Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

What is a good design?

Status
Not open for further replies.

spec

Well-Known Member
Most Helpful Member
ETO is a site for people who enjoy discussing electronics and, of course, there are many facets to electronics. But one major area is design.

Many times on ETO there is a lively discussion about which is the best design to meet a particular requirement. One person will post a design and someone may say it is good while another person will say it is bad.

Oddly, good design is not taught, at lest I have never met anyone who has been taught the criteria for good design. Sure there are courses that teach design (techniques) but not good design.

The software boys are always banging about good design, but when you boil it down they are just describing design techniques rather than good deign

So what is a good design? What are the criteria? I think I know, but what are your views?

spec
 
I designed a display panel... I was lucky that I designed it ( completely by accident) so its uses seem endless..

I had used a 40 pin pic and routed every pin so it could have future use... I seem to add and add more and more features.. There doesn't appear to be any digital / analogue cross talk... There doesn't seem to be any excess noise issues... I powered it with a known SMPS design that is very stable... The screen is connected such that large knocks / vibration give little issue and I can stick the complete display in 1 meter of water!!!

I didn't expect any of this... Doubt if I could do it again... I'm only on the third PCB revision!!

Luck!!!
 
I designed a display panel... I was lucky that I designed it ( completely by accident) so its uses seem endless..

I had used a 40 pin pic and routed every pin so it could have future use... I seem to add and add more and more features.. There doesn't appear to be any digital / analogue cross talk... There doesn't seem to be any excess noise issues... I powered it with a known SMPS design that is very stable... The screen is connected such that large knocks / vibration give little issue and I can stick the complete display in 1 meter of water!!!

I didn't expect any of this... Doubt if I could do it again... I'm only on the third PCB revision!!

Luck!!!

What you did encompasses some of the salient design criteria.

Not luck.:)

spec
 
When I started... It was proven on stripboard... Because of this tracks are kept short as possible as stripboard rearranging is my speciality.. the loadcell amp and the analogue inputs are on one side of the pic and all the digital signals are on the other... When I moved to home brew PCB's I kept the same format, so the very first professional board worked straight of the bat... I suppose the most important design aspect is the separation of analogue / digital components... I learned this from the complete bozo who laid out the last board for my last employers... And he was vastly more qualified than me... Oh and I also grasped the horrible design flaws of ground bounce very early on.... I spent an age wondering why a crystal wouldn't start oscillating.. A 4meg worked.. An 8meg worked, but the 12meg just wouldn't!!! Placing of crystals and stabilisation is the first lesson...
I hear the question " Why is my DS1307 time inaccurate".. "Because you strapped the crystal on crossing data lines..."

So on hind sight... Maybe I have a little idea about design..
 
Extending the discussion to "projects" in general.

They must:
1 Meet the clients specifications.
So that he maybe persuaded to buy another one.

2 Be delivered on time.
So that the client is happy. Also contributes to point 3.

3 Be with budget.
So that you can make a profit.

JimB
 
When I started... It was proven on stripboard... Because of this tracks are kept short as possible as stripboard rearranging is my speciality.. the loadcell amp and the analogue inputs are on one side of the pic and all the digital signals are on the other... When I moved to home brew PCB's I kept the same format, so the very first professional board worked straight of the bat... I suppose the most important design aspect is the separation of analogue / digital components... I learned this from the complete bozo who laid out the last board for my last employers... And he was vastly more qualified than me... Oh and I also grasped the horrible design flaws of ground bounce very early on.... I spent an age wondering why a crystal wouldn't start oscillating.. A 4meg worked.. An 8meg worked, but the 12meg just wouldn't!!! Placing of crystals and stabilisation is the first lesson...
I hear the question " Why is my DS1307 time inaccurate".. "Because you strapped the crystal on crossing data lines..."

So on hind sight... Maybe I have a little idea about design..

What you say is so true. A good layout makes the difference between a only works on Tuesdays circuit and one that is reliable and stays in spec. I have lost count of the number of times I have tried to get engineers to take layout rules into account, only to be told that I am pedantic.

We had one fellow at work, full of theories and simulators. He walked the walk and talked the talk, especially at meetings. One day he came to work on our project so I gave him the task of building a precision xtal oscillator. I also gave him a CD from the xtal manufacturers giving a circuit and set up instructions.

The next day he said that the CD was dud and would not load. I was going to try the CD in my desktop, but then I noticed a deep score across the CD. When I pointed this out he said, Oh yes that must have happend when I cut the wrapper of the CD with a scalple. We sent off for another CD and it loaded fine.

After a couple of days I asked him how it was going with the oscillator. He said that he had built the circuit shown on the CD, but the circuit was rubbish an did not work. I then went with him to the lab to see what the problem was.

I could not believe his circuit. For a start it was built on a breadboard. The amplifier was at one end with no decoupling. And the crystal was on two wires a full six inches long.

When I told him that his layout was wildly inadequate he got all pissy and implied that I was full of moon dust.

In just two days, his replacement built a beautiful oscillator that worked flawlessly for years.:)

spec
 
Last edited:
Unpacking a little bit, and getting down to brass tacks...

A good design:

Is economical to produce
Has a high yield in production
Gives the customer a high ratio of value to price
Works properly (i.e., meets its published specs) regardless of environmental conditions
Once in production, require little or no Engineering intervention
Is insensitive to component variations and aging
Doesn't rely on components meeting their "typical" data sheet specs
Is insensitive to interference (EMI)
Is well protected against ESD
Operates components well inside their maximum ratings, plenty of safety margin
Is reliable (i.e., low failure rate)
Is as simple as it can possibly be, but no simpler
Is easy to use, with a user interface that's as intuitive as possible
Has good user documentation
Has good design documentation
Has minimal susceptibility to component obsolescence​

I'm sure others can cite other factors that make for a good design, but those are the things that come to mind right now.
 
Unpacking a little bit, and getting down to brass tacks...

A good design:

Is economical to produce
Has a high yield in production
Gives the customer a high ratio of value to price
Works properly (i.e., meets its published specs) regardless of environmental conditions
Once in production, require little or no Engineering intervention
Is insensitive to component variations and aging
Doesn't rely on components meeting their "typical" data sheet specs
Is insensitive to interference (EMI)
Is well protected against ESD
Operates components well inside their maximum ratings, plenty of safety margin
Is reliable (i.e., low failure rate)
Is as simple as it can possibly be, but no simpler
Is easy to use, with a user interface that's as intuitive as possible
Has good user documentation
Has good design documentation
Has minimal susceptibility to component obsolescence​

I'm sure others can cite other factors that make for a good design, but those are the things that come to mind right now.

Now we are getting down to the coal-face core specifics of good design.:)

Just two more for the time being:

'component variety reduction'

and a biggie: 'safety'

spec
 
Last edited:
You often hear the word, 'overkill' used on ETO in reference to a design- anyone know what that means?

spec
 
You often hear the word, 'overkill' used on ETO in reference to a design- anyone know what that means?
I think sometimes it means "more complex than it needs to be." Other times, it seems to mean "has far more capabilities than what was requested."

I'm sure there are other meanings of the word, but those are the two that come to mind.
 
Doesn't rely on components meeting their "typical" data sheet specs
This fundamentally important design criteria has some serious ramifications.

Take a circuit board with a hundred TTL chips on it. The typical current consumption per chip may be 50mA per chip, but the worst-case current consumption may be 150mA.

Assuming typical current consumption, you get a 5V line current capability requirement of 50mA * 100 = 5A, but taking worst-case you get a current capability requirement of 100 * 150mA =15A. Which current capability target do you aim for?

Also, what to do if a component data sheet only gives a typical value or, worse still, no value at all for a particular parameter.:eek:

I am not expressing any view, just playing devil's advocate.:)

spec
 
Last edited:
This fundamentally important design criteria has some serious ramifications.

Take a circuit board with a hundred TTL chips on it. The typical current consumption per chip may be 50mA per chip, but the worst-case current consumption may be 150mA.

Assuming typical current consumption you get a 5V line current capability of 50mA * 100 = 5A, but taking worst-case you get a current capability requirement of 100 * 150mA =15A. Which current capability target do you aim for?

Good point. This is one of those times where "don't rely on 'typical' specs" has to be gamed a bit, and tempered with experience and judgement since it is very unlikely that all of the chips on a given board will draw the worst-case current.

Also, what to do if a component data sheet only gives a typical value or, worse still, no value at all for a particular parameter.:eek:
If a data sheet doesn't give any value at all for a particular parameter, I either don't use that part or I make sure my design will work over any plausible, non-absurd range of values for that parameter. In the case where only a "typical" value is cited, I'll usually guesstimate a range of values to work from in checking my design, provided the parameter is not critical. If it is critical, I simply won't use the part.
 
Good point. This is one of those times where "don't rely on 'typical' specs" has to be gamed a bit, and tempered with experience and judgement since it is very unlikely that all of the chips on a given board will draw the worst-case current.


If a data sheet doesn't give any value at all for a particular parameter, I either don't use that part or I make sure my design will work over any plausible, non-absurd range of values for that parameter. In the case where only a "typical" value is cited, I'll usually guesstimate a range of values to work from in checking my design, provided the parameter is not critical. If it is critical, I simply won't use the part.
An excellent answer- sometimes you have to be pragmatic.

In the case of the TTL chips, I regularly worked on racks which had around 1.5K total TTL chips and it was amazing that the total rack power supply load was close to the current corresponding to an average chip current consumption.

Talking about relying on data sheet parameters, I once came across a circuit where the designer was using the inverting and non-inverting input bias current of an opamp to generate an offset voltage.:arghh:

spec
 
Last edited:
Talking about relying on data sheet parameters, I once came across a circuit where the designer was using the inverting and non-inverting input bias current of an opamp to generate an offset voltage.:arghh:
Good grief... :rolleyes:
 
One point that should be made, and that often seems to be overlooked here, is that there is no single best design. There is more than one way to accomplish a task, and just because somebody is using a different approach than you, that doesn't necessarily make it the wrong approach.


Much like Ian said, in my own designs, I try to take a wider outlook than the immediate project. Adding a few extra pads to a circuit board and maybe a few extra components may expand the board's applications and allow it to solve future problems.
 
Hi JS,

One point that should be made, and that often seems to be overlooked here,
Nothing has been overlooked- we are having a discussion about design not a specific design.

is that there is no single best design.
Oh yes there is.:) But in some cases what you say is true: there are more ways than one to kill a cat without poking its eyes out with a sharp stick. But nobody has mentioned a best design. The subject of this thread is good design, not best design.

There is more than one way to accomplish a task, and just because somebody is using a different approach than you, that doesn't necessarily make it the wrong approach.
I can't see anywhere where anybody has said otherwise.

Much like Ian said, in my own designs, I try to take a wider outlook than the immediate project. Adding a few extra pads to a circuit board and maybe a few extra components may expand the board's applications and allow it to solve future problems.
Quite so. That is a tenant for good design. That is all that is needed to be said. Nobody is criticizing your designs.:)

So far everybody on this thread has been objective, lets keep it that way... please

spec
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top