Readers have expressed interest in seeing what exactly a Western Digital ‘Centaur’ drive looked like. I took a few photos of a WD95044-A drive, the larger capacity (40 MB) and newer variant of the WD Centaur family. Some photos were taken with the PCB removed in order to show what the chips on the PCB look like and what the drive chassis looks like without the PCB covering it.
Here’s what the drive looks like:
Its tank-like appearance is nothing like WD’s very modern Caviar drives. The Centaur looks like a mid-1980s drive, not something made at the end of 1990.
The top of the drive shows an old-school defect table, but with no defects listed. It’s not clear if that just means this particular drive had no defects or if the drive’s IDE controller remapped them; either is plausible.
Here’s what the bottom of the drive looks like with the PCB removed:
The black box of the stepper motor is clearly visible in the upper left. The big Nidec spindle motor stands out, and it is quite large, adding to the drive’s height.
Here are two further images of the bottom side of the drive shot from different angles:
The chassis is very “3D”, not at all flat. This is a half-height drive, wasting a lot of space for only 40 MB of storage.
Western Digital cared to put its name even where customers where unlikely to ever see it:
Here’s a detail of what I believe is a sensor that detects the “index hole”, which lets the drive measure its RPM and also determine the start of a track. There is an indentation in the side of the spindle motor housing. I believe that when the indentation passes under the sensor, it will trigger a pulse that serves the same purpose as an index hole on a floppy.
Here’s one side of the drive, with the stepper motor on the left. The drive has a very “mechanical” look, with lots of screws and bolts and stamped and cast metal. It does not look neat.
Here’s the other side of the drive. There isn’t really much to see:
And now the connector side. As was common with old IDE drives, there is no shroud around the IDE connector but there is a key pin. This drive is jumpered as a slave, with the jumper in the middle position. Note that the Molex style power connector is in the usual position but upside down. Fortunately it has a shroud that makes plugging the power connector backwards reasonably difficult.
Here’s the PCB; this is the boring bottom side that is visible when it’s mounted on the drive.
And here’s the much more interesting view of the top side of the PCB which has all the chips on it:
Not surprisingly, the core chips are almost all Western Digital’s own. The big WD42C22A ‘Venus’ is the same chip WD used on their ESDI/RLL/MFM controllers in the late 1980s and early 1990s. The larger chip next to it has a WD part number and (C)WDC ’90 but also (M)(C)i ’80, suggesting that it’s an Intel microcontroller with WD firmware in its ROM.
The smaller WD chips are WD10C22B, a self-adjusting data separator (supports MFM and RLL encoding), and WD12C20, for which I could find no documentation.
The Philips FCB61C65L-70T chip is an 8Kx8 SRAM with 70ns access time.
Here’s a better view of the connectors on the PCB:
The leftmost white connector visible in the photo above is attached to the stepper motor, the small two-pin connector in the middle to the RPM sensor. The largest connector goes to the spindle motor. There is another 12-pin Berg/DuPont style connector, not visible on the photo above, that links the PCB with the read/write heads.
Here’s the stepper motor:
Here’s one more view of the stepper motor. It says: WARNING REMOVAL OF THIS COVER WILL VOID ANY AND ALL WARRANTIES. I guess they really didn’t think the cover was safe to remove.
Amazingly, even though the WD95044-A is painfully slow with 28ms average seek time quoted by the manufacturer (just seek time, not access time!), it is much faster than its predecessor, WD95048-A with absolutely glacial 69ms average seek time.
Yes, those I found. Unfortunately I had a typo in the article (now fixed); the drive has a WD12C20 chip on it, not WD10C20, and I clearly got lost in the alphabet soup. The WD10C20 appears to be a predecessor of the WD10C22.
Hmm, what do we have here…
HA13441WD – spindle motor driver
UC3770A (x2) – stepper motor driver
32P541 – read amplifier
WD10C22B – data separator
WD42C22A – single-chip MFM/RLL/ESDI/IDE controller with DMA support
FCB61C65L – 8kB (16 sector?) buffer RAM
62-602050-101 – Intel microcontroller (8051?) with probably OTP ROM
That leaves only the WD12C20, which doesn’t seem to have any public datasheets, and the 186086-001 and 186622-001, which could be just about anything because Micro Linear made a lot of custom parts.
Huh, the spindle motor has the part number 186001-002 on it. I wonder why it’s so similar.