Talk:ST506/ST412

Capacity

What was the head/track/sector layout of the ST506? Rpresser 15:06, 20 June 2007 (UTC)

153 tracks per surface

4 surfaces (heads) per drive

10,416 bytes per track unformatted

typically formatted with 32 sectors per track and 256 bytes per sector.

The ST506 was a "dumb" interface so the user capacity depended upon the controller to which it was attached. The above comes from the ST506 Service Manual, May 1982Tom94022 21:10, 20 June 2007 (UTC)

Seagate or Shugart?

The ST506 and ST412 manuals that I have looked at (through bitsavers) indicate that the drives were Seagate.

It looks to me that Shugart drives were numbered with SA, and Seagate with ST, further evidence for Seagate.

I don't know when the change was made, though.

Gah4 (talk) 21:27, 17 January 2008 (UTC)

Shugart Associates [f. 1973], later Shugart Corporation is a different company than Seagate Technology originally Shugart Technology [f. 1979]. Alan Shugart was a founder of both. Tom94022 (talk) 17:00, 18 January 2008 (UTC)

Yes, but which company is responsible for the ST506 and ST412?

Gah4 (talk) 04:28, 14 February 2008 (UTC)

Seagate Technology Tom94022 (talk) 06:08, 14 February 2008 (UTC)

Cost

Out of curiosity, how much did one of these cost in 1980? I imagine they weren't available to the public, and so researching the cost would be hard, but perhaps someone reading this was actually there, in 1980, and can provide a source. -Ashley Pomeroy (talk) 23:21, 1 January 2011 (UTC)

The Seagate Sept 24, 1981 Prospectus gives the FY1981 revenue of $9.9 million for approximately 13,000 drives or an ASP of about $750. The claim then was that this was the first HDD priced below $1000 which probably was true although the Shugart SA1000 or the Memorex 101 may have beaten it below that threshold. Tom94022 (talk) 00:07, 2 January 2011 (UTC)

As well as I remember, for some years the usual high end microcomputer drive to buy was about $1000. The capacity increased over the years, but the price stayed about the same. The price of older drives would fall as new ones were introduced. Minicomputer drives, such as SMD, were more expensive. Gah4 (talk) 20:54, 28 April 2016 (UTC)

Interface limitations

The interface is limited in number of heads because the 4-bit head addressing can only express 16 combinations. The drive selection scheme is one wire per drive, so only 4 drives. And controllers only expect 5 Mbit/s per documentation. Bytesock (talk) 13:08, 28 April 2016 (UTC)

Limitations vs specifications

The paragraph was reverted because it ascribed as a "limitation" a mixture of some specifications and some unsupported limitations. The ST506 interface is indeed limited by the 4 drives select and 3 or 4 head select lines. If used according to the Seagate specification it can address either 8 or 16 heads, but one can use unused drive select lines as head bits going beyond 16 so the limitation is not 16. I know of know published 6.1 meter cable length limitation and it is not supported by the citation. The data rate is clearly not limited to 5 Mbits/sec since it is capable of 7.5 Mbits/sec when used in RLL and most likely even more since the original 5 MHz wasn't picked from any transmission limitation. Acurately written the sentence would be OR and since it didn't add a lot to the article, I reverted. Tom94022 (talk) 21:12, 28 April 2016 (UTC)

The specification does indeed limit what you can do with any interface. Of course, the interface can be changed circumventing the limitation, but then it would no longer be a ST-506 interface which is what the article is about. You clearly did not read the specification cited because the maximum cable length is specified as 20 feet (which is 6.096 metres - but 6.1 metres is close enough). A relatively simple calculation gives the theoretical maximum length as 6.9 metres so 0.8 metres is a reasonable safety margin. --Elektrik Fanne 16:28, 26 July 2016 (UTC)

The drive is designed around 5 Mb/s MFM, to satisfy bit error rate specifications. As with processor overclocking, one can use things out of specification. An important limitation is flux change density, and the RLL controller conveniently stores more bits with the same maximum flux change rate, but also has stricter requirements on flux change timing. I do remember much discussion on which drives were specified to run on RLL controllers, and which weren't, even if there wasn't any actual difference. Gah4 (talk) 03:49, 1 June 2018 (UTC)

I think it is more about margin than flux change rate or density. The bit detection window is 33% smaller in RLL than in MFM (hence 1.5 x capacity and data rate). Most drives (particularly stepper motor drives) shipped with a lot of window margin to allow for off track due to thermal and mechanical changes, so operating in home or office environment most ST 412 MFM drives would appear to initially work on RLL controllers but over time would have increased soft and hard error rates. A drive left on and working in an air-conditioned office might have a bunch of trouble when the air-conditioning failed. I think some vendors just selected their high margin drives and labeled them as RLL qualified but for high yield production the vendors had to improve the margin by doing things like changing formulation, flying lower, reducing off track, etc. Tom94022 (talk) 19:17, 1 June 2018 (UTC)

Average seek time

It is true that the longer seeks take longer because in part they require more step pulses. But is "The number of tracks increased the access time." as was once stated in the article true? The reference spec has the same "average access" times for the 612 track ST506 and 1224 cylinder ST412 so on its face in the reference the answer is no. A better question would be does the average seek time increase as the number of cylinders increases and which also on its face appears to be no. A more subtle answer would note that each additional cylinder requires one additional step pulse which in burst mode would require at least 25 more microseconds so adding 100 cylinders might add 2.5 msec. But some vendors measure average seek time from the last seek pulse in a burst so in that mode and with that definition there is no added time. And vendors of drives compatible with the ST412 interface but having more cylinders take into account the additional step pulses necessary for their additional cylinders. So there maybe no real world example where additional cylinders increase average seek time. On the whole I suggest the article is better off without this subtle distinction, if there is one at all. Tom94022 (talk) 18:18, 31 May 2018 (UTC)

It's generally true that more tracks increase seek time with stepper-actuated drives. However, the ST-412 introduced a 'burst' mode for stepping (note the manual's ** footnote) that was supposed to reduce the average seek time to that of the ST-506 – the burst stepping mode didn't really catch on (it required controller support), so in reality, the access time increased. The manual is not a good source on this but I don't have a better one. I don't insist on including the statement in the article, we can well leave it out. --Zac67 (talk) 19:55, 31 May 2018 (UTC)

• I think you are referring to mainly FDD interfaces which only operated in a "slow" mode. The ST-506 operated that way, but pretty soon afterward the HDD industry adopted the "burst" mode. The Shugart Associates SA-1000 which preceeded the ST-506 and upon which the ST-506 interface is based provided a burst mode. I think Seagate left out the logic to savea a few dollars. Tom94022 (talk) 22:29, 31 May 2018 (UTC)

I haven't thought about this for a long time. Do all drives buffer seek pulses, and then run the stepper motor at the appropriate rate, or directly run the stepper, like floppy drives? Since the distance moved is the same, it seems somewhat reasonable that the time is about the same. (Inertia, and all that.) There is then settling time. Reminds me that wikipedia likes secondary sources over primary, but I suspect any secondary sources just copied the primary. Gah4 (talk) 19:59, 31 May 2018 (UTC)

OK, from ST412 OEM manual, which also describes the ST506, the seek times, using the appropriate for each optimal step times, are the same. For ST506, you cut a jumper, and then vary the step times for optimal motor operation. For ST412, seeks are buffered by the drive, if sent faster than 200us. Or you can do slow (3ms) step speed on either one. Reminds me around 1989, I had a program to measure seek time for MS-DOS. I ran it on an actual IBM machine, and the message was something like "This disk is so slow, are you sure it isn't a floppy drive?" I have no idea which controllers did the ST506 fast seek algorithm. Or maybe it is done in software? Gah4 (talk) 20:38, 31 May 2018 (UTC)

I too haven't thought about this in a long time, but it is my recollection that burst mode was standard on all drives complying with the ST412 interface "standard" and that all corresponding controllers then burst at the maximum rate. And as drive manufacturers introduced new models with more cylinders in the same recording space they sped up the mechanisms so the average seek time got lower as tracks were added in the same space. Thus the ST251 with a ST412 interface and 820 cylinders has an average seek time of 40 msec. The article can probably be improved by noting the ST412 interface became the industry standard with a few changes such as the busrt mode standard. Tom94022 (talk) 22:29, 31 May 2018 (UTC)

Yes, it seems that burst mode came with ST412, such that the drive figures out the optimal step rate. The ST506 service manual mentions that it stops with two phases powered for faster settling, and so two steps per track. But ST506 has algorithm driven seek, though I didn't look up the algorithm. The controller has to vary the step rate following the algorithm. My first hard drive is (I think I still have it) a Miniscribe 6053, with I believe (after many years) a voice coil positioner. No stepper motor to limit the speed. It looks like with optimal controller the seek times are the same. That doesn't mean that such controllers were common. Gah4 (talk) 02:15, 1 June 2018 (UTC)

Since the reference source mentions algorithm driven seek, which can be much faster than fixed step rate, the article should mention that in the seek rate comparison. I don't know if the ST001 technical note is available, though, with the actual algorithm. Gah4 (talk) 21:45, 1 June 2018 (UTC)

uses 2 recorded bits to represent 3 data bits

The statement uses 2 recorded bits to represent 3 data bits seems strange to me. This is complicated, as bit has different meaning in digital electronic, computation, and information theory. What is actually recorded on the disk are flux transitions, and their positions are decoded to extract data (and clock) bits. Gah4 (talk) 21:03, 1 June 2018 (UTC)

It is and I need to fix it. MFM is a 1/2(1,3) code and the ST412 RLL is a 1/2(2,7) code. Since the minimum spacing of RLL is 2 zeros while the minimum spacing of MFM is 1 zero RLL can get 50% more data in the same distance between clock bits, that is 1001001 (RLL) fits in the same space as 101010 (MFM) and that's why the window is 1/3 smaller. Annother way to say it is RLL has three channel bit times in the same space as MFM has two channel bits. The physical minimum spacing and channel bit density are the same but coding gives 50% more capacity and I was trying to explain the difference in layman's terms. Looks like I got it wrong - if you can fix it go ahead. I'm thinking about it right now. Tom94022 (talk) 21:18, 1 June 2018 (UTC)

BTW channel bits are recorded NRZI Tom94022 (talk) 21:20, 1 June 2018 (UTC)

The only thing I can think of, is to explain it in flux transition density, but I don't have something to add yet. Gah4 (talk) 21:47, 1 June 2018 (UTC)

The density is the same, that is the minimum distance between two transitions is the same, its just that RLL has 3 clock times in that minimum while MFM has only two. How about something like:

• The limitations of the ST-412 interface are 5 million transitions per second on data lines, 16 heads, 4 drive units and a 20-foot (6.1 m) cable length. The standard channel code for the ST-412 (and ST-506) is MFM with one data bit per per transition for a data rate of 5 Mb/sec. The ST-412 RLL varient averages 1.5 data bits per transition for a data rate of 7.5 Mb/sec

It's the best I can get into without going into a MFM 1/2(1,3) code vs the ST412 RLL a 1/2(2,7) code. the point being the interface cannot have a higher transition rate than 5 MHz. Tom94022 (talk) 21:07, 2 June 2018 (UTC)

I forget now how to count bits for RLL(2,7), but MFM has data bits and clock bits, with a clock bit (transition) only between two '0' (no transition) data bits. (The current text mentions data bits, but not clock bits.) I believe that means on the average 3/4 transitions per bit time, or 4/3 bit/transition. Gah4 (talk) 07:03, 7 June 2018 (UTC)

If you link to the page it is explained - they both have clock and data bits and I didn't want to get into that here. Tom94022 (talk) 15:43, 8 June 2018 (UTC)

precompensation

The discussion about timing windows and RLL reminds me that there isn't mention of precompensation. This was usual for floppies, and extended to ST506 and such, and is described in the references. Gah4 (talk) 21:49, 1 June 2018 (UTC)

Actually it is sort of coverred in the article in the change in definition of the pin 2 from Reduce Write Current to Hd Sel 3 - another of the 506 -> 412 evolutionary changes and probably not worth elaborating on in the article. BTW @Gah4: did u notice we have been talking about this article off and on for ten years? Tom94022 (talk) 22:57, 1 June 2018 (UTC)

No, I hadn't thought about it being ten years. It seems to have been my 18th edit, and over a year after my first edit. Gah4 (talk) 23:11, 1 June 2018 (UTC)

It seems that there is a Write precompensation, but it gives much less detail that it probably should. Gah4 (talk) 23:11, 1 June 2018 (UTC)

ST-xxx confusion

The lead sentence now reads:

ST-506 is the designation for two related items: the ST-506 hard disk drive (HDD) and more generally the ST-412 interface, a connection to a computer system that evolved out of the original ST-506 interface.

which is a bit confusing to begin with, but then the next section goes on to talk about the "ST-506 interface" and the "ST-412 disk drive" and now I'm completely confused. Could whoever worked the ST-412 information into this article (or anyone) please give it a sweep for consistency? -- FeRD_NYC (talk) 01:37, 10 July 2018 (UTC)

The ST-506 and ST-412 were the first popular drives using this interface, so it was named after them. Without looking it up, I believe that the ST-412 has buffered seek, such that you can send pulses at a high speed, the drive counts them, then seeks as appropriate. The ST-506 doesn't do that. Controllers that allow for it might be give the ST-412 number, but otherwise they are pretty much the same. Many more drives used this interface, but it wasn't numbered after them. Gah4 (talk) 03:32, 10 July 2018 (UTC)

@Gah4: Thanks, but it still strikes me as confusing. Originally the lead only mentioned the ST-506, both drive and interface, and noted that the interface used for the ST-506 was subsequently found on many other drives, effectively becoming a de facto standard. That was clear enough, as it stood. But in the article's current state, the lead (green talkquote, above) makes it sound like the ST-506 is the designation for the drive, and ST-412 is the interface for that drive. So then going on to mention the "ST-506 interface" or the "ST-412 drive" seems contradictory, or at best out-of-left-field.

