Transient voltage issue

I think you've got the regulator drawn the wrong way round.

I had an issue with regulators failing some time ago (https://www.electro-tech-online.com/threads/regulators-failing.91068/)

It could be useful to extend the circuit to whatever is being powered from the 5 V supply and from the 45 V supply, as I am fairly sure that you wouldn't run a 5 V regulator from around 45 V unless there was some other reason to have a 45 V supply.

As this problem is happening on start-up, it could be caused by the power supply voltage going down briefly, so either the regulator or Q1 has a larger voltage on their output than on their input. That is what was causing the problem that I had. The possibility of reverse current is mentioned in the data sheet for the LM1117.

There are a few things that can cause the supply voltage to go down briefly during power up.
1) The mains input not being turned on cleanly.
2) The mains input being turned on near the end of the mains cycle so that the first cycle doesn't charge the capacitors fully.
3) The mains input being turned on near the start of the mains cycle, so that the transformer core saturates for about 1/4 of a cycle, giving no output voltage for that time.
4) The load on the 45 V supply being turned on before the voltage is high enough.

Oscilloscope traces during turn-on would be useful.

If you don't know what the problem is you could try:-

1) Use a simple 15 V zener instead of Q2 + zener. You don't need to accurately set the 15 V and there's not much power dissipation as there is a 5k6 resistor.

2) Add a diode on the emitter of Q1

3) Feed the collector of Q1 via a resistor. You can lose up to about 20 V without affecting the voltage of the 5 V supply.

4) Run the 15 V / 5 V supply from the centre tap of the transformer. It will need its own smoothing capacitor but it won't be affected by what the 45V supply does and there will be much less power dissipation.

Diver300 said:

It could be useful to extend the circuit to whatever is being powered from the 5 V supply and from the 45 V supply, as I am fairly sure that you wouldn't run a 5 V regulator from around 45 V unless there was some other reason to have a 45 V supply.

Click to expand...

The 45V is needed for the regulated power supply, which is the actual purpose of this circuit. The design calls for a maximum output voltage of 35 ~40V dc.

The 5V is used for a microcontroller, ina219 current measurement module, temperature measurement modules, LCD display, ...
When my circuit blew up in fireworks at the LM1117 I was astounded to find all electroincs components connected to 5V still to be in working condition.

Diver300 said:

As this problem is happening on start-up, it could be caused by the power supply voltage going down briefly, so either the regulator or Q1 has a larger voltage on their output than on their input. That is what was causing the problem that I had. The possibility of reverse current is mentioned in the data sheet for the LM1117.

Click to expand...

That is a possibility indeed. I am considering to connect an oscilloscope to the 45V line as triggerpoint, and to the 15V line as follower. And do some extensive power on/off exercises.

Diver300 said:

There are a few things that can cause the supply voltage to go down briefly during power up.
1) The mains input not being turned on cleanly.
2) The mains input being turned on near the end of the mains cycle so that the first cycle doesn't charge the capacitors fully.
3) The mains input being turned on near the start of the mains cycle, so that the transformer core saturates for about 1/4 of a cycle, giving no output voltage for that time.
4) The load on the 45 V supply being turned on before the voltage is high enough.

Click to expand...

Very interesting observations,, thank you! Maybe I should put a third oscilloscope channel to the base of Q1?

Diver300 said:

1) Use a simple 15 V zener instead of Q2 + zener. You don't need to accurately set the 15 V and there's not much power dissipation as there is a 5k6 resistor.

Click to expand...

Replace the current 5V zener with a 15V with cathode directly to the base of Q1?

Diver300 said:

2) Add a diode on the emitter of Q1

Click to expand...

Cathode to emitter? Anode where?

Diver300 said:

3) Feed the collector of Q1 via a resistor. You can lose up to about 20 V without affecting the voltage of the 5 V supply.

Click to expand...

No zener? Or just an extra resistor between collector Q1 and 45V_dc?

Diver300 said:

4) Run the 15 V / 5 V supply from the centre tap of the transformer. It will need its own smoothing capacitor but it won't be affected by what the 45V supply does and there will be much less power dissipation.

Click to expand...

Very smart! But extra rectifier, capacitor, and the GND of that rectifier will need to be connected to the existing GND, no problem I guess?

For starters I will connect my oscilloscope to the three points I mentioned above and do some tracing during extensive power on/off cycles. I will be back in a few days, I will be offline until Wednesday.

Edit: indeed I had placed the LM1117 wrong way round; yet in the actual circuit it is correctly placed.

earckens said:

Maybe I should put a third oscilloscope channel to the base of Q1?

Click to expand...

If you have more channels, use them. It can't hurt.

earckens said:

Replace the current 5V zener with a 15V with cathode directly to the base of Q1?

Click to expand...

Yes. Q2, it's base resistors and the capacitor also go.

earckens said:

Cathode to emitter? Anode where?

