US Doubler UltraSpeed Fastest Sector Reading Order

[+Nezgar]

The 1050 US Doubler is UltraSpeed capable, but does not have a track buffer. So generally, you will only get the maximum read/write speed if the disk is formatted with the special 5:1 UltraSpeed skew. However, by creatively rearanging the order you read the sectors in software, you can read regular disks at a higher speed as well. This probably also applies just as well to the Super Archiver, which normally behaves very similarly to a US Doubler.

I've noticed the sector copier CopyMate 4.3/4.4 does exactly this. It will try reading sectors in 3 different orders until the maximum speed is achieved. I've used RespeQt to analyze the sector orders of it's three different detection modes out of curiosity. Since I have a selection of disks formatted in various densities with various upgrades, I can determine which are optimal for each format. CopyMate 4.4 only supports single and double density, not ED.

									Reading order best for...
SD	1, 11, 4, 14, 5, 15, 8, 18, 9, 2, 12, 3, 13, 6, 16, 7, 17, 10	Disk formatted by Stock 810 or 1050
	1, 10, 17, 7, 16, 6, 13, 3, 12, 2, 9, 18, 8, 15, 5, 14, 4, 11	Disk formatted by USD, Speedy, Archiver (Chicago layout)
	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18	Disk formatted in USD UltraSpeed Skew

DD	1, 12, 5, 16, 9, 2, 13, 6, 17, 10, 3, 14, 7, 18, 11, 4, 15, 8	DD Disk formatted by USD
	1, 11, 15, 6, 10, 14, 5, 9, 13, 4, 8, 18, 3, 7, 17, 2, 12, 16	DD Disk formatted by Happy, Speedy or Percom
	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18	DD Disk formatted in UltraSpeed/Turbo 1050 Skew mode

The other utility that I know of that reads sectors 'out of order' for speed using a US Doubler is the SpartaDOS SCOPY command. (Specifically "SCOPY 9/16/86 ©1994 FTe") You manually tell it whether the source or destination disk is 'normal' by not specifying the /U (UltraSpeed) command line switch, which is supposed to read 'regular' disks in a fast way, or by specifying /U it will read sequentially... or so I thought:

SD	9, 2, 12, 3, 13, 6, 16, 7, 17, 10, 1, 11, 4, 14, 5, 15, 8, 18	Normal: ~1/2 rotation offset, slow reading Chicago skew disks (Those formatted by USD, Archiver, Speedy)
	5, 6, 7, 12, 13, 14, 15, 1, 2, 3, 8, 9, 10, 11, 16, 17, 18, 4	/U: ~1/2 rotation offset, [no SD turbo 1050 formatted disk handy to test]

DD	8, 15, 4, 11, 18, 7, 14, 3, 10, 17, 6, 13, 2, 9, 16, 5, 12, 1	Normal: Only fast with DD disks formatted by USD, slow reading Happy/Speedy/Percom
	2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 1	/U:  reads 1st sector last, slow reading Turbo 1050 std formatted disk

ED	15, 4, 18, 5, 19, 8, 22, 9, 23, 12, 26, 13, 2, 16, 3, 17, 6, 20, 7, 21, 10, 24, 11, 25, 14, 1	Normal: Stock 1050 or USD std ED disk fast, Turbo 1050 ED slow
	5, 6, 7, 12, 13, 14, 15, 20, 21, 22, 23, 1, 2, 3, 8, 9, 10, 11, 16, 17, 18, 19, 24, 25, 26, 4	/U: odd order maybe reads a USD skew disk faster?

It's odd that if you format a non-ultraspeed single density disk with a US Doubler, that SCOPY cannot "read it's own writing' at high speed... (Both being ICD products and all...)

 

Anyhow.. this information could be useful for anyone writing or hacking a sector copier, or incorporate into PC based 10502PC type programs like AtariSIO/atarixfer to read disks directly from a US Doubler/Super Archiver drive without a track buffer at high speed after determining the fastest option... Less overall spinning disk time and reducing wear can be valuable in imaging marginal disks.

