Designing and constructing a floppy drive from scratch

[Banned 05.11.2017]

Hi friends,

I've been wanting to design and build a floppy drive from (almost) scratch, the only things I will be using are the read/write heads, the motors/mechanical parts.

The rest I want to build from discrete logic, both digital and also analogue where needed.

I have a good theoretical understanding  of the physics of electromagnetism, but I'm not sure it will be enough.

Does anybody know of a book or online material on the engineering of floppy drives? Either 8, 5 or 3 inch formats. Preferably 8 1/4.

Thing is I really want to do this, and I will no matter what, BUT I don't quite know how a floppy works.

I see they use frequency modulation to write the magnetic data. What does this mean physically? I also know that clock data and data data are written together, so that a signal would look like a clock pulse, and then data after that. This means a low pulse on the clock, and then if there's another low pulse after that it's a 1, if not it's a 0. Does this qualify as FM ?  Is this what they mean by it ?

Also, the signal coming from the head, what does it look like ? Wiggly AC pulses right ? So I'll have to clean it up with a good ADC.

Thing is, I don't need to be compatible with existing formats at the beginning. As long as I can create a drive that can write and retrieve data in SOME format, that will be enough.

So all in all, I will have to take the head, create some digital logic to control it, and then send some pulses in some order to be written.

Can anyone reccomend me a text on floppy drive engineering? Does this even exist ?

[Alex Eisenhut]

http://www.ti.com/lit/an/snoa710/snoa710.pdf

[Richard Crowley]

So if you are using all the mechanical components from an existing drive, then you aren't really "constructing a floppy drive from scratch".
You are constructing a DRIVE CONTROLLER (the electronic bits) from scratch. 
Haven't you already started a rather extensive thread on this subject?  What is different about this one?
Weren't most of these questions already asked and answered?

Are you not able to find data online? 
You could learn most of what you are asking by simply getting a drive and a few discs and looking at the signal steam from the head on an oscilloscope.

The first generation of floppy drives were pretty primitive. Pretty much as simplistic as you would imagine.
In successive generations, they started improving storage density by attacking the problem from all angles.
Doubling the capacity by using both sides of the disc.
Using higher clock rates for reading/writing.
Compressing the data using RLL (https://en.wikipedia.org/wiki/Run-length_limited), etc. etc.

[alpher]

AFAIK there wasn't 8and1/4" floopy only plain 8", look for the Shugart floppy documentation.
Floppies are probably the best documented tech out there  .
FM, MFM, endec, etc. lots of info out there.

[rstofer]

The only drive I am familiar with is the 8" IBM 3740 style.  Yes, it used FM encoding and in order to capture the data there was a phase locked loop PLL that restored the clock from the stream.  This was the hard part.  Analog...

Then I ran the stream into a Western Digital 1771 Floppy Disk Controller chip, added some address decoding logic and, voila', a floppy for my Altair 8800.  Still have all of it.

But you said discrete...  Some time in the mid '70s, Motorola published a hardbound microprocessor applications manual.  In the book there was a floppy controller that used discrete TTL to recover the data instead of an integrated controller chip.  There was a LOT of logic.

Page 17 here gives the track layout:

http://bitsavers.trailing-edge.com/pdf/ibm/floppy/GA21-9182-3_Diskette_General_Information_Manual_Sep77.pdf

If you look at the index pulse, you will see it is followed by a gap then a sync field to lock up the PLL, then an address mark which must be compared in logic.  Basically, there are ID fields and DATA fields.  Lots of comparators to match the requested sector ID in the ID field with the ID just read from the disk.

I wish I could find that Motorola book, I used to have a copy, but I dumped it a long time back.  Silly...  I THINK this is the right book but I can't prove it:

https://www.alibris.com/Microprocessor-applications-manual-Motorola/book/4349637

It's pretty cheap (before shipping) so there isn't much to lose!

If you can find a 1771 datasheet, there will be a PLL circuit included.  There are other similar devices, maybe they will help.

[LaserSteve]

See the examples page here, and scroll down to the floppy decoder.. Your looking at the web site of a former designer for Rola, one of his specialties was floppy...

http://www.analog-innovations.com/SED.html

Steve

[BradC]

So all in all, I will have to take the head, create some digital logic to control it, and then send some pulses in some order to be written.

Can anyone reccomend me a text on floppy drive engineering? Does this even exist ?

This gives encoding methods, timing diagrams and a full schematic of the controller. It's probably a decent start point for at least looking at the timing, sync and encoding methodologies.

http://bitsavers.org/pdf/intel/iSBC/9800420A_iSBC_202_Hardware_Reference_Sep77.pdf

[BradC]

This is also about as basic as it gets for the Analogs, although it does use an MC3470 integrated read amplifier.



The Apple disk implementation was neat as Woz didn't use the index hole, so the on-disk encoding carried all sync data embedded, and the controller card and associated OS routines were a great example of elegant simplicity.

(edit) Huh, drawn by Dave Jones.

[Banned 05.11.2017]

So if you are using all the mechanical components from an existing drive, then you aren't really "constructing a floppy drive from scratch".
You are constructing a DRIVE CONTROLLER (the electronic bits) from scratch. 
Haven't you already started a rather extensive thread on this subject?  What is different about this one?
Weren't most of these questions already asked and answered?

Are you not able to find data online? 
You could learn most of what you are asking by simply getting a drive and a few discs and looking at the signal steam from the head on an oscilloscope.

The first generation of floppy drives were pretty primitive. Pretty much as simplistic as you would imagine.
In successive generations, they started improving storage density by attacking the problem from all angles.
Doubling the capacity by using both sides of the disc.
Using higher clock rates for reading/writing.
Compressing the data using RLL (https://en.wikipedia.org/wiki/Run-length_limited), etc. etc.

Wouldn't the data on the scope be changing too fast to be readable ? The data is not regular either and will be all over the place and changing too fast.

Would it be possible to see it by using some kind of triggering?

[Richard Crowley]

Wouldn't the data on the scope be changing too fast to be readable ?

The data rate from a floppy disk was very slow compared to what modern oscilloscopes will show.

The data is not regular either and will be all over the place and changing too fast.

But those changes will very graphically show you the differences between zero and one.

Would it be possible to see it by using some kind of triggering?

Sure. You would use various triggering and horizontal rates.
You would look at a few hundred ms sweep to see the overall layout of the entire track.
And even have the sector hole detector on another channel to see the timing relationship.
Triggering on the sector hole would be good for viewing the entire track. (Or viewing the first sector in the track.)
And you would "zoom in" to see individual sectors.
And then zoom in more to see how the ones and zeroes are written.

That schematic from Apple doesn't even show the sector hole signal because Woz was "flying blind".
He was either assuming the exact speed of the spindle, or wasting a lot of space with a "guard band".

[Banned 05.11.2017]

Wouldn't the data on the scope be changing too fast to be readable ?

The data rate from a floppy disk was very slow compared to what modern oscilloscopes will show.

The data is not regular either and will be all over the place and changing too fast.

But those changes will very graphically show you the differences between zero and one.

Would it be possible to see it by using some kind of triggering?

Sure. You would use various triggering and horizontal rates.
You would look at a few hundred ms sweep to see the overall layout of the entire track.
And even have the sector hole detector on another channel to see the timing relationship.
Triggering on the sector hole would be good for viewing the entire track. (Or viewing the first sector in the track.)
And you would "zoom in" to see individual sectors.
And then zoom in more to see how the ones and zeroes are written.

That schematic from Apple doesn't even show the sector hole signal because Woz was "flying blind".
He was either assuming the exact speed of the spindle, or wasting a lot of space with a "guard band".

Fantastic. Do you have experience working with floppy drives and/or controllers? Do you understand the process or writing and reading  the physical data onto the magnetic material ?

it would seem to me to be an "easy" process on the surface. By that I mean that the actual writing of data is simple, just a coil of wire around a C shaped conductor etc. What I am kind of confused about is why the clock data is written together with the data? If one writes the data only and uses an external clock what would be the problem ? Would the external clock go out of synch with the data? I'm not quite sure?

