Picture of the Dual PSU (17K)After deciding on the initial idea, the next step was to decide the basic requirements and specification of the unit. Based on market research and my experiences with commercially available units, here's what I came up with:

  • General

    Dual split outputs, primarily for op-amp based experiments. Permanently configured to give positive and negative rails rather than a conventional dual unit that has a series-tracking function. This enabled me to employ a topology that I'd devised to give very accurate tracking.
  • Outputs

    ±0 to 18 volts, ±0 to 2 amps. Automatic crossover between Constant Current (CC) and Constant Voltage (CV) with status LEDs
  • Voltage and current adjustment

    By separate course and fine controls for both voltage and current. In my experience this offers quicker setting and better resolution than 10-turn pots, though I accept that many prefer the (rather more expensive) 10-turn option.
  • Metering

    Dual 7-segment LED displays. Using devices with a 1999 count enables 10mV and 1mA resolution. Each display can display voltage or current from either the positive or negative supply.
  • Tracking modes

    Either the positive or negative supply can be the 'master'. This might be useful for quick testing - for example if the positive supply is set for 9 volts and the negative supply is at 12 volts, then changing the 'tracking master' mode will quickly change between ±9V and ±12V
  • Output switching

    DC outputs switchable - when output is isolated, the meters should display the selected voltage and current limit value, enabling the limits to be set safely before the circuit-under-test is powered up.
  • Physical

    I wanted to fit it into a 3U height, just under half-width. That fits nicely with some heatsinks that I've got, and my other kit.
  • Other

    Heatsink temperature should be monitored, and the outputs should be isolated in the event of overheating. This is because I don't intend to use it anywhere near full load continuously - so I can safely use smaller heatsinks than would otherwise be required for worst-case conditions.

During the 1992 summer break, I built the first prototype. It almost worked, but had several problems that weren't addressed until quite recently. As I only built the positive regulator, I didn't know for sure that my cunning tracking scheme would work accurately. Also, in my naivety I didn't plan earthing properly, instead trying to make 0V from thick heavy wire.... That's been consigned to history, and prototype no. 2 is shown below:

The (almost) completed unit

Here it is, working and looking pretty much like the original sketches. All it needs is labelling. Inside there is still some tweaking to be done, but it's on the shelf and in use. It's built in a standard box from RS - all steel apart from the aluminium front and back panels. Unusually for RS, it wasn't too expensive - about £15. Both the top and bottom are removable for access. It's reasonably well ventilated, which is just as well!

With the help of Paint Shop Pro, here's a labelled version. Sadly, the finished version won't look this good!

Completed unit (36K)

Most features should be obvious - I tried to make control layout logical. There are course and fine controls for both voltage and current, and green/yellow LEDs to indicate CV and CC modes. Pressing the Meter Mode switches changes the meter indication - the red LEDs in the display windows move to indicate the mode selection.

The tracking mode select is on the left. Normally both red LEDs are off, and each supply is independently controlled. Pressing the button lights the first LED and the positive supply becomes the master. Pressing it again moves the LED down, and the negative supply is now the master. However, since building the unit, I decided to change this to a 3-position toggle switch so that the tracking mode is remembered when the unit is re-powered. Tracking is very accurate, and when the unit is properly set up, it's quite rewarding to see both voltmeters indicate exactly the same reading at every setting of output voltage...

I realised that having the 0V terminal in green is slightly confusing, so recently I changed it to black, and changed the two negative black ones to blue. I could have provided 0V sense connections, but any losses in the 0V cabling would result in tracking errors. As the currents involved are relatively small, the +/- sense connections are probably unnecessary anyway...

The LED displays are a bit poor, particularly the left-hand "18" segments. Also, the strange Maplin display filters make it worse - if the LEDs aren't firmly touching them, then the filter blurs the image. I'm investigating better quality solutions that hopefully won't involve rewiring the panels... I'd also like to replace the small rectangular LEDs (and confusing labelling) with HP light bar modules, and produce a transparency that has the V/I labels on. I've discovered that I can print black in a high enough density on inkjet transparencies. The resulting printout is sticky, ie self-adhesive.

The power switch is illuminated - I've set it up so that it flashes to indicate a fault condition. Currently the only trigger is the heatsink temperature, but I'm planning a simple over-voltage detector. In fault mode, the DC outputs are isolated - they're switched by relays.

What's Inside?

Here's what you see under the top cover:

Top cover removed (82k)

Things are kind-of upside-down in there - the transformers are bolted to the underside of the chassis. The large board is the Logic board, and deals with all the mode switching and relay driving. It also contains the analogue switches for the meters. The 20-way ribbon-cable connects to the Control board underneath - all relay drives and meter signals go down there. The 34-way ribbon cable connects to the Meter panel - 12 wires go straight from there to the front panel for switches and LEDs - see later. The only other connection to the logic board is for the heatsink thermistor, just visible at the back. The inter-board wiring worked out quite well - it's always worth trying to use ribbons or similar to avoid having a birds-nest...

Note the double-insulation on the mains switch - safety must always come first!

View of the logic and meter boards (87k)

This is a close-up with both ribbon-cables removed. It shows the grey 12-way connector that connects to the front panel switches - it connects to a green wiring loom that you'll see later. The white Molex (just visible on the right of the 34 way latching socket) is the power input from the logic PSU. Power for the Logic panel is taken via the ribbon cable. The 20-turn pots are for meter calibration - there's enough slack on the ribbon cable to let you get to them. Standard 4052 CMOS switches deal with the signals.

This board was a bit tricky because of the limited height. To get the most height for components, there's a sheet of Paxolin for insulation underneath it, meaning that the clearance is less than 3mm. This board would've been much smaller if I'd used mechanical switching for the meter modes but they're so old-fashioned... I was also having problems fitting them on the front panel. Next time I'll be putting less in the box!


Here's the view with the lower cover removed. You can see the Control board on it's crude hinge mechanism. It works quite well in practice, which is just as well because there is still quite a bit of development to be done here. The 12VA transformer powers the logic, LED meters and the relays.

Bottom view (63k)

Things become clearer when you hinge up the Control board: You can see the Logic PSU board (left) with its connection to the Meter board, which is the other side of the aluminium sheet. That sheet is held to the front panel by all the controls, avoiding visible screwheads on the front panel. On the right is the main transformer (120VA) and above it is the rectifier assembly. That's held in by 2 screws accessible from above - the lump lifts out easily for repair once some Molex connectors have been undone.

Bottom view (105k)

You can see the other end of the 20 way ribbon cable, along with the 50 way front panel connection. This was possibly a bit of a risk, electrically speaking, but it seems fine in practice; I made sure that lots of parallel connections were used to carry the heavy currents. Normally, of course, you'd want the regulator output connections to be as short as possible for stability reasons.

This picture shows a different view of the Control board, this time with the larger ribbon cable removed...

Another view of the underneath (106k)

The yellow rectangular lumps are relays - the large ones in the middle of the Control board are output isolation, and also deal with the current meter source - when the outputs are off, they show the chosen limit value. The smaller ones on the right select tracking modes. I considered CMOS analogue switches for them, but relays are much easier, and I've got plenty of them...

On the left end of the Control board are the connectors for the power transistors and unregulated DC voltages from the Rectifier assembly. These are loomed together inside PVC sleeving for neatness, and to offer some protection from the heat in the area...

Front Panel Removal

The front panel is a complicated assembly, so I made sure that it's easy to remove from the chassis. The aluminium sheet supporting the Meter panel and Logic PSU is attached to the main chassis by two M3 spacers - one of which also secures the Meter panel. But, once you've undone them, disconnected the 50 and 34 way ribbon cables, and undone the 2 side screws, here's what happens:

Front panel assembly (83K)

The whole lot becomes free, attached only by the mains wiring. If required, you can undo the 2 small screws in the back of the switch, and unbolt the earth tag - then it's completely free. In practice, you only need to do this to get to the Meter panel, as the Logic PSU is easily removed in-situ.

Here's another view, showing the back of the front panel. You can see the loom required to connect the 50 way header to the front panel pots, 4mm posts, and the CC/CV LEDs (via left grey Molex connector). I won't doing this again in a hurry!

Another view of the front panel (98k)

The 5 volt rail supplies the LED panel meters, and the associated regulator generates about 3 watts of heat. It doesn't warm up too much because the front panel is an effective heatsink. I wanted to avoid too much heat in this area because the ICL7107 DPM ICs get quite warm. Also, while the separate voltage reference IC is quite good, it makes sense to not heat it up unnecessarily.

The mains earth 4mm post is connected directly to the mains loom - this is securely connected to the main chassis. The 12 way Molex loom splits into two, and connects to the output and meter mode switches, and the small board holding the tracking mode switch and LEDs. From there, the twisted white wires go to the power switch lamp.

Back Panel...

The back panel removes just as easily, should power transistor replacement be necessary. The only connections are the transistors (one connector), the thermistor (one connector) and the mains wiring (hopefully enough slack!)

Rectifier assembly 

Here's a close-up of the Rectifier assembly, showing the main smoothing capacitors and rectifiers (that could do with some heat sinking!). Apart from the screw-terminals for the AC input, all connections to this board are via the PVC-enclosed loom, bottom-left.

The aluminium-clad resistors are the current-sense devices. They don't generate much heat, but by clamping them to a heat sink, you can be sure that their tempco (already a respectable 50ppm/°C) doesn't appreciably affect the accuracy of the current measurements.

The smaller additional board was an afterthought. It generates additional rails of opposite polarity for the control circuitry, which get regulated down to 5V on the main board. Initially I simply used the opposite rail, but quickly realised that this introduced currents in the shunt resistors which affecting the accuracy of the current meters.

Note the use of LEDs - these are useful for 2 main reasons: quick visual indication of supply presence, and also as bleed devices. The output relays isolate the output as soon as the mains is switched off, meaning that almost no current is taken from the large smoothing capacitors - as a result, the green CV lights stay on for about a minute after you turn the power off!