Page 2 of 2

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Wed Dec 23, 2015 4:33 pm
by ZippyZapp
Thanks mr.vince for the detailed info. Just to be clear I am looking to use my KF to archive all my old disks. It is a great device. I do also use it to write back disks too so that they may be used versus the originals. Mostly Commodore 8-bit and Amiga but recently finding my old DOS disks too.

I realize it is not sold as a copier board, but the ability to write back to a fresh disk is huge, at least for my use.


Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Wed Feb 03, 2016 11:04 am
by anormal
Weak bits are always cause of debate (as i've seen in other forums), because i think, they are "strange"

JFYI (and i'll appreciate comments here from kryoflux people), i always wondered if this is true (never tried):
Weak data bits could be created by software, too. First approach is to
manipulate drive select/deselect bits in digital output register (3F2h).
For example, if you wish to create weak byte on drive A:, you should start
write operation, wait until desired byte will be transferred to disk drive
(not to FDC!) and repeat sending to digital register values 1Dh (select
drive 1) and 1Ch (select drive 0) while byte being transferred. Such
operation will modulate all data (including clock bits) written to disk
with square wave, moving them into uncertain region.
Second technique requires almost the same operation with diskette
control register (3F7h). Data rate switching will misplace data and clock
bits and deform them in shape, also moving into uncertain area.
Extracted from Copymaster 0.44a docs.

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Wed Feb 03, 2016 3:54 pm
by IFW
There is nothing strange about them, it's just a phenomenon caused by lack of flux reversals and AGC.
It is only strange if you can only see the high level access you have via PC FDC, not how the disk drive and magnetic data storage in general works.

There are other two other methods as well, but - again - you need to understand how the recording works, not what you see via an FDC.

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Thu Feb 04, 2016 1:46 pm
by anormal
Yes, as usually you are right. I think probably the best you have done with Kryoflux and previously Softpres, was teach us where the true interesting thing is, the surface, the magnetic fluxes, and how the drives work from it. I still see many post in other forums that ask question about sectors and possiblity of duplicating, etc ... when as you said, if you understand really the low level, the rest have much more sense. Of course the kF board is a very good collateral effect :D

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Thu Feb 04, 2016 3:26 pm
by mr.vince
Again, when you know what you are looking for, you could e.g. modify KryoFlux stream data and write it from there. We'll also continue to enhance the software, but as I wrote above... Weak bits are there on purpose. They are intended to cause issues, unless you have the original data description of the "gold master" (e.g. FreeForm script for Trace duplication device). The only problem is that to duplicate the disk you need to at least assume what was written (from what you read), as the data read back is very much different from what was written. The easiest way to create weak bits is to write nothing at all. An unformatted area (e.g. the splice between tracks) automatically becomes weak bits - it's pure noise that the drive won't read so to compensate it turns up the "volume" until the static becomes really loud (=generates flux reversals) which in turn are random.

It depends on the hardware if weak bits can be written on the target platform, most devices, e.g. the Amiga, use some kind of "hardware generator" for the bitcell size and will encode the data, so that becomes troublesome.

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Fri Feb 05, 2016 11:32 am
by spags
Ah, so that explains how that works. So really, the accuracy of the detected signal (amongst other things) is a function of the length of time since the last (set) of reversal(s) is detected. Presumably up to some agreed standard length of time, any reasonable and fit drive should detect the same sequence of reversals as any other fit and reasonable drive, then progressively get worse when starting to exceed that standard. So the amount and degree of that randomness will be affected by that length of time and how rapidly that gain control is adjusted.

Theoretically there would then be this "sour spot" from "within standard time" to "well out from standard time" where for a given disk, a badly written piece of code on different machines may sometimes correctly and sometimes incorrectly detect said floppy as being genuine or not. Or to put it another way, a specially crafted signal written to a floppy and interpreted by a specially crafted bit of code could be used to give some metric to describe the AGC characteristics of any given drive. Interesting, but besides being an AGC diagnostic feature it is probably of little practical benefit.

Thinking about that even further (and I know this sounds kind of obvious) then the concept of an "ideal" drive would be one that is made of ideal analogue electronics and ideal AD conversion and without AGC. So if there was such a thing as a digital head, that only read zero or one for any given ideal and digitally written floppy, then you could get an an exact duplication of the intended information on that floppy. So really the AGC was intended to counter for small variations in signal intensity due to the practical nature of the equipment being used.

Re: copy-protected 3,5" DOS Floppies & Kryoflux

Posted: Fri Feb 05, 2016 6:22 pm
by IFW

Keep in mind that the signal is analog, and threshold levels trigger the AD conversion.
The AGC/filtering affects the threshold.
You only have access to the D signal after this conversion, apart from the diagnostic A output of some drives.

In theory the drive density line should not affect the filtering during reading - only the coercivity during writing -, in practice most (all?) drives we've tried changes the filtering during reading in response to density line changes.
If you think about it, it makes sense, since you want a glitch filter, AGC/low pass filter as well as reasonable readability.
What is a glitch in a DD recording can be a valid flux reversal in HD recording.
If you can't see a flux reversal for say 8us in HD MFM, chances are you missed it, while in DD MFM it's just a valid 100 encoding.

The exact filtering seems to be fine tuned by every manufacturer/model according to their preferences for the density mode selected - so even a revision of a drive model may behave slightly differently.
Hence why certain drives have no problems reading e.g. Mac disks, while others would simply never be able to read certain encoding patters correctly, as the AGC would split them.

FYI as an example, there are professionally mastered C64 and Amiga games that use encodings that are randomly readable depending on the drive (actually, head) type... and I mean on the original machines :)
These disks very likely did pass verification during duplication so the problem never showed until the costumer returns started flooding in.
One example from C64 is the very first V-MAX! encoded disks, that exceed the limitations of the drive, and all of them have running replacement with an encoding within the limits.

Similarly, using more dense recording, such as long tracks (either as protection or just to cram more data onto the disk) has the side effect of altering the reliability of the reading if the target drive has more than usual speed wobble or the RPM is not close enough to the ideal value.
While the PLL can track such data (the Amiga PLL remarkably well), the parameters of the PLL are set with standard recording densities in mind - so once the rotational speed diverges and the densities change from the ideal it does have a cumulative adverse effect on readability.

Regardless, using long tracks was immensely popular by game publishers - since analog copiers did not have write pre-compensation that made at most a second generation copy of a disk completely illegible.