PCB Layout Process and Guidelines?
I'm currently in the process of creating a few PCBs using Sprint Layout and a few questions have come to mind. I'd like to get a few responses and points of view for PCB layout specifically for solid state design (which is why I posted here).
There are a few good nuggets of information scattered in various projects, particularly with respect to grounding and bypassing. I have been finding it difficult to get a high level overview of the subject with good generalizations.
Some of my most basic questions for those who are good at producing quality layouts:
1) What components to place first, where and why?
2) General workflow direction, from input to output, or from output to input?
3) What traces are laid out first, power and grounds or signal?
4) Choosing PCB size and shape, what are the primary factors in selecting PCB dimensions.
5) What components to place on the amplifier module and what should be on the PSU, such as fusing. If one places fuses in the PSU outputs is it still considered 'good form' to put a set on the amplifier PCB?
I welcome any advice and opinions pertaining to the above or otherwise that will help me down the right path when doing a design. Thanks.
Then I start layouting generally with the components in the middle and work my way towards input and output.
Ground should be as close to the signal circuits as possble. Use a ground plane if you can for signal ground only. Use only one point to connect the signal grounds to supply ground. Use star ground to connect the input- and output signal grounds together.
My 2 eurocents worth.
This is the one I normally point people at - simple, common sense from a very experienced designer...l
I am not a professional layouter, but I like to do layouts and learned a lot of somebody, whose profession it is.
Let me give you 2 different layout examples for the same (nearly) amplifier, the FETZILLA.
This layout has been done by Hugh Dean (I hope, Hugh doesn't mind showing his layout in this context):
This is how I did it:
Hugh has his "hand-writing" concerning layout, AlexMM has, I have, ...
(To read Hugh's and my name in the same sentence, ... !)
I am sure, there is no "best hand-writing".
If you do the layout carefully, if you try and keep the tracks as short as possible, if you adhere to some restrictions like running long tracks in parallel... : your own layout will do the job safely as well.
I will now try and answer your questions:
As I told you: I am a layout - amateur, and I still have to learn quite a lot.
Best regards - Rudi_Ratlos
Some food for thought so far, thanks folks. I have a some partially done, I'll post back what I have done for feedback when they are closer to complete.
I believe in starting with the case, then design the boards to fit the case... especialy important with smaller and portable designs. In power amps, heatsinks are often needed as part of the case design. Each heatsink would have an optimal spaceing to give attached output devices an equal share of the cooling capacity. So once you know how big the sink is and where the outputs should optimaly go, it gets easier to add the intermediary stages.
Where posible group parts into their functional properties, like CSS etc. this makes servicing easier to someone not familiar with the boards.
Some designers like our own Graham Maynard are realy good at drawing schematics which can be virtualy duplicated in terms of physical layout, without moving parts around.
Towards the input side, you want to keep traces short and take note of things like thermal tracking in LTP etc...
As for fuses, yes it realy depends on design.
I was trying to figure out why my mini aleph rails were consistantly lower than what simulating would suggest (which is an issue on an amp like that with limited voltage swing ability to start with), becasue I hastily assempled the PSU slapping two PSU pcbs in series (each which had onboard fuses with an esr of about 160m Ohm). Sadly the fuses are not in the ideal places to create a CRC effect with, and if one were to add them there they would not be in an ideal place to disrupt power fast after tripping, as there would then be a capacitor bank still between them and the amp.
I think good practice is to protect components in order of what they cost to replace, starting with the most expensive, your house, so fuse before transformer is of prima importance. Fuses after transformer has diffirent things to concider and sometimes are not usefull at all for protecting against anything but fire hazard. Download some fuse datasheets and become familiar with their real operating behaviours.
A fuse is not a switch which breaks when it reaches a set threshold. It is more like the SOA knee of a transistor which needs to be exceeded by one or other mechanism, either less power over more time, or more power over less, time, but time is always a factor.
1: Any component that requires special mounting like your power devices to a heatsink go on first.
2: input to output or vice versa? I think it's personal choice though I tend to go input to output.
