MattInX's Miscellaneous Musings Say what now?

The JVC 26-pin Hard Disk Interface (Part 1)

As I've mentioned in a previous post, my V86P is the higher-spec model containing a 20MB hard disk. Back in the mid to late 1980s, IDE was still relatively new and most systems were still using ST-506 interface drives, albeit using RLL rather than MFM to increase data rate and capacity. For example, I remember the Viglen Vig II 16MHz 286 we bought as our family PC back in '88(or so) had a 40MB RLL drive. Laptop form-factor systems were also relatively recent, with the floppy-based Toshiba T1100 being released in 1985 and the hard-disk enabled T1200 in 1987. Obiviously desktop drives of the era weren't ideally suited to laptop use, at that time 5.25" half-height was the dominant form-factor. The ST-506 interface also required two separate cables connecting the drive to the controller: a 34-pin control cable on which drives could be daisychained, and a 20-pin data cable from each drive directly back to the controller. These connectors alone take up almost 3.5" of space along the edge of the HDD logic board, and that still leaves power, which makes developing an alternate interface fairly logical.

JVC approached these issues by developing a hard disk in the same 3.5" form-factor being used by floppy drives, and a custom 26-pin connection back to the disk controller. Based on what I've found on-line, this series of drives was used in a number of portable systems, including:

• Data General One Model 2T
• Epson PX-16
• Epson Equity LT
• Epson Portable PC (Q150A)
• GRiD 1520
• GRiDCase 3 Plus
• Sharp PC-7200
• Toshiba T1200
• Victor V86P
• Wang WLTC Laptop

The drive construction is also clearly designed for use in a portable system, consisting of an outer housing with four rubber shock-mounts, one on each corner of the drive unit inside. At first, I thought this might have been specific to my laptop, but it seems that drives in other systems were also constructed the same way, so I assume this was part of the JVC design.

According to this post from 2001, an attempt was made to contact JVC to get information on the JD-3824G. From this communication we know that this series of drives was discontinued around 1991, and were an OEM product manufactured for use by laptop manufacturers. We also have the interface pinout (listed below), and the following drive specs:

• Capacity: 20MB
• C/H/S: 615/2/34
• Spindle speed: 2597rpm
• Encoding: 2,7 RLL
• Data transfer rate: 7.5Mbps
• Average seek time: 78ms
• Power rails required: +5V, +12V

Pinout for JVC original interface (26 pin)

Pin	Name	Function
1	GND
2	/READ	-Read Data
3	GND
4	/WRITE	-Write data
5	GND
6	RESERVED	Reserved
7	/SELECT	-Drive Select/+Power Save
8	/SHIP_READY	-Ship Ready
9	GND
10	/WRITE_GATE	+Read/-Write control
11	/MOTOR_ON	-Motor On
12	/HEAD_1	Head Select(+Head 0/ - Head1)
13	/DIR_IN	-Direction In
14	/STEP	-Step
15	/WRITE_FAULT	-Write Fault
16	/SEEK_COMPLETE	-Seek Complete
17	/SERVO_GATE	-Servo Gate
18	/INDEX	-Index
19	/TRACK_0	-Track 000
20	/READY	-Drive Ready
21	GND
22	+5V	+5
23	GND
24	+5V	+5
25	GND
26	+12V	+12

From this site, I've got a list of other JVC drive that probably have the same interface, but they give some more drive paremeters which are useful to know:

Model	Cylinders	Heads	Seek Time (ms)	Interface	Encoding	Capacity (MB)	12V-Peak (A)	12V-Avg. (A)	5V-Peak (A)	5V-Avg. (A)	Depth (mm)	Width (mm)	Hight (mm)	Weight (g)
JD-3806M		2		Unique	MFM	5
JD-3812M		2		Unique	MFM	10
JD-3824L	615	2	68	Unique(26pin)	RLL	20	0.44	0.16		1.2	144.3	104.6	28.8	535
JD-3824R		2		Unique	RLL	20
JD-3824T	615	2	65	Unique(26pin)	RLL	20	0.46	0.12		1.2	144.3	104.6	25.4	510
JD-3824TA	615	2		Unique	RLL	20
JD-3848H	615	4	45	Unique	RLL	40	0.6	0.12		1.3	144.3	104.6	31.4	575
JD-3848HA	615	4		Unique(26pin)	RLL	40

