Show by Label

Tuesday, July 18, 2017

My New Power Supply Design Project Part 3

Updated text and diagrams after I fully built and tested the supply.

The Auxiliary Power & Reference Supply

Enough with the simulation, I'll move over to the schematic capture of the complete project.
First of all the Auxiliary Power Supply and Reference Voltage.

Here is the schematic, using Diptrace:

There is really nothing special about this, but let me go through the design. I use a 9-0-9V AC transformer I salvaged out of a alarm radio a long time ago. I have no idea what the maximum current is, but that hardly matters. I'm only pulling a few milli-Amps from it. This transformer is located in the transformer box that I'll cover later.
The 9V AC as a minimum is important, because it creates enough head-room for the regulators, and helps with my "early warning switch off system" that I'll cover later.

The 9-0-9 V lines come in the power supply enclosure through a 3-pole DIN connector, like we use a few decades ago for audio equipment. From the chassis part, it will go to a connector on the main PCB. The above circuit is just that section of the main PCB.

Because I only draw a few milli-Amps, I use 1N4148 diodes for rectification, they can handle 100mA. Next to the filter caps, I use 18K bleeder resistors to quickly remove their charges. This helps with the turn-off/ back on of the supply. I use TO-92 low power linear voltage regulators to get plus and minus 5V. These supplies are used for the Opamps.

The +13V tap is used in my early warning signal that the mains is going down, and this voltage is used to turn the supply output off right away, to prevent mishaps at the DUT. As I mentioned in the simulation earlier, the trip point is at 10.6V, this is 3 volt above the regulation headroom margin for the 5 V regulators, so before they start to cave-in and cause problems with the Opamps, the output has been shut off long before.

I could have used the 5V supply of the positive regulator as the reference voltage for the volt and current setting. The PCB is made such that I can easily bridge the LM317L, and leave the series resistor in place. Because I want to be able to use this supply with a controller, I will use 12-bit DAC's to drive the volt and current settings. Using a reference voltage of 4.096V makes the settings and calculations easier, allowing 10mV steps in the volt setting. Also, the output stability and noise specifications are largely depending on the reference voltage source, so I'll build at least one supply with the 4.096V reference voltage. Note that the resistors for the volt setting and the current setting in the V/I control section need to be adjusted to get 30V and whatever Amps you want at the output.

I use a small trimmer to adjust the reference output to exactly 4.0960V. During my tests, this reference voltage turned out to be very accurate and stable with a fluctuation of only +/- 0.0002V over a period of several hours. This is a mixture of my 5 digit DVM (still in calibration) and the reference itself. I cannot measure the noise level other than with a scope, so I can't give you any numbers, but it's very low.

Transformer Box

As I mentioned before, the space on my "bench" is precious and limited. This is one of the reasons why I decided a long time ago to separate the bulky transformers and even the rectifier bridges and smoothing electrolytes to another enclosure, that I can put in the background. An other benefit is that I am much more flexible with this relatively expensive equipment. As an example, I added 4mm binding posts to the AC outputs as well as the rectified outputs so I can have multiple uses out of these parts. 

Right now I have three transformer boxes, of which I'll cover two here. One houses my 15-0-15VAC 3.3A transformer, rectifier and 2 x 10.000uF capacitors and a small 9-0-9VAC transformer. The second one houses my 12-0-12VAC 2A transformer, and also a smaller 9-0-9V transformer. Both transformer boxes will be used with this supply (I'll build two). The output voltage of the supply fed by the 12-0-12VAC supply will go to 30V, but the current will be a bit limited at higher voltages. At the maximum 30V DC output, I can pull 1.1A before the output starts to drop. At the realistic minimum of 1.8V at the output, I can draw just about 3A before the voltage starts to drop.

This is what is in the T2 12-0-12VAC 2A transformer box:

 And here is my T3 15-0-15 3.3A transformer box:

There is really nothing special about them, although I use a not so well known trick to get the most out of the relatively small capacitors in the T2 box. The little series resistors actually separate the two electrolytes. The first will take the full charge and discharge rectification smoothing cycle, and the second one is more of an effective filter due to those 0.6 Ohm resistors. The 9-0-9 AC transformer is feeding the auxiliary power section explained above. It is switched in parallel to the main transformer, and is connected to a 3 pin DIN chassis part. The main three outlets are going to 4mm banana sockets.

The 0V or AC transformer center tap is also brought out (as 12 or 15VAC), and that is used, together with the negative DC output, to create a separate and isolated 8V DC supply for the Volt and Amp panel meter.

