Wish: MAVIS computer recreation in Mode 7

[dp11]

It turns out these tapes aren't the same as mini dictaphone tapes.

[BigEd]

I suppose the shells are welded rather than screwed together?

[BeebMaster]

It turns out these tapes aren't the same as mini dictaphone tapes. :(

As in, they won't fit or play in a dictaphone, or as in they are more difficult to open so that the tape could be transplanted to another casing?

[dp11]

The tapes are sealed unfortunately. I've started to create a tape player for these tapes. Very early indications suggests that there is a good chance of recovering data of the tapes. A long way to go still.

[BigEd]

It would be excellent to succeed here!

[OneSwitch]

The tapes are sealed unfortunately. I've started to create a tape player for these tapes. Very early indications suggests that there is a good chance of recovering data of the tapes. A long way to go still.

So not standard mini, nor micro-cassettes? I wonder where they came from originally. Good luck. Would be amazing to get data from these tapes.

[BigEd]

I see there's a mention in Turing's Legacy by Yates, but that's not a book I have.

Thanks, I really appreciate you looking, even though I have seen both of those. The Turing's legacy book is a really good read, with loads on the National Physical Laboratory projects, when they had more government funding to do exciting imaginative things.

(Just to note, I now have a copy of the Legacy book.)

[dp11]

Could some let me know if this snippet sounds like real data from the tape ? The tape transport is entirely different so the speed could be wrong and the frequency response will be wrong too.

[BigEd]

(While you're here Dominic, can you measure the width of the tape? I think you said it isn't, after all, the same as a compact cassette. I wonder if the exact width will help in turning up technical documents.)

[BigEd]

It certainly looks more like data than non-data - here's the log(f) spectrum view from audacity:

[dp11]

Decoding this might be a bit of a challenge .

https://drive.google.com/file/d/1ek0OTH ... sp=sharing

[dp11]

@BigEd what is the name of that window in audacity ?

[dp11]

found it.

[BigEd]

I find the spectrogram helpful to distinguish different types of content - you can easily see the headers and data parts of the blocks in a Beeb tape - but I suspect what we have here is more of a pulse modulation than a frequency modulation.

The levels are low, but this has the look about it of data:

[dp11]

I've found a web page showing the difference I have the top tape. Now this might give a chance to find out about other systems that used the tape format.https://obsoletemedia.org/hp-82176a/hp- ... tte-shell/

[BigEd]

Oh, good find. Perhaps someone at the HPCC can help with the loan of a drive? There are many hp-41 fans and many completist collectors. They (we!) meet twice a month online.

The 82176A Mini Data Cassette was introduced by Hewlett-Packard in 1982 for use in its 82161A mini-cassette tape drive that was designed for use with the HP 41 calculator.

The drive could be mains or battery powered for portable use. Tape speed was 30 inches per second, and each tape had a capacity of around 130 KB.

The 82161A mini-cassette tape drive was later used with the 71, 75 and some other models that used the HP-IL (Hewlett-Packard Interface Loop) interface.

The 82176A Mini Data Cassette looks almost identical to the Philips-designed Mini-cassette, but there are some differences in the shell, such as the tape openings and a notch on the top, that prevent a standard Mini-cassette being used in the HP tape drive.

[dp11]

@BigEd Could you ask if anyone is interested in helping out?

I can't actually workout any really data. It sort of looks like a digital erase to me with block markers or *save block of zeros from memory

[BigEd]

I can certainly ask!

Meanwhile, see the article on p17 of the HP Journal for May 1983, about the 82161A Digital Cassette Drive:

[OneSwitch]

I wonder if any Ferranti technical documentation may remain in the likes of the National Archives: https://discovery.nationalarchives.gov. ... eedbec55c8

I would imagine the tape drive used would date to around 1978-1982. I would be surprised if the MAVIS computers weren't all using the same mini-cassette system when the trials started in 1979. Don't know if that helps at all with the HP side of things.

Nice work so far, Dominic, getting something from the tapes.

