12

I just came across this excellent article on cassette tape storage formats.

One thing confuses me though... in the "all digital" examples, like Figure 1C, what exactly ends up on the tape? That is, if I were to simply play the tape, what would I hear?

This article also poses the question as to why everyone didn't just use CUTS. It seems like the most resistant to noise, operates at higher speeds (including 2400bps in later versions) and can be played over a telephone! Yet following machines almost always used their own format, notably Commodore and Atari.

  • 1
    Please narrow this question down to ask one question; it's currently asking two things. You can move the final paragraph to a separate question if you like. Then the title can be edited to fit the question, and perhaps you could flesh out the paragraph to not rely so much on having read the article to know what it means (what's Figure 1C?). – wizzwizz4 Dec 13 '18 at 17:19
  • 1
    You can't start using CUTS until somebody has defined it, and by that time you have lots of tapes in circulation for your machine using a different format. So at best you have to support two formats not one, and somehow decide which format any particular tape is written in. "Don't change anything" is a simpler option to implement! – alephzero Dec 13 '18 at 17:26
  • 1
    CUTS pre-dates the formats used by Commodore, Atari, TRS-80, etc. By several years in fact. It also pre-dates MSX, which did use CUTS. – Maury Markowitz Dec 13 '18 at 17:30
  • 1
    The Commodore tape format goes way back to the PET in '77. It works the same across the entire line of Commodore products that support a tape drive, which probably has a lot to do with how strange it is. – mnem Dec 13 '18 at 18:05
  • 1
    Have you heard of the PixelVision? – MicroMachine Dec 14 '18 at 0:50
14

The simple reason is that interoperability was not a primary drive for this kind of storage, especially at the consumer level. Honestly, what's the point of reading a Commodore cassette on an Atari for 99% of the use cases?

I look at the hardware on the sample diagram and, you know what? It's a lot.

Most interfaces are a few bits of analog components like op-amps and some diodes.

UARTs? That's money.

Cassette was "cheap and easy". Some worked great, some were fiddly. While it's nice that some groups were able to send programs over radios to some machines, that was hardly a primary design requirement for most implementations. Local, cheap storage was the driver. Reliable storage is a bonus. Fast storage is extra double plus good, but lower on the list.

  • 1
    Of implementarions, the Acorn machines each have some sort of complicated automatic shifting and accumulation with built-in start/stop bit generation and semantic interrupts for byte received, high tone detected, etc. It's a lot. But the MSX is just a current level query 1-bit ADC-type thing, exactly like a Spectrum or a CPC, with appropriate polling loops. That's as to the advocated hardware and its suitability. I agree with all your points about the lack of value of standardisation when both machines and software are diverse and incompatible. – Tommy Dec 13 '18 at 19:19
2

in the "all digital" examples, like Figure 1C, what exactly ends up on the tape?

Varying levels of magnetization? SCNR.

But in all seriousness, that's all a magnetic tape will carry.

That is, if I were to simply play the tape, what would I hear?

On what kind of device? Different kind of recording means different amplifiers and heads as well (*1). Not to mention transport speed and tape material. Just because the compact cassette mechanics are used (*2) does not make any other part exchangeable or even compatible.

Now, assuming that a standard head is used when recording on standard spaced tracks of a compact cassette compatible mechanic; playing such back will result in ... well ... noise :))

What noise? Now that again depends as much on the recording format and speed the tape (the tape itself) and the amplifiers used. As combinations are endless, it's only possible do calculate some samples. First approximation here is to assume tape speed is constant (*3) CC speed is 4.7625 cm/s (or 1.875 inch/s). Next, lets assume a digital write density of 1600 bit/inch (*4) this gives us 1600 bit cells per inch, or 3000 over the length played in a CC player over a second.

This results in a first asumption of a basic 1500 Hz sound with encoding A and more like 3000 with G and somewhere inbetween with C. After all, the signal is no diferent from a port connected to a speaker and flipped the same way. Now, flipping that fast will result in reception of more than just a single frequency, as, depending on amplifier and speaker this will result in a PWM like aproximation (*5)

And at that point it's best to stop wild guessing, as the number of influences is, as described, rather high and whatever conclusion we come up with, it will be way off.

Now, beside all of the theoretical part here, there is a very simple way to listen to a digital recording on standard cassette: The TI 99/4 does already use a digital scheme. It's a Manchester encoding like H, at 1400 Bd. According to this (remarkable) site it results in an equivalent with roughly 690 and 1380 Hz.

This article also poses the question as to why everyone didn't just use CUTS.

Because they already had different solutions, often FSK as well? After all, CUTS itself was a company standard - in this case the one used by Processor Technology for the SOL as well as the Subsystem-B board.

Keep in mind, that article is a report about the Byte Magazin initiated meeting in Kansas City in 1976 (see the original Byte article), a competing magazine. Using the CUTS name was a way around that dilemma (see also this answer).

It seems like the most resistant to noise, operates at higher speeds (including 2400bps in later versions) and can be played over a telephone!

No it isn't. Using this format does not pose any recordinge advantage over many other FSK combinations. For all of them its all about selecting the right frequencies that need to be sufficient distinguishable as well as - and that's even more important, within the frequency range of the storage/transport media. Preferably frequencies where the media has it's highest sensitivity (*6)

Within this range any frequency pair can be used - as long as the decoder can seperate them. So 1500 and 2000 could have been used as well, even allowing more density.

Compact Cassettes had for example a usable range form ~20 Hz to 12 kHz (*7). Depending on the amplifiers and filters the usable range for FSK recording is somewhere between 0,2 and 8 kHz - ofc depending on the tape material used - after that it'll slip fast (Signal level shrinking while noise increasing). So for a computer recording a pair like 5 kHz and 7 would have made a better recording.

Ofc this would require more sophisticated software - or hardware filters.

For telephone connection the standard frequency range is 300 to 3000. Usability of CUTS for this is therefore just a nice side effect.

And so on. Bottom line: CUTS as defined has no specific advantage over any other FSK system when looking at the recording itself. It does have advantages due it's simple structure of the higher frequency being exactly double the lower and being slow enough even for under powered micros.

Yet following machines almost always used their own format, notably Commodore and Atari.

Add Apple and TI to this list. Still, many later on did follow KSC or an improved KSC (like Acorn). Somehow thats a feature more often found in non US computers (*8).

  • Commodore also inherited the format from MOS and compatibility over the various machines was more important (*9)

  • Atari in turn used 3995/5327 Hz for a way better security (detection rate) than KSC would support. Enabled as well due a filter for detection

  • TI's interface is already digital.

  • The Apple II uses also two frequencies, but for each only a single cycle is recorded. Thus somewhat between a digital and FSK code. A zero bit is half the size of a one, resulting in a variable length recording, with a speed between 1 KiBit when only ones are recorded and 2 KiBit for zeroes. Again compatibility as well as higher bitrate are reasons here.

Obvious, there are good reasons to not follow KSC - and foregiving compatibility in turn.


*1 - Ofc, head selection depends as well on the price range targeted.

*2 - To keep cost low.

*3 - Compact cassette tape transport speed varies with being slower at the begining of a tape and increasing over the length - all due constant angular speed while the radius of the pulling ...err... pully increases.

*4 - This might sound high for home computers, but it is rather on the lower end what tapes could do.

*5 - Real world heads, amplifiers and more so speakers will work on integrating the signal into a continuous signal, which then again is turned into discrete frequencies in our ear to be synthesised again .... many ways to slip of in prediction. (This BTW reminds me of a real nice and quite short program for the Apple II to play two sounds at once by calculating an envelope and realize it as PWM :)

*6 - That's for example why Ferrochrome (Typ III) tapes often where problematic with home computers. While being great in recording very low (Iron) and very high (Chrome) frequencies, the over all frequency run was quite different to other, more centered types - which in turn needs different amplifiers. Recorded/played on drives not made acordingly resulted in a distorted recording.

*7 - the upper frequency is defined head gap. Standard is good for 12 KHz. High quality decks use a smaller gap allowing up to 16 kHz recording.

*8 - Looks like the usual case of US style 'Not invented here' vs. European tendency to improve cooperation by following standards :))

*9 - Commodore took care from the begining to make BASIC programs transferable. Check this answer.

  • Concerning your note 3, tape transports try to maintain a constant speed across the heads by driving the tape with a constant-speed capstan and pinch roller. The take-up reel speed obviously varies and is handled either with a slipping clutch or a servo motor and a torque or tension feedback loop. The supply reel in a cassette system uses either a slipping brake or another servo motor to maintain the tape tension across the heads. – grahamj42 Dec 14 '18 at 6:34
  • Just getting around to this post. Raff, can you explain the Atari issue? Did those frequencies record/playback more accurately? And why those numbers? They seem odd, is this a harmonic or something? – Maury Markowitz Dec 21 '18 at 21:06
  • Yes, harmonics are one part. With a 3:4 relation, the first harmonic is at almost 16 kHz, thus way out of standard recording and at the upper limit for HiFi. These are hard limits due the head gap, thus any recorder will have sufficient filters (even more so the 410, but that's a different issue). Almost more important the frequencies are close to the middle of the range, an area most tapes should cover quite well. Next, the relation between recording rate (600 Bd) is roughly 1:6.66 and 1:8.88 thus at least 6 or 8 full waves are recorded per bit leaving much room for secure detection. – Raffzahn Dec 21 '18 at 23:34
  • Again a high ratio. There are many more mechanics Atari added to the 410 to make it one of the most reliable systems like filters, limiters and amplifiers. As a result the system (including the software) is already good to cope with variations of -40/+130% in tape speed. A quality mostly based on the frequencies. The Atari recording is (as well as the TI99s), for sure worth a seperate question(s) to get into details - just, @MauryMarkowitz don't expect me to give a quick answer over the next few days :)) – Raffzahn Dec 21 '18 at 23:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.