Panel Meter & Fan Supply
The power for the Panel Meter and the Fan is coming from this power section. In an optimal situation, both of these supplies should be separate and isolated from the main supply. Normally, this is done by added separate winding's on the main transformer. That's not so easy to do, and getting a transformer like that is close to impossible.

I need 12V to run the cooling fan, and I need anywhere from say 5..25V DC to power the panel meter.
The reason to keep these supplies separate from the main one is because they introduce a lot of noise. The fan with PWM activity and the display with switching noise.. You don't want that superimposed on the auxiliary and reference supply. Besides that, that aux. supply is floating on top of the main supply so you really don't want to mess with that.

Initially, I wanted to try the following method.
I wanted to use the 0V AV centertap of the transformer together with the negative DC main power output to create a lower value DC supply. Using the "raw" supply for that purpose is possible, but the voltage is most likely too high for normal regulators, and you generate a lot of extra heat to bring that voltage down. The extra rectification and filtering will help to isolate the two supplies somewhat.

So, this supply is rectified and goes to a 12V linear TO220 type voltage regulator. It will be mounted on the main heatsink. As the fan controller, I use a chip that I learned to use a while earlier, and although it is a little strange to figure out (for me), I still like it a lot. The TC648B is a dedicated controller for fans, and you can set the starting point, minimum RPM, maximum RPM based on an NTC input and a cut-off point for the fan as well. The resulting PWM is rather clean and all you need is a small MOSFET or transistor to drive the fan. I have configured it such that normally the fan is off, but will kick into life at about 30-35 degrees. I especially selected this rather low temperature, because it will take a while before the heatsink can be cooled down, especially when it is heating up quickly, and you don't really hear it in the beginning anyway.

The fan supply needs to be isolated from the main supply due to the induced noise coming from the fan and the PWM to drive it. 

The panel meter is a combined volt and ampere meter is fed by the unregulated 12V supply. I may need to add a small capacitor at the panel meter to further decouple the noise coming from it. The panel meter I selected is a type I really like, because it acts more like a DMM with switching decimals. It can display 0-33.000Volt and from 0 to 9.999mA and then switches over from 1.000 to 3.000 Amp. Perfect for the job.
In the first version, that was the basis for my PCB layout, I used this schematic:

I actually build a prototype for this section as well, and it seemed to work. I did not spend a lot of time on it, because it seemed so simple...
So I went ahead, and created a PCB layout and ordered the boards from OSHparc.

While waiting for the boards to arrive, I actually started to spend a little more time with the complete setup, and started to profile and measure everything. It was than that I noticed a problem, a BIG problem when I was measuring noise levels when the supply was in the lowest output settings. There was way too much noise  seeping into the supply.  So I had to redesign it, while the PCB is already in fabrication. Bummer!

While trying to keep the noise down, I actually saw that the panel meter I used was generating a lot less noise than previous panel meters I used before. I figured that if there was too much noise in the final product, I could still add some filtering to the supply afterwards, but not on the PCB anymore.

The biggest problem that I noticed was with the noise introduced by the PWM driving the fan. So I switched my attention to isolating that from the main supply. In these attempt, I used the DC-DC isolator method to electrically isolate the fan supply. Because the fan is rated above the single DC-DC isolator, I used two in parallel.

So my version 2 looked like this:

So I again started to measure things and found that even this setup produced too much noise. I new that the DC-DC isolators produced a lot of noise on the outside, but that was isolated, so no problem, I thought. I've used these devices before, but in other applications, so I thought I was safe.

I was wrong, it turned out that the DC-DC isolators generate a lot of switching noise on the input side, and that was something unexpected, and made these devices useless in this application.

Bummer! Another reset was needed.
The fan itself generates a lot of high frequency noise that no matter what I tried, I couldn't get it to a level as low as I wanted it. 

In my version 3, I have now resorted to the easy way out, using a separate 12VDC switched wallwart. The wallwart itself is a few feet away from my sensitive equipment, and where it comes in to the supply, I'll make sure it is quiet.

So here is the latest version (3) of this section.

Note that I added a Schottkey diode to protect the supply from input polarity reversals and I used a capacitor to filter out noise. Both are mounted on the DC chassis part.

After I got the PCB, I needed to isolate the fan controller section by some trace cuts and I also needed some cuts and wires to isolate and create the new panel meter supply. The LM340 on the heatsink was no longer needed, so I put a 78L08 TO92 regulator on the connector pins intended for the leads going to the LM340. 

After having populated the PCB and putting everything in the enclosure, I spend some time tuning the fan controller circuit, and I am now happy with the results. I dropped the slow start feature of the fan, because that uses PWM to slowly ramp the speed up. I don't wanted to avoid that noise. I figured that if the heatsink starts to warm up, it's better to make it run full blast, because of the inertia that the heat sink has towards the real temperature of the series transistor(s). 

So now, the fan kicks in at full blast, and terminates when no longer needed. The fan will kick-in at about 30 degrees C. This is all determined by the resistors setting the VAS pin (that sets the termination threshold) and the resistors setting the Vin pin that sets the starting point.

The panel meter supply looks like a bit of an overkill, but that meter will be on the front panel, close to the volt and current setting potmeters, so I want it to be as clean as possible, with no mains rectification hum present. These components are on the main PCB already anyway, so I may as well use them.

Series Transistor and Output Section

There are actually two versions for the two supplies that I will build. One is intended to be used with the T2 box, with limited output current at maximum output voltage, and one for the T3 box with the full 30V 3A.

This is the T2 version with only one series transistor:

And this is the T3 version with two series transistors in parallel to stay within the SOE by sharing the current, and spreading the heat.

Voltage and Current Control Section

Here is the main control section for the voltage and current control, and the output switch on/off circuit.

Note that I used two resistors in series to set the maximum voltage (R35 and R36) and current (R19 and R24) for this supply. This allows you to use various combinations. If only one resistor is needed, use a 0 Ohm one or a solder blob for the other. The current resistor selections in the diagram are for a 10K potmeter. If you use another one, I recommend a value somewhere between 1K and 50K, you need to recalculate the values. Initially, I used a good single turn potmeter, but after trying to set the current at milli-Amp settings, I switched to a 10-turn version.

My suggestion is to find the point where the maximum voltage, in my case 30V starts to drop when you apply more and more current. This will be the maximum current level the supply can deliver at 30V, and then select the resistor value for R19 and R24, so that the CC LED just comes on at the maximum value. 

Make sure that you first properly adjusted the reference voltage after a sufficient warm-up period before you try to figure out what the maximum current at the maximum voltage is. In my case, using my DC load, I could get just over 1.45A at 30V from my T1 box so I selected 120K + 56K resistors for R19 and R24 to get to 176K.

The value of the voltage setting potmeter is not critical and also not influencing the maximum output voltage. You can use whatever you have available but I suggest to use the same range as for the current potmeter.
Make sure you have normal red LED's (I use 3mm ones), not high efficiency ones, nor ones that use a large current.
Protection diodes D16 and D17 are not really needed. I put them in, and left them in, while I was testing. They protect the OpAmp input from stupid mistakes, but if everything is wired correctly, you don't need them. (their added capacity makes the supply a little faster without them)

Room for C21 was added for possible frequency compensation, but was not needed. Just don't install one. I also found that it did not really matter a whole lot if I used other OpAmps. I would stick to those that have a FET input, but otherwise you can try different ones. I used sockets for all the OpAmps, so I could play with various versions. Better is to solder U8 and U7 directly.

Thermal Protection

I changed the red LED to a higher intensity blue one, to better show the thermal cut-out event.

I used one of the many LM741 OpAmps I still have in my stash dating back to the 1970's. This is one of the few OpAmps that can be used as a comparator without introducing unwanted side-effects. If you want to know more about this, use Google, otherwise, use a real comparator instead.

Turn Off Protection

All the parts, except for those in the transformer box, and in the "On Output Terminals" and "On Heatsink" are on the main PCB. I finished a layout, using through-hole technology parts and SMD. All the resistors and capacitors, as well as the 1N4148 diodes are SMD, the rest is THT, simply because I have a lot of them. Plus, it makes it easier to replace them when needed. The Opamps are DIL versions and are socketed, also for easy replacement.

Here are some pictures:

This is the T2 transformer box.

This is my T3 box with the 15-0-15VAC 3.3A transformer.

Some of you will remember that in the early PC days, we had parallel interfaces to connect PC's to printers. As there was only one parallel printer interface, you could use a switch box to connect several printers to one PC. This enclosure is from such a box, all there was in it was one PCB with a rotary switch to connect two printers to one PC.

This is the main controller section with the bias supply, the volt and current controllers and the thermal switch.
I used good quality sockets to be able to select and switch the compensation networks.

This is the auxiliary supply with the 5V reference section.

Here is the 4.096V reference part, still on a bread board, which is how I typically start out and them transfer the circuit to a soldered protoboard.

This is the section that watches the raw auxiliary supply (the 13V tap) and connects to the bias totem pole to
quickly turn the output off and prevent glitches.

And here is the whole mess on my desk connected together. In the forefront is the Volt/Amp panel meter.

In Part 4 I'll show the final Power Supply and some measurements.


  1. hi paul
    would you say a 30vac 5A on the secondary would be a good transformer for this project

  2. Yes that would be possible.

    As I mentioned, the basic controller is capable of many different transformer configurations, in which case you may have to limit or change the maximum output voltage or current, but in your case, this is more than adequate for a 30V 3A output.

  3. I have been looking at your schematics and the complete schematic one is not so easy to read will it be possible to enlarge it or maybe split it in halve so it is easier to read
    maybe I missed it but I did not find a explanation for the glitch on and off part of the circuit

  4. mars, I realize that I need to redo the schematic and segment the parts to make it easier to see. That will be done after I have verified the PCB layout. This will take a week or so.

    The explanation for the power off glitch protection is in part-1.

  5. thanks for the reply
    guess the eyes are not what they used to be
    hope you have good results after you test the pcb as this project is quite interesting

  6. hi paul
    the small circuit named Disable Output that is connected to -5V is that connected to the on/off switch of the psu or does it go to a different switch
    also D19 and D27 near the output is that schottky diodes
    this is on the main circuit drawing

    1. That single throw switch is used to disable the output of the supply. One end of the switch is connected to -5V to be able to pull the bias away from the TIP142.
      D19 and D27 are for reverse polarity protection. They can be normal diodes, and can be Schottky types. They need to be able to withstand several Amps, and because Schottkey have a lower drop (0.3V instead of 0.6V), they can handle more power in the same package. I prefer those. I'm working on the final touches of my design and will be able to update soon. Stay tuned...

  7. This comment has been removed by the author.

  8. dear sir, may i ask one thing, there is a schematic to controlling voltage and current, it is only for 0..30v 0..1.5A?can i use it to controlling 0..30v 0..3.3A? thank you for the reply sir.

    1. after i reading carefully at the schematic now i'm understand how to adjust that schematic for limit current at 3A. i have 1 question sir, can i using 18-0-18 3A toroid transformer to replacing 15-0-15 3A transformer?and the last question is what type of that transformer?step down or step up?once again thankyou sir.

  9. I also have a one question more at panel meter and fan controller schematic. There is a terminal call NTC-2 (J16), what is meaning of V- TC648? what is the pin number of tc648 of V-? and also NTC-2? Thank you sir

  10. Those are the two connections to the thermistor.

  11. hi paul, i wanted to know if we need to increase the bias current for the 2 output transistors version. thanks and nice project

  12. Hi Fabio,
    You could, but you don't need to. I'm using the described circuit version with the two TIP142's, and that has been working fine.

    1. Fabio, here is my rationale for the 8mA bias current setting. The hFE (DC current gain) for the TIP142 has a minimum of 1,000. So to provide a 3A Collector-Emmiter current, it needs 3/1,000=3mA Base current. Using two transistors does not really change that, because each transistor will now only need to conduct 1.5A, so now a base current of 1.5mA is required. The total is still 3mA, well below the 8mA bias supply that is provided.

  13. Hi Paul!

    I only know a little about op amps, so here's my question.

    What is the purpose of using D11,C15 and R25 in the voltage control circuit? How do you calculate the values?

    (I see that type of configuration on the current control circuit too, but I guess they are the same thing, so you won't have to explain the same thing twice.)


  14. Thank you so much for sharing this great blog.Very inspiring and helpful too.Hope you continue to share more of your ideas.I will definitely love to read. best-whole-house-power-surge-protectors

  15. Hey Paul!

    Why are you using the LM317 with TL431? Why not just TL431?


  16. This comment has been removed by a blog administrator.

  17. Will you please provide a full schematic? All in one? maybe even the original files? You can send me personally through mail if you don't want everybody to find it. I want to make my own PCB with it with different DAC and ADC.

  18. Igbal, the diode across the opamps limit the excursion towards the power-rail, causing saturation of the output section and hence will increase the overall speed. The R-C network is used to tune the frequency response. Look at the simulation section for more information.

    I use the 317 in a constant current mode, supplying the TL431 to arrive at a very stable reference at minimal cost. You could do that with either one of them, it's just what I fancied at the time.

  19. Iqbal, my schematic capturing package (DipTrace) does not allow me to put everything in one schematic. It goes way beyond the number of allowable interconnects, and I already use the upgraded version. Besides, it will be impossible to see printed on A4 or letter size paper.

    By using the interconnect naming conventions across the different schematics, it does not matter for a board layout. I made my own layout from the schematic diagrams as they are published.

  20. This comment has been removed by a blog administrator.