Could Cassettes Have Been Faster

[toddtmw]

Okay. Atari Cassette decks were based on industry-standard hardware, but they were custom for Atari.

 

Would it have been possible to alter the deck to use a head that could access both the "front and the "back (Both the top half and the bottom half of the cassette tape) to double the transfer speed?

 

I realize this could have created all kinds of other problems, like if someone doesn't realize how it works, fills the "front" of the cassette, turns it over and then obliterates what they just wrote while filling the "back".

 

But aside from that, would it have been possible? Would it have made the cassette decks terribly more expensive? Could the Atari have been able to handle decoding both streams at the same time to where it would truly be twice as fast? 

 

Could speed have been doubled again if it supported stereo tapes?

 

Just thinking out loud...

 

-Todd

[zzip]

I would think doing "front and back" would require a special head that would make them more expensive to produce.

 

Stereo tracks..  maybe but it would have to deal with simultaneous data from two streams and would have to sequence them properly. 

 

One other thing that might have helped speed up transfers is longer blocks of data and/or shorter intervals between them on tape.

[+CharlieChaplin]

Well,

 

they could have simply changed the OS, so that tapes worked with 900, 1200 or even 2400 Baud, I think.

 

There are "software turbos" that do not require to change the data recorder at all, some of them simply save with a higher baudrate (800-1000 Baud, no special loader required, no additional hardware required), some save with a loader and even higher baudrate (900-1400 Baud, req. special loader, but no additional hardware required).

 

So, if the OS already had some of that code, loading at higher speed even on a standard (unmodified) recorder would not have been a big problem...

 

(Think that's also true for disk drives, they could have chosen 38k4 baud or 57k6 baud as standard data transfer rate, instead of 19k2 baud; but afaik, it was originally planned to load A8 disks via parallelbus, which they could not get working reliably for quite some time, so SIO was invented. )

 

[_The Doctor__]

Yes, they could have been faster, a change in the OS, a change in the cassette mech itself... no reason it couldn't have seek, find, any other manor of tape innovations of the day... indexing, backup, didn't someone half do it and call it a wafer drive? ADAM?

[TGB1718]

I always thought that if they had used the 2 heads, by using a switch to remove the Audio Track when it's not being used

and use the second head to provide a clock pulse, so you would have Data on one track and Clock on the other track, then

no need for two tone frequency, instead write the data at those speeds or higher, the clock pulse provides the sync, so even if

the tape gets stretched etc. with age, it would still work fine.

 

Just an idea :))

[Rybags]

A custom head might have added too much cost - though don't "auto reverse" decks have 2 pair of stereo heads?

 

That aside, a single track mates well to serial I/O.  Having 2 bits coming in at once would have meant Pokey would have needed 2 serial processing circuits, or just 2 Pokeys (the more likely solution).

2 Pokeys means  servicing 2 lots of IRQs for the data streams.  Fair enough, you'd expect the IRQs to occur fairly close together so you could probably alter the code to just check flags for the "other" input or output byte.

 

Ignoring extra tracks, the stock speed is sort of conservative and I did experiments in the day but don't think I got much better than about 20% improvement by increasing the bitrate.

By doing long blocks you can cut off a lot of time as well.

In theory, combine all 3, so an example 4-track tape at higher bitrate and bigger blocks would probably load about 6 times as fast.  So into the territory of competing with a disk drive.  But to do so you've probably increased the cost of the base computer by a fair chunk, and price was at a disadvantage from around 1980-1984 as it was.

[Mr Robot]

There used to be a camcorder that used standard audio cassettes. It recorded something like 10 mins of video on a c60 using the full width of the tape and a much faster tape speed. 
 

Anything is possible if enough resources are spent in the design. 
 

whether anyone would be willing to buy it if it cost near the same as a disk drive is another matter. 
 

running at low baud rates would keep the cost down and reliability up. 

[Faicuai]

7 hours ago, toddtmw said:

Okay. Atari Cassette decks were based on industry-standard hardware, but they were custom for Atari.

 

Would it have been possible to alter the deck to use a head that could access both the "front and the "back (Both the top half and the bottom half of the cassette tape) to double the transfer speed?

 

I realize this could have created all kinds of other problems, like if someone doesn't realize how it works, fills the "front" of the cassette, turns it over and then obliterates what they just wrote while filling the "back".

 

But aside from that, would it have been possible? Would it have made the cassette decks terribly more expensive? Could the Atari have been able to handle decoding both streams at the same time to where it would truly be twice as fast? 

 

Could speed have been doubled again if it supported stereo tapes?

 

Just thinking out loud...

 

-Todd

Cassette decks where really never meant to be main-storage devices. They were just a stop-gag solution.

 

They operate on a unidimensional storage-model, and (truly) speeding them up to serve "interactive" needs would require not only boosting effective data throughput, but special changes on reel-table, to make it more reliable, and to be able to wind-up or perform seek operations MUCH, MUCH faster and accurately.

 

You could get a WHOLE lot more out of disk-drives (or similar means) for every $1 invested on them. HOWEVER, tape deck's MULTIMEDIA capabilities (e.g. voice-and-data tracks) with SIO's own audio-line, would immediately appeal to learning / education needs. They would easily beat disk-storage due to lack of digital-to-digital audio-storage solutions.

