Hacker News Reader

Comment by danbruc | Hacker News Reader

I have always wondered if it would be possible to build a generic reader for magnetic tapes that essentially just scans whatever tape you throw at it and then it is up to some software to decode the content. I know close to nothing about the details of magnetic recording on tape, so maybe this is just not possible with reasonable effort, i.e. you could but you would need SQUIDs to make it work or something like that.

View on HN

You might be surprised at how the signals are actually recorded on tapes. For VHS tapes for example, it's actually a series of stacked "diagonal" lines of signal. Imagine a sequence of symbols like \\\\\\\\\\\, except at any given horizontal point along the tape there are about 5 stacked traces of signal. They're also diagonal on audio cassettes, but much less "slanted" relative to the tape's length.

I don't remember the precise terms for the various concepts, but hopefully that shows how specific to a given format the reader needs to be.

Technology Connections did an excellent video on this:

https://www.youtube.com/watch?v=KfuARMCyTvg

I'd recommend pretty much all of his video series to anyone in the HN crowd that's at all interested in analog media formats:

https://www.youtube.com/@TechnologyConnections/playlists

Yeah, that channel is where I learned it as well. I probably should have linked it.

Basically every video he does scratches that itch of learning a new thing about the world. The analog video series is indeed especially fantastic (laserdisc, VHS, B&W vs color TV, blu-ray, etc).

If a high resolution reader could “recreate” the entire tape signal in memory, you could do the format specific parsing in software. You could even identify the tape type based on the signal patterns

I've heard of someone doing this by connecting an ADC directly off the heads of a VHS player, but I can't find it now. The bandwidth of such a system is only about 6MHz, putting it well within modern SDR capability.

Certainly this is how the "applesauce" reader works: https://applesaucefdc.com/

hence the name: Vertical Helical Scan

That's a backronym. VHS originally was an abbreviation for "Video Home System".

Edit: also, the helical scan on a VHS tape can't really be said to be "vertical". A helical scan already implies a direction along the length of the helix, and "vertical helix" doesn't really make sense in the context of scanning down the length of a magnetic tape. Adding "vertical" to the actual term "helical scan" can really only be an attempt to back-fit an acronym to "VHS".

Audiotape is - or at least, Phillips compact cassettes are - recorded linearly. The recording method you're talking about for videotape is called helical scan, because the heads trace out a helix pattern. You might be confusing that with the head azimuth, which does alternate between fields on VHS to reduce cross-talk or something.

Some other bits of jank in the VHS spec:

- Because the tape is moving while it's being written, it stretches the signal out on the tape. This is perfectly fine for normal playback. But when the tape is not moving, the signal's now too wide for the playback head, and you can only read about half the picture. That's why your VCR had bars of static whenever you paused (unless you sprung for the four-head model)

- Audio is still recorded linearly, and you can't exactly chuck a linear head in an angled, spinning drum. So you have to put the audio head further away from the tape. And that distance is fixed; changing it means your machine is now playing audio out of sync with the video.

Also, there is an extension to VHS that lets you record audio along the video in the helical area, it's called VHS Hi-Fi and it improves the audio dramatically with the trade-off that any minor video glitches will add pops to the audio. Humans are way more sensitive to gaps in audio than video, after all.

I dumped a bunch of albums, and a few CD's, to Hi-Fi VHS. A while back, I stumbled onto a compatible VCR at a yard sale for a couple bucks.

Yes, those gaps are annoying as all get out. Tracking has to be dialed right in to recover the sound properly.

But, once you do that?

It's really great! Frequency response goes almost to zero, and up to 22Khz, and it's flat for most of that range.

One of the albums I recorded had a warp in it. When the actual vinyl was played on a good stereo system it was possible to see the speaker cones actually move in time with the warp. It's essentially a very low frequency signal.

Cassette does not reproduce that. I bet a good reel-to-reel system would.

Hi-Fi VHS reproduced it pretty much bang on perfect!

The best part was being able to add index marks! While recording, one could press a button and get one of those written to the tape. I had several tapes made with music I really like and could access pretty much anything on it quickly.

Fun stuff!!

I am at least aware of the diagonal recording on VHS tapes, that is why I assumed it might be hard do build a generic reader as you could not necessarily take advantage of the geometry or the speed of the tape or things like that. There is probably a good reason for the complexity inside an old VHS recorder.

Other tidbits I recall from what my Dad told me years ago. He repaired VCRs for a while.

The beta machines would phase shift the signal 180˚ every other frame when writing, and they use that for noise cancellation. VHS did four frames at 90˚ each, because of patents.

The read heads were on a spinning drum. Beta machines had an L loading system that would grab the tape and wrap it almost all the way around the drum with the read heads, exposing more tape at once to the reading mechanism. VHS had an M loading system that grabbed the tape on either side and pulled it up, covering maybe half of the drum, because of patents.

And supposedly VHS won out because Sony was begging a lot of money for patent licensing. But tape capacity was probably a factor, too.

One of the terms was interlaced, the nearly forgotten "i" next to something like 480i (versus 480p).

Individual magnetic lines of a VHS tape would only describe half a frame and the two halves would be interlaced back together.

Well, that's just how the underlying PAL/NTSC signal worked.

Right, and the fact that there were two different signal standards and VHS was a leaky abstraction that mostly just encoded the signal as-is all adds to the complication of any "generic magnetic tape reader" and hoping to get useful digital data back from that.