 

Determining which reading order is fastest for a given disk can also suggest the group of drive/speeders that could have originally formatted the disk.

[Rybags]

Should note though that special sector orders will sometimes be far worse on a normal drive or other type of non-track buffering mod.

 

And funnily enough - testing stock 1050 I found a slight improvement in speed by reading the sectors per track backwards 26 to 1.

[+CharlieChaplin]

Mycopier 1.c and 2.1 and Copymate 3.7 / 3.8 also do ED and as you told me both support the US-Doubler. Is the maximum USD speed of these programs slower than with Copymate 4.3 / 4.4 ?!?

 

I also have an external (stand-alone) formatter for US-Doubler ED diskettes, never used it, since I do not own the USD...

 

EDFORMAT.xex

FASTLOAD.BAS

FASTLOAD.txt

Edited March 9, 2019 by CharlieChaplin

[+Nezgar]

Should note though that special sector orders will sometimes be far worse on a normal drive or other type of non-track buffering mod.

Yes, but I was focusing on 'Standard' skew disks though, reading at UltraSpeed. An intelligent sector copier could also be written to handle reading a US Doubler 5:1 skew disk at 1x SIO by also reading sectors in another alternate order to compensate.

 

And funnily enough - testing stock 1050 I found a slight improvement in speed by reading the sectors per track backwards 26 to 1.

 

True, for 1x SIO. But only if the disk is formatted in a stock 1050, or other mod that uses the same default Atari interleave.

 

If the disk is formatted in ED by a US Doubler, (and maybe some others) a stock 1050 reading that disk will achieve the same faster speed reading forwards from 1-26. (at 1x SIO.) Same is true for SD.

[+DrVenkman]

I have a tangential question about formatting skews that I keep running into when one of these discussions is NOT going on, lol.

 

So I have two Atarimax Happy boards - is there a preferred utility for format them for DD use with a “better” or faster sector interleave? The SpartaDOS X FORMAT utility works fine for single and double-density formatting, in either Atari DOS or Sparta formats, but it absolutely refuses to allow formatting with anything but a normal interleave. I just get a message that the drive doesn’t support UltraSpeed or HIgh Speed. This is true even if I run the Happy utility to ensemble US emulation on the drive. (Pardon if I may be mis-remembering the exact terminology involved - I’m not at my Atari right now).

[+Nezgar]

Mycopier 1.c and 2.1 and Copymate 3.7 / 3.8 also do ED and as you told me both support the US-Doubler. Is the maximum USD speed of these programs slower than with Copymate 4.3 / 4.4 ?!?

 

I don't know MyCopyR 1.c, but MyCopyR 2.1, and CopyMate 3.7/3.8 only read sectors in sequential order, so UltraSpeed protocol is actually only faster if the disk was formatted with 5:1 UltraSpeed skew. Disks formatted by a stock 1050, or US Doubler in standard skew will read no faster overall than 1X SIO when reading sequentially like that. Each sector itself transfers faster, but there's just a longer delay to read the next sector, equalling out to the same.

 

CopyMate 4.3/4.4 have extra intelligence to try reading sectors in 3 different orders to achieve the minimum inter-sector lag with no track buffer in the USD.

 

I also have an external (stand-alone) formatter for US-Doubler ED diskettes, never used it, since I do not own the USD...

Those tools are for formatting disks with the faster UltraSpeed 5:1 skew, allowing disks to be read in sequential forward order (1-18, 1-26) at maximum speed without a track buffer. These disks will read slower in sequential order at 1x SIO in a stock 1050.

[+Nezgar]

I have a tangential question about formatting skews that I keep running into when one of these discussions is NOT going on, lol.

 

So I have two Atarimax Happy boards - is there a preferred utility for format them for DD use with a “better” or faster sector interleave? The SpartaDOS X FORMAT utility works fine for single and double-density formatting, in either Atari DOS or Sparta formats, but it absolutely refuses to allow formatting with anything but a normal interleave. I just get a message that the drive doesn’t support UltraSpeed or HIgh Speed. This is true even if I run the Happy utility to ensemble US emulation on the drive. (Pardon if I may be mis-remembering the exact terminology involved - I’m not at my Atari right now).

