Myarc HFDC Bug?

[+dhe]

I was able to run CFORM today, after install the win-driver today under GeneveOS7.

 

One trick it took me a minute to discover, the sub-menus don't pull down on a roll over, nor hitting enter, I had to go to the menu option - like Info and press the down arrow.

[Brufnus]

Does any of you know of any termination problems with the HFDC?

Many years ago I had absolutely no luck formatting an ST-251 until I became desperate and  tried shorting some of the pins on the terminator block, which would of course seem like a bad idea  - it turned out it formatted perfectly when I shorted the two outermost legs, and I've used it ever since, formatting it like that if necessary.

 

Now I have problems with a Seagate ST-1100 though (mentioned here before). I can format with 1072 cylinders with 1 head only if it's terminated properly, i.e. another drive with termination at the end of the cable and no termination on the 1100. If I terminate it however, I can't format with heads 1 and 2, but the remaining 6 heads seem to format properly, judging from the speed (with CFORM, MDM5 exits as soon as it try head 1).

 

I've tried different combinations and it behaves differently depending on these combinations - so I'm convinced it's simply a termination issue, not a fault with the drive or the HDFC as such (nor the other HFDC I possess, by the way).

Anyway, when formatted with 1 head and 1072 cylinders, it works flawlessly and I've even been successful with booting MDOS off of it. I would really appreciate being able to use all 8 heads, however.  

Edited July 31, 2022 by Brufnus

[+InsaneMultitasker]

8 minutes ago, Brufnus said:

Does any of you know of any termination problems with the HFDC?

Many years ago I had absolutely no luck formatting an ST-251 until I became desperate and  tried shorting some of the pins on the terminator block, which would of course seem like a bad idea  - it turned out it formatted perfectly when I shorted the two outermost legs, and I've used it ever since, formatting it like that if necessary.

 

Now I have problems with a Seagate ST-1100 though (mentioned here before). I can format with 1072 cylinders with 1 head only if it's terminated properly, i.e. another drive with termination at the end of the cable and no termination on the 1100. If I terminate it however, I can't format with heads 1 and 2, but the remaining 6 heads seem to format properly, judging from the speed (with CFORM, MDM5 exits as soon as it try head 1).

 

I've tried different combinations and it behaves differently depending on these combinations - so I'm convinced it's simply a termination issue, not a fault with the drive or the HDFC as such (nor the other HFDC I possess, by the way).

Anyway, when formatted with 1 head and 1072 cylinders, it works flawlessly and I've even been successful with booting MDOS off of it. I would really appreciate being able to use all 8 heads, however.  

Do both HFDCs have this same problem with the ST1100?  If yes, I would suspect the cable or the termination resistor is at fault.  Be sure that the termination is installed with pin 1 in the correct location.  As for shorting pins on the terminator block, if this works, I would suspect a broken terminator or the wrong resistance.  I do not see how formatting the remaining 6 heads works, unless the head select is actually formatting (or reformatting) the wrong head due to the termination problem.

 

If one HFDC works and the other fails, I would initially suspect U17, U18, or U9.  These three chips are part of the head select circuitry.  The 9234 is less likely to go bad but it too drives the head selection.  From a repair perspective, the 7406, LS373 and LS240 are the most common chips to go bad due to removing/connecting the cable while the drive and/or HFDC are turned on can result in damaged circuitry, especially when the cable and card are not keyed to prevent shorting out the pins.  I am not sure if the damage is instantaneous or over a period of time, but it does happen.  

[Brufnus]

Hi Tim,

 

Thanks for your reply. Yes, both HFDC's have trouble formatting the drive; however, they both work fine with other drives (and the same cables). I've also tried a different terminator block on the ST1100 (from another ST1100 PCB), but that makes no difference either.

I've never had any issues with termination as such, but I have to admit I've never gained the knowledge with respect to which orientation they have. I suppose pin 1 on the drive is at pin 1 at the cable as well, but how do I tell on the resistor itself?