Edited May 12, 2020 by Faicuai

[phaeron]

High-speed cassette operation is kind of alien to me as I grew up in the US with disks, but since returning to Atari decades later I'm learning how much more popular tapes were in Europe, which apparently was more cost sensitive. We know now that tapes could have run at much faster rates with minimal cost, both due to the abundance of third-party software-only and cheap hardware solutions on the Atari, as well as other computers that had faster tape interfaces. I was particularly fond of the CoCo's tape interface, which not only was much faster at ~1500 baud but also had a friendly file system built on top of it so you could load files by name.

 

Nevertheless, even the fastest turbo tape interfaces barely reach half the speed of the slowest disk drive, which no doubt limited the effort put into it.

 

[zzip]

14 hours ago, phaeron said:

High-speed cassette operation is kind of alien to me as I grew up in the US with disks

I had a 410 program recorder so high-speed cassette operation is alien to me too ?  I was in the US, but after begging our parents for the XL, we couldn't ask them to shell out another 300 or whatever for a floppy disk,  So we made due with the cassette for a couple of years until disk drive prices came down.

 

But in addition to being slow, it was also rather unreliable.   I can't tell you how many great Basic programs I wrote that I lost when I was unable to retrieve them from tape.   I also remember them being rather sensitive to bumps.   When my friend wanted to load a cassette game, we had to tiptoe out of the room and come back 10 min later because any vibration on the floor seemed to cause the load to fail.

[TGB1718]

15 hours ago, Mr Robot said:

running at low baud rates would keep the cost down and reliability up. 

But Atari Cassettes were probably the worst, they were so unreliable, I usually made 2 or 3 copies of anything I did

in the hope one would load.

[archeocomp]

I remember 800XL game sessions at friend's, waiting 20 minutes for F15 to load, with maybe 20 percent fail rate..

Things changed tremendously when friend upgraded his XC12 to Czech Turbo 2000 system, with load times four times shorter and reliability 100 percent.

[ChildOfCv]

Use a slanted, spinning head like a VCR does

 

I don't think that even modern tape drives approaching gigabyte per second throughput output to multiple tracks at once.  They just spin the tape a lot faster.  It may be possible, but the multiple magnetic fields could interfere with each other.

 

With special hardware, though, one could potentially increase the baud rate.  Tapes can fairly accurately store in the 10KHz range, possibly better.  Applying modem over phone line principals to the audio has the potential to increase bit density just from sending and receiving higher-pitched audio.

Edited May 13, 2020 by ChildOfCv

[Rybags]

For a time the IBM cartridges and many others used multi-track, single pass recording.  At first 18 then 36 bits in total (including parity, 1 bit per byte)

 

The big thing these days in fact isn't helical as I thought it might be, but linear serpentine recording, and it appears shingled mode used also.

I was in mainframe support when we migrated to the early 3480 type cartridges from the reel to reel tapes.

A 12 inch reel-reel could hold about 350 Meg on a 2400 foot tape (you could get longer 3600 foot tapes but they used thinner material).

The early 3480 were actually a bit less but typically could hit 450 Meg thanks to data compression.

 

But things have progressed hugely since then.  Today's 3592 drives are into the realm of multiple TB, 20 TB for the highest capacity ones.

 

Linear serpentine recording works by doing multiple passes with fewer heads that tracks on the tape.  On each successive pass the heads move to present the new data area to be read/written.

 

Note that shingled mode recording is used on some tapes and HDDs because write heads can't be made as small as read heads - so to pack more data in, the heads are moved a fraction of the area of the record head width and the next track written partly overrides the previous.  The advantage is fitting more data but the disadvantage is you have to rewrite the entire "previous" track on a HDD if doing random access writes.  Such HDDs are more suited to archival than for general storage.

[Mclaneinc]

I came from ZX80's so dodgy loads were a known but the load error ratio on the Atari just made sure I got a disk drive asap..

 

As for "could it have been faster", I say, how could it not have been..A Slug taking the bytes to the Atari one by one would have been faster..

Edited May 13, 2020 by Mclaneinc

[Rybags]

The load error rate - I can't remember it being that bad.   I learned quickly to use good tapes so the likes of TDK, Sony, Maxell were in use.  But I got a disk drive not too much longer after the computer, probably a bit under 2 years.

Then the funny thing - decades later phaeron tells us of the bug in the OS that causes a load fail one in every 1,200 block reads or whatever it is.

 

What was annoying is the stupid buttons on the 1010.  I returned the unit and used a borrowed 410 for a while, and the fix employed on the broken buttons is to put a screw in them.

But that's just shitty design.  Probably the worst of the XL products, probably the second worst thing Atari sold after the shitty resin filled power supplies.

[TGB1718]

Although it didn't make loading or saving quicker, I do remember a piece of software (name eludes me) that allowed you to

'Name' programs saved from BASIC or Assembler, this allowed you to save many files on the same side of

the tape as you could 'Load' them by name. e.g. SAVE "C:MYFILE.BAS"

 

Obviously this was only practical for small programs as the search was at 'Play' speeds, but I do know that Loading seemed

much more reliable as the program replaced the system "C:" device.