[iainfm]

I've been lurking on this thread for a while, hoping that I might be able to contribute to it at some point.

All I've managed to find so far is this video that shows the tape drive interfacing with an HP 41CX calculator.

It seems able to store some kind of index catalogue, and skip forward automatically to load the desired program on demand.

So, my main question is how was this drive interfaced with a bbc micro? I suspect it can't have been through the cassette port...

[BigEd]

My best guess from what we've seen so far is that the tapes Dominic has, which are from a MAVIS system, are the right physical form factor for the HP 82161A Digital Cassette Drive, and that the MAVIS and HP mechanisms will be the same. But it wasn't an HP unit that was used by MAVIS, and we don't know if the HP will be using the same magnetic format. I think it's likely the same format at bit level, and presumably that bitstream makes up a byte stream, but any record structure or file structure is likely to be different.

[dp11]

Just for reference this digital cassette recorder isn't quite the same http://www.tubedata.info/phunsy/Philips ... Manual.pdf. From the pictures the tapes are more like dictaphone tapes . But it shows there were commercial digital mini tape decks

[iainfm]

Thanks for that. The more I see of this drive the less convinced I am that the data can be recovered by 'analogue' means. I think it's going to take a proper tape drive to stand a chance.

It's not impossible though, at the end of the day the raw signal coming from the tape heads will be analogue in nature...but emulating how it changes from that into a clean data stream ain't gonna be easy.

Just my 2p.

[dp11]

Just some more tape photos showing the index hole in the tape and if you look carefully you can just see the 45 degree mirror used to bounce light through the tape hole.

[BigEd]

Just trying to collect some comparisons.

First, comparison of two (or three!) tape systems.

Philips' Mini Digital Cassette Recorder aka Mini-DCR uses Manchester encoding with a bit cell time of 167 microseconds. An OEM unit offers a 13-pin 14-position AMP connector. The cassette contains 35 meters of tape, 3.81mm wide. Tape speed is 10.6 to 18 ips. Data transfer is 6000 bits/sec, density on tape 330-560 bpi. Tape rewind time is 96 sec. There's a single motor. Start distance within 2 inches (or 2.6 inches with a change in direction, or 65mm of tape.) Stop distance within 1 inch. Tape head is single track, separate erase head. Cassette is manually flipped to access the other side. Capacity is 64k bytes per side using one huge record, reducing to 32k bytes per side as 128 records of 256 bytes each. Each record has one byte sync marker and 16 bit CRC. Writing and rewriting a block in between two others "may cause over-writing of th first part of the next data block." Inter block gap is 15 seconds. End of tape detected by motor stall. Calibration tape has 3kHz tone.

HP 82161A Digital Cassette Drive uses Manchester encoding with a bit cell time of 64 microseconds. Capacity up to 128k bytes. The in-drive firmware deals in 256 byte records, with a one-byte record ID and a checksum. The firmware formats the tape as 512 records over both sides of the tape (both in the forwards direction) thus using all 256 record ID values on each side. The write process is a series of flux transitions, driven to saturation. The read process sees a pulse for each written transition, either positive or negative, with the polarity captured at peak using a differentiator. A data record is about three times the length of an inter-record gap. The minicassette is 8 x 34 x 56mm with nominally 24 meters of usable tape 3.81 mm wide, allowing two tracks of data 1.45 mm wide. The head is two-track. There are two motors for the two tape hubs, no capstan. End of tape detected optically, using a 45 degree mirror in the cassette. Tape drive system module is removable. Designed in Corvallis, Colorado. (And needs to be an OEM part for this to be used in the MAVIS Mk2)

Comparison of MAVIS iterations

MAVIS Mk1 (1977) single prototype, using 6800 microprocessor and (probably) compact cassette storage. Built by NPL.