Click to expand...

I meant to put a diode in series with the emitter, so anode to emitter, cathode to the 15 V connection.

earckens said:

No zener? Or just an extra resistor between collector Q1 and 45V_dc?

Click to expand...

There would be no change to the zeners in this suggestion. The resistor would be as you said, between the collector of Q1 and the 45 V capacitors. I don't know what resistance or power rating because I don't know what current you circuit supplies but you could aim for 15 - 20 V across the resistor under normal operation. That would also prevent the current being too big if something behaves strangely during start-up.

earckens said:

But extra rectifier, capacitor, and the GND of that rectifier will need to be connected to the existing GND, no problem I guess?

Click to expand...

You don't need another rectifier. Just connect the collector of Q1 and the top of the R16 to the centre tap. You will need a capacitor between ground and the centre tap so that the centre tap voltage is held up during the parts of the mains cycle when none of the diodes in the rectifier are conducting.

It is very common to have a + and - supply made from a centre-tapped transformer, such as this:- https://www.bristolwatch.com/ele/webp2/78xx_bipolar.webp

Those have the smoothing capacitors connected to the centre-tap. I don't think that you want to change the smoothing capacitors for the main 45 V output, but you still need a smoothing capacitors for the 15V / 5 V supply. In the existing bridge rectifier there are two diodes connected between the -ve and the ac connections (as well as the two connected to the +ve connection). The two diodes connected to the -ve connection will do the rectifying that you need to generate around 22 V at the centre tap.

Here is the image from my scope:
1. yellow = rectifier output (45V dc) (10V/div)
2. turquoise = 15V dc line (5V/div)
3. purple: Q1 base (5V/div)
4. blue: 5V zener cathode (2V/div)
20ms timebase (ie @50Hz 2 phasecrossings per 10ms)

I guess the hiccups in the yellow line are the 50Hz phasecrossings?

Anyway, nothing special here.. but nothing blew up here neither

I guess I may have to do these power on/off cycles until something does blow up and then view the scope recording....?

The rectifier output only goes up near the peak of the 50 Hz supply, which happens every 10 ms which is what you are seeing.

None of the traces appear to have frighteningly fast rises.

I suggest that you change the oscilloscope timebase to 5 ms per division (or even 2 ms / div) because there's nothing to see after 20 - 25 ms.

It is also worth repeating the test a few time as it may depend on where in the mains cycle the power is turned on. It also may depend on what voltage the big capacitors have discharged to.

Of course, until you've found a cause, you can't rule out the possibility that adding the oscilloscope probes prevents the problem happening.

Diver300 said:

It also may depend on what voltage the big capacitors have discharged to.

Click to expand...

I make sure the capacitors are discharged before power on.

Here is a screenshot with 5ms/div timebase:

Diver300 said:

It is also worth repeating the test a few time as it may depend on where in the mains cycle the power is turned on. It also may depend on what voltage the big capacitors have discharged to.

Click to expand...

That will be for Wednesday, I will be offline until then (ie code for vacationtime ).

ronsimpson said:

View attachment 150045

Click to expand...

Unfortunately, the absolute maximum input voltage for the LM1117_5 is 20 V. The pre-regulator is still needed if the centre-tap is used.

Diver300 said:

Unfortunately, the absolute maximum input voltage for the LM1117_5 is 20 V. The pre-regulator is still needed if the centre-tap is used.

Click to expand...

Would a 3W 3.3v zener diode suffice as a replacement for the pre-regulator? The 500 ohm resistor ensures that there will aways be current flow through the zener (around 30mA) so the voltage drop should be close enough to the zener voltage at all times. I know zeners are not precise but precision is not required in this case so long as the drop is guaranteed to keep the voltage below the max rating of the LM1117. A 3.3v zener would dissipate 2.64W at the max current rating for the LM1117 (800mA). Do you think this would be a good idea?

I don't like the idea of regulating a voltage by the voltage drop from a zener, although it would probably work.

You won't need as much as 3 W, because you won't be able to get 800 mA from the LM1117 when there is a voltage drop across it of 14 V or so. That would be 11 W of heat generated, and that would need a big heat sink to handle.

The current taken will be what the circuit takes at 5 V.

I suggest a series resistor and a 15 V zener diode. The resistor value should be calculated to give a voltage drop of around 10 V at the maximum load.

Diver300 said:

Unfortunately, the absolute maximum input voltage for the LM1117_5 is 20 V. The pre-regulator is still needed if the centre-tap is used.

Click to expand...

The 78L905 has an input up to 30Volts, so wouldn't need a preregular.

That said, I often need to make 5V from higher rails. My typical preregular is an N-channel mosfet with a zener setting the gate voltage. As the mosfet takes effectively zero gate current, the +rail to gate resistor can be quite high as it only needs to supply enough current to bias the zener. This preregular only needs 3 components.