Power-One Power Supply Hackers Page

Visit the Blog for this project


Note:
To access the ExpressPCB .PCB and .SCH files for this page, use this .ZIP archive. Each .SCH and .PCB link below also links to the same .ZIP file.
-- Dave



Power-One has built open frame linear power supplies since the 70's. These workhorses have been built nearly the same way for so many years, and there are many available on the new and surplus markets. The design is simple enough, the technology is understandable to someone with basic electronics background. In other words they are ideal for hacking. What can be done with them? In addition to using them to learn about power supply theory, I have hacked many power supplies and find these to be fun and easy. I describe simple hacks from adding a meter or two, widening the voltage range, coaxing them to do adjustable current limiting. All these can be done with a few external parts. My grand finale is to eliminate the ancient uA723 regulator and replace it with a proper full range, precision voltage / current limit design with good precision monitors and even a PC to control and monitor it.

Modified, they can be used as lab supplies, battery chargers, or any place a lab supply is needed. They can even be scavenged for parts or better still, subsections, they can be re-packaged into a different chassis.

Here is a 24V 2A supply that I hacked years ago and have used as a lab supply ever since. I set the voltage range to 4V to 25V, removed a second board and drilled the chassis to make room for the AC line cord, switch and fuse, binding posts, and a Digital panel meter and 10 turn pot. I mounted rubber feet on the bottom to keep it from scratching up the bench and also to isolate its heat from the bench. If I did it again, I would add a separate panel for these additions and leave the original board intact. It provided useful +/-5VA outputs.
Hacked Power-One
      SUpply
Each Power-One supply has everything that a linear power supply needs: An unregulated DC power source, an output stage capable of dissipating some power, a regulator containing a reference voltage, a difference amplifier and a current limiter. After 35+ years they all still use the DIP version of the venerable uA723 Voltage Regulator. What else do they all have in common?
Condor and International Power also build nearly identical supplies. Other manufacturers may also. When you are searching for a supply on Ebay, use all the vendor names in your searches. The differences are subtle, if any.

Caveats:
Needless to say this page is not sanctioned by any power supply manufacturer. By disassembling and modifying a supply you will of course void the warranty, and you may never be able to get your power supply back to its original splendor. You may even blow something up. These things operate from raw AC line voltage which can be LETHAL. I *STRONGLY* recommend that you:

Insulate all AC wiring with heat shrink or electrical tape
Add a power switch and fuse of the correct value
Be careful when operating these or any power supplies

Regarding efficiency, linear supplies are quite inefficient, and therefore burn holes in the ozone layer, cause more coal-burning power plants to be built, and speed the decline of civilization. They are fine for low power applications or occasional use such as a lab supply. But if you're going to keep any power supply heavily loaded for long periods of time, please do the environment a favor and use a nice efficient switcher. Your electric bill will also appreciate it. While you're at it, replace your antique incandescent light bulbs with compact fluorescent bulbs. Each bulb saves more power than a typical linear supply wastes.

The Supplies

Here is a simplified circuit of a Power-One supply. This is a +15V, 3A supply similar to the HC15-3. Power-One Simplified Schematic
Here is the full schematic of the HC15-3 in ExpressPCB format as well as in .PDF.
Here's a photo of a HC15-3 below.
You can get ExpressPCB's excellent software for free at www.expresspcb.com.
Power-One HC15-3
      Power Supply

Most units have current sense resistors to enable current limiting. These are typically 2-3Watt power resistors, less than 1 ohm, wired from the output transistor's emitter to the supply output. Some supplies use 2 or more resistors in parallel to achieve low resistance and higher power ratings. The resistors are connected to the current sense pins on the uA723. These inputs are simply the base and emitter of an internal transistor. When the voltage from pin 3 to pin 2 exceeds about 0.65V, the internal transistor conducts, and pulls the output down. I discuss the details of this further on.

Higher current supplies use multiple output transistors in parallel. To achieve proper current sharing, each transistor has its own emitter resistor. Without the resistor, the hottest transistor would have the lowest Vbe which would cause even higher collector current, causing it to get even hotter. This effect is called thermal runaway and ultimately can cause failure of an output transistor as it hogs all the supply current. Emitter resistor help to balance these currents. When multiple transistor and emitter resistors are used, the current limit sense is taken from a single resistor with the assumption that the currents are all close to the same.

Most supplies have an extra 'boost' supply which is typically a half-wave rectifier and cap that provides a bit more voltage than the main supply. This is to provide the few extra volts that the regulation circuit and output transistor requires without increasing the main supply. If the main supply were simply increased, the output transistor would need to burn the power from that extra voltage drop. Lower voltage supplies simply double the V+ with an extra diode and cap. Some like the one below use a 'bootstrap' approach where a separate winding of the transformer provides a 7-10V DC extra voltage. This voltage is connected in series with the supply output, thus providing a voltage equal to 7-10V more than the output voltage.

Most are built around single-sided phenolic PC boards.The usual mounting method is via the TO-3 transistor mounting screws. The board can be removed by:
Unsoldering the pins of all T0-3s with a solder-sucker
Removing the screws that mount the T0-3s.
Carefully seeing where all the insulation hardware goes so you can replace it later.
Some newer units use socket pins for the TO-3s, so you don't need to unsolder the transistor to remove the board.

Remote voltage sense is provided on some supplies. Look for an extra pair terminals near the voltage output terminals. Remote sense is a "Kelvin" or 4-wire connection used to compensate for voltage drop in the power wiring by sensing the voltage at the load instead of at the power supply. A second pair of terminals is provided, and these are wired to the remote load. These terminals are used by the supply to measure the output voltage, so the + side is typically the voltage feedback to the voltage divider, and the - side is the V- or GND terminal of the uA723. Low value resistors are provided between each sense terminal and its power terminal so that the supply will still operate when the remote sense inputs are disconnected. The schematic of the HC15-3 in .PDF shows this.

Over-Voltage Protect is provided on some supplies. It is also called a "Crowbar" because it shorts the output of the power supply in the event of a fault that increases the output voltage too much. For example, on a +5V supply, the overvoltage is set to about +6.2V. This is implemented with an SCR and voltage sense circuit consisting of a Zener diode and resistors. When the voltage increases, the SCR turns on and shorts the output of the supply. If the supply is functioning it will current limit. If the supply's output transistor has shorted, the current limit won't function, and the AC fuse should blow due to a direct short on the power supply.

Some supplies use circuitry on the board for this. Some use an external board wired to the output terminals. When hacking a supply, you typically don't want it to ever crowbar, so I recommend disabling the overvoltage circuit. This can be done by simply disconnecting an external overvoltage, or by removing the SCR on the on-board types. The SCR is typically a TO-220 device located near the outputs. Unsolder and unscrew it, or clip its leads.

uA723

The uA723 (same as the LM723) voltage regulator has been around since the 60's. It is the core, the brain, the technological center, if you will of each Power-One supply. It contains all of the building blocks of a voltage regulator: voltage reference, difference amplifier, power stage, and current limiter. Below is a block diagram, courtesy of TI. Here is the uA723 full data sheet.

ua723 block diagram

This IC is very flexible and low cost. But it does have limitations. The voltage reference is nominally 7.15V. The minimum power supply to make this reference and the rest of the device operate is about +12V. The maximum voltage is +40V.

The inputs of the difference amplifier have a minimum (common mode) voltage range of +2V. This means that the output voltage cannot easily go below +2V without playing games. Most Power-one supplies operate these inputs between +5V, the lowest standard voltage output available, and the +7V reference. In fact there is a spec in the '723 data sheet that says that the + and - inputs must see no more than +/- 5V. So keep these signals between 2V and the 7V reference.

The '723 output voltage can be about 2.5V below the V+ and VC pins. Internally, the '723 uses a current source followed by two emitter followers and can output a maximum of 150mA. In Power-One supplies this is used to drive external TO-3 output transistors that boost this to a couple of amps. Most supplies use a 2N3055. The minimum beta (current gain) of a 2N3055 is about 25, so this 150mA could theoretically be amplified to about 25 x 150 = 3.75A in a single TO-3. It is a bad idea to operate the '723 at this maximum current, and 50 mA is more reasonable maximum. The maximum power of the '723 is about .75W, and at 150ma this only allows .75W / .15A or 5V of drop between the V+ and the output. Most power-one supplies operate at fairly high voltage drop.

For example, the +15V 3A supply that I have uses another transistor, a TO-220 TIP29 between the '723 and the 2N3055. Figure the TIP29 has an additional beta of 25. So the '723 provides 3A / (25 * 25) or only 5mA. Another supply I own uses a darlington output transistor type 2N6059 to provide the higher current gain required.

To identify if your supply uses Darlingtons or vanilla NPNs and assuming you can't just read the number off the TO-3 and find a data sheet, load the supply with a nominal load (~0.5 A) and measure the voltage drop from the base to emitter terminals of the TO-3. If you get 0.6 to 0.8V, it's a single transistor. If you get 1.1 to 1.5V it's a darlington.

About 2 or 3 amps is the maximum you want to draw from a single TO-3 without forced air cooling. A 2N3055 is rated for 60V, 10A and 115W, but 115W is only under ideal conditions: with an infinite heat sink and at 25C. Lower current, up to 5A, Power-one supplies have only the aluminum case as the heat sink. Larger supplies up to 50A often use additional finned heat sinks and encourage forced air cooling. Needless to say a 50A linear is a generally bad idea due to the power waste or hundreds of watts at full power.

I estimate the case heat sink to have a thermal resistance of about 3 degrees C / Watt without a finned heat sink and without forced air cooling. To keep the temperature rise to +50C (+25C ambient = 75C, quite hot!) , the transistor can dissipate 50C / 3C/W = ~17 Watts. At 3A, the transistor can only drop 17W / 3A = ~6V.

If you modify a supply and want to operate at lower voltages, you will drop much more voltage across the transistor and burn more power. Lets say you want to operate a +15V, 3A supply at +5V. At the same 3A, the output transistor will burn (15V -5V) * 3A = 30W more. The additional temp rise at 3C/W is +90C. Ouch! Without forced air cooling you will need to derate the current of the supply by 1/3 to 1/2. It's a bit counter-intuitive: this supply will output 3A at its full +15V, but at 5V only lower currents.

Keep in mind that the transformer / cap / diodes can still provide the full 3A at any output voltage. For that +15V 3A supply, they provide an unregulated +20V or so at 3A or 60W. The unregulated parts don't care if most of the 60W is being dissipated on the heat sink or in an external load. Well, they do care a bit since they also are mounted to the same heat sink, and when it gets hot, so do they.

Another way to manage more heat is to remove the T0-3 transistor(s) and move to a larger, external heat sink. The device can be removed and simply wired via the traces on the board to the transistor located within 6-12" away. Once you do this, the Power-One case is no longer needed: remove the board and transformer and mount them wherever you want. After removing all the parts, you could even saw the case in pieces to give you separate brackets for mounting the transformer and the board and/or heat sink. If you do this, there are things to watch for: the 2 (or more) transistor screws are the board mounting. This is convenient if you have a transistor and insulator, but not convenient if you moved them elsewhere. Since these transistor connections are still electrically hot, you need to use insulated washers to mount the board to anything conductive. Also on some designs, the 2 screw leads of the TO-3 are used as electrical connections on the board. Once you remove the transistor, these connections no longer exist. If you see traces on the board connected to both screw mounts, solder a wire from one screw mount to the other.

Increase the Voltage Range

Before you do the voltage range mod or the constant current mod, realize that both of these will cause a larger voltage drop across the power transistor at high currents. This is the main source of power and therefore heat in the supply. Get the transistor too hot at too high a power dissipation, and it will fail, usually by shorting its collector-to-emitter. This causes the supply to output its full unregulated voltage. Since you don't want this to happen, it is important to watch the maximum temperature at high loads, and particularly at low voltages. There are 2 basic ways to keep it cool: one is to apply forced air via a fan. The second is to limit the current at low voltages.

The voltage output of a Power-one is set with resistors and a trimpot. Older units use a big metal trimpot and newer ones use smaller plastic ones. Depending on whether the supply is rated for >7V or <7V, the circuit is wired differently. For Vout >7V, the +7V reference pin is wired directly to the + input. A resistive divider with the trimpot reduces the output voltage to 7V to be applied to the - input. The trimpot allows the output to be accurately set and allows for some adjustment range. (Sometimes a single supply is rated for either 12V and 15V operation and the trim is used to set the voltage. These are typically derated for lower currents for +15V operation than for 12V.)

For voltages <7V (usually 5V) , the Vout is wired directly to the - input and the reference is attenuated to the output voltage by a divider and trimpot.

The >7V supplies are easier to hack. One thing you typically want is a wide output voltage range. I like to use the 15V or 24V supplies at 2 or 3 Amps to build lab supplies. I want the output to go all the way to 0, but the '723 won't allow this easily. The '723 spec is +2V but I like margin and don't often need to go below +2.5V. To do this, attenuate the reference from +7.15V to +2.5V by adding two resistors. The resistor ratio is (7.15 - 2.5) / 2.5 or1.86 : 1.00. 1.00K to ground and and 1.87K to the reference will work. Now the output divider can go from 2.5V to slightly higher than its rated V. So for a +15V supply, lets push it up to +16V. The divider here is (16V - 2.5V) / 2.5V or 5.40:1 The large value is a pot or trimpot, so pick its value first and select the small one to match. For 16V, a 10K pot will draw <1.6ma of output current and <25mW. We probably want to use a 10 turn pot and these are readily available. One turn pots don't allow adjustment to 3 1/2 digits easily. Some people don't like 10 turns since in a panic they take a long time to turn down, so you could use a 10K, 1 turn in series with a another 1K in a coarse / fine arrangement. Set the fine pot to mid scale, adjust the coarse to get close, and then tweak the fine. The total 16V range here would be 11K, not 10K.

So 10K * 1 / 5.4 is 1852 ohms Use a 1.82K ohm 1%. Mount the pot(s) and the panel meter(s) on a scrap of sheet metal which can be mounted to the supply or mounted remotely. Remove the resistor in series with the PowerOne voltage trimpot and the low side resistor. Depending on your unsoldering skills you may remove the trimpot too. Replace the low divider resistor with the 1.82K, and the pot plus the other resistor with 2 wires.  Then run wires to the external 10K pot. Remember to wire the unused (CCW) terminal to the wiper for safety in case your pot wiper gets noisy and intermittently opens up. This will cause the supply to go to + infinity which is likely to do bad things to a load. This schematic of the HC15-3 shows the original circuit plus the changes.

If your new supply range is >19.99V, it cannot be displayed on a 3 1/2 digit meter at full resolution. So either use a 199.9V range and only have .1V of resolution, or add a range switch to the meter so at lower than 20V you can get 0.01V resolution. This is what I did on my first supply.

Current Limiting

In order to accurately limit current, it is generally required to measure the voltage across a low value resistor in series with the load. If this voltage exceeds a preset limit, then the supply outputs reduced voltage until the current drops to a safe value. Power-One supplies do this with a single NPN transistor in the '723. Its emitter is tied to the power supply voltage output and its base goes to a current sense resistor. In this way, if the voltage across the sense resistor increases to about +.7V, the transistor begins to conduct, reducing the output voltage. This is a crude current limit, intended to protect the power supply from a short circuited output load. However, this type of current limit causes the output transistor to dissipate the full voltage and current which is a large value of power, causing it to overheat and possibly fail. So an improved method is needed. This is called foldback current limiting. With this approach, when a supply goes into current limit, the current is reduced to a safe value, typically 1/2 or 1/3 of the limit value. removing the load allows instant recovery.

The '723 can be outfitted for foldback by simply adding a resistor divider to the current sense input such that as the output voltage drops, the current threshold also drops. This basic circuit is shown above. Another problem with this circuit is that it depends on an uncompensated Vbe voltage which varies about -2mV/ degree C. And also without varying the low-value current sense resistor, there is no easy way to reduce the current limit value. However, Power-One has come up with a clever solution, shown in the schematic of the HC15-3. Note the current limit trimpot, and how it is connected to the base of the output transistor, not directly across the current sense resistor. In this way, the sense resistor plus the output transistor's Vbe are being measured. These are compared to the trimpot voltage plus the 723's current sense transistor Vbe. Since the Vbe's kind-of cancel out, the voltage across the shunt is compared to the voltage across the trimpot. By reducing the trimpot's resistance, the current limit can be reduced also. The pull-down resistor and diodes provide current to the trimpot and tailor the foldback current limiting characteristics. Very clever.

However, for a lab supply, a precision constant current mode is more desirable than current foldback. In constant current, a control is used to adjust the maximum output current of the supply, and the supply will dutifully and accurately output this current at any voltage from 0V up to the voltage setting. To do this accurately requires an accurate measurement of current, a precision setting, and a control loop to take over the power supply when the current limit is reached. But the Power-One circuit can be easily hacked to provide a decent if not super-accurate current limiter. To eliminate the foldback limiting and allow constant current control, all that is necessary is to regulate the voltage across the trimpot by regulating the current through the trimpot. This can be done by replacing R6 and its 2 diodes with a 1mA constant current source to ground. Now, the 500 ohm trimpot will have a maximum of 1mA * 500 = 0.5V, the maximum voltage desired across the shunt resistor. By varying the trimpot down, the voltage across the shunt is also reduced. This circuit isn't perfectly precise, but can be set to any desired value from about 10% to 100% of the supply's current rating. It will drift a bit as the power transistor heats up though. But it's a lot better than nothing and can be used for many applications. I built this up and found that the Vbe of the current measuring transistor was a stable 646mV. But the Vbe of the 2N3055 varied quite a bit with current: from about 0.5V at 0 current to 0.8V at 2.5A current. The 0.5V came as a surprise to me. This caused the minimum current setting to be about 0.3A, not 0 as I had hoped. The shunt resistor in the HC15-3 was 0.12 ohms. With this change the current could be varied from 0.3A to about 2.3A. And it was pretty stable when in current limit.

For a 1mA current source, the National LM334 can be used. It is a resistor-programmable current source in a TO-92 package. A single 68 ohm resistor is used to set its current. You may have to mess with the 68 ohm resistor value to get the current ranges you want.

This is a crude variable current limiter. A precision one is shown below.

Output Monitoring

To monitor and adjust voltage, a DMM can be used across the output terminals. Current monitoring with a DMM can be done two ways: 1) Wire a DMM in current mode in series with the output, or 2) Monitor the voltage across the shunt resistor and solve ohms law.

Low cost 3 1/2 digit LCD and LED digital panel meters (DPMs) can be hard-wired to serve these functions. For voltage monitoring, use a 5V common ground (not a 9V) compatible DPM. Wire its input as a 19.99V or 199.9V range and set the decimal point accordingly. I like the All Electronics PM-128E for this application. It is flexible and cheap ($12.25). It needs a +5 supply at low current. I cheated on one system and used the +7.1V reference to power the DPM. It has worked for years, but I suspect that that DPM was specified to operate from +7V. A better solution is to use a 5V regulator such as the 78L05, powered from V+. A TO-92 regulator and bypass cap can be soldered right onto the DPM.

For current metering there are several choices. The simplest is to use an analog uA meter wired directly across the shunt with a scaling resistor to handle the 0.5V full scale. The problem with analog meters is that to get the readout scaled the way you want may require taking it apart and marking up the scale. Too much work considering that you still have an inaccurate (5%) current reading.

There are a few ICs on the market that allow current measurements across a shunt resistor, and then provide a nice ground-referenced voltage output. This can then be measured with a DPM or even the same DPM you use for voltage measurement if a selector switch is provided. Some lab supplies use this approach. Problem here is that most of these current measure ICs can't handle the wide range ( 0 to +20 or so volts) of a lab supply. Analog Devices recently announced (as of 7/07) one that should work. A differential amplifier or instrumentation amp (INA) can be used. Watch out for common-mode rejection and voltage range. This must accurately measure 0 to 0.5V across the shunt while its input common mode varies from 0V to +25V or higher. This needs CMRR in the 80+ dB range or resistors matched to 0.01%. Most monolithic INAs cannot handle both the high voltages required plus the need for the inputs and output to go all the way to ground. Most will need a negative power supply. -5V will do.

The voltage across the shunt can be directly measured by a DPM with its ground pin connected to the power supply output, but a floating +5V power supply is needed for this approach. Since many power-one supplies have a floating +V supply in the +7-10V range, this may be able to be regulated down to Vout + 5V and may work with some supplies. The HC15-3 is one example.

Accurate high-side current measurement is one of the gnarly problems in building a good lab supply. Many commercial designs put the shunt resistor in the ground path to simplify this problem.

Power Management and Efficiency

Power management is one of the tougher design issues on a linear supply. The goal is to minimize power wasted while still meeting all the specs for AC line voltage and frequency, and load. The unregulated DC voltage will drop as the AC line voltage drops. It will also drop as the load current increases due to transformer and diode losses. At no load, the ripple voltage will be low, but will increase at high current and at low (47 Hz) line frequency. A regulator has a 'dropout voltage' which is the minimum input voltage at which it will regulate. Typically this is due to the output transistor voltage drops (one Vbe + 1 Vsat ) = ~ 0.7 +1.0 , the shunt resistor drop, plus wiring drop, all at the maximum output voltage setting. For a +15v supply set to +15.2V, this is Vcmin:

+15.2V + 1.7V ( transistors) + 0.5V (shunt) + 0.3V (wiring) = 17.7V.

The negative peaks of the ripple voltage on should not drop below this voltage. On the high end, the '723 has a maximum V+ voltage of +40V. Its V+ is usually a higher (boosted) voltage than the Vc by 5-10V and is done simply to accommodate the additional 3Vor so of dropout that the that he uA723 needs.

Power-One supplies such as the HC15-3 are specified to operate from 104 to 132 VAC, 47 to 60Hz when jumpered for 120VAC. If you can accept a narrower input voltage and 60 Hz only, you can operate with a bit less dropout voltage margin.

Adjustable DC Load

To test power supplies, a load of some kind is needed. This can be as simple as a handful of different value power resistors. The gold anodized power resistors made by Dale and others work great. These can be screwed to an aluminum plate or other heat sink to keep their temperatures low. I keep some 1.0, 5.0 and 10.0 ohm, 10W and 25W values around. Put enough of these in series and parallel and you can load down a power supply. But I never have the exact right values and changing the load usually involved soldering. So I built an electronic load. This load applies a constant current to a power supply of between 3V and 30V. My original one used a simple power N-FET bolted to a big heat sink, an op-amp, shunt resistor and voltage reference. It was self powered from 20V down to about 5V and designed for up to 4A loads. Over the years it was upgraded to 10A. Now I use a surplus IGBT which will dissipate 10A at 12V or more (120W) with impunity. I think it's rated to switch 75A at 400V. At 120W, a fan is a must. The circuit is quite simple. Build the electronics on a little Radio-Shack proto-board and mount it on the biggest heat sink in your junk drawer.

The 1.25V reference diode, 15K resistor, and 10K pot develop a stable 0 to 0.5V. The single-supply op-amp and FET apply the voltage to the 0.05 ohm resistor. 0.5V across 0.05 ohms is 10A maximum. You can parallel multiple larger resistor values if you have trouble finding a 0.05 ohm 10W resistor. To build a 5A load, use a 0.1 ohm, 5W resistor instead. The only trick in building this is to wire the high current path with heavy gauge (16GA or more) wire and treat the 0.05 ohm resistor as a 4-wire device: heavy traces for the high current path, lighter wires soldered to the leads near the body for the voltage measure path. Here is the ExpressPCB schematic. and the .PDF. version. For extra credit, use the unused op-amp and a thermistor to detect when the heat sink gets hot and turn the fan on. The circuit as shown shouldn't be used at more than 18V or so since the full supply can be applied to the FET gate, These are usually rated for 20V max. By removing D3 and always using the external 12V supply, this limitation is removed and the voltage can go up to anything the FET can handle. Watch out for maximum power of the FET though. And remember to de-rate the FET power at high temperatures.
Load Circuit
Load
The top binding posts are for the load and the bottom ones for a current monitor via a DMM. The 3 resistors on the right are just spares. Note the big black IGBT in the background, this is in place of the FET in the schematic, but a couple of high power TO-220 or larger N-FETs in parallel will do fine. The big aluminum block is a surplus heat sink with its fins facing down. The brackets on the end keep it somewhat thermally isolated from the bench. With the fan, I have run this beast at 24V and 10A: 240W for short periods of time.

A real lab supply design

A real lab supply can output a precision voltage or current down to 0V and 0A. It has an accurate meter and nice front panel controls. This is more than a mere uA723 can provide. I had the brainstorm to replace the '723 with a small board of precision electronics. I unsoldered the '723 on this supply, and replaced it with a socket. Then the functions of the '723 were replaced with a radio-shack proto board via a 14 pin ribbon cable. Here is the board cabled to an old Power-One supply. Some minor components changes on the supply are needed to make this work on any supply. The divider resistors on the supply need to be scaled or adjusted (or bypassed) to provide +5V full scale. The foldback current sense circuit needs to be bypassed. Just remove all the resistors so that pins 2 and 3 of the '723 connect directly to the current sense resistor.

Lab Supply built from
      Power-One

This supply was originally designed to provide +5V at 3A and +30V unregulated at 2.5A I bought it expressly to build a lab supply out of it someday. On the +5V side (Output 1) I removed the overvoltage SCR and added one resistor to allow the output voltage to go down to +3V. More and more of my projects run off +3.3V.