MAVIS Mark II three prototypes, using Z80 micro using some physically smaller cassette. Made by Ferranti. Trials ran from October '79 to April 1980. As a microcomputer, it has 14k of ROM (PROM) and 16k of RAM. Also 128k of non-volatile RAM standing in for bubble memory (or possibly just 64k.) The PROMs contain the kernel, BEBUG the debugger, PLUM the interpreted language, and other packages. "A subset of PLUM forms a simple BASIC." Most or all of the ROMs are written in Z80 assembly language. Some applications are written in PLUM, some in Z80.

The cassette is described as holding 80 frames on each side of the tape, where a frame is a screen and presumably about 1k bytes. (Possibly 18 lines by 40 characters.) But applications can also be loaded from tape. The cassette is described as unreliable. The data handling interface is described as compatible with "any of the presently available phase-encoded digital cassette recorders." A block (or frame) can be accessed individually, or a file copied, or "a whole side read into the bulk store."

The higher level MAVIS software offers a FORMAT command and a TAPE INDEX command, which tells us something about the record structure on tape. Also there's UNSAVE which removes a file, so that perhaps implies a directory on tape. Similarly there's a file system in the non-volatile bulk storage RAM.

The teletext interface is described as using a two-chip solution from Mullard: TIC and TROM, with an external 1k byte RAM.

Info above from OneSwitch's scan of Julia Schofield's 1981 book.

[dp11]

Just for info the Mavis minicassette are 8 x 34 x 56mm with spool centre to centre being 25mm

Also from the same book

[BigEd]

...Perhaps someone at the HPCC can help ...

So, at yesterday's HPCC virtual meetup, Dominic and I both turned up and there was extensive chat about the MAVIS tapes. One thing we learn, by positive experiment, is that the MAVIS tapes do fit into the HP data cassette drive, and they do at least go through the motions of reading. However, plaintext data did not pop out.

It turns out that HP did use an OEM tape transport, so it makes some sense that it was also used by Ferranti a year or two earlier, and that Verbatim-branded tapes were available. (HP of course had their own branded tapes.)

It's notable but perhaps unsurprising that the technical article in HP Journal doesn't mention that the transport is an OEM part. It's also notable that HP's product is, it seems, entirely reliable, whereas Ferranti's efforts were described as producing an unreliable and indeed unsatisfactory storage solution.

More to come, probably, but not imminently. Probably.

[OneSwitch]

More to come, probably, but not imminently. Probably.

Hope so! Would it be possible to copy these tapes using the HP kit. Tape to tape style?

I'm giving a talk for the GAconf Europe event in April on MAVIS. I'll let people know that this sort of stuff is going on in the chat.

This might be of interest to some. Some new MAVIS art-work by my step-daughter Katie, with the MAVIS logo (1979 era Mk II) recreated by Brian Brasher.

https://www.dropbox.com/s/cpt171au2likc ... L.png?dl=0

[BigEd]

Thanks for the art!

I'm perhaps a bit pessimistic (generally), but my expectation is that the HP drives will be very useful for pulling the tape past the head in just the right way, and the head will produce the right signals to feed to the read circuitry, and that's all good. But I'm not very confident that the read circuitry will detect the bits correctly, because of the likelihood that the bit timing is different between Ferranti's design and HP's design. I think our best chance will be to capture the signals from the head, and post process them, and hope to decode them.

(If the HP system does correctly detect the bits, the next obstacle could be that the firmware within the drive is expecting a certain record structure, which again I would suspect will differ from the record structure used by Ferranti. If it does differ, the HP system may not successfully load a record into its buffer, and we'll find the data we read from the buffer isn't right, or isn't complete.)

It's very good, though, that the physical reading of the tape is likely to be OK: that's quite a difficult bit of mechanical engineering. We will have an effectively unlimited amount of computer power to decode the signals we will, hopefully, capture.

Speaking of which, Chris/scarybeasts has a tool to decode waveforms, in his case FM-encoded disc waveforms, but it's surely a similar problem:
https://github.com/scarybeasts/misc/tre ... ccsvpulses

(I'm saying 'we' here - it'll probably be Dominic and one or more HPCC members)

[daveejhitchins]

Would HP actually have placed decoding into the dive? I would have thought that the drive would just process the head output to digital format!

Dave H.