz sampling - yea, maybe that is relevant in the studio with a complex audio chain and mixing stages, but for 1:1 copying, nah, 48kHz or 44.1kHz is going to be adequate.

 

Not that I'm any sort of reference expert in this field.