Power supply power-on sequencer

This device reduces the amplitude of current peaks when you power many power supplies at the same time

Similar projects worth following
I have implemented power-on sequencing for other projects but this was often a weakness point in the system.
This time, I try a new approach which addresses the known shortcomings for medium-sized projects (200 to 2000W)
The cool part: no programming and a direct integration in the electric circuit protection !

This "idea" came on 20160511 and I'll use this someday so it's better that I document everything here.

Most low-voltage power supplies are switched-mode AC/DC converters and in some projects, I must use a LOT of them. For example, this is only one half of the 120 PSU that were used for #Mons2015 LED Screen ElectroSuper

To provide all the required juice, I build distribution systems such as this box:

But when you turn the power on, you have 120 PSU drawing insane amounts of current while they load their input capacitors. The inrush current is enough to make everything trip (I'll save you the conservative calculations).

This inrush problem has been addressed from the beginning by adding many local SSR near the PSU, controlled by their respective FPGA boards (a #WizYasep)

The above image shows a local distribution box that receives 16A and switches 4×4A. This is better but still not satisfying.

The "hack" that is the heart of this project consists in moving the SSR behind the circuit breakers, in the main distribution box. This increases safety and isolates the potential sources of faults.

The other "hack" is that the SSRs are mounted on a PCB that directly plugs into the circuit breakers ! It's much more convenient because this removes wiring, uses less space and is potentially safer.

Yet another "hack" is to go completely "analog": in the past I used programming (FPGA, I considered microcontrollers) but a simple delay line (resistor-capacitor-transistor such as in #Yet Another Electronic Lampyridae) is more than enough !

I just have to solve the power supply issue but a small wall wart is enough.

The system that I'm about to design has:

  • one mains power input
  • one small AC/DC low-voltage power supply
  • one "delay line"
  • 4 static relays (SSR)
  • Chain-in and chain-out signals (optocouplers)

They are placed on a board that has the shape of an electrical busbar that can directly plug into a row of circuit breakers. But instead of being a normal busbar, each output is individually switched on :-)

Several boards can be cascaded, to prevent simultaneous turn-on of independent circuits.


1. PSU
2. Latest considerations
3. Dimensions
4. Spark gaps

  • Spark gaps

    Yann Guidon / YGDES16 hours ago 0 comments

    Hmmmm another thing to ponder :-)

  • Dimensions

    Yann Guidon / YGDES2 days ago 0 comments

    The french circuit breakers have a very standardised format. I wanted to measure the actual parts but it's better to read the books and I found some informations. Crucially: the module widths are multiples of 18mm.

    I had to measure something anyway : the inputs are 4mm wide so the PCB's teeth should be around 3.5mm wide.

    There must also be some room for a copper busbar, probably behind the PCB.

  • Latest considerations

    Yann Guidon / YGDES3 days ago 0 comments

      Now that the PSU is selected, I'll address a few other aspects.

      My initial idea was to reuse the transistor-resistor-capacitor structure of the #Yet Another Electronic Lampyridae but the idea has evolved. I'll reuse the system I developed earlier, with a 74HC245 as a quad-amplifier.

      The '245 has 8 outputs but they are paired and paralleled (with one resistor per pin) to provide the required 20mA per SSR. This means 8/2=4 SSR (not 3 or 5).

      I have decided to use only 1 SSR per channel. I have wondered if paralleling two SSR was such a great idea. It saved on assembly but electrically, I didn't really test the idea. A single SSR can sustain 2A@220V under "all-weather" operating conditions and a dissipator is required for more. I think that I'll use a dissipator this time, for an additional reason : spacing. The SSR use some room/space, to ease convection cooling.

      Back to the 245:

      245 |     330
          |]---\/\/\---|     |
          |            |----[| SSR (1.6V, 20mA)
          |]---\/\/\---|     |
          |     330
      Each pin sends 10mA, so the resistor is (5-1.6)/0.01=340 Ohms (ok, 330 is great)

      The enable pin can be left active but can also be used for a global "disable". It's not the best way, though, as we'll see.

      The 245 is a simple amplifier, but the original system used the 74HCT245 which has a lower trigger voltage. How can I make a delay line with it ?

      Each output goes to the next input through a RC network. I want about 1 second so the RC ratio is about 1/1. For example 1µF and 1M Ohms. But there is something else to consider : the 245 doesn't have Schmitt trigger inputs. They are easy to implement with a 1M feedback resistor for example, so the RC's resistance must be lower. 10µF and 100K Ohms sound good.

      The original system used the 74HCT245 which has a lower trigger voltage. I must explicitly choose a 74HC245 to prevent the system from triggering too fast, or to remain active too long after an event.

         245 |     330
             |]---\/\/\---|     |
      outputs|            |----[| SSR (1.6V, 20mA)
             |]-*-\/\/\---|     |
             |  |  330
                |              |
      <--\/\/\--*--\/\/\--*---[| next 245 inputs
           1M       100K  |    |
                          |    |
                         --- 10µF

      So overall the system is mostly analog. All those resistors will create a mess though...

      So the idea now is to use a 74HC595 to properly isolate each output and avoid the weird hysteresis circuit. But the '595 requires a (slow) clock. That's one more circuit.

      I can make a clock with a pair of inverters (74HC04) but there might be a better approach: why not get the clock from the mains ?

      @K.C. Lee suggested (in a comment) that I "spy" the mains with an optocoupler. This allows the detection of short failures of AC current. This signal should be bandpassed to avoid eventual HF parasites.

      The optocoupler's output can then go to RC network, which will keep the /RESET signal low if no input is present, but also a 74HC4040 or 4060 binary ripple counter. In a european 50Hz network we get :

      1. 25Hz
      2. 12.5Hz
      3. 6.25Hz
      4. 3.125Hz
      5. 1.5Hz
      6. 0.7Hz
      7. 0.39Hz
      8. 0.19Hz

      A simple selector (or DIP switches) will select the desired frequency.

      The '595 Data input and Data output pins can be chained: an optocoupler will isolate the signal and each board can select between internal and external Din (another DIP switch).

      Note : the '595 is rated for a total current of 70mA so the resistors would probably be a bit higher, 390 Ohms probably.

      A red LED in series with the SSR's input will show the current flowing through it and give a direct reading of the sequencer's state.

      Once again I favour a "no programming" system, where it's so easy to pop a µC. I don't want to lose time with IDEs anymore and it's cheaper...

  • PSU

    Yann Guidon / YGDES5 days ago 4 comments

    I just solved the power supply issue.

    It is not a problem to get low voltage DC from high voltage AC but in my case, the power supply's characteristics matter.

    The current draw is expected to be in the 50mA range, at most. There are a lot of cheap power supplies but most cheap ones are in the 500-700mA range, suitable for USB. The power is too high, with rather large capacitors.

    When there is a power failure, these capacitors keep some charge for "a certain time", which keep the circuit on. The SSR will remain latched and if the power is restored immediately, a high current spike will occur, which is what we want to avoid !

    I found these modules, sold for a few buck only and rated for 5V 1W, that's 200mA.

    It's probably going to need some conformal coating though but we're almost done.

    However there must be a "fast reset" of the circuit : if the power is absent for more than 1/5s, the whole sequencer must be reset.

View all 4 project logs

Enjoy this project?



K.C. Lee wrote 5 days ago point

You can run an optoisolator off AC side using a cap-dropper for detecting missing power.
Use a proper safety rated X-cap.

You can also load up the power supply to overcome the capacitance.  :P

  Are you sure? yes | no

Yann Guidon / YGDES wrote 5 days ago point


1) I have optocouplers and I must implement a proper POR with brownout.

2) I'll load the power supply, at least to light a LED

3) the supply might be loaded with 3×10mA (from the SSR's internal IR LEDs), I must measure the discharge time.

  Are you sure? yes | no

K.C. Lee wrote 5 days ago point

You can also add power supply supervisor on secondary side to cut off your SSR when the voltage drops to say 4.5V instead of all the way to their minimum voltage.

Some of these supervisors have external reset input, so you can tie the opto output too.  I like TI TL77xx series because it also let you change the time delay..

  Are you sure? yes | no

Yann Guidon / YGDES wrote 5 days ago point

I might have one of those in a drawer. but only one so not a part I can base a small series upon.

The external input is a nice touch too :-)

  Are you sure? yes | no

K.C. Lee wrote 5 days ago point

I like the bipolar part (TL77 vs TLC77) as it has extended input range and can do crazy things like extending hysteresis and drive softstart.  :)  CMOS verion is down to 1.1V and lower power.

  Are you sure? yes | no

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates