How woud an experienced person "know" how to build the circuit?

[Beamin]

I'm asking a subjective question. Would you start by drawing out the main parts like the op amp microphone etc then figure out how to connect them? Or would you look up a basic circuit then start modifying it? I can read circuits but I can't really build them. The most I can do is take one and change a few parts so I'm trying to best figure out if memorization is better or experimentation. As far as the values go do you just guess at them and see what works when you build it or do you use a formula like e=IR for each one then build it and see if the values are too high low? If you were making this from scratch would you use an oscilloscope or just a simple multimeter to get it working (assuming you don't use the scope as a meter because you are that good)?

I realize there isn't one answer to this but it would interesting to see if there is a right way based on what I know about members vs a lazy way or more build it see what happens then work out the math vs do all your experimentation with formulas...

When learning in school do they teach you a whole bunch of circuits or do they tech you the parts in such detail that the circuits just become an exercise of the mind vs recalling a memory?

https://youtu.be/b0A_5H801rk

[ataradov]

Most of the circuits out there are combinations of a few basic building blocks that take inputs and produce outputs. Over time you learn those basic blocks, but it is fine to just look them up.

The rest is just matching outputs to the inputs.

[JS]

There are many different ways, most of the time you are good to go with standard configurations or app notes in datsheets, so no need to reinvent the wheel. When picking values, usually there is a cryticall part like input resistance/impedance for the first gain stange and all around that fall off to configure gain and poles. You don't need to memorize them, but knowing they exist and what to search in google is usually good enough to start. The things you use often ends in your mind anyway and the standard values to start playing around as well, ballpark figures so you don't ensld with reardo values ir impractical circuits.

Then once in a while you need a very specific application where you nees to go with something new, could be a special topology using conventional blocks like opamps or a 555 or a wired application of a particular component like exploding them as they weren't intended to be used. In those cases you nees to be more creative, an idea hits you ans you start to work around it, simulations are sometines helpful but quite some experimentation is needed to be sure the circuit behaves as intended in every scenario. You can't memorize those, you need to know what you are doing and be creative, think, when you are getting a shower or a coffee. Experience is a big part of this, knowing circuits and analysing special circuits and application gives a lot of tools for the task.

JS

[german77]

That depends, if you have already the components you figure out how to connect them. If you don't, you start with the basic circuit and modify it to your needs then search for components and make little adjustments to make it work.

In my case I often start with some components the most important and expensive ones. Then depending on the complexity I split the circuit into many basic circuits and expect an ideal input and output of the circuit. After that I adjust or add what its needed to work with the components I already have and do some quick simulations. Then I start joining those small circuits to make a big one. finally I start changing parts to make it simpler and add some circuit protection.

A good example should be the microSupply that Dave made. You start from a small and simple circuit then start adding complexity until everything works as you like.

For the values most of them are quick V=IR and transistor formulas like Ic=B Ib. Others require Logarithmic functions like the discharge time/voltage of a capacitor. Some are RLC formulas that are a little bit more complicated. And finally most datasheets have circuit examples. I often start with the circuit on the datasheet do the first two formulas by hand to get values to start with. Then start guessing comercial values for the other components until it looks good enough, it's faster than doing all the calculations with formulas.

[IanB]

I'm asking a subjective question. Would you start by drawing out the main parts like the op amp microphone etc then figure out how to connect them? Or would you look up a basic circuit then start modifying it? I can read circuits but I can't really build them. The most I can do is take one and change a few parts so I'm trying to best figure out if memorization is better or experimentation.

This is about the process of design.

You start out in your mind with an idea of what you want to achieve, and what steps are required to get there.

For each step, you need a functional building block that can perform that operation or function. In your mind, you will have by learning and experience a collection of possible building blocks you could use. You will pick one of them based on what is available, or by various other factors that help to limit your choice.

In this way, you build up the whole design block by block.

As far as the values go do you just guess at them and see what works when you build it or do you use a formula like e=IR for each one then build it and see if the values are too high low?

This is the next part of the design process. For each building block you have chosen, there will be specific rules for how to select component values for that block (for example, how to achieve the correct biasing, or frequency response, or stability).

Experimentation, or testing, is very important. Just because a design "works" on paper, it doesn't mean it will work if you build it. You have to take that final step of building it and measuring its performance before you can be comfortable you have it right.

When learning in school do they teach you a whole bunch of circuits or do they tech you the parts in such detail that the circuits just become an exercise of the mind vs recalling a memory?

Generally you will get to learn typical example circuits, and also details of how components work individually, and also the theory of how components interact with each other in large systems.

With experience you eventually learn how to find your way through complexity to quickly arrive at working solutions.

[tpowell1830]

Hmmm... don't have a reply to your question, but after looking at your previous posts and this one, you ask some very thought provoking, interesting questions. 

[IanB]

You start out in your mind with an idea of what you want to achieve, and what steps are required to get there.

For each step, you need a functional building block that can perform that operation or function. In your mind, you will have by learning and experience a collection of possible building blocks you could use. You will pick one of them based on what is available, or by various other factors that help to limit your choice.

In this way, you build up the whole design block by block.

Having now viewed the video, I observe that the building blocks required are a microphone bias circuit, a band pass filter, a signal amplifier, a peak detector, a voltage controlled oscillator, a power amplifier/speaker driver, and (optionally) a squelch circuit.

So this you would draw out on paper as a basic block diagram with each block labeled by function and the signal path between each block.

Having done this, you would then think about how to implement each block. Here you need to have some background knowledge based on your learning.

For the microphone input you might search "electret microphone" to see how to connect one to a circuit.

For the amplifier and peak detector you would immediately think "op amp" and would go through your cookbook of op amp circuits.

For the voltage controlled oscillator you might search for different ways to achieve this, or you might innovate and come up with something using creative thought (as might have been the case here).

And so it goes...

[dmills]

Also, a dirty little secret they tend not to teach in school.

Often correct order of magnitude is more or less good enough (And correct order of binary magnitude is almost always good enough), no need for 3 sig fig most of the time.
90% of the components you will ever place can be done by rule of thumb (low speed decoupling, some tens or hundreds of nF, just use what you have), pullups, 1k -10k or so, again it don't really matter most of the time, rail splitters, see pullups and make the two resistors equal.....

The real trick (that only really comes with experience) is knowing when some value actually really matters and you need to get your math on.

For most things there are standard blocks that everyone uses without much thought, and it is actually kind of rare to have to come up with something genuinely new (It is nice when it happens, but as you gain more experience it will happen less often because your mental library gets bigger).

For developing the mental library of fragments, the applications journals published by the various sand vendors are good reading, and if you can find some older books they can have some useful analogue in them.

Regards, Dan.

 

[joeqsmith]

I'm asking a subjective question. Would you start by drawing out the main parts like the op amp microphone etc then figure out how to connect them? Or would you look up a basic circuit then start modifying it? I can read circuits but I can't really build them. The most I can do is take one and change a few parts so I'm trying to best figure out if memorization is better or experimentation. As far as the values go do you just guess at them and see what works when you build it or do you use a formula like e=IR for each one then build it and see if the values are too high low? If you were making this from scratch would you use an oscilloscope or just a simple multimeter to get it working (assuming you don't use the scope as a meter because you are that good)?

I realize there isn't one answer to this but it would interesting to see if there is a right way based on what I know about members vs a lazy way or more build it see what happens then work out the math vs do all your experimentation with formulas...

When learning in school do they teach you a whole bunch of circuits or do they tech you the parts in such detail that the circuits just become an exercise of the mind vs recalling a memory?

https://youtu.be/b0A_5H801rk

For a profession, I suspect engineering by the seat of your pants will be looked down upon.   For a hobby, I am free to guess and try as much as I want.  I would think the point of school is not to see how well they can train a parrot but to teach you how to solve problems.

[dmills]

That is true, but the reality is that nobody working in the field has time to reinvent from first principles every time, and why would you.

Now if I am designing something small signal, wide bandwidth and particularly noise sensitive, then yea bring on the theory, and probably bring on the Spice (Cautiously, it is only as good as the models). Same for tricky filters where monte-carlo methods can be most helpful, but for 90% of what goes on most boards, close enough is close enough and that is a wide gamut.

The trick is knowing when you need to bring on the math and when you need the transmission line stuff rather then lumped element, with modern parts the need is actually increasing as the native edge rates become ever faster.

The point of the schooling is to help you to recognise when you hit the need for more careful consideration, but productivity comes from minimising the need for that careful modelling.

[Siwastaja]

Often, most of the design time is spent trying to find the combination of components that can do the job - especially if we haven't solved the same problem before and hence don't know where to start designing. This means countless parametric searches on Digikey, skimming through hundreds of datasheets just to find an IC to do (part of) the job. The bigger the part a single IC does, the better - we all dream about easy single-chip solutions. Then, you'll find out that the more integrated ICs, which look perfectly ideal on the first datasheet page, giving you high hopes, do more things wrong, or opposite to your specific requirements. So you are back to using multiple ICs, or typically need a combination of passive components, discrete semiconductors, simple ICs such as opamps or logic gates, and highly integrated ICs. At the same time, you need to make sure the components you choose will be available for production, or that replacements exist. Especially during the current component manufacturing/sourcing crisis, this has gotten more difficult.

While playing this game, the "schematic" changes all the time, based on these limitations on parts you can find. And once you find the right parts, the schematic is mostly fixed, through the combination of application notes describing "typical" usage, added with a bit of your own ingenuity, some "basic" building blocks you have acquired over the years in your mind, and by Googling how others have solved similar problems. And, sometimes it's a lot of tedious brainwork in the front of a simulator and/or soldering iron&scope trying things out, sometimes quite randomly.

Then, we throw in all the "rules of thumb" to save time, add some components randomly without doing due analysis, and sometimes we shoot ourselves in the foot in the process, requiring respin / aftermarket fixes or so.

[Beamin]

Hmmm... don't have a reply to your question, but after looking at your previous posts and this one, you ask some very thought provoking, interesting questions. 

Thank you others here have said the same thing. And even some others get annoyed at these questions because their brains are very linier. When I was in elementary school my gifted and talented class teacher thought us to "Think out side our comfort box". Try solving things that were hard or coming up with solutions where the obvious answer was none, or the status quo answer no one questioned. Doing this in business I was able to add lines of revenue we didn't have before or turn expenses into profits. You know you are doing it right when people say "I would have never thought of that" or "That will never work it has nothing to do with this." I also never use the word "can't", not even in my head, and subconsciously that leaves my mind open to explore solutions.   

[Raj]

Curcuit building depends highly on what you've been trying to make ,in my prospective.I ask the following-
Has it already been made?
if yes-Did it have a room for improvement?
         How can I improve it?

If no,what is going to be the purpose of it?
What parts/ already available modules/ parts of other people's circuit can achieve that
what circuit layout is best

then i'll see how i can bring power consumption,cost,interfearance and glitches down.
it'll involve -calibrations.
lots of measurement.
seeing and calculating where can I switch to cheaper and more common components. (this is where the main theory parts comes in).
trying to bring the number of different parts down. etc etc

 then I try to meat compliances


But all of the above is only when I try to make something for others.
If I am making it for myself, Money is non issue and I get batsh!t crazy,making multiple prototype boards.

But let me tell you.There's nothing new under the sun.
Most of the time,what we are doing is making the same things smaller/faster or with alternative parts,specially if the previous thing can't be made anymore because of parts being manufactured no more.

[tggzzz]

The answer is simple. You:

• read about standard components and building blocks, and understand how/why they work
• define what you want to achieve, numerically where possible
• read and research how others have done it before
• understand how well/poorly previous solutions match your requirements
• understand how well/poorly previous solutions have worked, and why they work
• think about how to divide your requirements into lots of smaller sub-units
• repeat 3,4,5 for the sub-units units
• work out how you are going to test whether each of the sub-units is functioning as you expect
• calculate the values of the critical components
• choose suitable components
• build
• test each sub-unit
• where something isn't performing as you expect, understand why and calculate replacements
• rinse and repeat 1-13 as appropriate

No, that isn't easy. It requires a lot of work over many years. Gradually you improve.

Note that the key points are research and understanding of the theory. Without those you are a caveman fumbling in the dark.

It also helps to have point 2 sorted out, although it can change over the course of time.

[Audioguru]

When I design a circuit I want it to work properly. Then I do not "guess" at parts values, or fiddle with many values until the circuit barely works when I build it wrong. Instead I calculate parts values with datasheet spec's of active parts and choose parts values that will allow the circuit to work properly when the spec's for an active device is minimum and maximum. Then every circuit I build works properly.

I have modified many circuits so that they do what I want and I have modified many circuits that do not work where the parts are wrong because their values were guessed or because the designer never used datasheet specs.

Minimum and maximum specs are important because when you buy a transistor you do not know if its current gain is low or is high, and each one is probably different. I design a transistor circuit so that it works properly with a transistor that has any passing current gain.   

[Raj]

Well.
I have to say,these days ,there's no fun in making electronics.Computer does most of the stuff for you and the stuff you do these days is just buy parts and get them together on a board and program it.
Back in the analogue and semi digital days, You would have to spend weeks diagnosing and fixing problems and since every device was a a different thing unlike today's smartphone which is literally everything fitted into a single unit, the demand for electronics engineers was pretty high.These days ,its more like either you are soc making Phd guy or nothing.

Now people like us have to spend more time thinking about how we can create new kinds of bulls#!ts like dongles,supporting peripherals that no one really needs (like gaming gear),internet of things,blockchains etc just t get a little bit of money from consumers.\

Either that,or you work in the industry of making industrial gear/research gear.

[Beamin]

A good example should be the microSupply that Dave made. You start from a small and simple circuit then start adding complexity until everything works as you like.

Episode # 221 lab power supply?

Does taking an ATX supply and adding a 5W resistor count? It has negative voltages and its free!
Can you take multiple ATX power supplies and put them in parallel to make more current? Or is the power supplie's overload circuitry too complex or finicky to do this?

I have built one power supply using a LM part but there wasn't so much to figure that out. I did figure out why the bridge rectifier connects the - sides to the + out though.


While on subject how can you send logic level signals from an arduino to another device (like a relay board with it's own power supply and MCU) when they don't share a common ground? I thought they would have to be joined together or use the same power supply to do this. Can you really just send one +5V lead to anothers I/O and have it work? I see things that look like this but it doesn'tmake sense like connecting one side of a battery.

[T3sl4co1l]

FWIW:

Notice that design works in permutation space.  That is:
1. We have some variable set of components.
2. Components have pins.
3. The list of connections between pins goes potentially as (pins)! (that's the factorial operator).  In a real design, the connections will be sparse, so the size of the space is on the order of, say, 4 choose 100 (that's the 'choose' operator).

Needless to say, the space is large, so you cannot memorize solutions.

That's fortunate for us engineers, who get paid to solve for points in that space.

We of course narrow down that space considerably, by applying electrical rules (any number of inputs can be connected together, and must connect to only one output; outputs cannot connect together), and using building blocks (amps, gates, current and voltage sources and sinks, switches, filters, etc.) to bring order to the mess (say, reducing a problem of 4 choose 100, to more like 3 choose 20 -- which is still pretty big to attack by brute force, mind you).

BTW, a "space" is a set of coordinates over some range.  A linear space might be, for example, an array of numbers.  3D space is defined by three axes, numbered over +/-infinity.  A permutation space is more specialized, but nonetheless is still just a set of coordinates.  If you assign a numbered net to each pin, then all pins that have the same number are connected together on that same net; if different pins connect to that net (even if it's the same number of pins), it's a different circuit.  So, different permutations are different circuits, and we have a permutation space.  A permutation space is different because it's exclusive: you can't have one pin connecting to two nets at the same time, that's silly, it's just one net all shorted together.  That just reduces to a simpler case.

So, any design approaches, algorithms, compiler designs, all that stuff -- anything that applies to a permutation space, can potentially apply to electronic design.

I don't know if that helps, but there's the joke about how mathematicians solve problems.  You see, they never actually solve any problems, they just restate the problem in terms of some other already-solved problem, and they're done.

Tim

[jmelson]

When learning in school do they teach you a whole bunch of circuits or do they tech you the parts in such detail that the circuits just become an exercise of the mind vs recalling a memory?

In EE school, they mostly teach you how to ANALYZE circuits.  This is a mostly "forward" process, reducing loops and nodes to simultaneous linear equations, reducing them and then solving them.  This is how circuit simulator programs work,  This is fairly easily taught, it is mostly mathematics.  Coming up with circuits is a lot harder, but without analysis, you are just blind.  With analysis, you can select likely generic circuits, write an equation for the generic response, and then figure out the component values to get the specific response (gain, frequency response, etc.) you need.  Then, stack these circuits together to get the complete function you need.  There are "cookbooks" of useful circuits that can be used as a starting point.

Jon

[Buriedcode]

While on subject how can you send logic level signals from an arduino to another device (like a relay board with it's own power supply and MCU) when they don't share a common ground? I thought they would have to be joined together or use the same power supply to do this. Can you really just send one +5V lead to anothers I/O and have it work? I see things that look like this but it doesn'tmake sense like connecting one side of a battery.

At the risk of patronizing you by going back to basics, you need need a potential difference (voltage) for current to flow.  Often circuits share a common ground as the 0V reference, but you could also connect the 5V line instead.  This would mean the grounds of the two connected devices are at different potential(voltage) and all signals are now referenced to the 5V line.  So if the signals are pulled to 0V from the transmitter (say an Arduino) then there will be ~-5V difference between the receivers reference, and the signal.  If the receiver had PNP inputs - this would in fact be the only way to send it a signal.
 
Often relay boards have opto-islators - so require current to drive the internal LED for it to switch the relays - which themselves are powered by a separate power supply. You could hook all the 0V (grounds) together, and use positive voltage signals.. OR.. connect all the LED anodes together, and use your signals for the cathodes. Of course that way the signal is inverted - it is active low, 0V turns on the relay.

The LED in the optoisolator doesn't care about other voltage references... you could hook up its cathode to 1000V and its anode to 1005V (via current limiting resistor!) and it still see's 5V.  Voltage is always relative, it is a "potential difference".

If you're talking about sending a signal with a single wire, then it can be done.. https://en.wikipedia.org/wiki/Single-wire_transmission_line
With ground (literally) reference: https://en.wikipedia.org/wiki/Single-wire_earth_return

About experience with circuits - my university covered a lot but wasn't particularly practical, I learnt far more the year after I graduated, learning different areas as and when I needed to know about them.  Essentially in a modular fashion (which is pretty much the basis of electronics, and is why such technology is far easier to understand than nature - it was created in layers and blocks rather than co-evolved).

I don't believe someone *must* know everything about Fourier analysis and filters before even attempting to design a filter, or know how to design a Zeta converter from scratch before using SMPS devices.  We live in a golden age where components are cheap, many subcircuits are available in modules, where some of the hard work is done for you, and the internet provides FREE information about pretty much anything.  Most software (calculators, PCB design, simulators both analogue and digital etc..) is either free or as limited freeware. 

If you have a problem than can be solved by electronics, then first see how others have done it, and ask yourself why they did it that way.  Then brainstorm your own methods and evaluation the pro's cons of each - engineering is all about compromise and a balance among many factors.  Flow or block diagrams help, which can then be broken down into more specific terms, then you can work on each part individually (sub circuits), start hooking modules up all the while testing and adjusting.  I don't believe I, or anyone else, gets formal training on "how to design a circuit" as its too vague, its mostly just common sense and experience.

[Beamin]

FWIW:

Notice that design works in permutation space.  That is:
1. We have some variable set of components.
2. Components have pins.
3. The list of connections between pins goes potentially as (pins)! (that's the factorial operator).  In a real design, the connections will be sparse, so the size of the space is on the order of, say, 4 choose 100 (that's the 'choose' operator).

Needless to say, the space is large, so you cannot memorize solutions.

That's fortunate for us engineers, who get paid to solve for points in that space.

We of course narrow down that space considerably, by applying electrical rules (any number of inputs can be connected together, and must connect to only one output; outputs cannot connect together), and using building blocks (amps, gates, current and voltage sources and sinks, switches, filters, etc.) to bring order to the mess (say, reducing a problem of 4 choose 100, to more like 3 choose 20 -- which is still pretty big to attack by brute force, mind you).

BTW, a "space" is a set of coordinates over some range.  A linear space might be, for example, an array of numbers.  3D space is defined by three axes, numbered over +/-infinity.  A permutation space is more specialized, but nonetheless is still just a set of coordinates.  If you assign a numbered net to each pin, then all pins that have the same number are connected together on that same net; if different pins connect to that net (even if it's the same number of pins), it's a different circuit.  So, different permutations are different circuits, and we have a permutation space.  A permutation space is different because it's exclusive: you can't have one pin connecting to two nets at the same time, that's silly, it's just one net all shorted together.  That just reduces to a simpler case.

So, any design approaches, algorithms, compiler designs, all that stuff -- anything that applies to a permutation space, can potentially apply to electronic design.

I don't know if that helps, but there's the joke about how mathematicians solve problems.  You see, they never actually solve any problems, they just restate the problem in terms of some other already-solved problem, and they're done.

Tim

[Beamin]

FWIW:

Notice that design works in permutation space.  That is:
1. We have some variable set of components.
2. Components have pins.
3. The list of connections between pins goes potentially as (pins)! (that's the factorial operator).  In a real design, the connections will be sparse, so the size of the space is on the order of, say, 4 choose 100 (that's the 'choose' operator).

Needless to say, the space is large, so you cannot memorize solutions.

That's fortunate for us engineers, who get paid to solve for points in that space.

We of course narrow down that space considerably, by applying electrical rules (any number of inputs can be connected together, and must connect to only one output; outputs cannot connect together), and using building blocks (amps, gates, current and voltage sources and sinks, switches, filters, etc.) to bring order to the mess (say, reducing a problem of 4 choose 100, to more like 3 choose 20 -- which is still pretty big to attack by brute force, mind you).

BTW, a "space" is a set of coordinates over some range.  A linear space might be, for example, an array of numbers.  3D space is defined by three axes, numbered over +/-infinity.  A permutation space is more specialized, but nonetheless is still just a set of coordinates.  If you assign a numbered net to each pin, then all pins that have the same number are connected together on that same net; if different pins connect to that net (even if it's the same number of pins), it's a different circuit.  So, different permutations are different circuits, and we have a permutation space.  A permutation space is different because it's exclusive: you can't have one pin connecting to two nets at the same time, that's silly, it's just one net all shorted together.  That just reduces to a simpler case.

So, any design approaches, algorithms, compiler designs, all that stuff -- anything that applies to a permutation space, can potentially apply to electronic design.

I don't know if that helps, but there's the joke about how mathematicians solve problems.  You see, they never actually solve any problems, they just restate the problem in terms of some other already-solved problem, and they're done.

Tim

   

Thats interesting and kind of how my mind works; to not think in the box but rather to define where the edges of the box are and see if problem lie inside or outside. Also good to see if a problem is too complex to figure out. When you say 4 chose 100 you mean there are four things that can each be connected up 100 ways? Would that be 400 or do they multipy together for some huge number?


What is this way of analyzing called? Seems like a step in the 5 sigma design process that works beautifully in real life.The five sigma can be used when buying a car or online dating or any number of things and can give you a huge edge in real life provided you can think out side that box non linerarly. Why doess my spel check broke?

[Bassman59]

Well.
I have to say,these days ,there's no fun in making electronics.Computer does most of the stuff for you and the stuff you do these days is just buy parts and get them together on a board and program it.
Back in the analogue and semi digital days, You would have to spend weeks diagnosing and fixing problems and since every device was a a different thing unlike today's smartphone which is literally everything fitted into a single unit, the demand for electronics engineers was pretty high.These days ,its more like either you are soc making Phd guy or nothing.

Now people like us have to spend more time thinking about how we can create new kinds of bulls#!ts like dongles,supporting peripherals that no one really needs (like gaming gear),internet of things,blockchains etc just t get a little bit of money from consumers.\

Maybe you're in the wrong field?

Either that,or you work in the industry of making industrial gear/research gear.

I am, and it's both fun and challenging. If it was easy, everyone would be doing it, and then it wouldn't be fun.

[IanB]

In EE school, they mostly teach you how to ANALYZE circuits.  This is a mostly "forward" process, reducing loops and nodes to simultaneous linear equations, reducing them and then solving them.  This is how circuit simulator programs work,  This is fairly easily taught, it is mostly mathematics.  Coming up with circuits is a lot harder, but without analysis, you are just blind.  With analysis, you can select likely generic circuits, write an equation for the generic response, and then figure out the component values to get the specific response (gain, frequency response, etc.) you need.

The complement to analysis is synthesis. Synthesis is less often taught and is harder to do, but is potentially more important when constructing or optimizing designs.

[tggzzz]

Either that,or you work in the industry of making industrial gear/research gear.

I am, and it's both fun and challenging. If it was easy, everyone would be doing it, and then it wouldn't be fun.

Precisely. Spot on.