Newbie Question : Board level design

You dont have to prototype first if you know what you are doing and your confidence is high that it will work. The PCB can be your prototype. In this case you have to make sure you use ISP (In-System Programming) or have a socket for the micro etc. Often you are forced to do a PCB because the circuit is too complex to breadboard or some other factor.

Experienced board designers must know all about signal integrity issues and are often intimitely involved in the design process. In many cases the design engineer will do the PCB layout.

Yes. But the PCB designer will need a complete schematic to work with, developed by the design engineer who knows all about what pins to use on the micro.

You use ISP In-System-Programming. Where this is not possible you might use a surface mount to DIP adaptor board so you can unplug the chip.

Yep, the datasheet.

You might want to have a look at my PCB Design tutorial:

Dave :)

The system engineer has to have some knowledge of both the microcontroller architecture and electrical requirements, so he can assign inputs and outputs to the microcontroller pins in a way that won't unnecessarily complicate the programmer's task.

As programming proceeds, it is quite possible that there will be changes to the schematic to accommodate the programmer's requirements. The actual PC board layout may not happen till fairly well into the project.

In most of my projects at work, I'm the project engineer, programmer, electronic designer, and PC layouter, so I get to argue with myself over all aspects of the job, and will often change pin assignments on the schematic to make the software easier.

App notes help, but with microcontrollers the programmer or project manager has to provide pinout assignments - which pins are used for which input or output signal.

Yes. In many companies, the board layout person just does board layout, and needn't know much about the function of the board - he just gets a schematic from the designer, and makes sure the board matches it.

Try to get a DIP version, even if you will use SM in production. For commercial products, it is common to make a few prototype boards before getting to the "real" board. Test areas, where SM parts can be mounted, can be put on these boards.

You need data sheets for all parts you will be using. Careful study of the datasheets should tell you how to interconnect the parts, and if the voltage levels and timing requirements are suitable.

--
Peter Bennett, VE7CEI  
peterbb4 (at) interchange.ubc.ca  
new newsgroup users info : http://vancouver-webpages.com/nnq
GPS and NMEA info: http://vancouver-webpages.com/peter
Vancouver Power Squadron: http://vancouver.powersquadron.ca

Usually you design the board, and after you order the bare board and the parts, while you're waiting for that stuff to arrive, you write the code. You *think* about the code before you order boards!

The same way tennis players know how to play tennis; lots of practice. But board designers usually do understand the CPU and the application pretty well; some of us write the code, too.

Not "just"; designers design, they don't copy.

First shot is good, and possible if you're careful.

Not recommended.

You don't. You read the datasheets carefully and lay out the real thing.

Datasheet and app notes. The gotchas are usually hidden in the appnotes.

Well, that's the job.

John

Hi, Upto now I've been building simple little projects on breadboards. Now I'd like to try my hand at designing a pcb board with an MCU onboard and all.

I have a couple of questions which have puzzled me.

1) Is the board designed first and then the software written for the MCU or is it the other way around. I had one old school electical engineer tell me the board is designed first whereas I thought the software and circuitry is prototyped in pieces first and then comes the completed schematic layout.

2) How do board level designers (who may not know much about programming the MCU) know HOW to layout the board? Do they just look at application notes from the manufacturer and lay things out and get it right on the first shot!?

3) Is it possible that a board level designer can layout a board without knowing anything about programming the chips onboard?

4) For surface mount chips (not in a DIP format where you can plop it into a breadboard for trial purposes), how do you go about trying them out before actually committing them to be produced on a PCB?

5) When you want to incorporate a chip into your design and hook it upto other chips onboard, what is the first thing you go searching for? The datasheet or...? How can you be sure it will work in harmony with what's already onboard (i.e. all the pin connections are correct)

I got a few more questions but I'll stop here for now.

Thanks

The board *could* be designed first if all of the MCU pins had exactly one possible function. This is often the case with micro-PROCESSORS; this is rarely the case with micro-CONTROLLERS. The schematic (not necessarily the PCB layout) is often a compromise between what's easiest for the software and what's easiest for the hardware.

The programmer may not be too involved with the translation from schematic to PCB layout. The design engineer (who may be the layout guy, as well) must be.

There are design rules and thumb rules for power busses, ground planes, bypass caps, trace widths and shape, etc., etc., etc.

Absolutely, if it is communicated to them what the purpose of each net and node may be.

In-system programming is your friend.

Yes.

The other datasheet.

--
Rich Webb   Norfolk, VA

There have already been lots of replies, but I'll throw in my few comments...

:: Hi, Upto now I've been building simple little projects on breadboards. Now I'd like to try my hand at designing a pcb board with an MCU onboard and all. ::

Do you have any actual layout tools in mind? Although they all require the same basic skills, they all differ in implementation and difficulty

:: I have a couple of questions which have puzzled me.

1) Is the board designed first and then the software written for the MCU or is it the other way around. I had one old school electical engineer tell me the board is designed first whereas I thought the software and circuitry is prototyped in pieces first and then comes the completed schematic layout. :: In a perfect world (rarely realised) the board requirements are stated first (Someone mentioned a project manager - that's one option). If the requirements of the board are clear, the electrical design requirements are at least clearer. As noted, unless the board is for a single purpose or perhaps just a general purpose (with I/O connector positions, for example) it is *never* a good idea to try and design the board without knowing what the software must do. Often, all the design functions (electrical, firmware software, layout) reside in one person. I don't know many electrical engineers who have not written significant amounts of code - they *want* to know what the code must do.

::

2) How do board level designers (who may not know much about programming the MCU) know HOW to layout the board? Do they just look at application notes from the manufacturer and lay things out and get it right on the first shot!? :: It is the task of the electrical engineer to guide the layout person (who may in fact be that same EE). We do schematic capture (with lots of notes), generate netlists and footprint requirements and pass them to the layout person with our notes. It is not unusual for the EE to sit with the layout person to deal with 'special' areas of the board. As to HOW to lay out the board - practise, practise, practise. A healthy does of aptitude helps, though.

We don't just 'look at app notes', although that forms part of the design exercise. App notes live in a perfect, isolated world. We design their parts into a larger scheme, which requires us to know how to adapt the information in the app note to our current requirements.

As to getting it right first shot, that takes a lot of practise (luck helps) and is usually a function of the complexity of the board, although classic neophyte problems abound for even the simplest of units. I had 5 in a row completely correct (i.e. the prototype is the shipping unit) for very complex boards (varied between really small and tight to big and hairy). That's the exception although we always try to get it right. It's the gotchas (which is why you should read the datasheets and app notes thoroughly) that catch you. ::

3) Is it possible that a board level designer can layout a board without knowing anything about programming the chips onboard? :: If you are talking about the electrical designer, I would say *no* except for the simplest of devices. All newer processors and controllers have multi-use pins in these days, and a thorough knowledge of what the code/system requires is necessary to assign the correct pins to the correct board functions. If you are referring to the PCB layout person, then they don't necessarily have to know details, although we have to convey rules about the circuitry (which could be high currents, fast transients, high speed systems etc).

::

4) For surface mount chips (not in a DIP format where you can plop it into a breadboard for trial purposes), how do you go about trying them out before actually committing them to be produced on a PCB? :: Others mentioned making a prototype board first - I have done that myself. On other cases, the SMD board *is* the prototype. For complex boards (or for analog sensitive boards) two different layouts will yield two different results, for an identical netlist.

::

5) When you want to incorporate a chip into your design and hook it upto other chips onboard, what is the first thing you go searching for? The datasheet or...? How can you be sure it will work in harmony with what's already onboard (i.e. all the pin connections are correct) :: The datasheets, application notes, tecchnical articles from the device manufacturer. I also call the local FAEs to see if they have reference designs.

:: I got a few more questions but I'll stop here for now.

Thanks ::

No problem

Cheers

PeteS

The first thing I do is write the manual. Then board layout, parts list, FPGA design, embedded software, test software.

John

The esteemed John Larkin wrote: "The first thing I do is write the manual. Then board layout, parts list, FPGA design, embedded software, test software. "

I completely agree - I wish more of my 'engineering managers' did.

For the elucidation of the original askee:

This is not merely a good idea, it's almost a requirement for any design if one wants to get it right. There have been many times I have heard (or have been known to comment / asked) "...but we thought the functionality was x". Alternatively "but we wanted to be able to use that pin functionality in a bidirectional way". More such comments abound.

A written spec is more than just guidance, it makes sure everyone involved understands what the unit will do, and just as importantly, what it will *not* do.

It takes a little time, but it takes more time to clear up misunderstandings later. The old saw that the further into the design/product process you are, the more expensive (in time and money) any change becomes is no more true than for this.

A week (or maybe more - a lot more, depending on the product) up front saves a lot of time in the back end.

Cheers

PeteS

[snip...snip...]

If I may make a suggestion that it would be a heck of a lot easier to follow your replies if you would employ the usual convention of prefixing each line of the embedded quote with a single character (usually a '>') in the first column.

I understand that you're posting through Google and that the Google usenet interface is egregiously bad. "Do no evil." Feh!

--
Rich Webb   Norfolk, VA

PeteS wrote [with hand-generated context]: . The esteemed John Larkin wrote: "The first thing I do is write the manual. Then board layout, parts list, FPGA design, embedded software, test software. "

------------ There have already been lots of replies, but I'll throw in my few comments...

------------ : :If I may make a suggestion that it would be a heck of a lot easier :to follow your replies if you would employ the usual convention of :prefixing each line of the embedded quote with a single character :(usually a '>') in the first column. : :I understand that you're posting through Google and that the Google :usenet interface is egregiously bad. "Do no evil." Feh! : Rich Webb

Getting blockquotes on Google Groups (The Easy Way):

't-click-the-reply-link-that-is-in-plain-sight+To-get-context-the-easy-way+on-Google+zzz+show-options+click-THAT-Reply-link

earlier this year I made my first MCU design, I used an off-the shelf stripboard designed for DIP ICs I first loaded the MCU with a simple program so I could drive the outputs and verify that the inputs functioned correctly and then after getting those parts working satisfactorilly I started on the software, because I kept adding features (and bugs) I exceeded the 1000 write cycles the chips flash could handle and then I was glad I had used a socket for it,

I think it would be hard to test the software without having a bouard to run it on.

once the designer knows which pins will connect to which board features the layoy task would be similar to any other,

as far as I can see, he need only know what the chip does, not how it has been implemented.

my brother solderes them to a dip-shaped carrier using 10A fusewire,

10A fuse wire is about the right gague for soldering to the terminals of surface mount ICs...

look at the chips data sheet see which pins can do what.... the 90s2313 I used had one pin with a PWM output, I used this to drive a loudspeaker for simple polyphonic sound. another pin had an async serial output capability I reserved this function for debuggging output...

Bye. Jasen

I find the board arrives entirely too fast. I haven't even ordered *all* the specialized parts yet (the assumption that everything's in stock fails with SMT parts), and the time was spent putting out the fires that began burning while the PCB-design push was underway.