Show by Label

Sunday, August 13, 2017

My New Power Supply Design Project Part 4

Before we dive in to the measurements, here are some pictures of the PCB and finished product.

As you can probably see, I used a combination of through hole parts with SMD. The reason is that I have a lot of THT parts, and I'm just starting with SMD.

Here are pictures of the finished supply for the 3A version:

The fan is mounted such that it will suck the air out. There are holes in the bottom of the enclosure, below the heatsink. It's a little hard to tell, but I added a number of holes in the top enclosure half to get as much airflow out as possible.

The parts on the backpanel are, from left to right, top to bottom:
Black negative supply, 3 pin DIN connector for the 9-0-9 VAC supply and the Red is the positive supply.
Yellow is the 12 or 15VAC or middle transformer tap for the panel meter supply.
Below it is for the 12VDC wallwar plug for the fan supply.
In the middle is the main supply fuse. (3.15Amp for the 3A version, which this is)
The two empty holes where used with the previous supply, but the thicker fan got in the way, so I can't use them anymore.

Here you can see how I mounted the two TIP142 transistors. In the middle is the NTC for the fan, it has a heatshrink to minimize effects of the airflow from the fan. This NTC is screwed into the heatsink with an M3 bolt.
Below the NTC is the 0.22Ohm shunt. The two green resistors are the 0.6Ohm balancing resistors for the TIP's. Sticking up are the 150 Ohm resistors in series with the Base. On top of the left TIP142 you can see the second NTC mounted on a lug, which is used for the thermal protection. It has a good heat transfer from the actual transistor die temperature.

Excuse the glue gobs for the LED's. I had dried out glue, which was needed to secure the 3mm LED panel holders. These holders suck, and the only way to secure them is to glue them. I ended up using super glue. Top right is the switch and the LED for the main power cut off.

And here is the frontpanel. I make a design in PowerPoint, and print it with a photoprinter on HP Premium Plus Photo Paper in the glossy type. I use double sided tape to secure it to the frontpanel, and carefully cut the openings with a small blade and very sharp knife.

Profiling and Measuring the Power Supply

Besides the obvious measurements, there is more information available that can be used to profile a lab power supply. After some searching around, I found a good application note that I'm following as much as I can, besides, it's a good read to get more information. The app note is from Agilent, number AN 372-1 and called Power Supply Testing.

One of the most important elements of any power supply is the dynamic load regulation, or as it is called in the app note, Load Transient Recovery Time.

In order to measure that, I first had to make some modifications to the pulsing of my Electronic Load. It did not have an adjustable base current load setting, so after adding that feature, the DC Load pulse has a DC offset capability. This is used to set a minimum load current, and the pulse itself raises it to the level you set with the output adjustment. You can follow the DC Load modifications in my post about that topic.

Before I took measurements, I twisted the leads to the DC Power Load as much as was feasible to reduce noise coupling. This is the setup I used for many of the measurements:

The DC load is connected to the supply output, and I use a 6digit bench multi meter to measure the output fluctuations. I use a scope to see any effects on the output itself.

When I finished the 3A version of the supply, my T3 box was still missing the 9-0-9 V transformer, so I did the following tests with the 1.5A T1 box first, and later with the T3 box.

Load Transient Recovery Time.

To  see the effect of the load transients on the output of the supply, I used a shunt resistor in series with the negative lead going to the DC Load. Accross the resistor is an (isolated) BNC connector that goes to another channel of my scope. The resistor is in a metal box to shield it from noise. The resistor is small (0.1 Ohm), so the volts across it are in the milli-volt region, so noise is a factor.

Here is a picture (taken with my new scope) that shows the effect of pulsing a 2.0A load at an output of the supply of just 1.8V. I used a setting of 1.8V, which realistically speaking, is the just about the lowest voltage you will encounter.  It is also the most strenuous on the supply because it means that the pass transistor is almost fully closed and the maximum voltage is across its C-E junction. It has to dissipate this large difference in heat.