 

As you found, the Happy 1050 "UltraSpeed Emulation" does not support the 'format custom' $66 command. You have no way to change the skew the drive uses when formatting. This is was generally not an issue since the skew is irrelevant on a buffered drive.

 

Enabling "UltraSpeed Emulation" from the happy utility menu achieves 2 things:

- on revision 1 boards, it actually enabled use of the USD UltraSpeed SIO protocol, otherwise only happy's own "Warp Speed" is supported

- on revision 2 boards, UltraSpeed is enabled by default, with buffered reads, and non-buffered writes. Enabling from the happy menu merely enables 'fast writes' which works around the bug that corrupts 3 bytes of any sector written at single density with UltraSpeed and slow writes.

 

Anyhow.. I think you indirectly asked about this before: http://atariage.com/forums/topic/257578-happy-810-resources/?p=4127516

 

Specifically, using the HAPPY.COM from RealDOS, or the USMULTR programs referenced in the above link, they upload a relocated US Doubler ROM to the Happy RAM, and now it will function identically to a real US doubler until next power cycle/reset, including support for the $66 custom format command - allowing format of UltraSpeed skew disks from SDX. Also, disks formatted with 'Normal' skew option will work a little faster (about 10%) in stock drives due to a slightly optimized default sector skew in the ROM. (What I call the 'Chicago' format http://atariage.com/forums/topic/279363-the-atari-810-revision-b-rom-has-finally-been-found-dumped/?p=4113840 )

[+DrVenkman]

Thank you, Nezgar!

 

I obviously need to spend more time understanding the ins and outs of the various drive upgrades, capabilties, etc. Thank you for laying it out clearly!

[+DrVenkman]

Specifically, using the HAPPY.COM from RealDOS, or the USMULTR programs referenced in the above link, they upload a relocated US Doubler ROM to the Happy RAM, and now it will function identically to a real US doubler until next power cycle/reset, including support for the $66 custom format command - allowing format of UltraSpeed skew disks from SDX.

 

Following up on this - I tried the USMULTR.COM program on the ATR in your link and it worked great! SDX formatted the disks I tried just fine with the UltraSpeed skew. Thanks!

 

(From the Happy 810 Resources thread, I had gotten as far as downloading the ATR and then forgetting I had it! Gah, it's hard to keep learning new stuff at my age. )

[+bob1200xl]

...or remembering stuff you used to know.

 

I don't think you can run a format such as is shown at the beginning of this thread - 1,2,3,4,5,6....sector format cannot be read with any known hardware at speed. It will read a sector, wait 1/5 of a second, read the next sector, and on like that. Not helpful.

 

How did we determine that skew format? Is this something new?

 

Bob

[+Nezgar]

...or remembering stuff you used to know.

 

I don't think you can run a format such as is shown at the beginning of this thread - 1,2,3,4,5,6....sector format cannot be read with any known hardware at speed. It will read a sector, wait 1/5 of a second, read the next sector, and on like that. Not helpful.

 

How did we determine that skew format? Is this something new?

 

Bob

 

Hi Bob, Let me try to clarify.. I know this topic is pretty esoteric!!

 

You are absolutely correct that a disk formatted with a sequential skew as you indicated would be very slow in practice. What I'm talking about above is the optimal order of reading sectors from a disk at UltraSpeed depending on the skew that was used when formatting.

 

Stock Atari 810 Rev C, Stock Atari 1050, and similar drives will format single density disks as follows:

1 3 5 7 9 11 13 15 11 2 4 6 8 10 12 14 16 18 (9:1)

 

The computer will request sectors sequentially, so for example starting with the computer requesting sector #1, by the time the drive has read it and finished transmitting it to the computer at 19607bps (according to Altirra HW reference manual), the disk will have continued rotating almost 180°, and by the time the computer requests sector #2, the placement of sector #2 8 sectors after #1 allows it to begin reading it with little delay.

 

When operating at UltraSpeed (52631bps *according to Altirra HW reference manual) that sector transmission to the computer can finish 2.7x faster, allowing the computer to then request sector #2 after it's rotated less than 90°, but still having to wait until 180° for sector #2 to come around.

 

A disk formatted with UltraSpeed skew puts sector #2 only 5 sectors after sector #1, instead of 9, allowing it to 'come around sooner' close to when it would be requested at UltraSpeed SIO rates. This allows a sequential read from the atari, ie 1,2,3,4, to be satisfied on the 5th sector after the last:

4 8 12 16 1 5 9 13 17 2 6 10 14 18 3 7 11 15 (5:1) (as per SpartaDOS Construction Set manual)

 

As Rybags noted earlier, this would also mean slower reading when running at 1x SIO due to the next sector always being requested too late, requiring a full rotation plus some for the sector to come around again.

 

Now, knowing the US Doubler can pick up the next sector as early as 5 sectors after the last, then on a disk formatted in a stock Atari 810/1050 we can see that the 5th sector after #1 is 11, 5th after 11 is 4, then 14, 5, and so on:

1 3 5 7 9 11 13 15 11 2 4 6 8 10 12 14 16 18 (9:1)

Then we can then read the disk most effectively at UltraSpeed by telling the computer to request the sectors in this order:

1 11 4 14 5 15 8 18 9 2 12 3 13 6 16 7 17 10

 

I hope this makes things more clear? All of the other sector reading orders in the first post are also optimized for disks originally formatted with various skews. Otherwise a US Douber in UltraSpeed mode will not effectively read a standard skew disk any faster when sectors are only requested sequentially by the computer. I should post a video of CopyMate 4.3/4.4 doing this, I think it's pretty cool to see in action...

[phaeron]

I had no idea that there was copy software that did this. This should work for reading any interleave at top speed for the current transfer mode, as it's basically just requesting the next closest unread sector. The gotcha is that you have to know the physical sector order on disk, which is not easy or fast to determine from the computer, and also need to be reading most or all of the track.

 

It would have been better if drives could just cache sectors as they pass by while looking for specific ones, but the FDC design prevents it. Supposedly GS/OS on the Apple IIgs does this as its floppy hardware is simpler and doesn't get in the way.

[+Nezgar]

Made a quick video to demonstrate.

 

 

PS: Hope you like the background POKEY tunes.

[HiassofT]

I implemented a similar solution in my sector copier for 1050 Turbo drives back in the late 1980es. I don't think I ever released it, was also hardwired to the 320k RAM upgrade I had back then.

 

Implementing out-of-order reading/writing eg in atarixfer wouldn't be a big problem, but determining the sector layout may be tricky. Either the user has to specify it or we'd need to start guessing.

 

For PC usage the opposite use case, reading a disk formatted with high-speed skew at standard speed would be the more interesting one. Standard 16550 serial ports can't operate at the US Doubler/1050 Turbo rates and the XF551 highspeed protocol is too timing critical to be used with USB serial adapters. And as the penalty for reading a disk with high speed skew at standard speed is usually a lot higher than the opposite, which will just fall back to standard speed throughput, it may he beneficial to some users.

 

This leads to an interesting question though. With XF551 and 1050 Turbo highspeed skew is fixed by the drive's firmware, but on US Doubler the sector layout is under software control so there could be some variations of actual sector layout, depending on the formatter.

 

Do you have any info on this, i.e. which parameters / sector layout do/did the USD formatters use?

 

PS: I'm not sure yet if I'm actually going to add this in atarixfer, that could be a rather deep rabbit hole, just thinking about scenarios for now.

 

so long,

 

Hias

[_The Doctor__]

that is unless you use the AtariMax SIO2PC usb version, it is timing critical hardware handles XF and others just fine.

Edited March 10, 2019 by _The Doctor__

[HiassofT]

that is unless you use the AtariMax SIO2PC usb version, it is timing critical hardware handles XF and others just fine.

The XF551 protocol also works fine with 16550 serial ports (not sure about Win/OSX, but at least on Linux with the AtariSIO kernel driver it's no problem).

 

When you have the AtariMax USB interface and APE/ProSystem you are pretty well covered, it's the other solutions that are interesting though, and for them using the AtariMax USB interface is not an option :-)

 

Dealing with 1050-2-PC mode and optimizing for sector skew is a pretty niche application nowadays. I guess not too many people do real floppy disk transfers nowadays and it could well be that the time spent on optimizing this will be longer than the total time saved by disk transfers :-)

 

so long,

 

Hias

[_The Doctor__]

Yep, I use the nick kennedy modified rs232 version, the atarimax rs232 version, and the atarimax usb version. depends on the machines I using at the time...

I am terrible, I have hundreds of disks...

[+CharlieChaplin]

I implemented a similar solution in my sector copier for 1050 Turbo drives back in the late 1980es. I don't think I ever released it, was also hardwired to the 320k RAM upgrade I had back then.

 

Hias

 

Hmmm, the copy program Track Copier by Arndt Bär also allows reading+writing a disk with turbospeed, no matter if it is formatted with a) normal/slow 1050 layout or b) fast/turbo layout. It will always read+write with turbospeed (68kbaud) as long as a turbo enhancement is present. All other Turbo 1050 drive copy programs do require the fast turbo format (skew) to read/write with turbospeed. So it looks like Arndt implemented the sector layout of both standard/slow 1050 and fast/turbo 1050 into this copy program. The name of the program is a bit misleading, since the Turbo 1050 does not read whole tracks at once (it does not have a trackbuffer like Happy or Speedy), but it simply shows onscreen when a complete track has been read or written after a while, instead of displaying single sectors...

 

Think I have seen a similar thing with the XF551 sector copy program on the german XF-Tools diskette (the file has a size of approx. 201 sectors, it includes a gamedos for 180k and 360k, a title-picture with DLI's, a built-in manual/text, some music, the sector copy program, etc.; would like to have only the sectorcopy program and the gamedos as stand-alone files). They implemented a sector table for standard and fast format and therefore also a 90k and 130k disk with standard format can be read+written very fast (with 38k4 Baud I think), as well as 180k and 360k...

 

(Both the Turbo 1050 and Hyper-XF can implement their Speeder via page 1 or page 6 and both use sector-skewing / sector-interleave; when one uses the original/unchanged Track Copier with a Hyper-XF, one can also read with ultraspeed, no matter if the disk uses standard/slow or fast disk layout - but only reading works, formatting and writing will fail...)

Edited March 10, 2019 by CharlieChaplin

[+bob1200xl]

Yeah... what you said.

 

Some of the original layouts showed 1,2,3,4,5...

 

That was my concern.

 

Bob

 

 

Hi Bob, Let me try to clarify.. I know this topic is pretty esoteric!!

 

You are absolutely correct that a disk formatted with a sequential skew as you indicated would be very slow in practice. What I'm talking about above is the optimal order of reading sectors from a disk at UltraSpeed depending on the skew that was used when formatting.

 

Stock Atari 810 Rev C, Stock Atari 1050, and similar drives will format single density disks as follows:

1 3 5 7 9 11 13 15 11 2 4 6 8 10 12 14 16 18 (9:1)

 

The computer will request sectors sequentially, so for example starting with the computer requesting sector #1, by the time the drive has read it and finished transmitting it to the computer at 19607bps (according to Altirra HW reference manual), the disk will have continued rotating almost 180°, and by the time the computer requests sector #2, the placement of sector #2 8 sectors after #1 allows it to begin reading it with little delay.

 

When operating at UltraSpeed (52631bps *according to Altirra HW reference manual) that sector transmission to the computer can finish 2.7x faster, allowing the computer to then request sector #2 after it's rotated less than 90°, but still having to wait until 180° for sector #2 to come around.

 

A disk formatted with UltraSpeed skew puts sector #2 only 5 sectors after sector #1, instead of 9, allowing it to 'come around sooner' close to when it would be requested at UltraSpeed SIO rates. This allows a sequential read from the atari, ie 1,2,3,4, to be satisfied on the 5th sector after the last:

4 8 12 16 1 5 9 13 17 2 6 10 14 18 3 7 11 15 (5:1) (as per SpartaDOS Construction Set manual)

 

As Rybags noted earlier, this would also mean slower reading when running at 1x SIO due to the next sector always being requested too late, requiring a full rotation plus some for the sector to come around again.

 

Now, knowing the US Doubler can pick up the next sector as early as 5 sectors after the last, then on a disk formatted in a stock Atari 810/1050 we can see that the 5th sector after #1 is 11, 5th after 11 is 4, then 14, 5, and so on:

1 3 5 7 9 11 13 15 11 2 4 6 8 10 12 14 16 18 (9:1)

Then we can then read the disk most effectively at UltraSpeed by telling the computer to request the sectors in this order:

1 11 4 14 5 15 8 18 9 2 12 3 13 6 16 7 17 10

 

I hope this makes things more clear? All of the other sector reading orders in the first post are also optimized for disks originally formatted with various skews. Otherwise a US Douber in UltraSpeed mode will not effectively read a standard skew disk any faster when sectors are only requested sequentially by the computer. I should post a video of CopyMate 4.3/4.4 doing this, I think it's pretty cool to see in action...

[+Nezgar]

Implementing out-of-order reading/writing eg in atarixfer wouldn't be a big problem, but determining the sector layout may be tricky. Either the user has to specify it or we'd need to start guessing.

Most, if not all of the common skews used have been well documented. The ones determined by firmware are all documented in the Altirra hardware reference manual. US Doubler UltraSpeed skews are documented in the US Doubler & SpartaDOS construction set manuals.

 

For PC usage the opposite use case, reading a disk formatted with high-speed skew at standard speed would be the more interesting one.

By counting the 8th or 9th sector after each sector to determine the next sector to read at standard speed. Atari's standard 810 Revision C skew & Stock 1050 skew are always 9 sectors after the previous, even when crossing the end-of-track gap. So, comparing with Atari's standard skew as a reference:

 

1 3 5 7 9 11 13 15 17 2 4 6 8 10 12 14 16 18 (9:1)

and the USD skew:

4 8 12 16 1 5 9 13 17 2 6 10 14 18 3 7 11 15 (5:1)

 

You can read a USD skew on a standard speed drive in this order:

4 2 8 6 12 10 16 14 1 18 5 3 9 7 13 11 17 15

 

The 'Chicago' skew used by the 810 Archiver, 810 Happy, 1050 US Doubler place the next sector 9 sectors after the previous, except when crossing the end-of-track gap, where it is 8. The gap is large enough that the 8th sector can be read just in time, resulting in 9.6% faster speed from my RWTEST calculations.

 

17 15 13 11 9 7 5 3 1 18 16 14 12 10 8 6 4 2 GAP (9:1 with GAP, but 8:1 when crossing gap)

 

This speed increase can be attained on disks standard formatted disks by reading sectors in reverse order which achieves the same effect, when not crossing the gap.

 

With this in mind, reading a USD skew disk a little faster in a stock drive can also be achieved by reading the 8th sector after the gap:

 

17 15 13 11 9 7 5 3 1 18 16 14 12 10 8 6 4 2

 

 

This leads to an interesting question though. With XF551 and 1050 Turbo highspeed skew is fixed by the drive's firmware, but on US Doubler the sector layout is under software control so there could be some variations of actual sector layout, depending on the formatter.

 

Do you have any info on this, i.e. which parameters / sector layout do/did the USD formatters use?

