PCB layout DC-DC convertors 1

[speeder]

Hello,
I am designing a board whose primary function is an amplifier for some very weak microvolt signals. Unfortunately due to space constraints I also have 4 5W DC-DC convertors on the same board which power external devices(not the amplifier circuit)
I have placed the convertors as far away from the analogue stuff as possible but I am not sure what to do with the ground(0V) plane.
I want to avoid any noise from the DC-DC convertors affecting the amplifier stage
The area underneath the analogue section .has a solid plane connected to 0V. I am told I also need to continue the plane underneath the DC-DC convertors or else the board may warp. Should I use 1 plane or should I split it so that the part under the DC-DC convertors is electrically isolated from the analogue 0V plane?
Thanks

 

[Madcow]

Here are some tips.I have used for dc-dc converters.

http://www.fairchildsemi.com/an/AB/AB-11.pdf

 

[electricuwe]

You should use two separated panes, connected only at one point. This way you can avoid circulating Hf-currents in the amplifier part.

 

[felixc]

I agree with electricuwe. Think about the kirchoff law to visualize where the current loops are. Make sure that the current that feeds the DC-DC converter, and those that are coming back from the load of the converter as well, use a gound path that does not go through your analog section.

 

[melone]

Those current loops are exactly what could prevent seperate ground planes from working. If the DC-DC converter is is electrically connected to logic using the non DC-DC ground plane (i.e. enable line from a microprocessor), the recirculation current MIGHT have to travel in a longer, higher loss path. Therefore, depending on the layout, could prevent proper operation.

I have always found that a thicker ground plane (lower impedance) and proper part placement / routing (taking into account current paths), has yielded good results.

 

[dspDad]

A ground plane is not for carrying power, it is for providing a controlled impedance in high frequency lines. Sort of the pcb equivalent of twisted pair. The use of a power plane is equally effective, as long as the power and ground are adequately bypassed. But if you don't run the return through a ground plane, the power and ground lines need to be immediately adjacent to each other, or on adjacent layers of the board. This is to minimize the inductance from the loop area.

Your analog signals probably don't need a ground plane if their frequency is low.

Your dc-dc convertors need the equivalent of a twisted pair supplying their power, and using a ground plane is an easy way to accomplish this, but not the only way.

Beneath and around the dc-dc convertors, especially if you are using a controller + transformer or similar, then controlling stray inductance is critical, and a ground plane is the easiest way to accomplish this.

Most of the dc-dc convertor chip manufacturers provide suggested layouts that keep feedback components and power in/power out leads from sharing the same current paths, and following their layout suggestions is usually pretty reasonable.

Noise can be coupled to your amps inductively, capacitively, by shared currents, and through the power supply. As well as ground planes, make sure the + and - have appropriate bypass caps close to each chip to reduce both high freqency and low frequency coupling, and consider an rc filter to supply clean power to the amps.

DspDad

 

[logbook]

Speeder,
What you don’t mention is bandwidth. The switcher will typically be operating at a fundamental frequency of 20kHz or above. If your circuit has a bandwidth of a few tens of hertz then most of the problems associated with the switcher will minimised.

By far the biggest problem with off-the-shelf switcher modules is the magnetic field they throw out. This couples into everything nearby and is effectively impossible to shield. I suspect that this will be the limiting factor in your design, once your ground plane and decoupling issues have been optimised. It is possible that rotating the module by 90 degrees, 180 degrees or 270 degrees will reduce the pickup.

 

[melone]

Ground planes aren't for carrying power???? How do you complete the circuit if the ground plane is conducting the current back to the power source? The reason ground plane are so effective is because their impedance is small compared to the "supply" side, and the losses are minimized.

 

[dspDad]

Ground planes are used to reduce loop area in lines carrying ac components, especially high frequency components such as those associated with square edges. Open loops define an inductance, and the inductive impedance is proportional to frequency. This results in common mode noise, overshoots and undershoots, and ringing.

Power is often carried by the ground plane, but power almost by definition is mostly dc, and is usually decoupled close to where it is being used to further reduce transients.

As such, power can be carried by any two wires, and the primary concern is the resistance and associated ir drops.

Ground planes usually have low resistance, and are effective for carrying the dc components, but the existence of ground planes was not necessary or especially useful until signals above 5-10 mHz began appearing everywhere. In your old ham radios at 2 meters there were no ground planes and they worked pretty well.

If you have a fast switching signal line on a board over a ground plane, virtually all of the return current flows in the ground plane directly under the lead, not spread out across the copper.

If you think of power return separately from where to slather on a ground plane, it is easier to solve the two problems independently.

The ground plane can also be used as a magnetic shield by placing the components on opposite side. The copper shorts out the eddy currents to some extent. The posts above have questioned the frequencies, etc. The specifics of how to solve or avoid problems depends on the frequencies, how much noise is tolerable, if there are other issues that are being optimized concurrently, (like cost.

DspDad

 

[speeder]

Thanks for all your suggestions!, The analogue signals in question are 1uS wide pulses with have a repetition rate of 300Hz, they can be as small as 10uV amplitude. The 4 DC-DC convertors run at 200KHz. The board must provide isolation from the incoming 24VDC supply therefore they are isolating type convertors. All 4 convertors have common mode chokes and differential mode filtering at input and outputs.
1 convertor provides +/-12V which is used on the board to power a low noise liner regulator arrangement to provide power to the amplifier stage. The other three convertors are not connected to the electronics on this board, their outputs go off to power a digital board and other sensors.
I could have one 0V plane covering the whole board but I am worried that the DC-DC convertors may couple noise into the amplifier stage through it. Because three of the four convertors are for powering the digital electronics I wanted to join their OV to the analogue 0V on the digital board only.
Is it wise to have the analogue 0V plane running under the DC-DC convertors but not electrically connected to them??

Thanks again

 

[dspDad]

1) Are your 4 converters synchronized?
2) Are the outputs tied together?
3) Since they are isolated, do they share a ground, or do some of them share a ground and others not?

The ground plane under the supplies should have at least a good capacitive connection or a direct connection to the supplies ground, otherwise they won't do any good for the supplies.

The other question is whether the analog ground plane (which may not be doing much anyway) should connect to the switching supply ground plane. If the analog section is isolated, then don't connect the ground planes together. If the analog and digital sections share a ground, the ground planes are already connected somehow (or need to be), but it is then important to split the ground plane appropriately to keep digital currents and analog currents from flowing through the same section of the board.

The signals you describe have all of their energy right in the middle of the noise components of your switchers. Tough problem. The switcher noise will probably be in the millivolt range. If they aren't synched, then you will have beat frequencies in the audio range. Hopefully your 10 uVolt signals have low impedance and differential inputs.

Good luck.

DspDad

 

[melone]

Using a single ground plane and paying careful attention to part placement and signal routing should yield a VERY acceptable result.

 

[speeder]

DspDad,
To answer your questions:
1) No, there is no facility on the convertors to do this
2) The board is split into two seperate sections The first section is made up of the first convertor, linear regulators and the analogue circuitry. The second section is made up of the other three convertors (outputs +24V,+3.3V and +5V), these three outputs are not used on the analogue board but go sraight out to power a seperate digital board and the 0V of these outputs are only connected together on the digital board. The two sections have no inter connections at all on this board other than the input supply to the convertors

3) All four convertors share the same +24V input supply, all four outputs are isolated from each other whilst on the analogue board only coming together on the digital board.

If I understand what you are saying correctly, if there is a ground plane under any convertor then in order for it to be effective the convertors 0V output must be connected to it?.

Thanks

 

[CircuitDesignExpert]

About your problem, I think the most important things is the ground plane of the analogue part(Part A) and "the analogue's linear regulator, digital part and all DC-DC convertors"(Part B) should be seperated. Since Part B circuit is not sensitive about the noise from the ground unless the frequency is higher then 66MHz. Thus, the analogue's linear regulator, digital circuit and all DC-DC convertors can be under the same ground plane in order to minimize the noise from each other. The main point is the connection between Part A and Part B should be careful. I think you can use "one point connection". That is for example, you can use one 0 Ohm resistor between the ground plane of Part A and Part B. Also, I recommend the power from the analogue's linear regulator to analogue's circuit shoule be added a "Ferrite Bead". It should be better. If it is worse, you can place 0 Ohm instead of Ferrite Bead. I hope it can help you.

 

[dspDad]

CircuitDesignExpert has the right approach.
I would additionally make sure the power and return lines to the dc-dc convertor feeding the analog section is separated from the power and signal paths used by the analog section.
Good luck.
DspDad