Re: Did Commodore cheat with the quad density floppies?

On Freitag, 4. Januar 2019 19:46:14 CET smf wrote:
> On 03/01/2019 17:29, afachat@gmx.de wrote:
> > GCR is NOT more efficient than MFM using the same bit frequency. In fact
> > MFM is more efficient (writing 8 cells @ 250kHz for MFM vs. 10 cells @
> > 250kHz for CBM GCR).
> > 
> > The reason that MFM can do this is that they actually use "half cells" at
> > 250kHz, but still ensuring that flux transitions are at least 4us (i.e. a
> > 250kHz cell) apart.
> 
> mfm needs 2 transitions for every bit, so 8 bits requires 16 transitions.

You are mixing up FM and MFM.
See https://extrapages.de/archives/20190102-Floppy-notes.html

> gcr stores 8 bits in 10 transitions. So clearly more efficient, but mfm
> can be clocked faster than gcr because it reduces the number of 0's that
> can occur in a row.

Actually it increases the minimum number of 0-bits from 0 to 1. That is why it 
can be written with the double frequency.

> 
> http://amigadev.elowar.com/read/ADCD_2.1/Hardware_Manual_guide/node0194.html
> 
>       8       FAST       Value of 1 selects two microseconds per bit cell
>                          (usually MFM).  Data must be valid raw MFM.
>                          0 selects four microseconds per bit (usually GCR).
> 
> When Amiga Technologies started fitting high density floppy drives to
> the A1200 someone put together a package that used gcr at the higher mfm
> clock, which (if you used high density media) was supposedly reliable
> enough to store more than the standard amiga low density (880k), but
> less than the standard amiga high density (1.76mb). I never tried it, I
> had an external dual speed drive that worked with standard format high
> density disks, but I think it worked out at approximately 1.46mb.

MFM speed with two us/cell still has transitions only at a distance of 4us or 
more due to the minimum 0-bit.

Recording GCR at the double frequency reduces the minimum distance between 
transitions to 2us due to the possibility of multiple consecutive 1s. This 
would probably go over the flux changes / inch rating of the media, depending 
on the media.

> It's here if you're interested http://aminet.net/package/disk/misc/HiDensity
> 
> Your "half cells" sounds like you're doubling the bit frequency, but
> then saying it's the same.

The misunderstanding here is probably due to the fact that "data bit cell" and 
"encoded bit cell" are all different between FM, MFM, and GCR.

Data bit cells in FM are 8us, and 4us in MFM, while there are no real data bit 
cells in GCR.
Encoded bit cells in FM are 4us, and 2us in MFM. GCR also has encoded bit 
cells of 4us.

André