Grounding a power supply and filtering.

[si2030]

Hi all,

Whilst this is to do with a power supply I have been working on with some great help from others on this this forum I wanted to be crystal clear about this part as its intimately connected with electrocution if screwed up... so a separate post.


GROUNDING
I have AC entering into the power supply enclosure with an earth wire (green). I have a rectified AC power supply with active being 320V and a ground.

I know I must attach the earth wire to the enclosure (metal).

With regard to the ground - should that also be connected to the enclosure? (I was thinking yes here).

Further, how do you minimise ground loops in the circuit - what should you do in the layout of the circuit?


FILTERING
I have an EMI filter line filter I was going to use.

View attachment 62214

I am intending to use this line filter but should I also use one of these (common mode/differential mode) as well? or do I not have to worry about this?

View attachment 62216

View attachment 62217

From what I can gather common mode and differential chokes differ with the way the coils are connected with the differential choke having one coil swapped around.

From the picture and having a look it appears both coils on these are wound in the same direction... (have I got this right?) So to create a differential choke all I would need to do is swap one of these coils around?



Kind Regards

Simon

 

[alec_t]

I have a rectified AC power supply with active being 320V and a ground.

That sounds lethal. Don't you have an isolation transformer?

 

[si2030]

I would have thought its as lethal as any off line switch mode power supply.

 

[davenn]

I would have thought its as lethal as any off line switch mode power supply.

well that depends on what is happening to that rectified 320VDC next ??

Normally in a SMPS the positive and negative from the rectifier are above ground potential, but are decoupled to ground via hi voltage disc ceramic caps ~ 1kV is common. Also it is normal to have MOV type device from each rail to ground as well for large voltage spike suppression.

I definately would NOT be grounding the negative side of the rectifier output, a real recipe for disaster

Dave

 

[ronsimpson]

That sounds lethal. Don't you have an isolation transformer?

I use an isolation transformer and fuses! Two fuses one on each power line wire.

 

[si2030]

Well let me be a little bit more specific... this is a 10amp adjustable 30 volt power supply with switch mode pre-regulator and LT1083 linear regulator. The Switchmode is using a half bridge into a switch mode transformer which in turn is fed into a linear regulator. There will of course be fuses in this design and all safety precautions required... I am not in this to electrocute myself.

Spice simulation at present.

View attachment 62248

As I have seen in other circuits, the rectified part in some cases is not usually grounded (although for simulation it has to be) however the circuit for the linear part is... some guidence on the initial post would be helpful.

 

[Nigel Goodwin]

Your diagram appears to short directly from live side to isolated side of the transformer - this must not be so - there must be no direct connection across the safety barrier.

 

[si2030]

So what you are saying is that there should not be any connection other than the transformer primary/secondary transformation - no ground etc... there are effectively two circuits only connected by transformers

The DC side should be grounded to the chassis which in turn is earthed but the AC rectifier + primary should have nothing to do with the rest of the circuit ground wise...

 

[MrAl]

Hi,

There are a couple issues that come up here.
The first is that we usually have a metal box with our circuit in it and we have 120vac or more coming into the box with a ground also, so that's three lines: H, N, G for Hot, Neutral, and Ground.
The second is that our circuit might be slightly noise sensitive so we'd like to use the nice metal case as a kind of shield for our circuit, but we dont want to connect one side of the line directly to the case even if it is the Neutral (a reversed AC socket wiring puts the Hot on the case rather than the Neutral).

The way this is usually handled is the Ground (not the Neutral) is connected directly to the case metal (the Chassis). The Hot and Neutral come in through a grommet not touching anything and go into a line filter that is made to handle the required power.

We have a Hot and Neutral going into the line filter and a Hot and Neutral coming out of the line filter. We are going to place two capacitors, both 0.0047uf, one from the Hot to Chassis and the other from the Neutral to the Chassis. But here it varies...some power supplies put the two caps BEFORE the line filter, and other power supplies put the two caps AFTER the line filter.

The two caps allow the circuit to get some shielding by the case for higher frequencies but dont allow too much line frequency current to pass.

This kind of arrangement will be found on many ATX style power supplies.

 

[si2030]

MrAl you are saving me here...

The spice modelling has the limitation that it wont converge unless all isolated nets have a ground... in a real world environment I recognise this is not a limitation. In the latest incarnation I have a situation where I used an IR2183 which connects the Mosfet drive circuitry directly to the low voltage circuit via a shared ground (see above) - the IC has a level shifter in it to handle the high side and doesn't need a special supply on the output... I think I understand now I must make sure the 320DC (rectified Australian mains) must not have the same ground as the low voltage circuit or, indeed, any ground (floating) and be completely isolated from the rest of the power supply circuit... the only connection being via the main switch mode transformer and its secondary...

