• Please review our updated Terms and Rules here

Improvising a power supply for a DEC VT100 serial terminal


Veteran Member
Apr 24, 2009
Canberra, Australia
The DEC VT100 terminal has an early switch-mode power supply unit that contains some proprietary or rare components that are hard to replace. In my 1981-built VT100, the primary winding of the main transformer seems to have failed, and finding an exact replacement would be near to impossible. Rather than trying to substitute for the failed component, I decided to build a replacement based on a common AT/ATX PSU.

I found an ATX PSU with a small form factor. A single Disk Drive power connector from this can supply the +5v and +12v rails at more than adequate current. A -12v rail can be found in the PSU's motherboard connector - Pin 14, usually blue wire.

Photos of the project in progress and a circuit diagram for the -23v rail are at



1. The VT100 requires a -23v rail to be used together with +12v to create a 35v potential in the circuit for writing setup changes to its Non-Volatile RAM. This had to be synthesized with a small circuit employing a transformer to deliver 24vac, a diode bridge rectifier, smoothing capacitor, and a 24vdc voltage regulator. The regulator is not just to protect against over voltage. It also reduces the ~24vdc coming out of the bridge rectifier to about 22.8vdc, close enough to the 23vdc spec.

The DC output of this circuit is wired with its nominal "positive" 23v to the common GND of the PSU connector, and its nominal "negative" to the -23v rail of the VT100 power supply connector.

Introducing a new transformer into the VT100 chassis needs care. If the transformer is located too close to the CRT, its pulsing magnetic field can induce instability in the screen display. The transformer needs to be located as far as possible from the CRT neck, while within the PSU chassis cage.

2. For the ATX PSU to be controlled by the VT100 external power switch, permanently short to ground the "power on" (or "power good") line in the ATX motherboard connector. It is a green wire at pin 16.

3. During testing, I discovered an undocumented trace on the VT100 Basic Video Board that links address line LBA6 of the Line Buffer SRAM directly to finger J of the PSU connector. This is nominally 12v GND, and in the original PSU configuration 12v GND floats independently of the 5v GND (which is the TTL logic GND). The AT/ATX PSU uses a single common GND, so if finger J is grounded it will hold the Line Buffer address line 6 permanently low and thus cripples the display.

When using an AT/ATX PSU for the VT100, it is ESSENTIAL that finger J NOT BE CONNECTED to the replacement PSU GND or to chassis earth.

The aim was to retain the original VT100 external AC plug, switch and fuse, and to feed PSU outputs into the original edge-connector socket for the power supply.

Fortunately, the original PSU circuit board with all in-circuit components can be unscrewed and unplugged from the external switch/fuse panel (see photos). I by-passed the voltage selector switch and left this external panel with simply a switched (fused) active and switched (unfused) neutral AC line, crossed by the original smoothing capacitor, plus earthed chassis.

From the ATX PSU I removed the cover panel, the IEC AC input socket and the external switch, because these functions are done by the VT100 panel. I selected one Molex drive connector with the +12v, +5v and common ground lines, and also separated out the -12v (blue) line from the ATX motherboard connector. I then cut off all unused lines. For safety, I trimmed the cut-off lines so that each voltage was of slightly different lengths, so that they could be taped up together safely without risk of shorting (photo).

Before taping, I soldered the stub of the green "power on" line to a ground line so that the soft-switch is permanently "on".

To substitute for the edge-connection fingers of the original PSU board, I made a "plug" from a piece of double-sided copper circuit board cut to size, using a hand engraver to outline the "fingers" on both sides. Though effective, this is a bit messy-looking. For a neater job, it might be possible to find a bit of commercial proto-board with appropriately-spaced fingers. The +5v and 5vGND rails have 5-6 fingers each, but one wide "mitt" representing each group of fingers saves a lot of cutting and soldering and works just as well (photo).

All GND fingers can be common, but remember that the "12v GND" on finger J should NOT be connected.

The five rails to be connected are +5, +12, -12, -23 and GND. Only five wires are needed.

Space is tight inside the VT100 chassis PSU cavity, so I soldered the plug fingers to about 20cm of colour-coded wires, with the other ends in one side of a pluggable screw terminal strip that could easily be reached from outside the chassis when the "edge connector" plug is in its socket. I could then put together the matching terminal strip socket with the outputs from the ATX and from the improvised -23v supply.

The ATX PSU could be slid between the brackets that had held the original PSU circuit board and locked in place via a couple of the original screw holes in the bracket.

The best orientation is with the ATX PSU fan pointing upwards through the VT100 cabinet, drawing air in from the bottom grille (photo).

The components of the -23v supply circuit mount on a small piece of perforated circuitboard, which in turn is mounted on a piece of structural board (I used 3-ply) cantilevered out from the brackets holding the new PSU (photo).

To keep the 24v transformer away from the CRT neck, I bolted it directly to the inside bottom corner of the VT100 PSU chassis cage. This required leaving the AC mains primary leads of the transformer long enough to reach the mains power input connector block at the other end of the cavity, and the secondary 24v leads long enough to reach back to the -23vdc circuit board.

With this construction method I could safely test that the right voltages were on all pins of the PSU output connector before plugging together the two terminal strips, then join the connectors and test VT100 function with the PSU still just outside the chassis (photo), and then when satisfied could push the assembly home with all wires inside and everything held in place by the original PSU brackets and panel.

Maybe one day I'll stumble across components that would allow me to restore the original DEC PSU, but in the mean time, I have a working terminal.

Very nice work! It's good to know that an ATX supply fits inside the available space. Also, the -23V solution is simple and does the job nicely. I'm sure that this will be helpful to others (me included) that sooner or later will have have a disaster like the loss of a winding on an unobtanium transformer.

Perhaps I should clarify that the AT/X PSUs for a standard AT/X case are likely to be too large to fit between the brackets of the VT100 PSU external panel. They can still be crammed into the VT100 PSU cavity, but a different construction would have to be used. In that situation, I would re-position the -23v transformer and circuit board between the panel brackets. It would make access to the connectors more difficult, but not impossible.

The smaller-size ATX PSU that I used is from a Compaq older Pentium compact desktop unit. These are pretty common and turn up as salvage often enough.

After many years, I have re-visited this post and realised that the link to photos was no longer valid.
In case there are any people still living with an interest in this work-around for a VT-100 power supply - the photos referred to in the original post are now found at this public link on Google Photos: