Nuevo formato universal para preservación de cassettes

[ICEknight]

¡Buenas a todos!

Venía a comentaros que actualmente en el Discord de preservación se está estudiando la posibilidad de crear un nuevo formato para la correcta preservación de las cintas de cassette (tanto de música como de juegos), digitalizándolas de la forma más perfecta posible para preservarlas tal cual, con cada una de sus pistas (cuatro para stereo o dos para mono) al volumen en que venían, sin interferencias ni variaciones en la velocidad y sin "mejoras" de ningún tipo.

Para ello estamos considerando utilizar un grabador con cabezal de 8 pistas, ya que permitiría realizar un par de lecturas simultáneas de cada pista en el caso de ser una cinta stereo, o cuatro de cada una en cintas mono. El software utilizado, en teoría permitiría detectar y eliminar imperfecciones en la superficie de la cinta y combinar las lecturas realizadas para obtener una especie de "master" de mayor calidad, que es lo que ya se está haciendo con los volcados de señal analógica de los LaserDiscs en el proyecto Domesday Duplicator. El control de la velocidad de la cinta se realizaría con la instalación de ruedas de medición en el reproductor pero, antes de comenzar a hacer modificaciones, debemos tener claro cómo conseguir leer la señal raw sin interferencias de ningún tipo.

Teníamos pensado adquirir un Tascam Portastudio 424/488/688 o Yamaha MT8X, a cuyo cabezal de 8 pistas poder enchufar directamente un Domesday Duplicator para así evitar interferencias, aunque sabemos que éste último posíblemente necesitará alguna modificación para leer la señal de audio correctamente. También hemos visto que el Tascam 238 Syncaset tiene salidas de señal analógica para cada una de las pistas, pero desconocemos si con ellas se pierde algo de fidelidad, en comparación con obtenerla del cabezal directamente (y si no hay ninguna ventaja, sería tontería tener que pagar de más).

¿Habría alguien por aquí con buenos conocimientos técnicos y/o acceso a grabadores de 8 pistas que estuviese interesado en colaborar en este proyecto? De primeras, necesitaríamos acceso a algunos de los grabadores mencionados para poder tomar muestras de la señal con un osciloscopio (disponemos de uno en Madrid) y enviárselas al creador del Duplicator para que nos pueda comentar las modificaciones que habría que hacerle (y también para comparar unos con otros). Y aparte de esto, cualquier sugerencia u observación acerca de algo que se nos pueda estar escapando será bienvenida.

(Entre los miembros del grupo interesados en sacar este formato adelante se encuentra el fundador de la Game Preservation Society en Japón, asi que la cosa va en serio)

P.D.: Para los interesados, también se está trabajando actualmente en la preservación de VHS a través del Domesday Duplicator: https://github.com/happycube/ld-decode/issues/16

[dlfrsilver]

Hello, beware, for the Amstrad CPC, the tapes were all recorded in mono. None of them were recorded or mastered in Stereo.

What matters is the recording quality. I use 64000hz minimum. The best quality commercially recorded tapes on Amstrad CPC/ZX and others possibly were the tapes from US GOLD like INDY 4, Strider II, Final Fight, etc. They used 96khz mastering equipment.

Other than that, the tape content need to be filtered......

[ICEknight]

Wew, this is quite old and some things have been discarded and found out since then.

What matters is the recording quality. I use 64000hz minimum. The best quality commercially recorded tapes on Amstrad CPC/ZX and others possibly were the tapes from US GOLD like INDY 4, Strider II, Final Fight, etc. They used 96khz mastering equipment.

Other than that, the tape content need to be filtered......

Nowadays we actually know that the official academic standards for cassette preservation are Stereo, 24 bit, 96000Hz, with the very minimum being Stereo, 24 bit, 48000Hz (so your 64000hz might be a fine mid-point): https://web.archive.org/web/20240130185 ... _bp_07.pdf

IASA-TC 04 recommends encoding to linear pulse-code modulation (PCM) with a minimum sample rate of 48 kHz, and for many purposes suggests transferring at 24 bit with a 96 kHz sampling rate. In fact, 24/96 has become the standard choice for audio preservation reformatting.

for the Amstrad CPC, the tapes were all recorded in mono. None of them were recorded or mastered in Stereo.

I'll have to check again with some non-Erbe Amstrad cassettes but I really don't think that statement is true.

A few months ago I've seen with my own eyes that many Spectrum tapes were indeed recorded with two tracks (Stereo) and many with just one (Mono). Also, there's some of them with two tracks but one of them (either Left or Right) being seemingly empty... which could be considered Stereo, too.

So some cassettes are fine to preserve as Mono but, for others, that would technically preserve 50% of their contents. Since not everyone has the means to check the number of tracks, it's better to just always dump in Stereo and then, if the tape does happen to have just one per side, just choose the one in best shape and leave the other (or keep it as a separate dump).

[dlfrsilver]

Wew, this is quite old and some things have been discarded and found out since then.

Some things are completely out of bound and totally useless, i will explain that further in my answer.

Nowadays we actually know that the official academic standards for cassette preservation are Stereo, 24 bit, 96000Hz, with the very minimum being Stereo, 24 bit, 48000Hz (so your 64000hz might be a fine mid-point): https://web.archive.org/web/20240130185 ... _bp_07.pdf

IASA-TC 04 recommends encoding to linear pulse-code modulation (PCM) with a minimum sample rate of 48 kHz, and for many purposes suggests transferring at 24 bit with a 96 kHz sampling rate. In fact, 24/96 has become the standard choice for audio preservation reformatting.

This is completely useless. If i push a single good reason, it would be this : I have numerised more than 2800 commercial tapes on
the Amstrad CPC 'ALONE', i am therefore an expert in this field, and i can say that throughout all the tapes i have dumped and preserved, and the quality of the tapes the follow statements :

1) We have only access to copies, not master tape or digital DAT.

2) Exactly like the floppies, if you raise the bar too high when recording the tape contents, the more higher you go with the sound rate (ex: 96khz), the more you bring 'dirt', 'noise', 'crap' with the actual bits of data, and that's something you absolutely don't want.

The only key here is to set the hardware for recording at the correct sound level with 44khz (which is more than enough, as 96Khz masters only appeared in the 90's. Before it was at best 44khz.

3) There are absolutely NO stereo tapes on the Amstrad CPC, for a simple reason : the Amstrad CPC use a tape drive that is only doing 'MONO'.
Recording Amstrad CPC tapes or even ZX spectrum (prove me wrong!) in Stereo, 24bits, 96Khz is just bloating the file recorded for nothing.
It doesn't bring anything like 'accuracy' or 'better sound'. The accuracy and the better sound cannot be achieved with this method.

For accuracy, precision in the sound with the least possible noise/crap/dirt, the recording hardware has to be set, and also the recording software, in order to get the best possible sound in 64Khz, 8bits, mono (sorry 16 or 24 bits sound did not exist in this era).

4) Regarding the WAV files bloated in "Stereo/24bits/96Khz", you get the exact same problem with the SCP format for the floppies : the SCP format for floppies is doing sursampling, and this leads to hidden errors in the flux, because the hardware at the time used for duplicated never went so higher. I have processed enough SCP files back to Kryoflux and also directly in IPF format to see that many SCP dumps out there are BAD, and people can't and did not even saw it !

5) "Sursampling/dumping" an old tape from the 80's or early 90's won't solve for free your noise/dirt/crap problems.

believing the opposite show that those thinking this have no clue or don't understand what dumping for preservation implies, or the whereabouts.

When using a software coupled with some hardware to dump tapes or floppies, people need to keep in mind that you're not getting datas the same way as when you read a floppy disk for Amstrad CPC by sector reading or a tape by read and then saving files on a disc or on another tape.

You get a full picture/photo of the disc content, this means all the flux constituting when gathered the datas (making then tracks or sectors), but also all the noise, shit, crap present on the disc.

When dumping without checking a disc, you ended with either blatant errors, Or hidden errors when you convert the SCP file to something else.

For the tapes, the same phenomenon appears : if your tape has a dead buffer (the small piece the band is sliding on), you will record crap/errors if you don't replace it, add to this your "let's go Stereo 24 bits 96Khz", and the end result will be just one simple thing : SHIT

The rule is the following :

1) set correctly your dumping hardware
2) set correctly your recording software
3) 8 bits computers like Amstrad CPC, Zx Spectrum, C64 are all using 'mono' recorded tapes. So go for Mono, 8bits, 64khz
4) Ensure the tape has no mould / Ensure that the Buffer on which the band slides is good (most of the time it's not, replace it by a new one)
5) With all those elements taken in account, you will be able to get a properly preserved tape content with a little bit of noise, that a real computer will read with no errors, and that a proper filtering/encoding software will be able to process in a clean file (ex: for the Amstrad CPC, the CDT file format, and then CDP file format, included in CSW2CDT).

I'll have to check again with some non-Erbe Amstrad cassettes but I really don't think that statement is true.

This is more than a statement, it's the conclusion of preserving +2800 commercial tapes for Amstrad CPC, including the ones with the hardest protection scheme.

My experience allow me to say that. I have 16 years old of it behind my back.

A few months ago I've seen with my own eyes that many Spectrum tapes were indeed recorded with two tracks (Stereo) and many with just one (Mono). Also, there's some of them with two tracks but one of them (either Left or Right) being seemingly empty... which could be considered Stereo, too.

Nope, you're completely wrong. all those tapes were mastered in mono at the duplication factory. I stand my case on that. You're seeing a second track because you asked your recording software to record in Stereo instead of mono, and then second hand it depend on the recording hardware used.

But no, most tapes were mastered in mono and duplicated in mono. You just want to bloat your files with crap, and even worse, the noise with mix with the actual data bits from the tapes = unsuitable result for preservation !

So some cassettes are fine to preserve as Mono but, for others, that would technically preserve 50% of their contents. Since not everyone has the means to check the number of tracks, it's better to just always dump in Stereo and then, if the tape does happen to have just one per side, just choose the one in best shape and leave the other (or keep it as a separate dump).

What you do is useless, you're bloating the file size for nothing, you take the risk of mixing noise with data, and the method used is completely wrong.

Preservation requires a more clever approach, you can't do preservation by just cramp up the sampling rate, use stereo and raise the amount of bits.
You need to get the data properly without all the 'traps for beginners' i mentionned above.

Lastly, i spent many year with CNGSOFT to make CSW2CDT, the best encoding tool currently existing for Amstrad CPC, ZX Spectrum and MSX, so you can mark my words !

[ICEknight]

Sorry to repeat myself but you are indeed wrong. It was the assumption of everyone in the game preservation communities that all 8 bit computer tapes were recorded in mono because they're read in mono. Thing is, nobody seemed to have actually checked how many tracks those tapes had, so I had to do it myself and...

I found out that there's at least three types of cassette recordings in our 8 bit computers:

• Monaural:
  
  Examples:
  - [Amstrad CPC] Master of the Lamps (Proeinsa)
  - [ZX Spectrum] Forbidden Planet
  - [ZX Spectrum] Basil the Great Mouse Detective (Erbe, Serie Leyenda)
  - [multiplatform] Whopper Chase
• Stereo but with only one track (left or right) recorded (which track, depends on the game):
  
  
  Examples:
  - [Amstrad CPC] Freddy Hardest
  - [ZX Spectrum] Deep Zone
  - [ZX Spectrum] La Aventura Espacial
  - [Commodore 64] Baloncesto
• Stereo with both tracks recorded:
  
  Example:
  - [ZX Spectrum] Lone Wolf - The mirror of death (System 4)

With that said:

• Does preserving all the tracks make a difference when reading from the 8 bit computer decks? No, since those read in mono.
• Does it make a difference when playing the Stereo file on a computer? Sometimes. Certain games will play only on the left or right channel. Others may have random stuff officially recorded afterwards or between blocks, which may or may not be in stereo. People nowadays don't know those things because they were never preserved like that.
• Does it make a difference to preserve the full audio instead of just preserving the game's code? Well, if the tape does have two tracks you'd want to ideally preserve those two tracks. If you don't want to, that's still good enough for the game to run and better than nothing.
• Does it "bloat" files? They sure are bigger than CDTs, but in a day and age when we're buying hard disks by the Terabyte, space is no longer an issue for preservation. Communities nowadays are archiving scans at 1200DPI 48 bits but, for public consumption, sure, regular web-sized JPGs are fine too.

But anyway, everything else you said you'd have to take to the guys at the Indiana and Harvard Universities mentioned here and those commissioned by the National Recording Preservation Board of the Library of Congress here, since they're the ones who set those standards, I was just relaying what their best practices for audio cassette preservation say and some of us are already trying to follow them when possible.

If anyone doesn't like what they say then, well... just don't follow them. Nobody's forcing anyone to do anything and I certainly don't intend to engage in any internet fights.

[dlfrsilver]

I stand on my experience, my discoveries. None of them, and you either, you never preserved as many tapes as i did.

Tapes were mastered in mono, because that's how the mastering machines were at the time. It's not a limitation due to the computers, that what the machines for duplication were.

Does it "bloat" files? They sure are bigger than CDTs, but in a day and age when we're buying hard disks by the Terabyte, space is no longer an issue for preservation. Communities nowadays are archiving scans at 1200DPI 48 bits but, for public consumption, sure, regular web-sized JPGs are fine too.

You just did not understood my point. My point is that it's useless and non-sense to dump a tape in "Stereo 24bits 96khz", because that's "sur-sampling", and you mecanically record tons of noise that mix then with the data bits from the tape.

Sursampling will hide errors, make the filtering of the garbage/noise/crap impossible, and finally make the encoding in a proper master format impossible.

I have some peeps who did that with CPC tapes. The signal of their WAVs was a complete catastrophy, and it was impossible to process them.

Before going the way of using Stereo 24bits 96khz, you have to make multiple settings to your recording hardware and software to get the best signal.
Best signal possible cannot be achieved when you raise the bar to Stereo 24bits 96khz.

BTW : scanning in 1200DPI 48bits is done in order to not only preserving a software paper parts, it's made to be able to reprint.

But anyway, everything else you said you'd have to take to the guys at the Indiana and Harvard Universities mentioned here and those commissioned by the National Recording Preservation Board of the Library of Congress here, since they're the ones who set those standards, I was just relaying what their best practices for audio cassette preservation say and some of us are already trying to follow them when possible.

We're not talking of audio cassettes, we're talking about Softwares stored on "Computer" cassettes. What eludes me completely is how they went to "stereo 24bits 96khz" for data software tapes without any care about the original material recorded quality, and the noise management.

Or the other possibility, is that you "passed by" some parts of their process. As i said, i have dumped and preserved tapes using extremely sensible and complex protection schemes, that no one before me had preserved or transfered. Those schemes cannot be preserved and extracted from tapes simply by using "stereo 24 bits 96khz". The noise level would be so high that a real CPC 464 for instance would fail loading that.

EDIT : i have read the PDF you gave the link for..... It's made for Audio mediums, not computer software tapes. For Software computer tapes, other methods must be used.