Considering the IR2183 is connected to the gates of the mosfets and also to the TL494 (no isolation) do I dispense with the IR2183 in this regard.... or... how can I use the IR2183 ensuring I still have isolation? optocoupler/ gate drive transformer or is there another more elegant solution to this...

Kind regards

Simon

 

[ericgibbs]

hi si,
The 'Common' symbol should be used, look at this example that uses that symbol and the 'ground' symbol, note the high value resistor.

E

 

[Nigel Goodwin]

So what you are saying is that there should not be any connection other than the transformer primary/secondary transformation - no ground etc... there are effectively two circuits only connected by transformers

The DC side should be grounded to the chassis which in turn is earthed but the AC rectifier + primary should have nothing to do with the rest of the circuit ground wise...

Yes - there should be no connection from the primary side - although commonly a 'slight' connection is made via filtering components, but these only pass insignificant current, and are specially selected safety components.

I would suggest you try looking at the circuit of any switch-mode design?. You might look at this one (on one of my websites), but note this is a class-II design as many are, R68 and C62 are the special safety components, and for a class-II supply are to provide a discharge path for static from chassis to the mains (and thence to ground).

 

[si2030]

Hi Eric,

Hey I had used the method of a high value resistor to simulate a floating supply but I had always used the ground symbol.. what is the common symbol in LT spice?

Si

 

[MrAl]

MrAl you are saving me here...

The spice modelling has the limitation that it wont converge unless all isolated nets have a ground... in a real world environment I recognise this is not a limitation. In the latest incarnation I have a situation where I used an IR2183 which connects the Mosfet drive circuitry directly to the low voltage circuit via a shared ground (see above) - the IC has a level shifter in it to handle the high side and doesn't need a special supply on the output... I think I understand now I must make sure the 320DC (rectified Australian mains) must not have the same ground as the low voltage circuit or, indeed, any ground (floating) and be completely isolated from the rest of the power supply circuit... the only connection being via the main switch mode transformer and its secondary...

Considering the IR2183 is connected to the gates of the mosfets and also to the TL494 (no isolation) do I dispense with the IR2183 in this regard.... or... how can I use the IR2183 ensuring I still have isolation? optocoupler/ gate drive transformer or is there another more elegant solution to this...

Kind regards

Simon

Hi Simon,

I just happened to notice that your feedback circuit isnt quite right yet. The feedback for this kind of regulation can not be linear, it actually must be non linear or you wont be able to regulate the input to the LT1083 properly. With a linear feedback the voltage across the LT will rise as the output rises and this would cause more dissipation than needed in the LT1083. You can check this in the simulation. It cant be turned down either or else the LT1083 wont get enough differential voltage near the low end of the adjustment range. The way it is supposed to be is the voltage across the LT is to stay somewhat constant, at say 2 or 3 volts. The network you are using in your previous post is made up entirely of resistors and that means it must be a linear network, so we have to fix this.

Your second problem is the isolation between the line and the output. You want isolation there as you indicated.

Taking both problems into account, it looks like the best solution might be to use an opto coupler for the feedback. The opto coupler will provide the needed non linear feedback because of the way we can connect it to the circuit, and also at the same time provide isolation between the LT section and the TL494 chip (and associated driver chip and MOSFETS and line). This means you wont have to use opto couplers on the output of the TL494 either which sometimes introduces an imbalance in output transformer primary currents.

The idea then is to connect the opto coupler input across the IN and OUT of the LT1083, possibly with an included series 1N4148 diode. The opto is to conduct when the input output gets to around 2 or 3 volts but not conduct much otherwise. This is the needed non linear function.
The output of the opto goes between the TL494 power +Vcc pin and the feedback input 1IN+ on your schematic, also with say a 1k resistor from 1IN+ to ground. The other input 1IN- goes to a reference voltage of maybe 2 volts developed from the reference of the TL494.

The operation then starts out by pumping up the input to the LT494 little by little with no feedback signal. As the input rises and the LT1083 output rises, the LT1083 starts to put out some current into the load, and this causes more of a voltage drop across the LT1083. Eventually the LT1083 starts to regulate, and this causes its input impedance to rise which means the input voltage rises yet the output doesnt rise anymore. This eventually causes more of a voltage drop across the LT1083 and that causes the opto coupler internal LED to conduct. When that LED conducts, the output transistor starts to conduct. When the output transistor starts to conduct the feedback on the input to the TL494 starts to rise. When the LED conducts enough that means the voltage across the LT1083 must have reached a particular level such as 2 or 3 volts, and since the opto transistor is conducting that eventually causes enough feedback to the TL494 to cause it to start to cut back the PWM pattern duty cycle, and this of course causes the input to the LT1083 to stop rising. This mechanism keeps 2 or 3 volts across the LT1083 at all times.

This isnt the way it is usually done however, as normally just a transistor is used here instead of an opto isolator chip. The difference is that the transistor can respond extremely fast when the input (to the LT1083) changes, but the opto coupler responds just a little bit slower. This means you'll have to check for stability and if it seems unstable you'll have to try a faster opto coupler.

 

[davenn]

deleted

 

[si2030]

Implemented but not quite working properly..

Hi MrAl others,

Thanks very much for that detailed reply... this is helping immensely.

I have now achieved full isolation between the mains section and the low voltage linear section.

I am however having some trouble with the way this is working.. its working but I can only manage a difference of just under 1 volt between the voltage thats entering the LT1083's and the voltage leaving them...

AS I understand it the optocoupler only turns on if the difference between the input and output is greater than some predefined value. When it turns on it sends the noon inverting input higher than the inverting input and so the error amp turns the TL494 output pulses off... thus letting the LT1083 input voltage to decay down to a point where the optocoupler turns on again however.. I cannot achieve a voltage gap bigger than 0.9 Volts between input and output.

View attachment 62303

Whilst this shows just one diode I have had as many as 8 diodes with the difference not budging from a difference of - 950mV.

I think this will work but I cant figure out how to make that gap bigger. The LT1083 requires at least 1.25 volts differential to work properly and I would have thought a safety margin in the vicinity of 2-3 volts would be more likely.... but I cannot get it to make this gap bigger.

I have also placed the simulation of this as an zip attachment if you and any others might run it...

Hoping you can see why this will only create a 950mv gap rather than one that is 2-3 volts...

Kind regards

Simon

 

[MrAl]

Hi Simon,

We would need a resistor in series with the diode(s). The value with one diode would be about 220 ohms but it partly depends on the current transfer ratio of the opto coupler. The output resistor of the opto might be better at 1k, but you should see if you get too much feedback.
The theory here is that the opto LED conducts a current:
iLED=(vDiff-(vD+vLED))/R
where
vD is the voltage of the diode (typically around 0.7v),
vLED is the internal LED voltage (typically around 1.2v),
vDiff is the differential voltage you want across the LT1083 (typically 2 to 3 volts),
and R is the resistor in series with the diode.
And the other part of the theory is that when the LED conducts this current iLED and if the opto's current transfer ratio is 100 percent then the output also conducts iLED amps and that current across the output resistor has to develop 2.5v for the feedback. So 220 ohms might be too low or it might be just right depending on the current transfer ratio and the output resistor equal to 1k.


Second, the reference needs to be raised to 2.5v for that feedback, so if you have 4k on top then 4k on bottom if it's a 5v reference.

In theory this really does work, it's just a matter of getting the simulation to work now. If you still get only 1v difference after these two fixes, then something else is definitely wrong with the circuit.
I might also suggest that to get this special type of feedback to work in the simulation you might get rid of all the unneeded parts like the current limit circuit and related, and especially the variable resistance and pulse source. Use a fixed resistance and fixed bias for the LT1083 and the simulation should go faster. Once you get this feedback working you can always go back to the full circuit.

 

[si2030]

Many thanks MrAl

Will start working on it a bit tonight (getting late) and all tomorrow...

Cheers

Simon

 

[ronsimpson]

I agree with MrAl about the resistor with the opto coupler. When current limit takes over there could be a large voltage across the LT1083.

 

[MrAl]

Many thanks MrAl

Will start working on it a bit tonight (getting late) and all tomorrow...

Cheers

Simon

Hi Simon,


You're welcome. It's nice to meet people that are interested in this stuff

The theory behind this feedback appear to be sound, as the following linearized model sim shows. With this model we make the switcher look almost like another linear regulator and do the analysis that way which of course is much simpler then. This can prove whether or not the feedback mechanism has a chance of working or not. If the linear model works then it should work with a switcher too, but if it doesnt work it will never work with a switcher either.
The linear model of the complete power supply does work with the required feedback so that means it should work with the real switcher too.

Here are the sim plots. The difference voltage is shown as 2.6 volts here. That was after making the diode series resistor 1k and the output of the opto couple resistor 4k. The control voltage (feedback) is also shown in blue.