DIY PS503A Rebuild Prototype

 This is the start of a DIY rebuild project for the Tektronix PS503A dual tracking power supply.

I once build this unit a long time ago, in the late 70's, using several original parts.

This version of the PS503A is a little strange because I can't find any information on the net about this particular unit, that uses 14-pin Fairchild 741 amplifiers, and I don't have the manual anymore. All the pictures that I could find use the more common 8-pin versions. In any case, I sold it to a collector together with all my Tektronix gear quite some time ago. Overview here.

An attempt to create my own dual supply many years later is described on my Blog here.

I'm showing this instrument because my plan is to use the same enclosure and the controls, but with a new PCB front panel. The PS503A does not have a display, but I'm going to use this one. After I build it, I made several changes to the way the display is used, hence the handwriting on the panel.  It is actually an accurate voltmeter with an LCD. I use a switch to show either the positive or the negative voltage. The switch above with the +/-V does that. The switch labeled Tracking will tie the two supplies together such that the voltage set for the positive supply will be tracked by the negative supply. This Master-Slave method is quite common.

Unfortunately, the above supply, that started as a study on using the standard LM317 and LM337 regulators works OK, but not great. The main drawback is that I used a transformer that only allows +/- 100mA at 30V and the current is not enough for many of my projects.

PS503A Dual tracking - really?

The PS503A was marketed as a dual tracking supply. It even says so on the front panel, but is it really?  Most people, me included, would interpret dual tracking as a master-slave configuration, by which one supply, the slave would "track" the voltage setting or changes of the master. However, the Tek engineers figured out a different way. They actually allow you to select the same, or separate output voltages on both supplies and use what they call the "tracking" to change the set voltages together by a percentage. Technically, this is not tracking but setting or adjusting together. Yes, both the output voltages are generated and set from one and the same reference, but still. In any case, with this method you can have for example an output set for +15V and one for -8V and change both voltages with one potmeter together in a relative method, by a percentage change. Great feature and very handy, but it's not dual tracking as I learned to understand it.

The goals for this DIY PS503A project

The goal is to build a standalone unit that can be placed close to a prototype to supply the required rail voltages. My intention is to use this supply when I'm working with Opamps that need a positive and a negative rail. 

As mentioned above, the rather special feature of the PS503A is that it not only has dual tracking, but also with two different voltages. The dual tracking in this case will supply a percentage change on both rails. More information can be found on the Tek Wiki pages here. 

The SG503A goes up to +/-20V and 1A, and I'm planning to see if I can increase the voltage to +/-25V, maybe not with a 1A output, but maybe half of that above 20V. A lot will depend on the transformer, it has to fit in the enclosure. Update: With the now selected 30VA transformer I'm limited to +/-20V, just like the original, but with a maximum current of 600mA. If you can accomodate a beefier transformer and use heatsinks on the power transitors and maybe for the bridge rectifiers, you can easily increase the current with this design.

The challenge with the original design of the supply is that the circuit needs Opamps that with one rail goes to the output limit plus some headroom, and also need a negative supply of several volts. This inherantly limits the maximum output voltage to an output that is several Volts below the maximum Opamp +/-rail voltages. The early 741 Opamps I believe had a maximum of 36V but in reality, could not be used reliably with this limit. The Tek engineers used a trick to use these Opamps in the circuit.

 

The H-P method

More modern supplies use another method, that is using a separate auxiliary supply for the Opamps and they regulate from the supply ground level up or down. My other Power Supply that I designed several years ago uses that method. That method was actually invented by a company called Harrison that was specialising in Lab Supplies. They were later purchased by H-P and became the division responsible for all power supply designs. The method eventually became known as the H-P method because they used it in many of their Lab Power supplies. This smart "trick" allowed engineers to build power supplies with very high (>100V) output voltages.

Applying this method to the PS503A would mean a significant demolition and reconstruction job that is beyond my plans for this supply.

Possible improvements to the PS503A

One of the items I would like to address is the over-voltage protection. The original over-voltage protection for the PS503A uses a Zener with an SCR to cut the output if it is above about 24V. That would be a protection for catastrophic failure conditions.  In earlier versions, like the one I originally built, the SCR would actually short the raw supply rails with a small resistor to the common such that it would blow a fuse. That's pretty dramatic, and that method is generally called a crow-bar circuit. In the latest revisions of the supply however, the SCR now cuts the drive to the output transistors, and will also discharge the output capacitor. This is a clever trick because it  protects the DUT in case the power transistors fail. 

There are several components that do not like a dramatic disappearance of one of the rails, it could actually damage the chip. So my plan is to see if I can couple the fault mechanism to both rails at the same time. However, after thinking some more about this, I did not pursue that and will not add that for this prototype.

The other improvement I would like to add is a tracking overvoltage protection, that would remove the output if the actual output voltage level is only a Volt or so higher than the set voltage, to protect my precious circuits much earlier and at lower voltages, which is especially critical for 5V and below rail voltages.

Building a simple prototype

To better understand the circuitry and try some of the enhancements or changes, I built the positive supply section and the voltage reference section on protoboard. I did that in several steps to test them out before adding more complexity.

Here is the complete positive supply section, working and functional with some improvements already. It does not have the over voltage protection yet. That was to be added later.

What you see here on the left is the original voltage reference section with an alternative I want to use. Next to it the dual +/-9V references (for both supplies) followed by the voltage controller and the output section. The red test lead in the middle is connected to the output terminal, which is the right side of the green 0.6 Ohm current sense resistor. To the left of that are the two output transistors. On the top right edge of the first board you can also see a device on a carrier which is the new current source.

The circuit on the protoboard on the right is the current controller. The red LED is lit, because I adjusted the output current with my Dynamic Load and dialed in the tripping point of the Current Limiter, such that the supply is now in the Constant Current mode. The original PS503A uses a little incandescent light bulb (18V 26mA) to show an indication of the output level voltage. I changed that circuit slightly and now use a yellow LED, and you see that lit on the right hand side. More about that below.

Improvements

As a reference to what I did and will be doing, you can use the original schematic, published by Tektronix of the latest revision.

This can also be found here, when you look at the manual on the right hand side of the page. 

The quality of the picture is not great, because it is a scanned copy of the manual.
Here is the positive supply section I prototyped and will be addressing below:

The reference supply

While I was building and testing the prototype, I noticed a few items that I wanted to address. First of all, I found that the reference supply is not as stable or resistant to sudden temperature changes, although it uses the classic form of temperature compensation. The Tek engineers used a common trick by using a diode (CR24) in series with a Zener (VR24) to counter the voltage change due to temperature. (one has a positive coefficient, and the other has a negative effect) However, that does not work well with sudden changes, like a draft. It also takes quite some time to warm-up to a stable reference voltage. When I used my DMM on the reference voltage and briefly blew some of my breath over it, the voltage jumped before it slowly normalized again.

Granted, this will not happen this drastic when everything is warmed-up and inside an enclosure.

But, it's also not as stable as I would like to have it (see below), and worse, the +33V supply for this circuit will vary with the output load as I describe later. The 33V rail voltage can drop to 24V with a maximum load. The REF01 voltage reference I will use will be better suited for these changes, but I may want to add a small circuit to keep the voltage constant. I need to test that later. The voltage reference chip that the engineers at the time most likely didn't have access to, or it was deemed too expensive, or deemed to be too good for the intended applications for this supply. I will most likely make both options available on the PCB, if there is enough room. Update: There is not so I dropped that circuit.

Here is the schematic (in KiCad) for the VRef and Tracking circuit. On the left in the box the original voltage reference circuit and in the smaller box the new one.

The Vref of 9V, or in my case 10V is fed to two Opamps, in a configuration where one is a voltage follower with the same output as the input, U1, to provide +10V for the negative supply, and U2 that changes the polarity to a -10V reference voltage for the positive supply. Yes, you're reading that correctly,  it's not a mistake.

 

By flipping switch SW1, the fixed reference voltage is now made variable and that sets up the dual tracking feature. The two trimmers are used to calibrate the output of the supplies to the maximum output voltage, plus a bit of overhead room.

Here is the measurement that I took from the original reference voltage output.

This is after a warm-up period of about 15 minutes. It takes a very long time to settle. Note the span of 14.3mV over this 1hr period. This is not all together bad, but it also means that the output will swing with it, amplified by a factor of two. (10V reference amplified to get a 20V output)

The real test is to measure the output of the supply and include all components that could contribute.

Here is how that looks for a small snap-shot:

Not impressed with this, I used a REF01 10V reference chip to replace the original 9V Zener diode reference, and this is the result measuring the output again:

The swings look much more dramatic, but the span is only 3.8mV, a 4x improvement and this was after a rather short warm-up with a new reference that did not have a burn-in period yet.

These measurements need to be repeated with the circuits on a proper PCB and within an enclosure, but the result is a good enough improvement.

I also checked the line voltage regulation because the supply voltage changes from 33V down to 24V with the maximum load. The specification for the REF01 lists 60-100ppm/V but I see no change.

Lets have a look at the various other building blocks of this instrument.

The positive supply

This schematic already shows my latest changes in it, explained next.

When you look at this schematic, you have to realize that the +/-VDC rail, that is actually shown as +/-33V in the original, and the +/-27V rail are not constant values. They are only accurate when there is no load on the output. Both voltages will go down with increasing loads and can get as low as 24V! See the section about the transformer selection below for details. The +/- 27V rails were only added as a limit to protect the 741 Opamps, and provide a maximum, but the voltages will get below that value with higher output currents.

Please note that this is still under development so can easily change. I will do my best to update the schematics as I'm progressing, but don't use them yet to build your own, give me some time to complete it first. There is now a 2.7R resistor from the cathode of Q15 to GDN, and D29 is now a 1N5401, a 2A part instead of the 1A original. Update: With the 600mA maximum, they can change back to 1A devices.

The improvements I already made

The Opamps

Instead of one of the first available Opamps, the 741 Opamps that were used in the earlier generations, I selected the TLE2141 Opamps. Mainly because I have many of them, and they can stand a 44V supply maximum, which makes it more reliable than the originals that went kaduk quite often. Even so much so that lower (+/-27V) protection rails were added to the PS503A design, and even that did not stop all the failures. This could also be contributed to the fact that the early 741's were not that reliable stemming from earlier manufacturing processes. I could have use the TL071, but that also has a rather limited maximum supply voltage specification. As a minimum, during the prototyping phase, I will be using the more rail tolerant TLE2141. It also allows me to raise the +/-6V2 voltages to +/-8V by using regulators. Update: I'm now using 6V4 rails. I have some 36V uA741CP on order so I can try the more modern versions as well (with adjusted rails). There are modern versions of the 741 available that have a 44V rail maximum, but they were not available through my preferred sources.

When you use the relatively slow 741's, you dont need frequency compensation to prevent instability or oscillations. My plan is to test them when I have a complete instrument on a PCB.

The current source/bias for the diode OR and output transistor

This step is a pre-cursor requirement for increasing the supply voltage and is better because it removes the effect of the rail voltage swings due to the load as well.

If you look at the original schematic, I replaced the bias setting resistor (R87) to the Base of the driving transistor Q14 from a 3K resistor to a 5mA Current Sink. This will make the diode OR circuit and Base current independent of the raw supply and the variable current flow when the output voltage is changed from the minimum to the maximum. In the 20V version, that point moves from 9V to about 25V, and hence the Base current changes for Q14.

The just about standard method is to use a current source and not a resistor. In my previous design for a power supply, I used a typical circuit to create a more stable current source. 

Below is that circuit. It used a red LED as a rather stable voltage source to bias the PNP transistor Q2. The Emitter resistor R4 sets the current, R15 sets the LED current. The capacitor of 220uF was added as a startup glitch protection, because it delayed the bias setting until the other supplies where stable.

In that supply design, I used LED's as the diode OR because they were used on the front panel showing the CV or CC mode activity.

There are many other possibilities to create a current sink/source, but I opted to use a dedicated LTC chip (LT3092) and two resistors to precisely set the current. I already had two of them in my stash, waiting to be used. I tried it out and it works really well. By using the formula in the datasheet, I could easily select the resistor value for the 5 mA current I wanted. This is a little higher than the original current that varied between 1.7mA and 4.9mA using the resistor, but that is not very critical.

The Vref for the current limiter

Because I'm now using a different VRef circuit, I needed to change the Voltage reference for the positive supply current limit circuit. In the original, that voltage was tapped from the VRef circuit (connection between R28 and R29), and then fed through a diode (CR78) to the base of Q80. I simply adopted the circuit from the negative supply and used that for the positive supply as well. This adds R41, Q11 and R40 in my schematic.

I noticed that by changing the unregulated supply voltage, the collector voltage of Q3 supplying the CL potmeter is changing only a few mV between 33V (no load) and 25V (maximum load), keeping the voltage across the CL potmeter quite constant at 750mV for the 400mA output load setting, but needs to be checked again when I will go to a higher output voltage from 20V to 30V, or the full 1A current. It may need a voltage regulator or current sink. Update: Not required.

The Voltage output indicator

Because I don't want to use a light bulb for the Voltage indicator (nice but old school), I'm using an LED that many users of the PS503A had to change to as well because these tiny bulbs are now hard to get. I opted for a yellow LED, and that required a value change for two resistors, R14 from 5K6 to 15K and R15 from 510R to 160K in my schematic. Both values depend on the LED current, so may need some tweaking with different LED's. I'm simply using one that I already have, but is without specifications. In my case, the LED is just off with a zero Volt output and will be bright at 20V. When the output is switched off with SW2, the current from R15 is disconnected and the LED is off to show you that there is no output voltage present. Just like the original. When the overvoltage protection kicks in, the output goes to zero Volt and that also turns off the LED.

Decoupling the Opamps

I also added decoupling capacitors for the Opamp rails.

Sziklai output transistor pair

The last change, for the time being, is the change of the Sziklai output transistor pair. In the original, the power transistor was in the TMXXX mainframe and is not so easy to find. I selected a combination of the BD139 and the D45H11. I had a few left from my Walt Jung voltage regulator design, and if Walt found this transistor good enough, who am I to disagree? This selection may change however when I have a real setup that will allow me to do some power stress tests later on. 

If you want to know more about this Sziklai configuration and the difference or pro & cons compared with a Darlington pair, use Google.

There are two main differences that I care about. First of all the input voltage for the first transistor is 0.6V for the Sziklai and double that or 1.2V for the Darlington combo. The other difference is the thermal stability that is reportedly better with the Sziklai, and the two transistors do not need to be on the same heat sink, which was important for the original PS503A with the output transistor located in the TM5XX mainframe. The original TM5XX mainframes used the TO220 versions of the 3055 and 2955 transistors. More information about these transistors and many possible replacements can be seen here.

Resistor value changes

Although I purchased all the required (odd) resistor values, I may change several of them to the E96 range. The reason Tek used these sometimes odd ball values, like 1K05 or 1K62 and 976 Ohm is that in those days, through-hole metal film 1% or better resistors only came in these values (not even in E192). 

The BOM will show some of the alternatives.

The over voltage protection

I waited to describe this circuit because I wanted to test and understand it first and also see if I could improve on it.

After adding the circuit to my prototype I finally figured it out. It turns out that I totally overlooked the presence and clever functionality of CR66 in that original circuit. (D29 below)

Let me explain it with this circuit section inside the box.

D29 is now a higher current 1N5401. Update: With the 600mA maximum, I can change back to the original 1N400X.

The output voltage is tapped by R13 and the 24V Zener D28 that go to the Gate of the SCR. That voltage is filtered by C47, and R42 pulls the Gate low to avoid false triggers and provides the bias current for the Zener. 

So, if there is a voltage on the output that exceeds the Zener voltage, the Gate of the SCR gets pulled-up and is triggered. When that happens, the Anode of the SCR will be shorted to the Cathode at GND level, pulling the Bias of the diode OR also to GND and thereby removing the drive to the Base of Q14 and that in turn switches off the output transistor Q13. So far so good.

However there is also the 50uF output capacitor C2 that still dumps it's capacity into the DUT. With a short, this current can be many amperes, probably a lot more than you think. Keep that in mind. The Cathode of D29, which function I initially overlooked, gets pulled down and conducts and that shortens the output through D29 and the SCR to GND. Great solution, however, there is even more to it. 

In case the output transistors malfunction, like an E-C short of Q13 which will be the cause of the over voltage, they will no longer react to the removal of the Bias to the Base of Q14. Due to the circuit design, when there is no limiting drive from the diode OR, the output will automatically slam to the unregulated voltage rail level. In these cases, D19 will come to the rescue again and will also short the unregulated 33V rail to GND and that will blow the main fuse. 

 

So this clever D19 trick takes care of all possible overvoltage causes and protects the DUT. Hats off to the Tek engineers, I just didn't get it at first.

I added a 2R7 resistor to the cathode connection of the SCR to limit the shorting current a bit. This is also in the negative supply, so why not. 

The tracking overvoltage protection

Before I can start to work on a PCB layout, I wanted to investigate if I can make the over-voltage protection "tracking" so it will not just watch out for a serious overvoltage crisis at the output, but also protect against a certain voltage delta above the actual voltage setting when the voltage regulation fails.

This is especially important when you're dealing with low voltage devices, because the original over-voltage protection will only activate when a voltage in access of 24V is detected on the output. That could happen when there is a failure in the output section, the Sziklai pair as an example. Unfortunately, by the time the over-voltage protection is activated, your DUT could already have died from a voltage overdose.

The tracking over-voltage protection will activate when approx. 1..3V above the set voltage is detected, so when your working with 1V8, 3V3, or 5V devices, they will hopefully survive. 

Here is my current implementation in the lower box below for the moment:

I'm using U13 as a comparator to determine the tripping point. The output of the voltage regulator Opamp U7 is used as the reference voltage for U13. The output of U7 has almost the same voltage as the output, so varies between 0..20V. I'm also tapping the output terminal and drop that voltage by two diodes in series to create a delta between the output and the set voltage. C51 and R52 are there to provide a bias for the diodes and also a filter.

When experimenting with it, I noticed that the protection would activate with the voltage setting potmeter fully counter clockwise (0V). This was caused by the fact that the output of U7 was going slightly negative in that position, and that would trigger the comparator. Adding a small resistor (R58) in series with the voltage setting potmeter lifted the output to less than 100mV positive, and that solved that problem and still provided an over-voltage trigger. The value of this resistor will also depend on the Ohms range of the wiper of RV10 in the CCW position. It may become a calibration point later.