Output 2 contained the transformer winding, rectifier diodes and filter caps only. It used a wire to bypass all the regulator circuitry and connect the + unregulated supply directly to the output. Note that all the regulation components are missing. I removed the wire, added a .25 ohm shunt resistor, bypassed the trimpot, and added a TO-3 NPN transistor. I also added a 100uF 50V output capacitor. The remaining circuitry is on the perf board and is shown on this .PDF schematic. This is the first prototype of this circuit. It works as follows:

There are two control loops, one for voltage and one for current. Each control loop uses a measurement circuit, and an integrator. A power stage drives the output. Diodes are used to select either the voltage or the current integrator to control the output, depending which is in range. Either integrator can pull the output low if the output tries to go above its limit. The current sense consists of a differential x10 amplifier to bring the 0.5V signal across the current sense resistor to +5V full scale. Its common mode rejection needs to be very good, like 80dB in order accurately measure the small voltage across the shunt while the output voltage varies from 0 to full scale. To measure CMRR, unload the power supply (0 current) and see what the meter reads. Then vary the power supply voltage and see how it changes. The ratio of voltage change to meter change is CMRR. To build it with readily available parts requires that the ratio of its 2 pairs of 10K / 100K resistors be matched precisely, better than 0.1%. You can do this with a 4 1/2 digit DMM which will resolve the resistors to one part in 10,000. Or wire up a whetstone bridge driven from a 10-12V power supply, and use your 3 1/2 digit DMM on the mV range for the null measurement.

You also need the op-amp to have good CMRR and be able to handle the full unregulated voltage. (See the LT1013 below.)

The voltage sense circuit consists of a simple voltage divider. A goal of both these circuits is to scale both signals to 0 to 5V (or whatever reference voltage) full scale. That way they can be measured precisely with a simple ground-referenced DMM or A/D. That way the current and voltage adjust pots also operate off the same reference voltage.

It did work OK, but it had a few problems. The current sense amplifier output cannot go all the way to 0, and neither could the supply output. This is due to the LM358 inability to pull its output to ground. Also to switch from voltage to current control and back, the integrator outputs need to slew the full power supply voltage of about +40V in order to take control. This takes time and in response to a large current change, the output will droop or rise (kick) while this is happening. Also LM358As used are only rated for +32V operation and I was using them at +40V. So they are probably not long for this world.So I changed the op-amps to LT1013s which can handle 44V. The ability to go to 0V is pretty important. since when you short the output with a current meter to set the current, the loop must regulate with 0 output volts. So I added 2 series diodes between the regulator and the TO-220 driver transistor. This did the trick. On this early proto I was wiling to live with the current measure not going all the way to 0. A 1K resistor from the current amp to GND would help a bit here. But at full current (5 V out) the op-amp would dissipate 5mA X 40V = 200mW and get hot. Not terrible, but not ideal.

One fix for the 0 voltage and current problems is to build a low current negative (-5V) supply for the op-amps. Another is to bias the current measurement opamp so 0 current is actually +0.5V or so, but this complicates the A/D or DPM design. The integrators and voltage amp can be operated from a lower supply voltage, like +12V to reduce their slew times. Then this lower voltage can be amplified up to the required output voltage by an additional amplifier. But the current sense amp and the output amp cannot operate from lower voltages. Fortunately there is a nice part, the LT1013, a reasonably priced, precision, dual, single supply opamp that can handle up to 44V. Also I want an LED to tell me when the supply is in current limit mode. The difference between the voltage at the output of the integrators tells this and a comparator will do the job.

I don't mind hand-wiring a simple first proto to prove out a concept, but when it gets complicated with support circuitry, or I need to build more, not so much. So my plan is to build an ExpressPCB to do this. The first version will prove out all the analog stuff and will probably use an LCD DPM to do the monitoring.
Here are the original schematic, PDF Schematic and PCB artwork so far. I added a +12V and -5V supply, limited the range of the integrator opamps to +12V, changed the op-amps that run off -5 and +35 to LT1013s which can handle up to 44V. The design supports either a single DPM with a switch to allow it to read either Volts or Amps, or 2 DPMs, one each for current and voltage. It supports an LED to show when it is in current limit mode. The layout uses just 1/2 of an ExpressPCB. The right half can be a copy of the design (if you need 6 boards) or can be anything else you can fit in the area. ExpressPCB only allows 350 holes on a mini-board, and this design is right up to that limit. If you add anything you may have too remove something also.

For the +12V regulator, I used a LM317 adjustable regulator. These can handle the higher input voltage. The LT1013 with a -5V supply is specified for 44 - 5 = +39V. This is fine for a +15V supply but a +24V supply will sometimes have its raw DC as high as +40V. Then with a line surge, bad things could happen. A 5V zener in series between the V+ and the op-amp supply would do the trick.

There is an effect called "integrator wind-up" where a control loop takes extra time to respond to a change because the integrator has gone off scale in one direction and then needs to integrate all the way in the other direction in order to regain control. With a power supply like this, let's say a load is applied that causes the current limit to take control. When in current limit, the voltage integrator goes to its + full scale. Then if the load is suddenly removed, the current integrator will integrates up until the voltage integrator comes back down to take control. This takes time and causes a jump or overshoot in the output voltage. There are many strategies for eliminating integrator windup. One is to simply reduce the output voltage range of the integrators. The integrators need not rise above the maximum control setpoint of the supply. The integrators are powered by the +12V supply. It may be desirable to reduce this voltage in order to reduce wind-up. So having an adjustable regulator here might be a good thing.

When multiple output transistors / shunt resistors are used, one could simply sense current from one shunt and multiply it times the number of shunts. But this approach won't be very accurate. How to measure the current from several shunt resistors and sum them up while still maintaining the resistor matching required for high CMRR? I didn't want to build multiple high CMRR amps so I came up with a simple resistor network to manage this. The magic value is 500 ohms. By adding a 500 ohm resistor to each 10K, the CMRR is still well balanced. To sum 2 shunt resistors, replace the + side 500 ohm resistor with two 1Ks, one for each resistor. For 3 shunts, use 3 x 1500 ohms. for 4, use 4 x 2.00K. All these resistors in parallel equal 500 ohms. All these are available in 1% except for the 500 ohms. But 499 ohms is close enough. Even better, these multiple resistors can each be sky-wired directly to the shunt resistors and only their common terminal needs to be returned to the 10K of the current measure amp.

When you are done building this dandy controller, it may be apparent to you that the power-one supply is not all that necessary. All we're really using of it is the transformer, rectifier, TO-3, heat-sink (case) and filter cap. These parts can be bought or scrounged separately, mounted to sheet metal pretty easily and the Power-one board can be pretty much eliminated. Advantages of this approach? Since you need to derate the Power-one current spec by about 2/3 to 1/2 when building a lab supply, the transformer, diodes and cap are over-designed, meaning overly large. The transistor is OK, but the heat sinking of the power-one case will be too light. Adding a real finned heat sink and mounting it on the rear of your package will be a big improvement.

Power Supply Dynamics

The dynamic or transient response of a power supply is design dependent. Raw Power-one supplies are pretty simple. Their frequency response is primarily dictated by the output filter capacitor plus the uA723 compensation capacitor. Complicated a bit by the fact that the output transistor can charge the output but not easily discharge it. The supply is designed to provide a steady DC controlled by a trimpot, so as long as the output tracks the trimpot during adjustment, all is good. A lab supply has a much wider range of adjustment. And a programmable lab supply can jump from one voltage setting to another pretty quickly.

Another aspect of transient behavior is the response of the current limit circuit. On a power-one, the speed of the output to drop during current limit and to recover after isn't real important. However you don't want the output voltage to overshoot much when the output current is removed. With a lab supply intended to operate regularly in both constant current and constant voltage modes, the circuit should be able to switch modes cleanly. The circuit that does this is sometimes called a 'crossover' since it automatically switches from constant current to constant voltage. One simple way to build this is with two integrators, one for voltage and one for current control. Then the output of each integrator is simply analog 'or-ed' with two common-anode diodes and a pull-up resistor. The lowest integrator voltage limits the output voltage. If the voltage output is higher than the voltage setting, then the voltage integrator slews down to reduce the output. Same with the current integrator. Lowest integrator wins.

This works well except for one important case: When the unit is in current limit mode, the voltage integrator goes to it's maximum value, set by the integrator's supply voltage or by some limiter circuit if present. Then if the current load is quickly removed, the current integrator slews up to increase the output voltage. But the voltage integrator can't begin slewing downward until the output is above the setpoints. The net result is that it takes time for the voltage integrator to respond and the output overshoots it's setpoint. The 2-integrator circuit I built causes about a 1V overshoot. This extra 1V can damage sensitive circuits. Not good. The resistors in series with the integrator caps help. They add a "P" or proportional term to the loop so an instantaneous change in an output ( I or V) causes a quicker response of the integrator output. A "D" or derivative term can also help here. But it cannot be corrected for all cases.

A more elegant solution is to use a single integrator with a 'crossover at it's input. The charge on the single integrator controls the output. The single integrator either integrates voltage error or the current error with some simple decision logic. The voltage error normally controls the integrator. But if the current output is above the current setpoint, the current error controls it. This can be done pretty simply with a 'precision diode' or absolute value circuit.

Packaging a lab supply

Packaging homebrew projects is always a challenge. You want them to look decent but don't want to put infinite work into the case. Trying to jam your latest creation into an off-the-shelf box can be frustrating. One part won't quite fit so you need to go to the next bigger size box which now is mostly filled with air. The box looks, well, off-the-shelf. For lab projects or prototypes, any old box or even no box can be used. In the early days I built boxes from wood and sheet metal. Plywood for a quick-and-dirty, oak and plexiglas for something that lives in the house, and teak for the boat. But unfortunately, the seventies are over. One advantage of wood, is that it can be machined with a table saw or router, it can be screwed into. It can be painted or varnished. Lately I've been looking at "King Starboard" and other plastics. Starboard is being widely accepted as an alternative to wood on boats since it can be machined like wood and never needs refinishing. It's available in marine stores in mostly white, and on-line in black, gray, white and off-whites. Plexiglas (lexan is good for machining, acrylic is too brittle) is also available in clear and other colors. Contact your local plastics dealer or buy scraps on Ebay.

My box concept: 2 vertical sides of 3/8" or 1/2" plastic, machined with slots for the top and bottom, and and threaded for screws for the front and rear. The other panels are 1/16" or so sheet aluminum either painted or left natural. The front can be 0.060, 0.090 or 1/8" aluminum to give  a sense of strength. 1/8" aluminum rack panels are one way to go. These can be obtained in a nice brushed aluminum finish for a price. Heights are standard 1U/2U/3U ... sizes. The front of this design overlaps the sides, top, and bottom to hide the plastic from the front. I'd round those sharp corners. Use contrasting black socket head (allen) screws and it will look pretty decent. Here's a .pdf drawing of what I have in mind. Visio is great for seeing if things will fit. Measure the components of your project, draw them as simple shapes, then arrange them to fit the enclosure, or change the enclosure size to fit the objects. You can do a nice 2D dimensioned drawing with Visio.

The real deal

Here is the final HC15-3 power supply. This one is designed to output up to +18V at 2A.




For the low voltage HC15, I was able to build a nice simple 4 op-amp circuit. Two 8 pin DIP op-amps, a +5V regulator, and a -5V converter fit nicely on a small RS breadboard. The design does 0.0 to 18V at up to 2A. The transformer and board came directly from the HC15-3. The DPM is a Modutec 0-2V model.

The Mods to the HC15-3 board are as follows:

Remove the uA723 regulator and replace it with a nice machined-contact IC socket.

Bypass the boost power supply. This requires removing R1 (220 ohms) and adding a wire to connect V+ to VC.

The clever current limit was bypassed: remove the current limit pot R5 and CR7, then wire U1 pin2 to the + side of the current limit resistor R2.

The voltage divider was bypassed: remove R9 and R10 and adjust R8 for it's minimum value.

Remove the output transitor Q2, bolt it to the heat sink (with proper insulating washers) and run wires from the board E-B-C connections to the transistor on the heat sink.

Check your wiring. Make sure:
The power board + Sense pin is connected to the uA723 pin 4
Pin 3 is connected to Vout+
Pin 2 is connected to the + (input) side of R2

Without the uA723 installed, power it up and check that V+ and VC are both about +23V

After turning off the AC power, you'll need to discharge the cap with a 10 to 220 ohm resistor from V+ to VS- (GND) to safely proceed.

Instead of a crimp-style DIP cable, this time I built a my own cable from a 14 pin DIP header and six #24 wires. All you need is V+ (12), IN- (4),V- (7), ILIM+ (2), ILIM- (3) and OUT(10). When soldering to a DIP header, put it in a socket to prevent the heat from melting the body.

