How do you learn to design like a professional?

[Cerebus]

1) Practical differences in real components
Placing filter capacitors as close as possible to IC power pins (just happened to have a lab demonstrator mention it one time)
Choosing one type of capacitor over another
Which parameters in datasheets are stable vs. which are likely to have a large uncertainty, change with temperature or frequency etc.
If you need to amplify a signal are you going to just automatically go for an omp-amp unless there's a reason you need discrete transistors?

For the umpteenth time on here I'm going to recommend "The Circuit Designer's Companion" by Tim Williams, ISBN 0 7506 6370 7

From the introduction:

Electronic circuit design can be divided into two areas: the first consists in designing a circuit that will fulfil its specified function, sometimes, under laboratory conditions; the second consists in designing the same circuit so that every production model of it will fulfil its specified function, and no other undesired and unspecified function, always, in the field, reliably over its lifetime. When related to circuit design skills, these two areas coincide remarkably well with what engineers are taught at college − basic circuit theory, Ohm’s Law, Thévenin, Kirchhoff, Norton, Maxwell and so on − and what they learn on the job − that there is no such thing as the ideal component, that printed circuits are more than just a collection of tracks, and that electrons have an unfortunate habit of never doing exactly what they’re told.

This book has been written with the intention of bringing together and tying up some of the loose ends of analogue and digital circuit design, those parts that are never mentioned in the textbooks and rarely admitted elsewhere. In other words, it relates to the second of the above areas.

Its genesis came with the growing frustration experienced as a senior design engineer, attempting to recruit people for junior engineer positions in companies whose foundations rested on analogue design excellence. Increasingly, it became clear that the people I and my colleagues were interviewing had only the sketchiest of training in electronic circuit design, despite offering apparently sound degree-level academic qualifications. Many of them were more than capable of hooking together a microprocessor and a few large-scale functional block peripherals, but were floored by simple questions such as the nature of the p-n junction or how to go about resistor tolerancing. It seems that this experience is by no means uncommon in other parts of the industry.

2) An actual process to follow
Do you make sure to simulate everything beforehand?
Do you get an over-engineered, way too expensive version as a prototype first and then begin replacing sections with slimmer designs?
Is there a general rule for choosing between buying an IC to accomplish somethings vs. making your own circuit? How complex does it have to be to justify just buying a premade IC?
When selecting components do you have certain transistors you default to for certain tasks even if a slightly better one might exist?

I've never seen anybody go through designing a circuit, and while I can and have designed circuits I have no idea if they're "good" solutions or flimsy ones. Mainly, are they robust enough?

I think a good summary is: What is the difference between the textbook and the real circuits? And what are the general rules of thumb and shortcuts you use when choosing a circuit design?

I know that's a very broad question, and if it could be answered in a forum post electrical engineers would basically be replaced by programs in a year, but any advice would be great.

I think an anecdotal approach might be helpful here, I'll write up the process around a circuit I'm in the middle of designing, that indeed is not yet finished. It'll take a little while, so I'll make it a post of its own.

[tggzzz]

For the umpteenth time on here I'm going to recommend "The Circuit Designer's Companion" by Tim Williams, ISBN 0 7506 6370 7

From the introduction:

Electronic circuit design can be divided into two areas: the first consists in designing a circuit that will fulfil its specified function, sometimes, under laboratory conditions; the second consists in designing the same circuit so that every production model of it will fulfil its specified function, and no other undesired and unspecified function, always, in the field, reliably over its lifetime. When related to circuit design skills, these two areas coincide remarkably well with what engineers are taught at college − basic circuit theory, Ohm’s Law, Thévenin, Kirchhoff, Norton, Maxwell and so on − and what they learn on the job − that there is no such thing as the ideal component, that printed circuits are more than just a collection of tracks, and that electrons have an unfortunate habit of never doing exactly what they’re told.

This book has been written with the intention of bringing together and tying up some of the loose ends of analogue and digital circuit design, those parts that are never mentioned in the textbooks and rarely admitted elsewhere. In other words, it relates to the second of the above areas.

Its genesis came with the growing frustration experienced as a senior design engineer, attempting to recruit people for junior engineer positions in companies whose foundations rested on analogue design excellence. Increasingly, it became clear that the people I and my colleagues were interviewing had only the sketchiest of training in electronic circuit design, despite offering apparently sound degree-level academic qualifications. Many of them were more than capable of hooking together a microprocessor and a few large-scale functional block peripherals, but were floored by simple questions such as the nature of the p-n junction or how to go about resistor tolerancing. It seems that this experience is by no means uncommon in other parts of the industry.

That's a useful book; I have the first edition (1991). Just as no university course can cover all topics required in industry, no book can cover everything. And any course/book that attempted to do that would fail miserably.

Once the basic theory and practice of circuit design is under the OP's belt, they can consider getting TAoE and x-Chapters, Pease's "Troubleshooting Analogue Circuits", McConnell's "Rapid Development" (software specific).

Like everybody else, any professional should read the comp.risks usenet group, also available (and searchable) at https://catless.ncl.ac.uk/Risks/ Why? Becauses it induces humility and an appreciation of how things fail.

[Ice-Tea]

Don't know why you guys complicate things so much. The difference between hobbyist and professional?

The hobbyist looks at the typical values in a datasheet. The professional looks at min/max. 

[Cerebus]

As promised a narrative of a circuit design, something I'm still working on.

