Triple outputs 160W Lab PSU -- EPSUX3 version 2 !


As promised in the previous thread, this is here a new
start for an "upgraded" version of the unfinished v1 lab power supply.

A big piece of the job design has already done, but many work still
to be done before to have a "ready to be build by anyone" serious project
without any malfunction.
Another important point, for this new design i've take care as possible
to choose low cost and easy to find part in order to give the better chance
to build a DIY high performance PSU without paying more than a commercial

So i will post here regularly the project progress.

First, the key specifications of the EPSUX3v2 laboratory power supply:

- Small and light weight design.
- Triple isolated outputs, 160W total power available.
(2 x 0..20V 0..3A + 1 x 0..12V 0..3A).
- Wide input voltage main 85..265AC 50/60Hz.
- High-efficiency design.
- Current or voltage regulation available down to 0V.
- Voltage or current displays for each outputs.
- Mixed switching/linear outputs regulations .
- Very fast transient response and low noise.
- thermal sensor
- Symetric output mode with only one setting (master/slave mode).
- Output disable switch (no need to turn off the PSU).
- Magnetic part easy to build (DIY friendly).
- Low cost and easy to find components.
- Designed with EMC and safety considerations.

The EPSUX3 will use same enclosure and front panel design as the V1.
So, a picture of the EPXU3v1, just for remember of how does it look:

Many improvements has been made in the new specs.
The input AC/DC convertion is done now by a PFC
and LLC converter, allowing more power and higher

The output stage use now a buck step-down converter
followed by a new linear regulator from LinearTech, the LT3081.
For this output stage, i was inspired by this article in LTmag.

The presence of a linear regulator will improve drastically the
output transient response and avoid the need of high value
filtering output capacitor that is undesirable for a lab PSU.
Linear regulator reduce also the switching noise
from step down pre-converter.

You can see bellow the Beta version of the EPSUX3v2 full synoptic :

Here, a more detailled synoptic of the output stage block :

Today, i have done some job on the PFC stage (based on NCP1654 IC from Onsemi).
It seem working fine from 0 to 220W and for 90 to 250Vac at input.Cool !
A small picture of the prototype "on the fly" board ;

I love to build design like that... :)

I will come back soon.

Last edited:

Some news of the project progress.
I worked on PFC optimization, trying various inductor (home made and commercial).
Working with high efficiency is much more easy from a 230V main (EU) than 110V..
I get about 97%@200W with 230V main, but hardly 94% with 90Vac.
The PFC design is now mainly clean, just need to test others MOSFETs
for optimizing cost/efficiency (i receive parts these days).

I will work next weeks on LLC that works, but i want to do some changes.
I want to try to replace the gates drive transformer with a silicon driver.

I worked also lately on the CPLD design and all parts speaking with.
From the previous design, i replace the ADC (ADS7816) and the analog mux (DG408)
by a 8 inputs low cost ADC, the MCP3208. This solution is much less expensive and need much less space.
I also replace all digital potentiometers (AD8402) by a dual SPI DAC, the MCP4802 ( that are also less expensive).

I used a prototype board to test each new parts to test functionality and validate the CPLD code.
In the previous design, the digitals potentiometers controlled a voltage divider that giving
a non-linear output range from digital code. Now, a big improvement has been done using
a new control method. I use the control pin from the LT3081 regulator that are drive
by the dual output DAC (1 output for voltage setting and one for the current setting).
This control scheme allow a linear variation of the output with the DAC code.
The DAC work fine and the ADC is not yet functional but it's a question of days.

In parallel of that, i built a prototype board to test the output regulator section (buck+LT3081 hybrid regulator).
The design is mainly inspired from the LT journal note, but because their design use a very small
and hard to solder buck IC, i replaced it by a TPS5450 regulator in SOIC8 package, more DIY friendly.
The hybrid regulator work fine with this buck IC. The very precise voltage and current settings allow
and real square V/I regulation. Another interesting function is it's ability to regulate the output voltage down to 0V.
This require a small negative current to be sourced on the output.
A small negative voltage generator allow this (-8mA).

Currently, i don't have done serious measurements on the regulator (like noise and transient response).
The LT note describe mainly the same topology and there is no reason that i find very different results from us.
Anyway i will done some measurements these weeks.

You can see below some pictures illustrating the current work from each part.
The questions and remarks are welcomed.
I will be continue to post project news to let you informed. :)
Regards, and to soon.


The CPLD DAC/Display and ADC proto board :

The EPM570 eval board

Both connected together

The output regulator board TPS5450 + LT3081

The LT3081 zoom ( schottky bridge rectifier for the the 3 outputs of LLC on background).
Hello Carpin,

No, the final design will use 3 digits display. I think it's good enough.
Also,the front side of PSU is not so large...:)
The prototype circuit board use 4 digits, because the ADC is a 12bits model and i want to test it in full range.
In final design, the last digit is omit and that avoided the blinking of the last digit.

PFC and LLC Test.


I share with you today the current project advancement.
I worked on the PFC characterization and the LLC design
Then i tested the whole PFC and LLC.
So, i describe here the PFC results, and the job done
on LLC design (with performance results).

1/ PFC measurements results

You can see the PFC efficiency results chart with 90V and 250V main with associated losses.


The peak efficiency reach 98% for 250V main, and about 95% for 90V main.
As for boost converter, lower is input voltage higher is input current.
This is why losses increase at low input voltage.
This final PFC design use NCP1654 PFC controller from ONsemi, a CoolMOS MOSFET
from infineon, and a SiC (Silicon Carbide) diode also from infineon.

2/ LLC construction and design

I had build the "on the fly" board for th LLC, and test it this week.
The more time consuming step is the design of transformer that is the key
part in LLC design.

This prototype work fine and i get 200w output without issue.
I have made efficiency measurement of LLC alone, and with the PFC front-end.

I recall that the main advantage of LLC converter is to allow very high efficiency
and low EMI signature because.The reasons is because primary switches are turn-on
when Drain-Source voltage is zero (soft-switching) and output secondary diodes
are turn-on at zero current.These reduce drastically switching losses .
Another important point is that the current in transformer winding is mainly
a sine wave ! This is due because we use it at the resonant frequency of LLC tank.

First, you can see below a picture of the LLC prototype board.
I know, it's not yet a beautiful finished board ! ;)


You can see from left to right ;

LLC controller(SOIC8) / half-bridge driver(SOIC8) / Primary MOSFETs on heat-sink/
Orange and gray resonant capacitors / LLC transformer / twelve diodes for 3 full bridge rectifier/
filtering capacitors.

You can see below an real illustration of the current waveform in LLC tank for no load
and full load conditions:

Yellow - LLC Transformer current
Magenta - LLC Output voltage channel 1
Blue - Primary switch gate drive.

1/ Condition : Pout= 210W, Vin = 390Vdc.

2/Condition : Pout= 0W, Vin = 390Vdc.

As you can show on pictures, the primary current is near sine wave with very low harmonic contain.
With no output load, the current is more triangular and is not zero.This is because at no load we see
only the transformer magnetizing current.
We can show also that the switching frequency is not the same !For no load it's about 131kHz and 122kHz
for full load conditions. The frequency is the used means of LLC to regulate the output voltage.
In fact, the voltage gain of the resonant tank change if you are near or far the resonant frequency
and the controller and feedback loop change the switching frequency to get the desired voltage at
actual load condition.

As i said, on of the main advantage of LLC is high efficiency, so i
tested the LLC from zero to more than 200w to draw the efficiency chart.
Each output of LLC deliver about 25V.For easiest testing, in connected these 3 outputs
in series to use only one load and single V/I meter.

2/ LLC measurements results

You can see below the real efficiency i got with this prototype board :

These result speak from himself...
The peak efficiency reach more than 96%, and it's always 90% from only 2W.
I don't know other topology allowing higher results at this power level ! :)

So, it's important for the whole design that each parts work with highest efficiency,
because all loss will be dissipated in th "small" enclosure, and all must be done to
avoid excessive temperature rise inside !
At this time, i am not really sure if a little fan will be necessary or not.

2/ PFC+LLC combo measurements results

The LLC has been tested first with a 390V lab PSU, but after extended tests, i connected
it with the front-end PFC.

That give this:

Now, the real global efficiency of the combo has been done using high performance power meter.
The results is below for low and high main voltage (90-250V).

Despite this very good results, it seem important to take care abut heat management when the PSU
will be heavy loaded at low line voltage (90v), losses could reach 22W.
This point will be look for the current project.

3/ The next step ?

I will done some others measurements on LLC, like load regulation and residual voltage.
I must think about auxiliary 12v to powered the PFC/LLC controller (i use external PSU from now).
I must work also on output buck regulator and LT3081 control with PLD board.
Some news again in next week..To follow :)

Comments welcomed !..


Spirtos, jcga and the others before ; thank you for your very friendly comments. I appreciate.
I'm a little surprised about the lack of questions despite many views of the thread.
Maybe all is already clear ? :)

This past week i worked on the digital section of the project.
Now, i can read the 8 channels ADC (MCP4802) without problem and display
the value read for each channel. The CPLD code for writing/reading the IC is Ok.

The ADC is a 8 channels type, and 6 channels are used to read the current and voltage of each PSU outputs ( 3 outputs x 2 measures = 6 measures).
So, 2 channels are free to use.
I use one to allow temperature measurements and protect PSU from
internal overheating in case of missing airing.
The temperature sensor is a NCP9700 low cost SOT23 sensor that
give a 10mV/°C sensitivity with 500mV offset.
The sensor is already wired in my proto board.
I will use it to switched-off all outputs if a threshold is crossed.
The threshold values (high and low for hysteresis) will be programmed
in the CPLD.
Another free ADC input is free, but i will probably find soon a way to use it...

The next step, is to build a small voltage amplifier to drive the setpoint input
pin of the LT3081 from the DAC output (the setting voltage of LT3081 is equal to wanted output voltage, so we need to amplify DAC output).
So, i will be able to set and measure voltage and current with my CPLD board and regulator boards connected together.
I think i would do a small movie to demonstrate this first "real" use.
Then, i could start to make some measurements and verify each needed functionality.

To soon...

I'm a little surprised about the lack of questions despite many views of the thread.
Maybe all is already clear ? :)

Hi Frex,

in my case, I'm more an "analog" man so even if I understand your detailed and clear explanations it's beyond my design capabilities.

What is sure is that I try to optimize (read performance - capabilities - size ratio) my lab instruments and your project perfectly target my goal !

So I wait patiently until the project is complete to build one

Jean Claude

PS: PM sent
Hi Frex,

in my case, I'm more an "analog" man so even if I understand your detailed and clear explanations it's beyond my design capabilities.

What is sure is that I try to optimize (read performance - capabilities - size ratio) my lab instruments and your project perfectly target my goal !

So I wait patiently until the project is complete to build one

Jean Claude


I too am rather analog man ! (in micro-volt range ) :)
PSU design (switched or not) is a very analog design !
And a very good analog understanding is crucial to design it.
Maybe the more difficult step is transformer because that scares many hobbyists !
About he high voltage, i agree that many care must be taken to avoid lethal risk.
I always use isolation transformer when i design off-line SMPS...