3: I lay on power / ground initially because it's best if it's minimally disrupted and it HAS to be there. Resistors make fine jumpers over power traces. The video boards I did were mostly 4 layer with a solid ground plane. The other middle layer was primarily power routes or if needed, signal routes that will never need to be altered. I've not tried 4 layer for audio but I'd expect it to be very good if not better.
4: I try for rectangular boards with as few routes as possible as they cost less. 'Routes' are special cuts / cutouts for oddly shaped boards. Size often has mechanical requirements for chassis or card cages. Or, leave enough space around power devices so it doesn't 'cook'.
5: Any devices that cause a lot of mechanical stress are better chassis mounted if possible.
First, the very high current surge/spikes that naturally flow in the transformer secondary - through rectifier diodes - through first barrage of filter caps. I don't want any of this on my pcb unless it's out at the edges like Hugh does it. If I can't afford the space to do it his way, then I want it off the pcb.
Either way, I want the current loop area that these large current charging spikes flow around to be as small as possible so they don't couple and radiate into my small signal circuits. This means I want the secondary of the transformer - recifiers - caps to be physically very close to each other with all wiring twisted pairs. The twisted pairs should carry counter-flowing current so that their magnetic fields overlap and cancel out as much as possible.
So if you put any part of the psu on the amp board, you want the rest of the psu mounted very close to the amp board itself. You also want to ensure you have thick & wide traces on your pcb where these large spike currents are flowing.
The second thing I worry about is high frequency noise generated by the rectifier diodes 'switching' on and off. I follow Hugh's recommendation and use soft recovery diodes (that's not the same thing as slow recovery). And if possible, in-line chokes. You don't want these chokes anywhere too close to your amplifier as they do leak magnetic flux. Just like output inductors - you don't want these anywhere close to each other or your small signal circuits either.
And a small afterthought - power supply filter caps - electrolytic caps don't like heat - keeping them away from the hot output stage will lengthen their lifetime. It's an exponential relationship - a drop in temperature has a significant benefit.
Anyhow, I'm not in a position to talk much from experience, these are just the guidelines I give to myself.
In terms of pcb design:
I haven't made many so again I'm not an expert in any fashion. But here are some of my thoughts:
a) I like to keep the main power rails close together so that countervailing current flows are adjacent - minimizes area of loop for magnetic coupling. I also like to keep them short and fat and away from the small signal parts of the circuit. Most everyone else likes to separate the power rails around opposite edges of the pcb - the opposite to my philosophy.
b) Think about where the big currents flow through the load. I find that in most cases I don't have to worry about large currents flowing through the GND on the amplifier pcb - those big currents all happen off board. I place my star gnd at the speaker gnd connectors. On my amplifier pcb the big current flows occur only in the + and - rails.
c) I haven't used a ground plane but may try it in my next design. I do know that a ground plane is a complex animal - it is NOT an equipotential once you move away from dc currents. If you place a ground plane on the one side of your pcb and then run signal traces on the other side, the return currents for these signals will tend to follow the path of the signal traces as they return through the ground plane. You need to consider a ground plane as behaving like a mirror image of the signal traces with all the consequences of cross-talk. The higher the frequency the more this is important. With no ground plane consider carefully where return current flow back to the gnd connector on the pcb - shared ground traces are OK where currents are small, but any larger currents warrant their own trace.
d) Place the 'must have' parts first - usually you have no choice about the power devices. You may also want to place trimmers and connectors in places where they can be accessed. You may need mounting holes.
e) I then partition the board into areas - high current power, low current signal, feedback paths etc. and then layout into these areas. I am careful near any sensitive high impedance points such as the feedback node and input (i.e. the LTP on amplifiers that use them) as these are very prone to capacitive coupling of unwanted signals - you can end up with an oscillator instead of an amplifier this way.
f) If the layout is not pleasing to my eye then I will change it. Don't ask me why, but it seems an important part of the process for me.
p.s. use large pads for through-hole parts, easier to work with and less likely to lift up when you have to replace the part in question because you blew it up !
|All times are GMT. The time now is 09:16 PM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio