Mind-boggling schematic of power supply.

[pgib8]

I came across the schematic of a power supply at my work. The "engineer" that designed it doesn't work here anymore and as far as the company is concerned the design is fine. There was an "anomaly" that prompted my attention. I spent several days trying to figure out what was going on to no avail but had to return back to my project. Meanwhile another person added some bandaids, which caused another problem, and another bandaid was added. As I mentioned, the company thinks the design is fine but personally I think there is something seriously wrong with it. It's really been boggling my mind, so I'm sharing it because I'm really interested in what some of you might have to say about it.

The only connection this board has to the outside world is AC mains. The first confusing part is that the 3.3V for the microcontroller is at the same potential as the NEUTRAL from the AC mains. For whatever reason this needs to be the case.
In the grand scheme that would put 3.3 at Ground, and it would put GND at -3.3V. That's the first layer of confusion to me. Because the board has no other connections, I decided that it's safe to ignore this part and look at it from a local perspective of the board, so that GND is Ground, and 3.3V is 3.3V and so forth.
On the board you will see a 4V rail. This is dedicated to a wireless communications module. Apparently during the design, this module wasn't transmitting but now the module is updated and transmitting, bringing us to the "anomaly".
So what happened is that when the module is transmitting it will draw about 1A from the 4V supply. At that same time, the 3.3V would dip down to about 1.5V, and recover as soon as the module stopped transmitting.
The finger was immediately pointed at RF interference, and while possible, I was skeptical of that. Something tells me there is a different issue that is with the design itself.
When I looked at the 3.3V regulator, the 12V input was staying strong but the 3.3V output dipped. I concluded that maybe it's not the 3.3V that's coming down, but rather the GND coming up.
I started trying to figure out the return currents and it was really confusing to me.
If you look at the schematic you will see two switchers, one creates the 12V, the other the 4V (from the 12V). For the return current I imagine there are 4 basic paths depending on the combination of which switch is on or off. In addition there are capacitors but I think they are quickly drained to a lower level when the 1A is drawn from 4V.
Perhaps one of the issues is that when the switch for the 4V is off, the inductor L39 is supplying the current. The return current must go through D31 before reuniting with the source at L39. This may be causing GND to rise.

Btw. I noticed that Q1 was in series with U3, instead of U3 driving the gate of Q1. This doesn't have anything to do with the issue here but that tells me that the engineer may not have known exactly what he was doing; so I think it's likely there are other mistakes.

 

[Arouse1973]

Wow that's a lot of capacitance on the output of the regulator. Make sure the regulator is not current limiting when charging up those output caps
Adam

 

[AnalogKid]

To start with, I don't understand why you think the +3.3 V output is connected to AC neutral. There is no connection between the AC neutral and any of the DC output voltages. You can draw a zig-zag vertical line starting above the center of C163, down through the middles of T1 and C16, over to the middle of U2. This is the isolation barrier between the primary and secondaries. C16 is a capacitor, so there is no galvanic connection back across the isolation barrier. It must have X and Y ratings. Also, there is no ground symbol anywhere on the primary side.

The finger that pointed at RF interference is almost certainly wrong. Look at the return current paths in the board layout. Also, what happens to the 12 V input at the 3.3 V regulator power input pin 1 when its output is sagging?

What is the input voltage? The input capacitor stack is rated 800 V, kinda high unless the input is 440 Vac. Q1 might be a strange form of a cascode circuit to protect U3.

ak

 

[Arouse1973]

Just above the neutral input it say 3.3 V? Maybe that's why...dont know

 

[pgib8]

Look at the return current paths in the board layout. Also, what happens to the 12 V input at the 3.3 V regulator power input pin 1 when its output is sagging?

That's what I said, what about them? I think the 12V changes a little (I have to find out) but it definitely stays plenty above the minimum input voltage of regulator. Regulator will dip even when little or no current is drawn. If anything the the problem exists slightly less when more current is drawn from regulator (again return path?).

Also, What is the input voltage? The input capacitor stack is rated 800 V, kinda high unless the input is 440 Vac. Q1 might be a strange form of a cascode circuit to protect U3.