In the lower output voltage ranges, the overvoltage detection is about 1V higher than the reference and goes up to about 3V at the maximum output, which is a nice side-effect thanks to R52.

There is one caveat however. If your DUT has a sizeable capacitance or if you're using a battery, rapidly decreasing the output voltage could trip the overvoltage protection. You need to be aware of this, or you can install a switch to disable this circuit. Later testing with the completed instrument will verify the functionality some more and determine any required changes.

For the first prototype, I've added jumpers so I can disconnect this protection.

The negative supply

This supply is almost a mirror image of the positive supply.

Note that in this schematic, I made an error by switching the polarity of D22 and D23. They need to be "turned around".

I copied the current limiter reference voltage circuit around Q10, D21 and Q4 and also use that on the positive supply, because I use a different main voltage reference for the current limiter. C18, on the Base of Q9 is not in the original circuit, but I've added it here too, just in case it is needed. D34 is now a 1N5401. (Update: No longer needed with the 600mA maximum)

The overvoltage failure protection looks a bit weird. It fooled me. David Hess, an expert I got to know on the EEVblog and has several PS503A units, had this to say: 

"The cathode of the SCR is tied to the *negative* output via power diode D24 and the base of Q12, so the gate of SCR Q17 must be pulled *positive* relative to the negative output to trigger, which means the trigger current comes through D35 and R55 from ground, which is the positive part of the negative output." 

(I translated his Tek references to mine using the schematic above)

In order to test the overvoltage failure protection and the overvoltage tracking protection, I built the negative supply voltage controller on protoboard as well. It's on the lower protoboard.

I was pleasantly surprised that the design of the tracking overvoltage protection worked well right away for the negative supply as well, this does not happen to me very often (;-))

The Power rails

Apart from the reference voltages, there are actually three different dual rail voltage pairs in the original, the unregulated +/-33V, the "regulated" +/-27V (this is only an overvoltage protection for the Opamps above 27V, but this can drop to 25V depending on the load) and the regulated +/- 6V2. Regulated means simply set by a Zener diode. The original for the +/- 6V2 rails used large 750 Ohm 2W resistors and 6V2 Zener diodes, but I don't like that at all so I'm simply using voltage regulators during the prototyping.

During the prototype phase, I'm using PTC fuses, in the final version, I will test the functionality and if required, replace them by glass fast blow fuses or a MOSFET to make the overvoltage failure detection work. The final value will be dictated by the maximum current output.

When I was doing the layout I found that there was not enough room to add frequency compensation components. They could be required when you use fast Opamps.  To also allow the 741 Opamps, or your preferred favorites, I added LM317/337 TO-92 regulators for the +/-8V rails so they can be adjusted with just one resistor to any voltage you desire, or need, for your Opamp selection. I will initially select a +/-6V4 rails to accomodate the uA741C versions I ordered.

Transformer selection challenges

I have now selected the main transformer and this is the process I used.

I have a nice shielded dual 24VAC version with 30VA ratings that should have a current capability of 625mA per winding and I know it will fit in the enclosure. But, is the 24VAC sufficient? 

The TM5XX mainframes have two 25VAC windings per slot that supply 0.5A or 25VA for the normal compartments, and supply 1A or 50VA for the high power slot in the TM504 or even 1.2A or 60VA for the two high power slots in the TM506. The TM504 list the windings as 25VA or 50VA, but the schematic diagram for the TM506 shows all the windings as 24.8VAC. That's a bit reassuring.

With a transformer of dual 24VAC and 30VA we have enough power for an output of 20V and 600mA, but only about half of the 60VA that is needed for 1A outputs. That specification will need at least a 60VA transformer. I found and ordered a 25VAC 60VA one that may just fit.

Increasing the output to 30V 1A will need as a minimum a dual 33VAC 80VA transformer and I have my doubts if that will fit in the enclosure I want to use. I need to look for a suitable transformer but for now, I have put the 30V output enhancement on hold. Another option is to limit the current to 500mA with 30V in which case a 33V 40VA transformer might work and might fit.

I need to make a decision soon on what I'm going to use because it dictates the PCB I'm going to design next.

So what are the effects of my 24VAC 30VA transformer?

I put together a prototype setup to test the parameters and already made some changes to the schematic..

What you see on the protoboard is the bridge, a bleeder resistor on the bottom, the main capacitor, a 1 Ohm series resistor, changed from the 2R2 resistor I wanted to use, and a 2,200uF for the second capacitor, replacing the 1,000uF I originally wanted to use. The 2R2 got too hot with 1A and the 1 Ohm was sufficient. The increase in the second capacitor dropped the ripple quite substantially, also making up for the reduction in the resistor. The second capacitor and the 1R resistor might not be needed for the 600mA version, but I leave it in as an option for the 1A version.

While playing around, I created a spark even when the power was off, because without a load, the voltage will stay up for a long time. I experimented with a few different values (also in series) with SMD and THT resistors for this bleeder circuit and settled on using a 2W 3K3 resistor that did not get hot during a10mA discharge current. Note that the original PS503A does not have that bleeder at all. It may not be needed with the full circuit, it could drop the voltage fast enough by itself, time will tell, but during the prototyping it's a safety feature.

Here is what I measured on the output of this configuration using my Dynamic DC Load:

No load 33.7V

100mA 30.5V 300mV ripple

200mA 28.7V 360mV ripple

300mA 26.8V 410mV ripple

400mA 25.5V 470mV ripple

500mA 24.2V 520mV ripple

600mA 23.0V 560mV ripple.

What these measurements show is that the unregulated voltages seem to be too low to support the +/-27V rails, right? Well, this stumped me for a little while, but after some more thinking about this, it dawned on me. The +/-27V rails are only there to protect the old 741 Opamps from voltages above their limits. When these rails drop with the output current load, they still function fine, and as long as there is enough headroom, the supply will continue to function just fine. More modern 741 versions have a 44V maximum specification and with them you don't even need the 27V "rail" clamping. I'm going to leave them in to allow more flexibility selecting Opamps. (741 variants come in 30, 36 and 44V ratings)

To have enough regulation headroom, the maximum current that is usable with my transformer is 500mA. 600mA is possible, but I'll keep the max at 600mA. Here is why.

When you're adjusting the Current Limit (CL) with a normal potmeter travel, it will be difficult to precisely set the tripping point in the 1A version. I'm going to borrow the high voltage feature from the original version, but move the switch to the front panel to manually set the current to 300 or 500mA or 600mA or 1A using the same method. This will make the percentage setting of the CL potmeter more straight forward and is a lot safer for low current devices.

Supporting a 1A output

A preferred local supplier has a toroid transformer that has dual 25VAC windings supporting 1.6A, meaning it is an 80VA unit and should be even better than a TM506 high power compartment. With a size of 93.5 x 40mm it just fits in the enclosure. If it fits with the components located on the bottom needs to be seen.

Here is a picture of both transformers positioned in the enclosure to give an idea about the size.

To the right is the new 2x25VAC with 1.6A with 80VA. It is significantly larger and weighs just over 1Kg. The one on the left weighs 532gr. The heatsink will have to be mounted on the outside of the back panel, if the back panel itself is not sufficient to have enough room for the PCB. I will use the room that the 80VA will leave as the size for the PCB.

The 80VA transformer produced the following results:

0 load 36.90V

250mA 34.7V 480mV ripple

500 mA 33.9V 570mV ripple

750 mA 31.5V 720mV ripple

1000 mA 30.2V 880mV ripple

1250 mA 28.3V 1.1V ripple

1500 mA 27.7V 1.3V ripple

The higher voltages will have to be accommodated for all circuits that use that rail.

 

This transformer will be sufficient for a 20V output at 1.5A, 30V output at 500mA and also 25V at 1A.

This transformer is a bit overkill for what I want so I will continue this project myself with the 30VA transformer, which will be limiting the output to 20V with 600mA.

Power rail schematic

Below is the updated Power Rail schematic for the 30VA transformer. If you select another transformer, only the fuses need to be replaced by higher ratings, but the circuit will also support higher currents and voltages. There will be resistor value changes for both regulators in the current limit circuit.

We now know that the  +/-33V labels in the schematic diagrams are there to show the voltages with no load, and the same counts for the +/-27V "clamping" rails. The +/-8V rails are unaffected by the load. When making the layout, and while considering to use the original 741 Opamps, I decided to use LM317/337 adjustable regulators so we're not fixed by the slightly too high 8V rails.

I added a second main filter capacitor of 2,200uF separated by a small 1 Ohm series resistor to better filter the unregulated raw rail voltages. This will become more important when optionally, the output will be increased to a 1A current and/or at the maximum output 25V or even 30V. If that is the case, the 2.200uF could be replaced by another 4.700uF to keep the ripple low. The higher voltages will create larger currents for Q7 and Q8 and you may want to change the resistor values of R209 and R31 to keep the current for the Zener diodes in check.

The circuit around Q5 and Q6 is a startup protection circuit to make sure everything powers-up correctly and at the same time. It also functions as further protection if one of unregulated rails develop a fault. This is something I need to verify myself. It's not on the protoboards. Thinking about this circuit some more, I think it should be moved after the fuses, so if one of them blows, the other supply gets turned off. Aain, I need to test and verify this.

The transistors used in the +/- 27V rails may change once I know the total current consumption, which is unknown to me at the moment.

Note that there are several different GND symbols used in the circuits indicated by a number. Eventually, they all come together at the Common output connection, but are kept separate from each other to create a "Star" GND. You can easily see that when you study the PCB pictures of the original PS503A on the Wiki site.

The Voltmeter

This circuit is quite simple but in order to switch the display voltmeter inputs around so we can measure the positive and the negative supplies, the power rails for the meter have to be galvanically isolated from the rest of the circuits, hence the small additional transformer. The minus input from the voltmeter is internally connected to the minus supply voltage of the meter/display, so that's why this is needed.

As a change from the earlier design, I now switched to an even smaller transformer, used a full bridge and deleted the 8V regulator. This is what I used in the earlier supply and worked well.

Next steps

Now that I have the inner workings understood (at least I think so), and verified most of the enhancements with the prototypes, I think I'm ready now with a first attempt on designing a PCB. With that comes all the sorting out of the components and footprints, so that will take some time.

I have the layout of the main board done and I'm now working on the front panel which will be a PCB as well. The layout of the controls on the rather busy front panel required a few changes to the main PCB, so I'm going back and forth to make sure it all fits the first time. I hope...

There are also some changes to the circuits and I will post the latest schematic diagrams, and screen shots of the PCB's below.

Here is the completed layout of the PCB.

I'm using TO-92 packages for the transistors so I can easily swap them out for other types if required. When that is done I could switch them out with SMD versions with a new board revision.

And here is a 3D view of the front panel - note that it will have a black solder mask:

There is a ground fill on the front and back to act as a shield and will also add some sturdiness and produce a very nice half-matt black finish. All the holes are plated through with a larger surface on the back to properly ground all potmeters and switches. The four LED's have a normal 3mm hole. I decided to not implement a LED indicating the low current setting (300mA). It's obvious from the switch setting, and the front panel is already very busy.

Another thing you may miss is the switch to activate the dual tracking feature. I found an original potmeter with the switch at a site that sells Tektronix parts, and I purchased one. Not having a switch is not that cumbersome, because you can turn the potmeter fully CW (fully up in the diagram) to the 100% setting, and that will actually be the same as using the switch.

Here are the latest schematic diagrams with the latest changes. The updated Voltmeter section is just above.

Not many changes, but I did ad a jumper to deactivate the tracking overvoltage circuit in both supplies.

I'm now using LM317/337 adjustable regulators to create +/- 6V4 rails instead of the fixed +/- 8V ones, to have more flexibility selecting Opamps, in particular the 36V uA741C's.

At first I had room on the PCB to implement the original reference circuit, but I found out that I needed more room close to the front panel for the current potmeters. Eliminating that board space made it too cumbersome to move that circuit someplace else, now that the rest of the layout was already done, so I took it out.

Mechanical construction

Below is the old front panel that will be replaced by the new PCB. The paper copy shows the dimensions of the main PCB inside the enclosure, with now plenty of room for the potmeters and the three 4mm binding posts in the front.

