Chapter 5: What’s inside

After the dim bulb testing, I am now reasonably confident in the power supply’s ability to handle 110 volt AC input directly. But it’s not quite ready yet for the full 230 V mains voltage. So, thanks to a colleague of mine who let me borrow his 110 volt transformer, I can now move on and worry about the mains later. I haven’t got a mains power cord anyway.

First thing, though, I let it run for half an hour just spinning the fans to make sure all caps have reformed. Both 5.1 and 12 volt outputs are bang on even with no load.

Now, before connecting the electronics to the PSU and powering the system up for the first time, let’s do a little teardown to see how it’s built, what’s inside, and whether it looks OK.

Left side

The section above the power supply hosts the 3.45 V regulator module, which I mentioned earlier. The remaining space is occupied by the System module, which includes (left to right):

The rest is populated with the ICs for the TURBOchannel controller, memory controller, cache, and ROMs.

A quite substantial amount of logic is implemented on one-time programmable logic devices such as the AMD PAL22V10 and TI 20R4/20R6/20L8 series in square 28-pin packages. This is likely a testament to how early the DEC 3000 AXP series was in Digital’s Alpha programme, before a full complement of large scale integrated ASIC chips for that platform had been developed.

This may even hark back to the ADU, or Alpha Development Unit, — the first Alpha-based computer, which was built for software development at Digital in quantity of 35. Note how abundant the little square chips are on its CPU module:

Below is one of the four identical memory motherboards (MMB). Each one is labelled 54–21141–01 B01 and 50–21140–01 B1. In addition, they bear short individual markings (left to right): YWP–C, AACC–2, GS–2, and AACC-2 again.

The memory modules are proprietary 100-pin SIMMs, 32 bit wide plus ECC. Each one can hold between 2 and 32 MBytes. This system is fitted with 4 MB modules, P/N 54–21139–CA, also labelled 5021138–01 B1 and 5421139CP B01. Each one is populated with ten Fujitsu 814400A–70 ‘fast page mode’ DRAM chips on one side.

Two sides of a memory module belong to different memory banks. In this configuration, only banks 0 and 2 are populated, and the empty sides of the SIMMs correspond to banks 1 and 3.

Here is a close-up view of four sections of the system module, part number 54–23153–01. As usual, click to enlarge.

Right side

Under the right side panel there is room for four SCSI hard disk drives. My machine came with one RZ28 drive, a 3.5 inch model with a formatted capacity of 2.1 gigabytes. There have been different variants of the RZ28, made by different manufacturers.

Quantum model DSP3210
Part number RH20E–AW
5400 RPM
SCSI-2 fast narrow 50 pin interface
10 MHz 8 bit SCSI bus
2148 MBytes formatted capacity
16 heads (surfaces)
3045 cylinders
Variable capacity per track
Track density 3256 TPI
Max. linear density 64,000 BPI
(1, 7) RLL code

Source (bitsavers)

Located at the rear of the machine is the I/O module, P/N 54–21147–02. It plugs into the System Module via the brown connector shown here at the top.

Going from left to right, we can see an AM7990 “LANCE” Ethernet adapter, AM79C30A ISDN and Audio controller, socketed Dallas DS1287 real-time clock, two dual-channel serial communication controllers from ZiLOG, two big TURBOchannel interface ASICs, and two NCR 53CF94–2 SCSI controllers. The I/O module also provides three more TURBOchannel slots (on the other side of the PCB) in addition to those on the System module.

The coin cell battery inside the DS1287 RTC is certainly dead. I’m not going to deal with it now — I want to make sure the machine is working first. Being sealed inside the RTC package means the battery hasn’t leaked onto and corroded the PCB — a common plague of vintage computers.

That’s it for the teardown. Nothing looks broken, burnt, or leaky. Time to clean away the dirt and dust (remarkably little of it!) and get ready to try and start the system up.