I think one major network in the US is 720p60, but the vast majority of broadcast transmissions globally is 1080i25 or 1080i30*

Of course sadly so much content is produced at 1080p24 nowadays for that “film” effect it’s rather meaningless for display purposes.

Online of course it tends to be p50 or more commonly p60 (even in 50hz countries). YouTubeers tend to be able to afford more lights and clearer microphones than big budget productions too.

* 30000/1001

I don't know about DAT as an audio format but DAT in its data storage format "Digital Data Storage"[0] is also helical scan.

[0]: https://en.wikipedia.org/wiki/Digital_Data_Storage

That sounds like a similar problem to reading unknown floppy formats. I'll bet you that actually scanning the flux (I think that's the term) image of the tape without shredding it is a more difficult mechanical problem.

"Find the lines" seems to be a common image processing technique. If you had the 2D data couldn't you virtually scan along them with that?

If you "scan the magnetism" for the whole width of the tape -- which such a generic reader for magnetic tapes would have to do anyway -- you can parse out such things after the fact.

    adg.
    .beh
    ..cf

Even if your input is "a..db.gec.hf", that should be able to postprocess out later, for any possible angle, given enough resolution.

It's a lot like decoding music from vinyl grooves based on a 2D image scan of the vinyl.

That's going to take an extremely high resolution to avoid cross-talk. For a linear tape with multiple tracks you just need to have a few small heads at different heights.

This is pretty much the only way 5.25" floppies are read today with modern equipment. While you can still get USB 3.5" floppy drives there are no USB 5.25" floppy drives. Modern PCs don't have the correct drive headers anymore, either, and even if they did, they wouldn't be able to read Commodore or Apple II floppies.

However, there is hardware and software called GreaseWeazle that actually reads the magnetic flux transitions on the disks to extract the data within.

https://github.com/keirf/greaseweazle

The greaseweazle still looks like a connector for an existing 5.25" drive, is there any hardware you know of that replicates the drive itself?

> The greaseweazle still looks like a connector for an existing 5.25" drive

That's because it connects to an existing 5.25" drive

> is there any hardware you know of that replicates the drive itself?

Why would you build a device to emulate a device that already exists, when all you need to do is communicate with it?

Because nobody's made new 5.25" drives for decades and eventually they will all be gone?

I don't want to have to buy used gear off eBay, and wouldn't need to for 3.5" disks or drives - you can actually find people producing new stock.

Why doesn't 5.25" hardware get the same treatment?

GreaseWeazle is great (I own one!), but it's far from "the only way today".

Not that long ago, motherboards still had a floppy controller. These computers do mostly still work.

So do the ones from the 80s, for that matter.

>USB 3.5" floppy drives

To anybody considering: Don't bother. They can only read/write the most standard IBM PC format. No flux streams.

Instead, get any old floppy drive and a greaseweazle at about the same cost. You'll be able to read and write all sorts of formats, and recover data from damaged and otherwise unreadable floppies.

For VHS there's the VHS-Decode project that's trying to do something similar. I don't think they work by scanning the tape itself with some custom hardware, but rather by recording some kind of internal RF drum head signal inside the VCR itself. Then they have a software stack to decode the VHS signal into actual video. Looks pretty cool, but I've never used it.

https://github.com/oyvindln/vhs-decode

TechMoan has used a reel to reel player as a version of this for audio tape. A few obscure formats only fit in rare specialty players, so he's simply wound their internal tapes onto reels and digitized them that way.

This seems quite possible in principle, despite the different helical scanning, tape widths, speeds and whatnot. A device at the British Library archival research did similar for wax cylinders and old disc records, scanning the physical surface with a small wavelength laser interferometer. AFAIK it generated an obscene amount of point cloud data.

My thought is that, if you could develop a sensor capable of sampling magnetic alignment at sub-micrometer resolution it would also generate unbelievable amounts of data that would first need compressing or decoding into AV signals.

This idea sounds similar to "software defined radio"... but for tape reading. Interesting.

This system is similar for old analog audio recordings:

https://irene.lbl.gov/

https://blogs.loc.gov/loc/2012/03/unlocking-sounds-of-the-pa...

I remember one of the curators mentioning that they had a ton of WWII records for things like daily newscasts which due to wartime supply issues were made from less durable materials, and since they weren’t commercially valuable the owners hadn’t spent time transferring many of them to newer media.

> My thought is that, if you could develop a sensor capable of sampling magnetic alignment at sub-micrometer resolution it would also generate unbelievable amounts of data that would first need compressing or decoding into AV signals.

Hard disk drive read head sweeping across the tape? Hard disk track widths are well below 1 micron so presumably the potential for high resolution is there?

Nice idea. I think you'd have a problem with shedding. A HD head is designed to fly/float above a clean platter. Old tape is icky, flaky mess of brown dust and wobbly plastic that would foul up the head PDQ. But rather than scanning maybe you could encapsulate a bunch of old HD densely together in some kind of epoxy/glass to make a high-res tape head?

In theory you could have some sort of multi head or rasterizing system scan every bit of the tape and construct a 2D map of the magnetization. It would have to be 2D because many tape formats have multiple tracks, not even in parallel. But decoding that data would be a nightmare that made optical character recognition look like child's play.

Really? Couldn't you just 3D model the player and run the simulated tape through it? Pretty amazing if there's something a 70s mechanical device can do that a computer can't do easily.

I wondered the same. We have software-defined radio now, what's stopping us from having a software-defined magnetic media reader? You do need to scan the tape a very high spatial resolution though.

Reply on news.ycombinator.com