The bottom side of the board with the bridges and output transistors, will be mounted flush on the metal back panel for cooling. To the right is room for the mains input receptacle and filter. The main transformer will be mounted on the top half of the enclosure, hanging down so to speak. If the cooling of the back panel alone is not sufficient, I will add a heatsink mounted to the back panel on the outside of the enclosure. At least that is the plan.

Since making this picture, I made a few more layout changes by moving the reservoir capacitors on the right side towards the edge of the board, now possible because I used a smaller transformer for the LCD display.

The two boards are now uploaded to my sponsor PCBWay, and I hope they will accept to sponsor this project as usual. They did, so the two boards are now in production. As soon as I have the boards, I will continue with the project.

The boards arrived in record time (less than a week) and look great as usual.

It took a little to find some free time, but they are now almost completely populated.

I'm still waiting for a few missing parts, but I'm almost ready to start testing.

The front panel in the background is also mostly done.

Here is the back side of the front panel with the wire harness. 

Most of it is in position, although not secured in place yet. I need to verify the proper switch settings with the front panel text, and may need to rotate the switches and also the potmeters so the contacts are out of the way.

The tracking potmeter and the two current limiter potmeters are not installed yet because I needed to glue the four LED's in place and needed the room. I also need to solder the wire harness to the LED's before I can mount the three missing potmeters. I also need to wire the pos-neg switch to the display and output terminals. All these wire connections will be on the back of the panel. The main switch will be wired last.

It's a very busy looking back of the front panel with a very tight fit and not much wiggle room left.

Another improvement?

While waiting for the PCB's, I looked at a way to replace the two rail fuses (F1, F2) by something else. During the prototyping phase, I will be using PTC fuses, but they are pretty slow to react. Glass fuses will be much faster but are destructive, you have to open the unit to replace them. I've looking into using MOSFET's that will be driven by the two fault conditions. They are much faster, and are non-destructive. I prototyped part of this circuit, and it works really well.

As soon as I have the prototype working on the new PCB I will experiment some more with that possibility.

First Power-up

First of all, I took a look at the rails to see if they were all there. To test that, I used my Lab supply and connected both channels to the 24VAC transformer inputs. That allowed me to keep an eye on the current consumption, in case something is not right.

Well, there are several things not right, so I will have to go through the issues. I know it worked on the breadboard so that's assuring. A few things were added based on the prototype so there can be many things. I'll come back later when I have sorted it out.

What went wrong and got me stumped for a while was that I forgot that I swapped the driver transistors around, so the front is facing the output transistors. I did that positioning to avoid crossing of the traces. between the two transistors. Because of this mistake, the output supply would not regulate well, and a few times I had 33V at the output. Not knowing what was going on, I disabled the over-voltage protection by removing the resistor in series with the 24V Zener diode. Unfortunately, a side effect was that I blew the voltage controller Opamps that I still had as the TLE2141. I replaced them with the original LM741C Opamps. With the Opamps shorting out, I also had two casualties with the +27V rails transistors, even though I replaced the original 2N2222 with a 2N2216, and the 2N2907 with a 2N2905. During this period these transistors got too hot and I even had to use heatsinks on them.

After rotating the driver transistors around, all was still not well but cool. Eventually I will swap the 2N2219 and the 2N2905 out with the original versions. 

The other problem was caused by yet another goof from my part. I used the +/-6.4V rails to power the LT3092 current sinks instead of the +/-33V rails. That limited the maximum output voltage. I could not easily connect them to the +/-33V rails, where they should have been connected to, so I took all the parts off the PCB and added the original 3K resistors on the back of the PCB, connecting directly to the output transistors. If/When I will go through a new layout, I will add the possibility to select either option.

I also found that I swapped around D22 and the D23 LED in the negative supply,  yet another dumb mistake but easy to fix. 

I then found that the two yellow voltage indicator LED's that I used, where from different models. The one I tried with the prototype (trying it for both supplies) worked fine with the 160K resistor value for R15 and R22, but the LED I used for the negative supply was nowhere near as bright, so I eventually selected a 91K resistor that made it look similar, but not the same. I will need to order two new yellow LED's to make them the same.

I had a few other connections to potmeter and switches swapped, but they also were easy to fix by changing the order of the leads in the plug headers.

So, with all that out of the way, I could finally start to work with the supply.

The maximum output levels of both supplies were calibrated to 22.0V to give it some extra headroom. 

I used a 150R/1W resistor to quickly test the current limit circuits, by touching the output connectors. When I was happy with all the controls now working and the potmeters turning the right way, it was time to use my Dynamic Load to test the current limit minimum and maximum ranges. I always like it when I can activate the current limit in the lowest setting, but that's not the case here. At the output set for 20V, the positive CL only cut in at 10mA. This is easy to fix by lowering the value of R50 and R34 from the 680R value to a lower one. The negative supply now also cuts in at about 10mA. (the first step of my Dynamic Load)

I then performed the maximum current test. In the highest range, I measured 1.2A, the result of the 1K05 resistors that were in the original schematic for the "hi" current mode. This maximum can be changed by using another value for R35 and R51. I will not be able to get that high a current with the transformer I selected, so I need to recalculate the values for R51, R59, R35 and R65 for both the 300mA and 600mA current limit ranges. I'm now using 510R for the 600mA range, and another 510R that will go in parallel for the 300mA range. With these values I will actually have just over 600mA and 300mA, so a little head-room.

I changed the resistor values for the minimum setting (R34 and R50) , but could not get them to turn-on the CL mode with the potmeter at minimum. They still need a few mA to light-up.

Testing the limits

While still using my Lab Supply, I performed another maximum current test, but with the maximum output voltage of 20V. My Lab supply can only go up to 32V so a tad less than the transformer would provide. However, I set the outputs of the Lab supply for a maximum current of 1.25A. With 32V DC applied at the 33VAC connectors results in a raw supply of 30V, due to the bridge, the PTC fuse and the 1R series resistor. Using my Dynamic Load, I measured the output voltage set at 20.0V with my DMM while increasing the current to see if or when the output voltage would drop. For both supplies, I could go to the maximum 610mA limit and the voltage stayed at 20.0V. At the maximum 600mA the raw voltage went down to 29.5V. This means that the raw supply input components only contribute with a 0.5V drop at the maximum current. With the maximum 600mA output, the power transistors that were mounted isolated on the 1mm thick back panel where getting warm but not hot at all.

Using the 30VA transformer

When I found everything functional, I swapped the Lab Supply with the 30VA transformer so I could profile the maximum voltages and currents again. Now that I have the transformer in place, I also have power for the display. Because the display is actually a DMM, it is very precise.

Positive Supply:
With the transformer, the raw voltage is +32.6V at no load. The ripple at the +33V test point is 4.6mV. The +27V test point measures +26.1V, the +6.4V test point measures +6.43V.

At the 300mA output current level, the raw voltage drops down to +27.27V with a ripple of 22mV. The output is still +20.0V. The 27V TP reads 25.45V. 

With the 600mA maximum current, the raw voltage drops to +23.7V, the ripple is 38mV and the output voltage is still 20.0V. There is a 3mV ripple on the +27V test pin, and the voltage at the 27V TP reads 22.78V. The 27V supply drops a little too much for comport, but otherwise, everything is good.

Negative Supply has an issue:
With the transformer, the raw voltage is -32.5V at no load. The ripple at the -33V test point is 4.6mV.
The -27V test point measures -26.1V, the +6.4V test point measures -6.47V.

At the 300mA output, the raw voltage drops down to -27.3V. The ripple is now 22mV. The problem is that the output voltage drops from 20.0V to 19V and it now has a few mV ripple. It starts dropping from the 20.0V level at around 220mA. At that level, there is no significant change on the -27V supply, it is still -25.55V.

With the 600mA maximum current, the raw voltage drops to -23.8V, the ripple is 34mV, which is the same as with the positive supply but the output voltage is only -15.8V, so we have a very weird issue here. It works with the Lab Supply, but not with the transformer. 

My first hunch was the tranformer or the AC to DC conversion components. Swapping the secondary leads to test a phase issue showed nothing. Even swapping the secondaries from the positive to the negative side showed no difference. So it's not the transformer. The DC voltages between the supplies all measure the same. What is going on here? Well, at the point where we loose regulation, I see a 50Hz mains ripple effect on the raw supply, that comes through on the negative output. The positive supply is totally devoid of this ripple all the way to 600mA. It sure looks like I have a bad reservoir capacitor in the negative supply, that did not show up when I used my DC Lab Supply. De-soldering them and measuring them with my DMM showed no issue though. I even measured the bridge, no issues. Weird!

I found the cause.
It's the startup/imbalance protection circuit around Q6. The Base of Q6 is getting the ripple and with a higher current (dropping raw voltage) it will eventually draw the diode OR down and pass the ripple on to the output. When I disabled the link to the diode OR, everything is fine. So what is the root cause? A bad Q6. In the tester it shows a diode between C-B, no E. Replacing the 2N2907A transistor solved the problem.

The output voltage now stays at -20.0V all the way until the maximum current of 600mA is reached, no ripple. Problem found and solved!

Testing the Dual Voltage Tracking

This was easy to test. When both voltages are at the same level, you really have a dual "tracking" all the way up and down, almost the traditional way. But, remember, this is not a master-slave configuration and not really tracking, so you can also do the following. When you set one voltage to say 15V and the other to 20V, the "dual" tracking will vary the voltages as a percentage. BTW, if you have the potmeter without the switch, you can still use it as in "without dual tracking" or disabled, just always remember to rotate it fully clock wise.

We have a working unit!

Note that on this picture, you can clearly see the difference between the two yellow LED's.

Testing the over-voltage circuits

I already know they work when I had the earlier issue with the swapped
driver transistors, but wanted to verify them now that we have a fully working system, using the transformer.

Testing the tracking over-voltage circuits

Keep coming back regularly, I will be adding or changing information as I go about designing the rest of the project.

More later....

Github project repository

There will be a dedicated Github project when I have finished the design, that will have all the information you need to build one yourself and I will also enter the project in the Shared Project section of my sponsor PCBWay.

The DIY SG505 mains power supply

This post will detail the mains related portion of the SG505 generator. I decided to split the whole power section in two, the quiet parts stay with the main generator, and the mains related and otherwise inherently noisier parts will go into another enclosure, all by themselves.

Think of it as a dedicated substitute for the TM5XX mainframe the original SG505 would have as its home.

After a few revisions, this is the final version of the supply, Version 4. (I'm showing it here so it shows-up on the title page of the Blog post)

The V3 Mains Power Supply

In an earlier version, the power supply was on the main SG505 generator board. This circuit  is inherently noisy and will no longer be in the same enclosure as the generator. 

This is the schematic of the new version I'm now working from. There are not too many changes, but I added ferrite coils in between the larger reservoir caps, and I used a common mode coil at the output. I also added a few more capacitors to quiet the thing down as much as possible.

The dimensions of the board are such that it can go in any kind of enclosure, and mounted on studs through the 4 mounting holes.

Discovered an issue

After trying the new setup, I discovered that I made a mistake...

A wimpy transformer

With the generator working, I started on the counter section. However, when I flipped the power switch, the generator almost stopped working, the output dropped significantly. To make a long and embarrassing story short, I misjudged the amount of power that the counter circuit draws on the supply. I measured it at 120mA, and was under the impression that with the 50mA for the generator I was within the power budget of the transformer. We're not because there is a significant voltage drop of the raw supply, dropping the headroom for the 40V regulator, and then also for the shunt regulators.

Here is the culprit, the VPP28-180, a 2x14VAC @ 180mA. It's not beefy enough. I had that one in my stock so used it. I should have know better, and I should have tested it better.

A new transformer will require a turn of the board, but I'll wait to be sure. Extra sure. This time.

To continue with the rest of the assembly and testing, I took off the 12V regulator parts from the main board, so they would not interfere and I could supply an external 12V to power the counter circuit.

The rest of the generator worked fine, so I could spend some more time on the power supply fixes that resulted in a board turn and a version 4.

Power Supply V4

To counter the lost head room when the current goes up, I tried a 15-0-15VAC transformer that I have in stock and is much beefier with 800mA. I used one of the spare V3 PCB's to build-up a new version, and also made some changes, reflected in the new schematic.

I used larger filter capacitors with 2200uF/63V, the only ones that I have with this voltage. The 1000uF/50V that I used before is a little tricky now with the higher input voltage. I ordered 1000uF/63V, and also 3300uF/63V to give them a try later. 

The second 75mA fuse is gone. The larger load and the filter capacitors cause an in-rush current that is blowing the fuse. I now use a 500mA PTC instead while testing, it will be lowered to 200mA after my testing.

When using the new 30VAC transformer, I measure the ripple at a load current of 200mA (using my Dynamic Load), and I only see 400mV, the raw voltage is 42.5V, which is not enough head-room for the regulator to provide 40V. 

When I increase the current to 240mA, the voltage regulator drops out of regulation.

At a load of 160mA, the raw voltage is 43.2V and this is the expected current draw of the instrument as far as I can tell. I need to measure that again, but I need to put the parts that I removed from the 12V circuit back on the main board in to do that. That requires a disassembly of the instrument again, and is next on my list, but not today.

I added the 12V regulator components on the main board and started testing again. The generator works fine when the counter is off, but when I switch it on, the positive 16.5V supply drops, this is due to the 40V dropping a bit, and that is because the raw unregulated voltage is dropping too much for the 40V regulator to continue to provide a steady 40V. That was to be expected because there is not enough headroom. 

The result of this investigation is that even a 2x15VAC is not enough, even though the transformer should have plenty of power.

Here is what I measured at the DC unregulated side with that supply:

Open : 49V, Generator 46V, Generator + Counter 42.7V: not enough, another wimpy transformer.

Using a beefier version

Next try is with a VPP36-560, a 2x18VAC transformer, another one that I have in stock. It was a left over from the Curve Tracer project. I don't need the 560mA it provides, but this transformer type is also available in a 280VA version.

When I measured this transformer, it showed these results:

Open : 60V, Generator 56V, Generator + Counter 51V.
The ripple is still 200mVp-p but we have plenty of headroom now.

With the unregulated input this high, the 40V regulator must now have a good heatsink, which was not required with the lower voltages. It's easy to create some room on the PCB to add one though.

With this transformer, there is no voltage drop anymore, and the other good news is that the generator continues to function very well with the counter on or off.

New power supply, counter off:

And below is with the counter on.

 
Note that this FFT is not as nice as the previous one, it has some hum and I had to readjust the H2 but this time could not make it disappear. I also found that the Vernier adjustment changes the H2 distortion level a bit. Looking at the circuit, that seems to make sense as it is in the feed-back loop of U1 where also the AGC is used as a summing-input. I'm speculating that the hum could be the result of the 40V regulator needing to work so much harder due to the significantly higher raw input voltage.

There is virtually no change in the noise floor (-128dBV and also in the harmonics). The 12V section is now quiet enough although for very clean measurements, you could switch it off, just in case.

THD is now 0.00033%. 

Adding a pre-regulator

With the extra head-room that the 2x18V transformer provides, I can now also implement a pre-tracking regulator. Because the job is then split in two, with one working on the line regulation, and the other on regulating the voltage output. The load is shared, so I can use two much smaller heatsinks.

Besides, the split will also create more separation between the mains related hum and the output voltage. The U1 pre-regulator provides about a 5V head-room for the U2 voltage regulator. Only 4 additional parts are needed for this pre-reg circuit.

Below is the text-book (from the LM317 datasheet) implementation of the pre-regulator, or tracking regulator.

(this is not the final version, have a look at the Github site (link below) for the most current information)

With the Manhattan style modification to try it out, and the long flying leads to the transformer, I can't see the benefits in the FFT, but my hope is that it will, once everything is on a proper PCB. 

I can even reduce the size of the heatsinks you see above, the regulators are not getting hot.

As soon as the ordered filter capacitors come in, I can try the 1000uF versions and see if they are adequate.

In the mean-time, I'm working on the new layout for the power supply.

The layout has provisions for the VPP36-560 transformer (11,62 Euro's at DigiKey) the one I already have and will use, and the smaller VPP36-280 (8,62 Euro's at DigiKey) which should be sufficient.

I also measured the supply with the 2x1000uF/63V and also the 2x 2200uF/63V for the main filter capacitors C6 and C7. The ripple voltage measured across C7 with the 1000uF was 320mV and with the 2200uF 200mV. Both values are adequate but the PCB will support both types.

Happy with these results, I uploaded the updated power supply V4 and the updated front panel Gerber files to PCBWay for production under our sponsoring arrangement. (even though I screwed-up twice, nice of them)

It will usually take about a week for them to arrive.

Final results

The boards from PCBWay arrived in the usual very good quality and I build-up one of them. Check them out for your own projects: https://www.pcbway.com/

Everything worked flawlessly, so I adjusted the output to 42.00V and connected the generator board. After a few minutes, while calibrating the voltages, the power dropped to zero. It turned out that I used a 100mA PTC for F2, and that is getting too warm and cuts out. I'm now using a 200mA PTC that I planned to use earlier, and that works fine.

I calibrated the output voltage again but now measuring the Vreg on the generator board, to account for a few 100mV power losses, and then proceeded to calibrate the +17V and the -17V. 

I could not adjust the +17, it stopped at about 16.4V, however, the -17V worked fine. This is due to a Vreg that is a little too low for the balancing to work, so I raised the Vreg voltage to 43V and all was well. To add a little bit more headroom, I'm now setting Vreg to 44.00V.

The heat sink I selected for the main voltage regulator (U2) is a tad too small, because the device gets a little warmer that I hoped for. I can still touch the tab without burning my finger so there is no real issue, but still. The U1 pre-regulator is only warm to the touch so fine.

Below is the now completed enclosure for the power supply.

The front panel is the back panel from the SG502 that I'm using here as the front panel.

And below is the setup of my instrument bench. 

The top shelf has the actual instruments. From left to right: the DC Dynamic Load, the SG505, my frequency counter, and the GPSDO on top of the master clock, the rest is out of sight) BTW, these are all instruments that are listed on my Blog. Below that shelf is room for the supporting stuff, like the power supplies.

This concludes this project for the time being.

Here is the overall project Github with all the design files: 
https://github.com/paulvee/DIY-Rebuild-of-the-SG505

Here is the Shared Project listing on the PCBWay website:
Shared Project DIY SG505 Power Supply