So based on all that information, it looks a lot like a modified ST-506 interface with TTL signalling for /READ and /WRITE rather than differential pairs, and a handful of additional signals that seem a lot more suited to a disk installed in a laptop. I can confirm from playing with my system that the disk does not spin up until the controller initializes it, and the head automatically parks after a couple of seconds of inactivity (I'm assuming that's what /SHIP_READY is to signal). Using TTL signals for /READ and /WRITE was probably deemed acceptable because you weren't going to have more than one hard disk in a laptop, and the disk was going to be mounted very near the controller due to space constraints so the system didn't have to cope with the long cable lengths the ST-506 spec permits (up to 20'). I'm also guessing the /SERVO_GATE has to be asserted in order for the drive to act on /SEEK. The Reserved pin probably isn't used, but it's worth checking to confirm.

From a drive emulation perspective, it's important to identify how the data is encoded when reading from/writing to the disk. Given the sparsity of information from JVC, the natural thing to do would be to go in the opposite direction and look at the disk controller. In the V86P, this is a small daughterboard that plugs into a pair of pin-headers on the motherboard. The 26-pin disk interface cable is then attached to this board, with a connector on the other end that plugs into the drive.

At first glance, there's not much on the board, but by the late 80s this was all pretty standard stuff found in a large number of desktop systems. The three OMTI chips on the board looked promising, and I do recall coming across OMTI chipsets on disk controllers in the past. Googling the part numbers (20509, 20513, 20527) didn't get me terribly far, but once I started looking at datasheets on bitsavers I was able to find several that sounded promising, with names like "Programmable Data Sequencer" and "2 of 7 RLL VCO Encode/Decode Chip" even though the model number didn't line up with what was on my chips. Included in that archive were some sales brochures, highlighting the range of disk controller ICs. Right at the top of the first page was a promotional photo showing the chips, and there amongst them were the 20509, 20513, and 20527! Skimming through the rest of the sheet showed no other mention of them though, but it did mention several chips for which there were datasheets. I started reading through the "OMTI 5027 2 of 7 RLL VCO Encode/Decode Chip" datasheet on the principle that if my drive is indeed RLL, the OMTI chip used on my board is probably similar. Imagine my surprise when I got to the pinout page and it showed the silkscreen on the package as "OMTI 20527"! Based on this discovery, I was able to identify (and get datasheets for) almost all the ICs on the controller:

• OMTI 5090 (marked 20509B) - IBM PC Bus Interface
• OMTI 5055 (marked 20513B) - Memory Controller and Programmable Data Sequencer
• OMTI 5027 (marked 20527D) - 2 of 7 RLL VCO Encode/Decode Chip
• Sharp LU800V1M - CMOS 8-bit Single Chip Microcomputer (ROM-less) (SM803/A without ROM)
• Fujitsu MB8464A-15L - 8x8k 150ns SRAM
• Fujitsu 27C256AP - 8x32k CMOS PROM
• JVC BU3167K - Unknown

Given this is a JVC interface we're dealing with, I suspect that the JVC chip is handling the signals that aren't there in the ST-506 interface, such as Power Save, Ship Ready, Motor On, and Servo Gate. I think one of the next things I'll look at doing is ringing out the connector and seeing which lines are present on the pins of that IC.

The use of the OMTI 5055 and 5027 would seem to confirm that this is using standard 2,7 RLL encoding with ST-506 signalling - why else use off-the-shelf chips that implement it? That should make emulating the drive a reasonable goal (although probably not one for this challenge). I talked to David Gesswewin, developer of the MFM Emulator at https://www.pdp8.net/mfm/mfm.shtml and whilst his emulator doesn't currently support RLL, he said we could take a look once I've got some more information and some hardware to work with so long as I'm happy to write some code.

I've put together a simple breakout board with a pair of 26-pin connectors wired together, and a 20-pin header in the middle bringing out the 16 signal lines to allow for probing with a scope or logic analyser. I've also put a provision on the board to take +5V and +12V inputs and jumpers to select whether the drive 26-pin header gets power from the host or the breakout board for each of the three power lines. The PCB is away being manufactured, and fingers crossed will show up in a week or two.

The next things I plan to look at are:

1. Try and identify the motherboard header pins that connect to the OMTI 5090 - I know what the 5090 pinout is, so that will let me identify the pins on the motherboard
2. Try and identify a partial pinout for the JVC BU3167K based on the 26-pin header and surrounding chips
3. Build up and check the breakout board when the PCB arrives
4. Hook up a logic analyser and try to get some traces to confirm the hypotheses abou ST-506 signalling and the proprietary signalling