 

I don't recall ever having disconnected the cables while any of it was powered on, but on the other hand I won't exclude the possibility there's a flawed chip somewhere. However, since both HFDC's work with other drives, I guess it has something to do with the ST1100 itself, or perhaps the HFDC termination which is flawed but works with other drives despite of this. I know the termination is important, but I've often witnessed it working nonetheless with improper termination (and sometimes it even ONLY worked like that...) c".)

 

You're right; the head select could be erroneous of course; I have no way to tell if it's actually formatting the heads the controller tries to access. I MAY format with 8 heads with the termination installed though, but then the available sectors doesn't match what it should be, since CFORM runs fast through the upper 7 heads (as if it merely skips the tracks rather than actually formatting them). I can't use MDM5 in this case, since it merely formats head 0 and then exit as soon as an attempt on head 1 is made. 

And lately, when I e.g. create a directory after such a format, it only creates garbage characters in the directory name and throws an error message.

 

There should be termination on the HFDC's as well, right? Can I somehow verify these - resistance measuring or something? 

Edited July 31, 2022 by Brufnus

[+InsaneMultitasker]

The HFDC has not on-board termination, beyond whatever the chips provide.  I have encountered instances where adding a floppy drive caused the hard drive to fail with symptoms similar to a termination issue, and the fix was to replace the LS240 and/or LS374.  Since both HFDCs respond in the same manner, I am inclined to think the HFDCs are ok. 

 

Please refresh my memory - are there other hard drive models that are working properly with one or both HFDCs?  

 

Also, we must keep in mind that you are trying to format an ST1100 (9 heads) with unmodified HFDCs, correct? For the unmodified HFDC, the write precomp output is tied to the same pin as head select line 3.  This is almost certainly interfering with the write operations, which may also explain why only one head works.  If the write precomp is selecting HEAD3, the only valid value for the drive would be 9.  

 

The heads are represented base 0, so we select the 9 heads as 0,1,2,3,4,5,6,7,8:

 

HHHH

3210 - head select bit position

0000 - head 0

0001 - head 1

0010 - head 2

0011 - head 3

0100 - head 4

0101 - head 5

0110 - head 6

0111 - head 7

1000 - head 8

 

If enabling write precomp sets the bit for H3, the drive will only see 1000 and 0000 as valid heads.  0000 corresponds to the first head, which is what you said is formatting.  In MDM5, try selecting 7 heads with a write precomp of 0.   Per the specs, these larger drives provide their own write precomp. Setting to 0 might hold H3 low, allowing you to format 0,1,2,3,4,5,6,7.  This is conjecture until proved.   

 

ST1100 termination appears to be 100ohm.  See this link for drive details:

https://stason.org/TULARC/pc/hard-drives-hdd/seagate/ST1100-84MB-3-5-HH-MFM-ST506.html

 

[+InsaneMultitasker]

@Brufnus I need to confirm if it is "reduced write current" or "write precomp" that should be set to 0.  I will attempt to locate my notes and schematics later today. 

[Brufnus]

6 minutes ago, InsaneMultitasker said:

The HFDC has not on-board termination, beyond whatever the chips provide.  I have encountered instances where adding a floppy drive caused the hard drive to fail with symptoms similar to a termination issue, and the fix was to replace the LS240 and/or LS374.  Since both HFDCs respond in the same manner, I am inclined to think the HFDCs are ok. 

 

Please refresh my memory - are there other hard drive models that are working properly with one or both HFDCs?  

 

Also, we must keep in mind that you are trying to format an ST1100 (9 heads) with unmodified HFDCs, correct? For the unmodified HFDC, the write precomp output is tied to the same pin as head select line 3.  This is almost certainly interfering with the write operations, which may also explain why only one head works.  If the write precomp is selecting HEAD3, the only valid value for the drive would be 9.  

 

The heads are represented base 0, so we select the 9 heads as 0,1,2,3,4,5,6,7,8:

 

HHHH

3210 - head select bit position

0000 - head 0

0001 - head 1

0010 - head 2

0011 - head 3

0100 - head 4

0101 - head 5

0110 - head 6

0111 - head 7

1000 - head 8

 

If enabling write precomp sets the bit for H3, the drive will only see 1000 and 0000 as valid heads.  0000 corresponds to the first head, which is what you said is formatting.  In MDM5, try selecting 7 heads with a write precomp of 0.   Per the specs, these larger drives provide their own write precomp. Setting to 0 might hold H3 low, allowing you to format 0,1,2,3,4,5,6,7.  This is conjecture until proved.   

 

ST1100 termination appears to be 100ohm.  See this link for drive details:

https://stason.org/TULARC/pc/hard-drives-hdd/seagate/ST1100-84MB-3-5-HH-MFM-ST506.html

 

Yes, I actually considered that, too - either boot off another HDD or from RAMdisk and detach the floppy entirely.

 

I did cover pin 2 on the 34-pin connector on the drive in order to avoid the H3 issue. I've also tried setting write precomp. to 0 in CFORM, but I don't remember if I did it in MDM5 as well. I'll that as well as your suggestion about selecting only 7 heads.

 

I've used many different drives on the HFDCs, all have been working fine - ST125, ST225, ST251, Rodime 3855, IBM 0665, Olivetti, Tandon 262, NEC D5126H and even a Kyocera RLL drive.

 

Another thing is - I do have a couple of ISA controllers, including two Seagate 11M, so if I could find an old pre-486 in my collection, I could test it there with SpeedStor or something.

 

Anyway, that 100ohm termination resistance should be measured between the individual legs?

You're right; these drives seem to handle write precomp internally thus freeing pin 2 for the head select instead. I've never used more than 6 heads on the HFDCs, however.

[Brufnus]

9 minutes ago, InsaneMultitasker said:

@Brufnus I need to confirm if it is "reduced write current" or "write precomp" that should be set to 0.  I will attempt to locate my notes and schematics later today. 

Okay, thanks - actually I think it's reduced write current, now that you mention it. But I could set both to 0; I believe the newer drives will handle both just fine.

[+InsaneMultitasker]

1 minute ago, Brufnus said:

Okay, thanks - actually I think it's reduced write current, now that you mention it. But I could set both to 0; I believe the newer drives will handle both just fine.

I agree. try setting both to 0.  Per the ST1100 spec:

 

Write Precompensation
---------------------
RLL write data must conform to the conventional RLL encoding rules
and must write precompensated by 12 ns per the industry standard
RLL data write precompensation rules. Write precompensation is re-
commended on all tracks; however, write precompensation must be
performed on tracks greater than Track 300.
 

No write precompensation is required for MFM.

 

Pin 2 in the command cable (Head Select 2(3)) is defined as Reduced
Write Current in the Seagate St412 interface. The SWIFT controls
write current internally, and thus the line is not needed for this
function and is used for head selection.

 

[Brufnus]

8 minutes ago, InsaneMultitasker said:

I agree. try setting both to 0.  Per the ST1100 spec:

 

Write Precompensation
---------------------
RLL write data must conform to the conventional RLL encoding rules
and must write precompensated by 12 ns per the industry standard
RLL data write precompensation rules. Write precompensation is re-
commended on all tracks; however, write precompensation must be
performed on tracks greater than Track 300.
 

No write precompensation is required for MFM.

 

Pin 2 in the command cable (Head Select 2(3)) is defined as Reduced
Write Current in the Seagate St412 interface. The SWIFT controls
write current internally, and thus the line is not needed for this
function and is used for head selection.

 

That settles it. I'll try these things then - measure resistor resistance, disconnect the FDD, use MDM5 with 7 heads, 0 at RWC and precomp, and see how it turns out.

 

Thanks for your help!!  

[+InsaneMultitasker]

The attached scanned instructions (for the 16-head modification) state to set write precomp to 0; RWC is disabled and the control line is repurposed for HEAD 3.

 

For the unmodified HFDC, the circuit and ST1100 specs suggest we can set RWC to 0 to hold "HEAD 3" control line low. Total heads must be 8 or fewer.  I am not 100% certain that the HFDC EPROM DSR respects the RWC setting but it is worth a try!

 

[Brufnus]

12 hours ago, InsaneMultitasker said:

The attached scanned instructions (for the 16-head modification) state to set write precomp to 0; RWC is disabled and the control line is repurposed for HEAD 3.

 

For the unmodified HFDC, the circuit and ST1100 specs suggest we can set RWC to 0 to hold "HEAD 3" control line low. Total heads must be 8 or fewer.  I am not 100% certain that the HFDC EPROM DSR respects the RWC setting but it is worth a try!

 

Thanks, Tim!

 

I tried once again when I came home from work yesterday, although I haven't had the FDD disconnected yet there's no difference yet.

Anyway, I could eventually get rid of the H3 signal either by modifying a cable or by covering pin 2 (which I already tried).

 

I just got a theory, though. Since CFORM quickly runs through heads 0 and 1 as if it encounters an error and simply skip the tracks, and then seem to format properly from heads 2+, perhaps it could be something with the H0 signal, which for some reason doesn't reach the drive properly?

 

Perhaps I could modify a cable so that H0 and H3 are swapped, and H1 and H2 as well, in order to reverse the entire head select, so that head 9 effectively becomes head 0 and head 0 isn't used at all. One thing that concerns me with respect to modifying (one of the) HFDCs is that it probably could cause problems with other drives whidh indeed require the RWC signal...? c".) But as the text suggests, one would have to forward that signal by some other means.

 

 

 

Edited August 1, 2022 by Brufnus

[+InsaneMultitasker]

You could modify the HFDC and yes, it would affect other drives that require the signal.  I don't think swapping the pins around is going to be of much benefit, and since other hard drives work, H0 should be just fine on the card.   If you confirm that the drive works on an old PC card, then I you could try some of your other ideas.   however, if it were me at this point in time (personal opinion!) and one model wasn't working, I would set the drive on a shelf and use other models that are known to work   

[Brufnus]

9 minutes ago, InsaneMultitasker said:

You could modify the HFDC and yes, it would affect other drives that require the signal.  I don't think swapping the pins around is going to be of much benefit, and since other hard drives work, H0 should be just fine on the card.   If you confirm that the drive works on an old PC card, then I you could try some of your other ideas.   however, if it were me at this point in time (personal opinion!) and one model wasn't working, I would set the drive on a shelf and use other models that are known to work   

I agree, the card itself has to be sound. I do feel bad about giving up though, since it's a rather high capacity and silent drive, which also seem to be the fastest I've ever had for the Geneve. But perhaps you're right... 

for what it's worth, I really appreciate all your efforts and suggestions, thanks a lot Tim! You're indeed an asset for our tiny community!

[+InsaneMultitasker]

55 minutes ago, Brufnus said:

I agree, the card itself has to be sound. I do feel bad about giving up though, since it's a rather high capacity and silent drive

Others may yet chime in, and maybe there are ideas that have not been explored.  That is what makes the forums 'fun'. 

 

[Brufnus]

13 minutes ago, InsaneMultitasker said:

Others may yet chime in, and maybe there are ideas that have not been explored.  That is what makes the forums 'fun'. 

 

Yes, you're right. Actually I just got another - simpler - idea. I have a strong feeling it's head 1 or it's platter surface that causes the issue.

A way to test that could be to swap H0 and H3 only; then when it sends 0000 it selects head 0, then when it sends 0001 it thinks it's addressing head 1, but reversed it would give a 1000 and select head 9 instead - then the rest of the heads will receive the proper signals as usual.

I'll modify a cable like that and see what happens - if I blow up the entire house, I'll know what I did wrong.  