The DPM is a 1.999V range unit with a +5V power supply. A voltage divider measures the V out and divides by 10.0. I use the gain adjustment on the DPM to calibrate the voltage meter range. Make the supply DPM agree with an external DMM. Then to calibrate current, short the output through a DMM set to 2A or 10A. Set the current limit pot to about 1.5A, and adjust R20 till the DMM and the DPM agree I used a fixed value for R20 but a 5K trimpot would be better.

The voltage adjust is a nice 100K 10-turn pot. The current pot is 1 turn only. I was too cheap to use another $10 10-turn pot. The heat sink is surplus, intended for a big Apex power amplifier module and re-drilled for a TO-3. The sheet metal was scraps, and the sides are 1/2" King Starboard in a beige color.

Here's the schematic in .PDF and in Express. It's a simple 2 integrator design that uses a -2.5V reference voltage. The adjust pots balance against fixed resistors that control the full scale voltage (R14) and current (R18) values. The integrators use GND as their + input, and output up to +3.3V. The gating diodes add +.7V to make +4V maximum. This is multiplied by the 5X amplifier to make +20V max. The output stage needs 2 Vbe drops (1.4V) + .25V for the shunt or about 1.7V. So the max output is about 20 - 1.7V, just over the required 18V. In reality the LM358 integrators will output a bit more voltage since their output is being pulled high by the diode and 10K pull-up.

The Current limit LED works reasonably well tied to the Voltage integrator and +5V. In current limit mode, the voltage integrator loses control and goes low. The light is a bit dim when the voltage out is above about +15V, but otherwise works fine.

Since the max V is only +18V and the max I is 2A, the DPM design is simple. The ranges can be 19.99V and 1.999A. it means you can't read over 1.999A though, the meter goes into over-range. But 1.999 is close enough to 2.000.

03/08 Update: VI Design / Battery tester

The power supply was so successful I decided to go to the next level: a computer controlled voltage / current source and sink. This is sometimes called a VI. It can sink or source current, be adjusted, turned on and off so it can charge or discharge (test) any battery. I worked at an ATE company, Teradyne and we built very high quality VIs for semiconductor testing. A proper VI used in ATE probably has many decades of current ranging from micro Amps to Amps and and is a very precise instrument. They are generally 'four-quadrant' devices that can sink or source current at + or - voltages. This one is 2 quadrant, no negative voltages. And it has just one current and one voltage range.

A simple LabView program controls it. No Power-One components this time. 0 to 20V, +3A source or -5A sink. The control loop is similar to the one on the PowerOne hack, but uses an additional integrator to control the sink current. I use a small National Instruments USB data acquisition (DAQ) unit to control and measure it. The DAQ has 2 0-5V DACs and 14 bit ADCs. Mine is the 14 bit USB6009 (but the 12 bit USB6008 ($149) will also work fine. The output stage uses two big TO3 transistors on a heat sink. One NPN to source current and one PNP to sink it. Here's the schematic in ExpressPCB format. and in PDF

If the DAQ had three D/As I would use the third to independently control sink (negative) current, but with just two, one DAC is used to control both source and sink at +/- the same setting. So far it hasn't been much of a limitation. The first DAC controls voltage.

One limitation of this system is that this DAQ (like most low-cost USB DAQs) is grounded to the PC, so the load must essentially be floating. Not a problem with a battery, but as a power supply or load, the grounding can cause errors if there are other paths to ground. Dave's first rule of analog problems: "It's always the ground."

The control circuitry is built on a large RadioShack board. Behind the transformer is another small RadioShack board with the rectifier diode and filter cap. The output connectors and the shunt resistor are on the right. The NI USB DAQ is on the upper left. The big white thing on the left is a lithium battery being tested. The white twisted wires on the bottom is the not-so-safe AC input. It is fused and insulated however. The transformer is about 30VAC, 3A. the final unit will have the AC in and a cooling fan on the rear, hopefully a DMM or two on the front, and the ability to operate it stand-alone (via knobs) or remotely via the LabView DAQ.

The DAQ is about 0.5% accurate and I wanted better, so I calibrated it. Unfortunately NI doesn't support any type of calibration for these cheapie DAQs, so I was on my own. I measured zero and full scale voltage on the ADCs and DACs and calculated the appropriate corrections. Works great and gives be about +/- 0.1% voltage accuracy. But every channel and voltage range needs calibration, and when you change DAQs, the numbers all need to change. I only use one voltage range and 2 inputs so the task wasn't bad.

VI Photo

Here is a screen shot of the LabView application running. It can operate as a normal VI, or can charge or discharge a battery, terminating on a specified condition. The top plot is the instantaneous voltage and current. The bottom plots are voltage and current long-term trends. There are two reset buttons to reset either the time or the accumulated charge. This indicator reads + when the VI is sourcing current and - when it drains current. Here is the LabView code.

For charge termination control, many batteries want to be charged at a constant current until a voltage limit is reached, then monitor the current until it drops to a certain value. This works well for Lithiums, lead acid, and ni-cads. Be careful not to exceed the current and voltage limits on batteries or they will be damaged or explode. A proper high current charger or tester would also monitor temperature and time and terminate if either of these exceeded limits. Exercises left to the student.

For discharge testing, the VI is set to apply a constant current load and the LabView program monitors the voltage. LabView terminates when the voltage drops below a threshold so as not to damage the battery. LabView also accumulates the current every second to arrive at the battery capacity. I accumulate mA-Seconds and scale it to display mA-Hours.

Lithiums are the pickiest of all. Charge voltage is generally 4.20V per cell, not 4.25. Over charge and they die. Discharge below 3V and they die. Too much charge current and they die. Look a them cross-eyed and they die. Don't balance the voltages properly and they die. Or worse, explode.

I have taken apart many old laptop batteries to scavenge their 16450 cells. With careful disassembly and very careful unsoldering, it can be done. The key word is "old". New ones are too valuable to take apart, and old ones lose so much of their capacity that they are not useful. So this has been an un-fruitful endeavor. Finally I found some nice surplus oblong-shaped lithium cells from Saft, about the size of three 16450s. They test at about 5000mAH and are currently powering my bike light. .

Screen Shot

12/09 Update: Boat Anchor or Audio Amplifier?