All USD UltraSpeed formatters that I have looked at so far lay down the sector order as provided in the US Doubler & SpartaDOS construction set manuals. I've yet to see one that deviates: (Single Density example only here for simplicity) If there is any variation, I expect it might only be in the starting sector order... ie whether sector 1 is placed at the beginning, sector 4th sector (as documented) or the 9th sector (180 degrees).

 

4 8 12 16 1 5 9 13 17 2 6 10 14 18 3 7 11 15 (5:1)

 

The 1050 TURBO optimizes this a little bit further by accounting for the extra delay when crossing the end-of-track gap, where it is 4 sectors after instead of 5, I don't have my notes handy with the actual sector order to document authoritatively, I'll look it up later, it's also missing from the Altirra Hardware Reference Manual 2018-08-12 edition. But I do recall that the disks formatted by a 1050 Turbo could be read/written at 52K Ultraspeed without blowing revs on a US Doubler.

[HiassofT]

Most, if not all of the common skews used have been well documented. The ones determined by firmware are all documented in the Altirra hardware reference manual. US Doubler UltraSpeed skews are documented in the US Doubler & SpartaDOS construction set manuals.

Thanks a lot for the detailed info!

 

I already read through the info in the Altirra hardware refereence manual but didn't know about the info in the US Doubler and Spartados manuals - I'll have a closer look at them too.

 

so long,

 

Hias

[+Nezgar]

I already read through the info in the Altirra hardware refereence manual but didn't know about the info in the US Doubler and Spartados manuals - I'll have a closer look at them too.

There's not a whole lot to glean there, its just 1 page (page 150, PDF page 161) where the sector orders are stated:

https://archive.org/details/SpartaDOS_Construction_Set_1985_ICD/page/n161

 

The takeaway is basically that there's really only 2-3 common skew possibilities if one was to write a program to find the optimal order for any given disk. The fastest method would probably be to time the latency between the first few sectors, rather than an entire track to match up with well known skews.

[Jeffrey Worley]

On 3/9/2019 at 2:15 AM, Nezgar said:

The 1050 US Doubler is UltraSpeed capable, but does not have a track buffer. So generally, you will only get the maximum read/write speed if the disk is formatted with the special 5:1 UltraSpeed skew. However, by creatively rearanging the order you read the sectors in software, you can read regular disks at a higher speed as well. This probably also applies just as well to the Super Archiver, which normally behaves very similarly to a US Doubler.

I've noticed the sector copier CopyMate 4.3/4.4 does exactly this. It will try reading sectors in 3 different orders until the maximum speed is achieved. I've used RespeQt to analyze the sector orders of it's three different detection modes out of curiosity. Since I have a selection of disks formatted in various densities with various upgrades, I can determine which are optimal for each format. CopyMate 4.4 only supports single and double density, not ED.

									Reading order best for...
SD	1, 11, 4, 14, 5, 15, 8, 18, 9, 2, 12, 3, 13, 6, 16, 7, 17, 10	Disk formatted by Stock 810 or 1050
	1, 10, 17, 7, 16, 6, 13, 3, 12, 2, 9, 18, 8, 15, 5, 14, 4, 11	Disk formatted by USD, Speedy, Archiver (Chicago layout)
	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18	Disk formatted in USD UltraSpeed Skew

DD	1, 12, 5, 16, 9, 2, 13, 6, 17, 10, 3, 14, 7, 18, 11, 4, 15, 8	DD Disk formatted by USD
	1, 11, 15, 6, 10, 14, 5, 9, 13, 4, 8, 18, 3, 7, 17, 2, 12, 16	DD Disk formatted by Happy, Speedy or Percom
	1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18	DD Disk formatted in UltraSpeed/Turbo 1050 Skew mode

The other utility that I know of that reads sectors 'out of order' for speed using a US Doubler is the SpartaDOS SCOPY command. (Specifically "SCOPY 9/16/86 ©1994 FTe") You manually tell it whether the source or destination disk is 'normal' by not specifying the /U (UltraSpeed) command line switch, which is supposed to read 'regular' disks in a fast way, or by specifying /U it will read sequentially... or so I thought:

SD	9, 2, 12, 3, 13, 6, 16, 7, 17, 10, 1, 11, 4, 14, 5, 15, 8, 18	Normal: ~1/2 rotation offset, slow reading Chicago skew disks (Those formatted by USD, Archiver, Speedy)
	5, 6, 7, 12, 13, 14, 15, 1, 2, 3, 8, 9, 10, 11, 16, 17, 18, 4	/U: ~1/2 rotation offset, [no SD turbo 1050 formatted disk handy to test]

DD	8, 15, 4, 11, 18, 7, 14, 3, 10, 17, 6, 13, 2, 9, 16, 5, 12, 1	Normal: Only fast with DD disks formatted by USD, slow reading Happy/Speedy/Percom
	2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 1	/U:  reads 1st sector last, slow reading Turbo 1050 std formatted disk

ED	15, 4, 18, 5, 19, 8, 22, 9, 23, 12, 26, 13, 2, 16, 3, 17, 6, 20, 7, 21, 10, 24, 11, 25, 14, 1	Normal: Stock 1050 or USD std ED disk fast, Turbo 1050 ED slow
	5, 6, 7, 12, 13, 14, 15, 20, 21, 22, 23, 1, 2, 3, 8, 9, 10, 11, 16, 17, 18, 19, 24, 25, 26, 4	/U: odd order maybe reads a USD skew disk faster?

It's odd that if you format a non-ultraspeed single density disk with a US Doubler, that SCOPY cannot "read it's own writing' at high speed... (Both being ICD products and all...)

 

Anyhow.. this information could be useful for anyone writing or hacking a sector copier, or incorporate into PC based 10502PC type programs like AtariSIO/atarixfer to read disks directly from a US Doubler/Super Archiver drive without a track buffer at high speed after determining the fastest option... Less overall spinning disk time and reducing wear can be valuable in imaging marginal disks.

 

Determining which reading order is fastest for a given disk can also suggest the group of drive/speeders that could have originally formatted the disk.

Scopy CAN read it's own writing in Ultraspeed mode.  If a disk is in regular skew, you give no arguments to the command:

 

Scopy A: O:TEST.SCP 

 

Will copy the source disk using ultraspeed mode even though it is in regular skew, to the file in the O: drive named TEST.SCP

 

If the source disk is a US skew formatted disk, you give the /U argument, thus:

 

Scopy A: /U O:TEST.SCP

 

Scopy will copy then at high speed.

 

To copy a non-skewed disk to a skewed destination:

 

SCOPY A: B: /U

 

To copy a Skewed disk to a non-skewed destination:

 

Scopy A: /U B:

 

To copy a skewed disk to a skewed destination:

 

SCOPY A: /U B: /U

 

Basically, Scopy assumes the source is not skewed and reads the sectors out-of-order to produce a high-speed read that sounds as if the source is actually skewed.  It cannot detect if the source is actually skewed and thus if the source is skewed and it is not told so, it reads the sectors out-of-order, producing a non-skewed input that reads as though it were a non-skewed disk being read in proper order.  It is weird but that's the dope.

 

 

Best,

 

Jeff

Edited June 14, 2020 by Jeffrey Worley

[+Nezgar]

@Jeffrey Worley My point was that SCOPY when reading in a disk formatted with "regular" interleave without the /U switch, it actually requests the sector numbers in a non-sequential order to achieve the fastest read of the disk with a non-track buffered US doubler at 54Kbps. (Similar to the behaviour of Copymate 4.3 in the video example I posted above) The problem is it assumes the disks were formatted using Atari's "Standard" 9:1 interleave for single density (for instance) which works great, but the US Doubler ROM formats single density with a slightly tighter interleave compared to a stock Atari 810 or 1050, and SCOPY slows way down trying to read a such a disk formatted with a US doubler without /U.

 

So, that is where my comment comes from that SCOPY cannot "read it's own writing' at high speed". Odd both being ICD products and all.

 

(ps: no need to quote the entire long post... 1 line would have probably been sufficient)