440vac sounds about right. Don't worry about Q1. Probably a rookie mistake but shouldn't have anything to do with.

I did neglect to point out the A3.3 link, which does in fact exists between Neutral and the regulator's output region (A3.3 and 3.3V_CPU). I've never seen this before but apparently it is required this way.

What interests me is that if suddenly 1A is drawn from 4V supply. The caps will only support this for a split second. Let's focus on the current that is making its way through L39. Option 1, the switch inside U20 is open (off), the current is sourced by L39 solely, because: L39 is opposing the attempted decrease in current, by building up a voltage at its terminals to oppose the force that is trying to change current. In this case think of the inductor as a battery, generating it's own power. The minus terminal will be on the left, and positive on the right. In this case the inductor continues to push current through the circuit. The inductor creates its own voltage in substitution to keep the current going the same way it was. As the inductor is "sucking" current from the left side, this void will be filled by current coming in from D31. This is the return path. Due to a voltage drop across this diode, one stands to reason that GND may become elevated.

Question: If GND rises, doesn't that mean the overall potential of L39 rises, so that it's output rises, which in turn wouldn't that raise GND even further? Until the switch closes (on) at U20.

Once the switch is closed, now directing 12V towards L39 that has about 4V across, trying to increase current through the inductor. The inductor opposes the change in current, again by building up a voltage that is working against the force trying to change current. Now the left side becomes more positive in regards to the right (kind of like 2 batteries placed together but opposing each other, one hampering the other).
The potential on the left side will be about 12V (little less). Under this condition D31 is blocking. Where does the return current go now?

I take it must go through T1 (pin 8). You can see how this might get confusing. Also consider the two states at T1, with the switch Q1/U3 open vs closed. During one of those states, the secondary winding of T1 (8 to 10) is surely opposing a change in current in such a way that it will be more positive on pin 8 in regards to 10. Isn't that the reason why D3 exists?
If T1 is in such a state that pin 8 is more positive than 10, but in this case D3 is blocking. Perhaps this makes T1 angry and it build up even more voltage, so that GND rises.

Question again, if this is even true, but if GND rises, then the 12V would rise too? Like the whole system is floating up. Floating up from what? I think this is where Neutral starts to come in.
I couldn't decide on whether it made sense to consider Neutral. In reality it is most likely tied to earth ground, and represents the actual 0V (thus GND actually being -3.3V). But then again, would it make any difference as it's the only link, like if Neutral was at 4000V or for that matter at -4000V, would it even make a difference?

And finally, let's consider the 3.3V regulator U8. To refresh what it looks like inside:
The only real current path in the regulator is from pin 1 to 5 (12V to 3.3V). Not the other way, and not to GND. As far as I know the GND connection is only used as a reference. The 3.3V output is regulated in regards to GND. The connection from 12V to 3.3V is made by a FET. It's not the kind that's either fully on or fully off, but it's kept in the middle to vary its resistance. If the output gets too high, it will increase resistance "burning off voltage" and if 3.3V gets too low (or GND comes up, right?) it will decrease its resistance and let more from the 12V side pass through.

If we think back to how this whole system is floating around Neutral, well the only link to keep that in check is this resistance in the regulator. If somehow GND were to rise in respect to Neutral, then the regulator would reduce resistance, letting more from the 12V go through and pull the whole system back down. Only problem is if there is no current drawn from the regulator, or too little current, it may not be able to do that.

You see I am struggling with this circuit, trying to understand the consequences of it. Personally to me it appears very complicated and I'm sure I'm wrong about some of my assumptions too.

 

[Arouse1973]

You are correct that the output voltage will rise for a rise in common potential. The feedback resistors are chosen to give say 5 volts out with a common of 0 volts. If this rises by 1 volt then the output will rise 1 volt.

This has nothing to do with the inductor, it's the feedback network. If the common point has too high a resistance then this will cause the regulator to keep adjusting it's output voltage as it switches on and off causing large variations of output as the feedback voltage goes up and down.

Thanks
Adam