 

[Brufnus]

Oh, by the way... I have a drive from which I can boot and list the contents of the root directory - and even execute files from the drive, also from subdirs, IIRC... but I ca'nt read the contents of the subdirs, that gives a device error.

Is there any way to fix this?

Edited August 1, 2022 by Brufnus

[Brufnus]

For what it's worth, I just managed to salvage a Tandon 262 which died on me some months ago. Much can be said about MFM drives, but they¨re pretty sturdy, that's for sure!  

[Brufnus]

On 7/31/2022 at 9:22 PM, InsaneMultitasker said:

The HFDC has not on-board termination, beyond whatever the chips provide.  I have encountered instances where adding a floppy drive caused the hard drive to fail with symptoms similar to a termination issue, and the fix was to replace the LS240 and/or LS374.  Since both HFDCs respond in the same manner, I am inclined to think the HFDCs are ok. 

 

Please refresh my memory - are there other hard drive models that are working properly with one or both HFDCs?  

 

Also, we must keep in mind that you are trying to format an ST1100 (9 heads) with unmodified HFDCs, correct? For the unmodified HFDC, the write precomp output is tied to the same pin as head select line 3.  This is almost certainly interfering with the write operations, which may also explain why only one head works.  If the write precomp is selecting HEAD3, the only valid value for the drive would be 9.  

 

The heads are represented base 0, so we select the 9 heads as 0,1,2,3,4,5,6,7,8:

 

HHHH

3210 - head select bit position

0000 - head 0

0001 - head 1

0010 - head 2

0011 - head 3

0100 - head 4

0101 - head 5

0110 - head 6

0111 - head 7

1000 - head 8

 

If enabling write precomp sets the bit for H3, the drive will only see 1000 and 0000 as valid heads.  0000 corresponds to the first head, which is what you said is formatting.  In MDM5, try selecting 7 heads with a write precomp of 0.   Per the specs, these larger drives provide their own write precomp. Setting to 0 might hold H3 low, allowing you to format 0,1,2,3,4,5,6,7.  This is conjecture until proved.   

 

ST1100 termination appears to be 100ohm.  See this link for drive details:

https://stason.org/TULARC/pc/hard-drives-hdd/seagate/ST1100-84MB-3-5-HH-MFM-ST506.html

 

I think you're spot on with the drive seeing only 1000 and 0000 as valid heads. If I enter 9 heads in CFORM, it seem to format with heads 0 and 8 properly, but runs much faster through the tracks when attempting a format with heads 1-7. And, of course, if something triggers the MSB of the head select, then everything after 0000 will become 1000 or above - which of course are invalid head selections.

 

Now I've consulted my old drive list and it turns out almost all the drives I possess should handle 0 RWC/WPC, so I'll modify at least one of the HFDCs and see what the result will be then. A good idea could be to insert a switch between the outputs from 5 and 12, so that it's possible to revert if necesary.

 

By the way, it does look to me as if there actually IS a termination SIP on the HFDC...? c".)

[Brufnus]

One HFDC is modified, but it hasn't made any difference. Alas... 😕 I may have done something wrong, but I believe I did the mod correctly. The controller still works, anyway. Perhaps I should give up on that drive, at least for now. 😕 

[Brufnus]

After continued testing, I've discovered something that really puzzles me.

 

When the drive sits alone and terminated, I have 3V on pin 2 all right... but when it's in the middle of the cable, unterminated, the voltage drops to almost 0V on that pin... as if the last one on the cable is responsible for that drop. Could this be caused by faulty termination resistor packs?

The voltage on pin 14 is as high as 3.65V; hopefully that is within permitted specs...

Edited August 22, 2022 by Brufnus
typo

[Brufnus]

Here's something some of you may find interesting. I'm about to order an MFM emulator which uses a beaglebone single board computer for emulation.

It can both read and emulate existing MFM drives, but also create custom empty images - that is, allow one to set different drive parameters such as testing with >8 heads and >1024 cylinders, etc.

 

I plan to create images of our entire club archive, but I've had some issues with the PC in question. However, with the flexibility and capacity the emulator will offer, I can do it natively instead. These archives will be distributed afterwards; perhaps someone might find the contents useful. c",)

 

Anyway, the emulator in itself will, of course, present several advantages, such as better reliability, swapping of images, less power consumption, less noise (!) and, arguably, higher speed due to the lack of physical head positioning etc.

 

I'll share my experiences, once I have it up and running.

Edited September 9, 2022 by Brufnus
addition

[+9640News]

2 hours ago, Brufnus said:

Here's something some of you may find interesting. I'm about to order an MFM emulator which uses a beaglebone single board computer for emulation.

It can both read and emulate existing MFM drives, but also create custom empty images - that is, allow one to set different drive parameters such as testing with >8 heads and >1024 cylinders, etc.

 

I plan to create images of our entire club archive, but I've had some issues with the PC in question. However, with the flexibility and capacity the emulator will offer, I can do it natively instead. These archives will be distributed afterwards; perhaps someone might find the contents useful. c",)

 

Anyway, the emulator in itself will, of course, present several advantages, such as better reliability, swapping of images, less power consumption, less noise (!) and, arguably, higher speed due to the lack of physical head positioning etc.

 

I'll share my experiences, once I have it up and running.

Please keep in mind the only way to use >8 heads is a physical hardware modification to the HFDC.  It does not matter whether it is an emulator or real hardware. 

 

I'm sure the one you have is cheaper than the DREM, however I would do a fair amount of investigation to confirm it emulates the 9234 chipset.  It is not the same as some of the WDC chips out there and specific modifications to the underlying code had to be made for the DREM.

 

Your setup would be a lower cost alternative for the HFDC emulation, however I might offer the suggestion with so many people having TIPI's, you may want to build a complete TIPI folder and subfolder set of the files as well, etc.

 

Beery

[Brufnus]

3 hours ago, 9640News said:

Please keep in mind the only way to use >8 heads is a physical hardware modification to the HFDC.  It does not matter whether it is an emulator or real hardware. 

 

I'm sure the one you have is cheaper than the DREM, however I would do a fair amount of investigation to confirm it emulates the 9234 chipset.  It is not the same as some of the WDC chips out there and specific modifications to the underlying code had to be made for the DREM.

 

Your setup would be a lower cost alternative for the HFDC emulation, however I might offer the suggestion with so many people having TIPI's, you may want to build a complete TIPI folder and subfolder set of the files as well, etc.

 

Beery

Yes, I did the modification on one of my HFDCs. If using the emulator on the other one, it will just have to live with the 8 head limit for now. 

There's a list on the website specifying which controllers work, and the HFDC is listed as fully functional with the emulator.

 

You're right about the TIPI; I don't have one myself but if I can somehow distribute any stuff in a TIPI-friendly way, I'd be happy to do that.

Edited September 9, 2022 by Brufnus

[RickyDean]

9 hours ago, Brufnus said:

Here's something some of you may find interesting. I'm about to order an MFM emulator which uses a beaglebone single board computer for emulation.

It can both read and emulate existing MFM drives, but also create custom empty images - that is, allow one to set different drive parameters such as testing with >8 heads and >1024 cylinders, etc.

 

I plan to create images of our entire club archive, but I've had some issues with the PC in question. However, with the flexibility and capacity the emulator will offer, I can do it natively instead. These archives will be distributed afterwards; perhaps someone might find the contents useful. c",)

 

Anyway, the emulator in itself will, of course, present several advantages, such as better reliability, swapping of images, less power consumption, less noise (!) and, arguably, higher speed due to the lack of physical head positioning etc.

 

I'll share my experiences, once I have it up and running.

Yeh, I ordered 3 of these boards about 2 years ago, but life has been in the way, as well as money.