I just obtained a Power-One  F-24-12-A. This giant, full rack width supply is 24V at 12A  or 28V at 10A. It has nine TO3 devices and 4 heat sinks to dissipate / waste all the power it needs. It also specifies a 50CFM fan to fulfill its destiny as a space heater. I makes little sense to use a 300W linear supply when a switcher would be 1/10 the volume and 85% efficient vs. maybe 60% efficient.  But free was free. The last time I looked there were a handful of these beasts on Ebay for ~$50.
The beastIt would make a pretty poor lab supply. A proper lab supply of more than 100W uses some kind of pre-regulator to reduce the voltage drop and power dissipation in the pass element. Many such as the original HP supplies used SCR stages to pre-regulate and as such are very efficient as well as clean. I love my HP 6286A.  20V at 10A and it is barely warm when outputting high currents at low voltages.

My thinking is to use just the unregulated section of the F-24-12-A to power an audio amp. It would have to be rewired for +/- operation. Since the unregulated section was about 30-35V, a decent voltage for an audio amp, this could work. But alas, the transformer is not center tapped. It is tapped, but only to supply either 24V or 28V raw. A jumper on the board is used to select the tap. So it cannot output full wave DC to both a + and - supply. Too bad. And the filter caps are three at 13,000 uF 50V. They would be fine for an audio amp, but three? You need either 2 or 4. And the diodes are TO3; first time I have ever seen diodes in a TO3 case. To build the diode bridge, there are two different TO3 diode part numbers, one common anode and one common cathode.

A toroidal transformer is best for an audio amp. Toroids throw off much lower magnetic fields which can induce hum in low level audio stages. But this freebie is staring at me, daring me to give it new life. I wonder if this transformer's windings are bifilar, two windings in parallel and can be somehow separated? Sure enough, they are bifilar and the windings can be easily separated. Check it out. Use solder wick to get every speck of solder off the leads. Then use small pieces of heat shrink pushed under the transformer insulation to prevent shorts.

Pre-hackAfter desoldering

There are two 28VAC windings in parallel,  in series with two 3VAC windings in parallel. The 3V is added to the 28V to provide extra volts for the 28VDC option. This means that the transformer can provide either 28VAC, 31VAC, or if you wire the 3V in series with the 28V, but with reversed polarity,  28-3 = 25VAC. And there are two of each of these that can be rewired in series for a center tapped configuration. Cool! The AC options are 25, 28 0r 31V (50, 56, 61 VCT). DC is 1.414 * the AC, minus about 1V for the diode. The loaded supply is about 3 volts less. I measured this with a 3A DC load. Then you lose about 4V in the filter caps due to ripple and the output transistors of a typical amp design , the power into an 8 ohm speaker is about (((VDC - 4) / 1.414 )^2) / 8. This .pdf schematic shows the final design with the transformer connection options. Here is the ExpressPCB .SCH File.

VAC
Wiring    
VDC
No Load
(1.4* VAC) - 1
VDC w/ Load
Power estimate into 8 ohms
(((VDC - 4) / 1.414) ^2 ) / 8
25VAC: 50VCT
28V - 3V
+/- 33 +/- 30
40W
28VAC: 56VCT
28V, no 3V
+/- 37
+/- 34
53W
31VAC: 62VCT
28V + 3V +/- 41
+/- 37
64W

The windings are good for about 350W, so a guess at the total audio output is about 250W. I have wanted to build a multi channel amp based on the National LM3886 for years. The LM3886 is a decent 50W part, so I'll probably go with the middle voltage option. I'll saw the PC board and the bracket just to the right of the 2 caps. That will give me a transformer, the diodes and two caps, just the raw DC supplies,  and will remove all the regulator stuff. The board is mounted to the chassis via the TO3 hardware, so leave the diodes and maybe the next row of TO3s just for mounting.

The unit came apart fairly easily. The 4 heat sinks and 7 of the 9 TO3s just unscrewed. The 3 big caps just unscrewed with an allen wrench. The two  TO3 diodes were a challenge to unsolder, They have big leads and the board holes are a tight fit. After sucking away most of the solder on top, I had to pry the part off one lead at a time. Two big soldering irons, one on each lead would have helped. 

I used a Sharpie to mark where the copper cuts and jumps need to go. The heavy copper layer was too thick to easily cut with an Xacto so I used a utility knife to make 2 cuts and then peeled off the copper strip between the slices. For wire I used #16.

Here is the board prior to being cut in two.
Pre Surgery
And after, showing the board top and bottom. After cutting the traces and adding the wires, make sure to continuity check every connection.
Topbottom

Here is the unit all built up. I kept the screws of the right-most TO3 transistors to use as convenient mounting screws (1) and output terminals (3). You can determine if the output is to be grounded or not at the chassis. Use an insulating shoulder washer for the ground screw to keep is isolated. Remove the insulation to ground it. Typically in an audio amp up want to ground the power supply at the amplifier, not the power supply. This requires using the original TO3 insulators under these terminals.

Mark the TO3 diodes to maks sure you get the right one back in the right place. You can use a DMM on diode range to identify their polarity. And DO NOT FORGET to install C3 backwards from the marking on the PCB: + side goes to GND. I should have crossed out the original "+" on the board.

With no 3V windings connected, the AC voltage is 28VAC. Output voltage is +/- 37V, no load, and +/- 34V with a 3A DC load. Ripple is about 1.5Vp-p at 3A. It's all ready to power a nice audio amp.

For the chassis, after removing the heat sinks, board and transformer, I marked it and cut it with a band saw. A jig saw or a hack saw plus patience would also do it. The heat sinks are mounted to pressed-in studs which remove with a light hammer blow. With the lighter load of an audio amp, the diodes should generate little heat and so no heat sink is required. The chassis alone is a fine heat sink. You could also remove the left end plate since it is unlikely to be used for mounting, and it provides no added strength. You could also kill the chassis bracket completely and just mount the board and transformer to your chassis. You'd need heavy spacers and longer screws for the transformer. You'd need to replicate the mounting holes, but the original chassis could be used as a fine template.

The three unused terminals on top of the transformer are for the original boost supply. This 8.5VAC winding could be used to power some additional logic or analog circuits. I removed the diodes and the cap that they were connected to.
Final Thing

Here are the leftover parts, not including the other half of the cut-off chassis. The chassis, transistors, big cap, and remaining 1/2 board could still be built up as a lab supply, but would need a transformer and diodes. By the way, the 6 removed T03 series pass transistors are 2N3773s, 140V NPNs. 4 of them  would make an nice audio amp quasi-complementary output stage. The .22 ohm ballast resistors and heat sink hardware can also be used.
Box o parts


Visit the Blog for this project

Back to Dave's Home Page

This page was last updated 7/24/10