The yellow trace (top) is the actual load pulse measured over the series shunt of 0.1 Ohm. The pulse is about 2V which means a 2A load. The blue trace is the transient measured on the supply output.
In the upper screen shot, the load is applied, which means that the current demand jumps from about 100mA to about 2A. It takes the supply about 30 uSec to react to this load change. At first the output goes low (blue trace), and then the regulation kicks in. This could be improved by adding a larger output capacitor, sometimes a few 100uF's is used, which is commonly the case.

Note that the output level of 1.8V after the initial transient is dropped a bit. I measured this to be 25mV with a more sensitive measurement. At a 1A load, the drop is only 10mV.

In the above screen shot, the load is removed, so the voltage over the shunt goes from about 2V (2A) to about 100mA. The blue trace is the measurement on the output of the supply. The output shoots up 300mV for a duration of 12uSec before regulation kicks in again. Note that this pulse is also an effect of the output capacitor dumping it's charge. This is why I keep it as small as possible. I personally believe that more damage can be done by the output capacitor dumping it's charge into the DUT, then the small sagging as the result of adding a load very quickly.

As you can see here also, when the load is removed the output level raises by about 50mV.

If you are looking at this for the first time, the pulse shooting up above the set output level may scare you. However, this is a normal behavior, and because the duration is so small, at about 12 uSec, it will cause no harm to the attached circuit.

The good news is, that the 300mV peak is about the same regardless of the output level. When I put the scope input to AC coupling, I can measure the pulse across the output range. With a 1.8V output (with AC coupling) the scope reports a pulse of 200mV, at 3.3V 195mV, at 5V 190mV, at 10V 186mV, at 15V 177mV and at 20V 173mV, etc.

For these measurements, I'm using the T2 transformer, because I don't have a 9-0-9VAC transformer for the T3 box. Initially, I used an old 6-0-6VAC transformer. The headroom for the 5V aux. supply is too small, and I need to change the zener diode value in the supply switch off transient circuit. This will stop me from interchanging these transformer boxes, so although thr supply is configured for 3A, I'm using the T2 transformer, not allowing me to go up to the full specifications at this moment.

One caveat though.

At this moment, I'm not sure if I'm looking at the maximum edge speed of the DC Load, or the Power Supply. The 555 Timer pulse that is driving the MOSFET's in the DC load has a rise time of 500nSec. I need to profile the DC Load a bit more to see where the limitations really are. For starters, it was never designed for speed, that requirement and modifications were added later on. For now, the Power Supply Transient Recovery Time looks pretty good to me, even though it could be limited by the DC Load. The Agilent app note mentions that this is typically measured in milliseconds or microseconds, so with 30 uSec rising and 12uSec falling, I can't be that far off.

Load Regulation
We already saw an example of the load regulation, where the output dropped 50mV from a set level of 1.8V, when a 1.5A pulsed load was applied.
I'm not sure that this is the correct way to measure load regulations, so for this measurement, I'm going to use my DC load in the static mode (DC only), and just dial in the current at various voltages..

The load regulation I measured is as follows. Although the panel meter is very precise (more later), I'm going to use my 5 1/2 digit DMM as a reference.

At an output of 1.8004V with no load, a 1 Amp load, dropped the output voltage to 1.7975, a drop of 2.9mV, or 0.16%
With a load of 2.0A, the output dropped to 1.7967V, 3.7mV, or 0,20%.

At an output of 24.998V, with no load, a 1 Amp load dropped the output to 24,997V or 1mV, or 0.004%.
With a load of 1.5A, the voltage dropped to 24,995 or 3mV or 0.011%

Current Limit Characterization
The Current Limit Characterization showed that this is a true Constant Current supply, as opposed to a less precise Current Limit supply. The set output current by the DC Load stays constant when the voltage is changed from minimum (at CC) to the maximum. The current deviates less than 1 mA (the least significant digit did not change) with a 5mA, 100mA, 500mA, 1A and 1.5A loads, while the output voltage is going through the range. Once set, the current limit staid stable.

PARD (Periodic and Random Deviation) Measurement
At this moment I don't have the ability to really measure this.
I can only use my scope, AC coupled, at the most sensitive vertical setting (5mV/div) and measure the noise on the output. I first established the "noise floor" of my measurement, by connecting both the tip of the probe and the ground lead to the negative output. The scope then tells me I have a 12 mV p-p noise level in the peak-detect mode.

Measuring directly on the output, I get 10mV p-p at 0V output,  and I get approx. 14 mV p-p at all levels from 1 to 30V.

Turning on (Startup) Transient
The output of the supply should not have any transients when it gets turned on by the main power switch.
Normally you should not really do this, but if there is a brown out event, the supply may turn off and on again.
There is a provision in the bias circuit (the 220uF capacitor) for the series transistor, that will delay the bias current, in effect holding off the series transistor from conducting.

Here is a screen shot of the turn on transient at an output voltage of 4V (bottom trace) with a 500mA load. The top trace is the raw positive auxiliary supply. The output gets turned on when the aux supplies are in regulation. There is no ringing or overshoot on the output.

Turning off Transient
If the supply gets turned off by the main switch, or a brown-out, there should be no transients at the output, to prevent damage to a (sensitive of expensive) DUT.
I concocted a little circuit that gets triggered by the sagging auxiliary supply, before the main regulators that are driving the Opamps come out of regulation. As soon as the unregulated supply gets below 10.6V DC, the bias for the pass transistor gets turned off. You can see that in the simulation result shown in part 1.
Here is the real thing:

The top trace is the raw positive auxiliary supply at about 15V.
Note that when that voltage is dropping to the 10.6V trigger level, the bias is cut and the supply output is gone before the series regulators are running out of headroom at about 7V. This results in a very clean turn off at the output.

Line Regulation (Source Effect)
I don't have the means (a variable transformer) to test this, but because our electric grid in the Netherlands is very reliable and stable, I'm not worried about this aspect, although a big factor is the transformer I use. The T2 transformer I used is at the limit, especially drawing higher currents at a high voltage, because the supply can get out of regulation. The good news is that the CV LED will go out as soon as the series transistor is out of headroom, so you have a visual warning.

Overvoltage Shutdown.
This is not implemented.

Short Circuit Output Current
I tried this three different ways. First by the Electronic Load by setting it to the maximum current draw. Then I used my DMM in the 10A current mode and put that across the output.
Finally I used a lead to short the output.
In all cases, the supply went straight into CC mode with no side effects.
I also did the "sparky" test by rapidly shortening the output and drawing sparks. No problem!

I did a little of that already, but only for relatively short periods of time. After a sufficient warm-up period, the output stayed at the set voltage level to within about 1 milli-Volt. This is without the 4.096V reference in pace.
From the 5V regulator, I used a series resistor of 10 Ohm and a 47uF Tantalum cap to create the reference voltage.

I just did another measurement, now with the 4.096V reference and the fully build supply. The supply was set at 5.0025V, and 4 1/2 hours later it is still at that level. So it stays within a 100micro Volt. I did not make a plot over time, still need to learn how to do that with my new scope, but I checked the value regularly. Only the last digit (100micro Volt) is sometimes flipping. For a power supply, this is more than adequate.

Measuring with the T3 box main transformer

OK, same stability measurement. Output at 1.8048V, with a 3A load, the output dropped to 1.7995, or 5.3mV, so a regulation accuracy of 0.29%.

With the output set at 20.005V, and a 3A load the supply is on the verge of current limiting, at which point the output is 19.750V, or a drop of  255mV, so a regulation accuracy of 1.27%.

With the output at 30.009V, the supply is getting out of regulation just over 2.6A.
At 2.5A, the output is at 29.910V, or a drop of 99mV so the regulation accuracy dropped to 0.33%, still respectable.

I should point out that the inability to get the full 3A at 30V is more a function of my transformer than the supply itself. Although I also should point out that the heatsink is not large enough to supply 3A for longer periods. The enclosure has no room for a larger one, but that was part of my constraints.

I seldom if ever need that much current, and if I do, I have other means. This supply will be used most of the time below 20V and well below 1A. Mission accomplished as far as I'm concerned.

I'm expecting this question so let me be frank and up front. The current PCB has some issues with the fan supply and the panel meter supply, as I pointed out earlier, and needed some surgery with trace cuts and wires. So, although I was initially planning to, I'm not going to publish the Gerber files, so please don't ask. I also used a lot of THT components, and that makes the board size larger than it needs to be, and therefore more expensive.

With the details in the schematics I provide you can build your own, it's not that difficult and not that critical. Just make sure you have a main center ground point to avoid loops, and keep things apart that needs to be separate.

Building the second supply for the 1.5A version
In the meantime, I finally found some time to scrap the old 30V 1.5A supply kit connected to the T1 box, and started building-up my second supply. I did some more work on the PCB trace cutting and isolation of the Fan Supply, so it's much cleaner.

With the exception of the maximum current resistors, all the parts are equal on this version. Obviously, I only needed one TIP142 series resistor. Look at the diagram in part 3 for details.

After carefully soldering all parts on the PCB and creating wire harnesses from all parts not on the PCB, I wired it all up and everything worked as it should. As mentioned in the testing section above, I now have a 30V at 1.45A version as well.

I hope you enjoyed this journey as much as I did.

If you like what you see, please support me by buying me a coffee:


mars said...

hi paul
are the drawings below the updated ones since you have built the pcb and tested as i see the transistors in the on/off glitch (changed to Bc 847 and 857)and load section(below Tip 142)changed to Bc847 and the mosfet changed to 2n7002

mars said...

also the zener changed from 10V to 7.5V

paulv said...

Yes mars, these are all the final components. BTW, you can use any general purpose transistor for the NPN and PNP types, I just happen to have these. I normally use my trusty 2N3904 and 2N3906 workhorses, but I didn't have them in SMD. The Zener diode changed in tune with my aux. supply changing from +/-8V to +/-5V.

mars said...

hi paul
did you have yor pcb made with 1oz or 2 oz of copper and are you now using a 6Vac transformer or stil the 9Vac transformer for the aux supply that feeds the 7805 and 7905

paulv said...

Mars, I used 1oz copper.
The aux supply transformer I use is mentioned at several places.

mars said...

does the ground of the aux power supply only go to the pos of the output or does it go to all the other grounds as well
this is not very clear to me as on the drawings all the grounds are tied together

paulv said...

All the grounds are connected. The only exception, and it is clearly marked, is that the fan supply ground is not ground, but the -Vin.

mars said...

what pcb size did you end up with

paulv said...

73.5mm x 57.2mm

Unknown said...

waiting the complete series schematic diagram...:) Thank you before...

paulv said...

Everything is in part 3.

mars said...

hi paul
what would you say is the lowest zener value that can be used in the Turn Off Protection circuit (say you want to use 6v instead of 12v)

paulv said...

I thought I explained that. Maybe not. I'm not sure that I understand (6V instead of 12V) This should help: The series transistor must be turned off before the aux supply gets out of regulation. The headroom for the aux supply series regulators is approx. 2.5-3V, so you need to stay a few volt above that for the cut-off. You also need to add 0.6V to the Zener voltage to get the turn-off voltage, due to the transistor VBE. If in doubt, simulate it like I did.

mars said...

i want to see if i can get the circuit to work without using the 9-0-9 aux transformer and have built a circuit that gives me 5-0-5vdc not using the transformer or the rest of the 5v circuit and that is why i asked if a lower voltage and lower zener can be used for the Turn Off Protection
in the thermal protection circuit do you also see 7ma sinking in the opamp when the circuit is on

Unknown said...

Really im not clear and understand how to build this psu, because the schematic is divide into several part at part 3 topic. Hope someone kindness want to draw the schematic in full schematic without separated the schematic. Thank you

Anon said...
This comment has been removed by a blog administrator.
mars said...

hi Paul
as a matter of interest where did you find a 0.6R 2W resistor for balancing the tip142

paulv said...

I had them in my stash, some dated to the 60's and 70's.

mariof said...

I apologize for poor English.
Electronics is my hobby.
You have done a very solid job.
I've built a trial version, it looks very good.
I think to switch the main transformer windings to make less heat on the transistor. Have you tried this solution? What difficulties can I expect?

paulv said...

Thank you Mario!

You can certainly use a relays or a MOSFET to switch the transformer winding's to a lower voltage, ie. use only one half of the secondary winding's.

As a matter of fact, you can even put the winding's in parallel to get more power (Amps) out of the transformer as well.

All you really need is a comparator circuit that drives a suitable relays, or a MOSFET transistor. I don't have a circuit for you right now, but I'm sure you can find them on the internet.

If you care to look at my Curve Tracer post, I'm using a relays to switch two winding's in series. This comparator is more complicated, because it has to work on a triangle waveform, so it needs rectification before the level can go to a comparator.


mariof said...


Thank you for your quick response.
Now I'm testing the idea you gave me, it looks good.
I read that you bought DMM VC8145.
Also, I want to buy the same.
Are you satisfied with Vichy VC8145?

I wish you everything Good

paulv said...

Hi Mario,
Yes, I'm very satisfied with the VC8145, after I disabled the annoying beeper that is. The unit is very simple to use, precise and stable. Recommended!

mariof said...

Hi Paul.
Thank you for your recommendation.
VC8145 is already at my place.
I checked it with the AD548 voltage reference.
The difference by 1 or 2 digits occurred in the fourth place after the decimal point (eg 7.5065V instead of 7.50664V), it's probably a good result :-)

How did you turn off the buzzer?

paulv said...

Just took the buzzer out. I actually destroyed it while de-soldering, so be careful if that's what you want to do.

mars said...
This comment has been removed by the author.
Johan Holstein said...

Is there a way to buy a PCB ?
Best regards,
Johan Holstein

paulv said...

Hi Johan,
I did not make my boards available for ordering through OSHpark because you'll have to make several modifications to make them work with the updated schematics. I have one PCB available for you for €10,00 which was my cost price and I also have the supply from this post available for sale for €60,00. Remember that the Power unit is not included.

paulv said...

The PCB and the supply prices are without postage/shipping because I have no idea where you are located.

Unknown said...

I really enjoyed reading through this whole process! I am also a hoobyest after work i play around and fix things.... I have been trying to look for plans on a controlable psu for my lab that could work for what i want with bo luck. If you have extra boards for sale msg back if not GREAT JOB!

paulv said...

Unknown, I have one pcb for sale.
€10,00 without postage costs.
Let me know where you live and I can calculate that.
I can setup a Paypal friend payment request.

MaT said...

If i buy PCB from you do you add complete parts list?

paulv said...

MaT, I have a complete working supply for sale. One of the supplies in this post. I can sell that with the power box that contains the transformer, or without. This box is pretty heavy, so it depends where I need to ship it to.

I can also sell you a bare pcb with all the required components if you want. Let me know what you decide to do, and let me know what country you are in so I can estimate the shipping costs.

If you decide to go forward, send me a note with your email address.

MaT said...

I cant find your email. my is legolasster at gmaildotcom . Im from EU country :) Send me a message so we can make a deal.

Unknown said...

Hi Paul,
Happy new year.

Have no idea if you are still active on this post or not.

How do you regulate the voltage? I do not quite understand the voltage regulation part, it looks like it's an open loop implying that there is no active regulation? Am I correct in my assumption?

Thank you for your development in this project.


paulv said...

Hi Charles, I tried to explain the voltage regulation in part1 :
If you didn't already, please have a look. The output is compared and regulated against a fixed and stable reference voltage.

Unknown said...

Hi Paul,

My apologies, I MUST have skipped over it, I have all the tabs open will go back and read carefully.

Again thank you.

J. Thomsen said...
This comment has been removed by the author.
J. Thomsen said...

Hi Paul

Thank you for publishing this. Nice design. I learned a few things :)
I hope to implement this as a fully symmetrical design (Tip142/Tip147) to give fx ± 30V.
I would like to to be tracking, that is, share one reference for pos/neg channels.
But I can't figure out how to make the reference voltage with two floating controllers.

Can this be done?

Jørgen, Denmark

paulv said...

Hi Jørgen,
Making it truly symmetrical can be done. Relatively easy with only the voltage parts tracking, but more complicated when the current is thrown in as well. By the time you're done with it, there will be two separate supplies with one acting as a slave and one as the master. I did something like that with the earlier 0..30V supply also posted on my Blog. Have a look there for the details on how I solved that.