Also what is the PLL used for?

[glarsson]

That schematic from Apple doesn't even show the sector hole signal because Woz was "flying blind".
He was either assuming the exact speed of the spindle, or wasting a lot of space with a "guard band".

Very few floppy systems used a sector hole (hard sectoring). Most floppy systems only used the index hole to indicate the track start (soft sectoring). However, both systems "fly blind" as no speed information within a sector or track is provided to the controller by the drive. The sector hole only provides a way for the drive to verify that a disk was inserted (assert the ready signal) and to lay out the tracks in a neat way (all starting at the same "time"). The Woz method makes no difference except that the tracks start at a random point.

[Richard Crowley]

Fantastic. Do you have experience working with floppy drives and/or controllers?

Only the general principles.

Do you understand the process or writing and reading  the physical data onto the magnetic material ?  it would seem to me to be an "easy" process on the surface. By that I mean that the actual writing of data is simple, just a coil of wire around a C shaped conductor etc.

Yes, that is basically it.

What I am kind of confused about is why the clock data is written together with the data?

Because, otherwise there is no clock to read the data with.

If one writes the data only and uses an external clock what would be the problem ?

You can't depend on a cheap floppy drive (or even a gold-plated military/aerospace drive) to stay "in sync" with any external clock.  That is completely beyond we we can expect here in the Real World.

Would the external clock go out of synch with the data? I'm not quite sure?

It is guaranteed to go out of sync.

Also what is the PLL used for?

To recover the clock!

[dexters_lab]

have a search around on the Shugart floppy interface as documents will often include theories on FM and MFM encoding, track and sector synchronisation etc so should form a good primer for making something yourself.

The members of the Kryoflux community forum are all highly knowledgeable about the nuts and bolts of floppy drives so i would recommend heading over there as they often have to deal with strange floppy disk formats

floppy formats can be very simple, the 8" floppies from a piece of video equipment i have been looking at recently have no sectors and rely on just a single track of data with some synchronisation bytes so you don't have to have the complexity of track sectoring

[TK]

That schematic from Apple doesn't even show the sector hole signal because Woz was "flying blind".
He was either assuming the exact speed of the spindle, or wasting a lot of space with a "guard band".

Very few floppy systems used a sector hole (hard sectoring). Most floppy systems only used the index hole to indicate the track start (soft sectoring). However, both systems "fly blind" as no speed information within a sector or track is provided to the controller by the drive. The sector hole only provides a way for the drive to verify that a disk was inserted (assert the ready signal) and to lay out the tracks in a neat way (all starting at the same "time"). The Woz method makes no difference except that the tracks start at a random point.

And it makes it possible to use both sides of the floppy disk by just cutting the write protect notch on the other side and inserting it flipped.

[rstofer]

Speed information was included by way of the Sync bytes (inside the Gaps) that locked up the PLL.  The PLL was the heart and soul of the IBM 3740 disk drive.

http://deramp.com/downloads/floppy_drives/FD1771%20Floppy%20Controller.pdf  Page 13

http://info-coach.fr/atari/hardware/_fd-hard/AN-505.pdf

Page 4 of this document shows one approach for creating the PLL which, ultimately, results in FDClock and FDData to the controller chip.

https://archive.org/details/Western_Digital_1771-01_Application_Notes_19xx_Western_Digital

[Banned 05.11.2017]

Speed information was included by way of the Sync bytes (inside the Gaps) that locked up the PLL.  The PLL was the heart and soul of the IBM 3740 disk drive.

http://deramp.com/downloads/floppy_drives/FD1771%20Floppy%20Controller.pdf  Page 13

http://info-coach.fr/atari/hardware/_fd-hard/AN-505.pdf

Page 4 of this document shows one approach for creating the PLL which, ultimately, results in FDClock and FDData to the controller chip.

https://archive.org/details/Western_Digital_1771-01_Application_Notes_19xx_Western_Digital

How many clock cycles does the PLL have to synch to the signal? Also, what clock is there to use before the PLL has caught up with the clock in teh floppy ?

