Problem with different grounds using dual output transformer

[edrean]

I'm not even sure what to call the question that I have, but here goes...

I am using a dual output (secondary) transformer (2 x 8v). The two transformer outputs are two independant windings (eg. four output pins, thus no center tap). I want to keep my 5v circuit separate form my 3.3v circuit, so I use one secondary winding to power a 7805 regulator and the other secondary winding to power a 3.3v switching regulator circuit. Each winding first goes through its own diode bridge before going to regulation. In summary - two secondary windings - two diode bridges - two regulators.

Now my processor runs off the 3.3V and it switches transistors which switches relays which run off the 5V. So where do I connect the transistor's emitter? To ground that comes from the 5V's winding or the 3.3V's winding? Does it matter?

Let's say I want to use my processor to interface to a computer's serial port. The MAX232 runs off 5V, but the TX and RX lines come from the processor. Which ground should be connected to the MAX232 GND pin? The one that comes from the 5V's winding?

How does this work?

 

[ericgibbs]

hi,
Connect the two 0V lines of the regulators together so that you have a common ground for the 3.3V and 5V

 

[edrean]

Thanks for the quick reply.

What would be the implication if I didn't connect the to grounds together?

The reason I am doing this is because I want to keep the grounds separate. I don't want any noise that gets into the 5v ground to affect my 3.3V ground. However, I assume that if I connect the two grounds together right at the bridge rectifiers then anything from the 5v part will not propagate into the ground of the 3.3v part.

 

[ericgibbs]

Thanks for the quick reply.

What would be the implication if I didn't connect the to grounds together?

The reason I am doing this is because I want to keep the grounds separate. I don't want any noise that gets into the 5v ground to affect my 3.3V ground. However, I assume that if I connect the two grounds together right at the bridge rectifiers then anything from the 5v part will not propagate into the ground of the 3.3v part.

hi
You cannot drive a transistor base from one supply to control the action of the transistor when its 0v line is not connected.

You can tie the two -Vdc on the bridges together. I assume you have the +V lines all decoupled.?

Also if the grounds do give a problem you could use a low voltage drop 3.3Vreg on the 5V line.

 

[edrean]

You can tie the two -Vdc on the bridges together. I assume you have the +V lines all decoupled.?

I (hopefully) have the decoupling sorted. Here is the first part of the schematic. It shows the transformer and the way that I wired it to the bridge rectifiers. It also shows the two large caps right after the bridge rectifier. Also, right after the 7805 I have another 0.1 uF cap (5V to 0V). I also have a 0.1 uF cap right at my processor voltage input pin.

 

[ericgibbs]

I (hopefully) have the decoupling sorted. Here is the first part of the schematic. It shows the transformer and the way that I wired it to the bridge rectifiers. It also shows the two large caps right after the bridge rectifier. Also, right after the 7805 I have another 0.1 uF cap (5V to 0V). I also have a 0.1 uF cap right at my processor voltage input pin.

hi,
I would recommend a 100nF across the 4700uF to help improve the high frequency noise filtering.
Depending upon the current loading and voltage drop on the 4700uF the bridge diodes are going have a fast/ high current pulse when recharging the 4700uF's

Again dependent upon the loads the 3.3v and 5v are powering you will require a larger value decoupling cap than 100nF across the regulated outputs, say a 470uF or so in parallel with the 100nF

Eric

 

[JimB]

If you need complete electrical isolation between the 3.3v and 5v supplies, you will have to use an opto-isolator to couple the signal paths between the parts powered from 5v and parts powered from 3.3v.
Otherwise the 0v side of the 3.3 and 5v must be connected together.

You mentioned noise on the 5v supply, where do you think that will come from?
The relays should be OK if fitted with diodes across the coil to catch the back-EMF, and the relays themselves should provide good isolation from whatever they are switching, unless it is something very agressive.

JimB

 

[crutschow]

Another method to isolate noise between two grounds is to use a small inductance to connect the two grounds together. For high frequency noise even a ferrite bead may provide sufficient inductance.

 

[edrean]

Depending upon the current loading and voltage drop on the 4700uF the bridge diodes are going have a fast/ high current pulse when recharging the 4700uF's

Thanks for the advice Eric. My processor setup will probably draw 200mA max. If all the relays are on you could probably add another 200mA. So there should be no reason to exceed 500 mA overall at any one time.

The bridge rectifier can handle 1A, while it has a surge overload rating of 50A. The transformer can provide 0.875A on each secondary winding. Is this OK for the charging of the 4700 uF cap?

If you need complete electrical isolation between the 3.3v and 5v supplies, you will have to use an opto-isolator to couple the signal paths between the parts powered from 5v and parts powered from 3.3v.

Opto-isolators is a good idea.

You mentioned noise on the 5v supply, where do you think that will come from?
The relays should be OK if fitted with diodes across the coil to catch the back-EMF, and the relays themselves should provide good isolation from whatever they are switching, unless it is something very agressive.

This circuit is a redesign of a previous one that failed miserably. The relays are used to switch 250VAC coiled external relays which in turn switch 3-phase contactors which switch 3-phase pump motors. Each time I switched the relays off my processor would reboot. I tried various things and in the end it seems that putting MOV's on the AC side of the relays solved the reboots. It should be worth noting that I had groundplane covering my entire board. The relay noise probably coupled to the ground plane and caused a voltage drop in my processor. Needless to say, I am now doing as much as I can to eliminate as much noise as possible coming from wherever.

I am willing to throw caps around my board like confetti, but I don't want to overdo it unecessarily.

Another method to isolate noise between two grounds is to use a small inductance to connect the two grounds together. For high frequency noise even a ferrite bead may provide sufficient inductance.

Interesting. What magnitude of inductance are we looking at here? I have some of those axial inductors (those that look like resistors) lying around. They are 100 uH, but I suspect that may be too much...

Thanks for all of your help. I am sucking all this up like a sponge at the moment.

If I connect the two grounds together I will do so only right where they leave the two bridge rectifiers. The shortest route (lowest impedance path?) for noise on the 5v's ground would then not be past the processor on the 3.3v side.

This is also the first time I am designing a mains power supply (mains-transformer-rectifier), so I am keen to learn where I am going to get caught out.

This piece of hardware will potentially be used in noisy environments (motors, radio signals etc), so I will be using an aluminium enclosure. The enclosure will be wired to the mains' earth. I have a question about this also, but posted it under a separate topic (https://www.electro-tech-online.com/threads/connecting-enclosure-to-earth.115082/).