If ST-506 can refer to either a drive or interface, and ST-412 can as well, I think that should be made much clearer in the article, most especially in the lead. -- FeRD_NYC (talk) 04:19, 10 July 2018 (UTC)

The ST506 and ST412 refer to both a drive and an interface; the ST412HP is an interface only - they are all closely related so they probably belong in one article. Probably we should rename the article to ST506/ST412 and redirect ST506, ST412 and ST412HP to this one article. BTW, Seagate did not use dashes; it was the ST506 not the ST-506. I'd try to rename the article but it is heavily linked and I'm afraid I would break to many links so if someone else is more confident please do so.

I will take a hack at the lede. Tom94022 (talk) 05:55, 10 July 2018 (UTC)

I think the latest revision is much improved, my thanks! -- FeRD_NYC (talk) 06:19, 10 July 2018 (UTC)

Inadequate lead section

I hate that I'm being That Guy again, but I'm about to flag the lead section as inadequate, based on the current article content. The issue, specifically, is that the lead introduces the two hardware designations, but then says nothing about them or why they're notable. Instead it digresses into a technical detail about encoding differences, which absolutely does not belong in the lead section. -- FeRD_NYC (talk) 22:09, 24 July 2018 (UTC)

Go ahead and flag if you wish but you can do more for Wikipedia by editing so why not try and add to the lede what u find missing. IMO the encoding differences are appropriate to the lede since the explain the distinction between the ST412 and ST412HP. Tom94022 (talk) 06:26, 25 July 2018 (UTC)

I've added a sentence to put ST506 in a better context. --Zac67 (talk) 10:46, 25 July 2018 (UTC)

I think this canard needs to be challenged, even long after the fact. Without considering opportunity cost, there's no way to know that any particular user dropping a quick flag is less productive for Wikipedia. Protestant neoliberalism is sometimes defined as the belief that somewhere out there someone is shirking. The corollary being that if you're not working in plain sight on this very page, you're almost surely contributing nothing at all, therefore you are self-evidently best employed contributing more here before you bugger off to shirkville.

However, on Wikipedia, it turns out that shirking is not is not the preferred opiate of the masses who show up to make useful edits, and almost all of us have a set of articles we can improve directly and efficiently, and another set of articles where we can detect the fault, but not efficiently rectify the fault, and so for these opportunity cost dictates the global utilitarian action as flag and faff off to greener pastures. — MaxEnt 20:31, 16 September 2020 (UTC)

I would like to add that the image shown in the top right corner and presented as "ST506 with cover removed" is not an ST-506, or a Seagate drive at all. The drive presented is actually an NEC D5126. It has the incorrect spindle, the wrong filter, an internal head amp, non-twinarm heads, the incorrect placement of the head cable, the wrong PCB, the wrong stepper motor, and no large brass interrupter. It is also open at the head end. This drive is very clearly an NEC D5100 series mechanism, most likely a D5124 or D5126. Sadly, I am not skilled enough with editing to figure out how to change it myself. FAMICOMASTER 07:06, 21 MAY 2023

@FAMICOMASTER: Agree it is likely not a Seagate ST-506, see [1] but it might be a later version; designs do change. There is no editing skill required, just find a public domain image, in WikiMedia or upload one there, and replace the link in the article to the correct image. I'd do it but I'm traveling right now and limited in the things I can do Tom94022 (talk)

@Tom94022: The image linked is most definitely an ST-506 or Texas Instruments TI-5, a licensed copy of the ST-506. The drive pictured in the article is absolutely not a Seagate mechanism at all, of this much I am certain. While designs can change, the entire mechanism most certainly will not. I cannot figure out how to edit the header of the page like this, only subsequent sections individually. FAMICOMASTER 20:57, 21 MAY 2023

@FAMICOMASTER: The 4th tab along the top of the page, "Edit," allows one to edit the entire article. FWIW I have not been able to locate a reliable public domain image of a Seagate ST506. Tom94022 (talk) 04:18, 22 May 2023 (UTC)

CDC-506

According to the following manual, Control Data had a compatible interface which it called CDC-506.

http://bitsavers.trailing-edge.com/pdf/cdc/discs/wren/77738035F_Wren_II_Disk_Drive_OEM_Manual_Nov85.pdf

106.69.65.79 (talk) 03:35, 15 March 2020 (UTC)

As the manual says "The CDC506 Interface is compatible with the ST506 Industry Standard." The manual is dated 1985 while the ST-506 shipped in 1980 so the CDC name change doesn't seem particularly relevant to this article. Tom94022 (talk)