[amyk]

Looking at the (freely available) ECMA standards on floppies would also be a good idea.

58,59,69 (8")
66,70,78,99 (5.25")
100,125 (3.5")

[Banned 05.11.2017]

Looking at the (freely available) ECMA standards on floppies would also be a good idea.

58,59,69 (8")
66,70,78,99 (5.25")
100,125 (3.5")

I think it would be a good idea if you stopped posting on here since you indicated that you don't think I have the "aptitude" to take on such projects.

[Richard Crowley]

How many clock cycles does the PLL have to synch to the signal?

That depends on how good the PLL circuit is. Early discrete-implemented circuits were pretty slow by modern standards.  Later generations of specialty IC PLLs were much better, and that was a key contributing factor to the increase in capacity of floppy (and later hard) drives.  Of course, if you really want to make a PLL out of discrete components, you have a major project on your hands. Good luck getting anything functional before next year.

Also, what clock is there to use before the PLL has caught up with the clock in the floppy ?

A common solution was to insert some fixed amount of ones before the data starts to "lock-in" the PLL.
You really have a lot to learn about encoding and decoding asynchronous data.  Designing discrete circuits to read and write magnetic media (whether tape, floppy, or hard drive) will make designing a CPU look like child's play.

[gnif]

Looking at the (freely available) ECMA standards on floppies would also be a good idea.

58,59,69 (8")
66,70,78,99 (5.25")
100,125 (3.5")

I think it would be a good idea if you stopped posting on here since you indicated that you don't think I have the "aptitude" to take on such projects.

I think it would be a good idea to be respectful before you get banned.

You should note that the questions you are asking show your level of understanding and ability.

I am no EE, I dabble in electronics, hack on things, etc.. Yet even I understand the need for an encoded clock with the data (clock drift, thermal expansion of the media, thermal expansion of the read heads, belt slip, someone bumps the drive, just to name a few). Even a much more more modern and precision device in comparison (CD/DVD/BR) encodes timing and also tracking information in a sub channel on the disk.

People are pointing you at information that will answer your questions, go and read it, appreciate they spent their time trying to give you a leg up. This forum is very friendly to those willing to put the required effort in to learn.

[Banned 05.11.2017]

How many clock cycles does the PLL have to synch to the signal?

That depends on how good the PLL circuit is. Early discrete-implemented circuits were pretty slow by modern standards.  Later generations of specialty IC PLLs were much better, and that was a key contributing factor to the increase in capacity of floppy (and later hard) drives.  Of course, if you really want to make a PLL out of discrete components, you have a major project on your hands. Good luck getting anything functional before next year.

Also, what clock is there to use before the PLL has caught up with the clock in the floppy ?

A common solution was to insert some fixed amount of ones before the data starts to "lock-in" the PLL.
You really have a lot to learn about encoding and decoding asynchronous data.  Designing discrete circuits to read and write magnetic media (whether tape, floppy, or hard drive) will make designing a CPU look like child's play.

You really have a lot to learn about respecting other people before they tell you to FUCK OFF. Hope you get it now.

I have a master's degree in mathematics and I am doing a master's degree in microelectronics. Magnetic media are irrelevant today. I am doing it for fun. I am 1000 levels above your old ass.

And oh, about designing a CPU being child's play, go tell that to Intel. I bet you yourself couldn't design even a simple 4bit CPU.

Here's a simple serial tape read/write circuit that works: https://diyhacking.com/make-uart-cassette-tape-interface/

That should show you that getting started in designing a simple circuit to read magnetic media is not as hard as your pea sized brain thinks.

Best Regards

[Banned 05.11.2017]

Looking at the (freely available) ECMA standards on floppies would also be a good idea.

58,59,69 (8")
66,70,78,99 (5.25")
100,125 (3.5")

I think it would be a good idea if you stopped posting on here since you indicated that you don't think I have the "aptitude" to take on such projects.

I think it would be a good idea to be respectful before you get banned.

You should note that the questions you are asking show your level of understanding and ability.

I am no EE, I dabble in electronics, hack on things, etc.. Yet even I understand the need for an encoded clock with the data (clock drift, thermal expansion of the media, thermal expansion of the read heads, belt slip, someone bumps the drive, just to name a few). Even a much more more modern and precision device in comparison (CD/DVD/BR) encodes timing and also tracking information in a sub channel on the disk.

People are pointing you at information that will answer your questions, go and read it, appreciate they spent their time trying to give you a leg up. This forum is very friendly to those willing to put the required effort in to learn.

You are no EE ? Your profile photo does show that. You look more like a clown.

You know nothing about my own efforts so you better shut your little faggot mouth up as well.

If you are no EE, then don't try to teach those who are, and try to make yourself sound like you're better, because you aren't. You are just a little loser.

It's easy to say you understand the need to encode a clock together with data when you yourself has heard it from others, and has never even designed a relevant electronic circuit yourself, as a "dabber".

This forum is for those willing to answer questions, and not for those willing to take other's questions and mock them, otherwise you might make them angry, and they will tell you to go bury your head into a shit hole, where it belongs.

[Banned 05.11.2017]

How many clock cycles does the PLL have to synch to the signal?

That depends on how good the PLL circuit is. Early discrete-implemented circuits were pretty slow by modern standards.  Later generations of specialty IC PLLs were much better, and that was a key contributing factor to the increase in capacity of floppy (and later hard) drives.  Of course, if you really want to make a PLL out of discrete components, you have a major project on your hands. Good luck getting anything functional before next year.

Also, what clock is there to use before the PLL has caught up with the clock in the floppy ?

A common solution was to insert some fixed amount of ones before the data starts to "lock-in" the PLL.
You really have a lot to learn about encoding and decoding asynchronous data.  Designing discrete circuits to read and write magnetic media (whether tape, floppy, or hard drive) will make designing a CPU look like child's play.

You surely know nothing about PLL's. I have built many PLL's before out of discrete components, and I did it in less than a week. It seems your old brain is too slow. I feel you buddy.

[gnif]

Welcome to ban land. You are one of those idiots that believes that because they have a 'Masters' they are better then the rest of us. My work speaks for itself, in this community and others, you clearly have no idea who you are talking to.

You are no EE ? Your profile photo does show that. You look more like a clown.

Ok 

You know nothing about my own efforts so you better shut your little faggot mouth up as well.

How old are you? You're coming across like a child.

If you are no EE, then don't try to teach those who are, and try to make yourself sound like you're better, because you aren't. You are just a little loser.

You're not going to get far in this world in any industry. I bet you clash with just about everyone you feel is beneath you. I don't need to try to make you think I am better, nor do I really care if you or others think if I am, it's not important to me, just as it's not important to most people, we all grow out of that need when we leave adolescence behind.

It's easy to say you understand the need to encode a clock together with data when you yourself has heard it from others, and has never even designed a relevant electronic circuit yourself, as a "dabber".

If you had any brains at all you would know that a oscillator is never perfect and has drift, it's a very simple principle and limitation, and one of the first things an EE student learns... you don't need to be a rocket scientist to know that if you set two clocks to the same time they will gradually lose sync with each other.

This forum is for those willing to answer questions, and not for those willing to take other's questions and mock them, otherwise you might make them angry, and they will tell you to go bury your head into a shit hole, where it belongs.

Your questions were answered, you were given detailed information to research. If you want to get angry that we wont do your homework for you, then that's your problem. Grow up and stop acting like a little kid.

And oh, about designing a CPU being child's play, go tell that to Intel. I bet you yourself couldn't design even a simple 4bit CPU.

You need to get a better grasp on the English language. He did not state that designing a CPU like yours was so easy a child could do it. He stated that designing a CPU like yours is very simple in comparison to what you are asking. Found your website by the way... not that impressed for someone that has a 'Masters'.