As part of something I'm designing (for myself as a hobby project) I have a need to measure the timing of a digital signal to a greater resolution than my system clock will allow. The system clock is 10MHz but I want to measure the time between two subsequent input pulses to better than the 100ns resolution that this clock allows.

What I'm measuring is the time between successive 1 Pulse Per Second (PPS) pulses from a GPS timing receiver. The timing receiver in question itself uses a 12.504 MHz and can only select one of those clock edges to output the PPS signal on, this means that the timing of the physical output signal only has a resolution of 39.98ns. However, internally it can calculate the timing of that signal to higher precision. The receiver has an accuracy of around ±15ns and a quantization error of ±20 ns from the clock edge selection. It will output, as serial data, the difference between when the PPS is physically output and when it should have ideally been output. This correction factor is available as a single precision floating point number, which is good enough that we can presume that it doesn't dilute the precision avalable internally.

My objective is to measure the actual transition of the PPS signal with sufficient precision and accuracy that I minimise how much I dilute the innate accuracy of the GPS timing modules. Obviously sampling at my 100ns clock would result in terrible dilution of the precision available from the GPS timing receiver. So I need a way of measuring the signal with greater precision and finer resolution.

Now, I had two choices here. Stick with the digital domain or switch to the analogue domain.

Sticking with the digital domain would be simple, but would require using a much faster clock and much faster digital electronics to be able to count at a rate fast enough to capture meaningful resolution. Any compromise in clock speed will reduce the precision available to me. My current choice of digital circuitry is a Lattice MachXO2 FPGA, picked becuase it's a cheap part, available in QFN packaging so doesn't require an exotic PCB and a cheap breakout board is available. These come with an integral PLL that I could use to multiply up my 10MHz system clock. The XO2 will run a 16 bit counter at 324 MHz and a 64 bit counter at 161 MHz, timing for a 32 bit counter isn't offered in the datasheet. A guesstimate says that I could run a 32 bit counter at 200-250MHz giving me a timing resolution of 4 to 5ns. That's an unattractive resolution compared to the inate uncertainty of ±15ns of the timing receiver, adding something like 33% more uncertainty.

There are several techniques in the analogue domain for 'time interpolation' that can be used to measure the difference between when the PPS pulse arives and the next clock edge available in my digital domain, using this as a vernier indication against the main clock. This is where previously having spent time pouring over other people's designs comes in. From looking at and studying the designs for the HP 3458A multimeter, several of HP's scopes, some of HP's timer/counters and Tektronix scopes I'm aware of several possible designs for time interpolators.

There's a fantastic method using a "Triggered Phase-Locked Oscillator" and digital vernier interpolation used in the HP 5370A Universal Time Interval Counter. It's capable of producing 20ps single shot time resolution but sadly requires the use of custom hybrid circuits to make the "Triggered Phase-Locked Oscillator". So that technique is out.

Most of the other time interpolation methods are variants on the dual slope integrator. They charge a capacitor from a constant current source for the duration of the pulse to be measured, and then discharge the capacitor uisng a much weaker constant current source. This has the effect of stretching the input pulse by the ratio between the two current sources. It's a fairly simple thing to count clock pulses between when the integrator capacitor starts discharging and when it reaches a quiescent 'between measurements' condition using ones slow clock, a comparator, a counter and some sort of state machine to manage the whole process.

How good you can get this dual slope technique to work is a combination of two things, how high you can push the ratio between the two current sources and how stable the circuit is against temperature induced drift and all the other things that cause an analogue circuit to drift over time. Getting a ratio of 1:1000 shouldn't be too difficult. Making reasonably stable current sources in the region of 1uA and 1mA is quite practicable.

So this is the route I went down. The schematic for the current state of play is below. It has got to the "checking it out in LTSpice" stage. It hasn't yet met the real test, which is to prototype it and see how the whole design holds up in the real world. I'm happy with it so far, but I'm not holding this out as an example of good design just as an example of what the design process looks like and some of the decisions that have gone into it. I'm sure the professionals can pick plenty of holes in this particular circuit, when it comes to electronics I am only a humble hobbist after all.

What you're looking at is a circuit designed to take an input pulse that is 100-200ns long, and output a pulse that is approximately 1500 times longer at 150 to 300 us. The length of the output pulse is directly proportional to the length of the input pulse ± some constant. Both the ratio of proportionality and the constant offset will drift with time and temperature. With my 10MHz, 100ns, system clock that gives me 100ns/1500 = 67ps theoretical resolution.

Linearity of the pulse stretcher, as designed, for short pulses is terrible, moreover, producing short calibration pulses for it would be impractical. If we tried to measure a 1ns pulse directly we'd get rubbish results, instead the pulse to be measured is used along with the system clock to gate a flip flop in such a way that the pulse fed to the pulse stretcher runs from the rising edge of the input pulse from the GPS receiver to the second rising edge of the system clock following, so we effectively add 100ns to every pulse that we measure - we can get rid of that in software further down the chain.



(That's a bit difficult to read. If you open the image in it's own window it's a lot better.)

The whole thing is built out of quite standard building blocks. There are current sources/sinks, there's a differential amplifier doing service as a threshold detector and level shifter. There's a differential pair operating as a current switch. There's an integrating capacitor, a clamping diode. The slope amplifer and comparator on the output are quite conventional and are probably underdesigned at the moment. Once the first prototype is built up I'll want to take a careful look at the anti-windup provisions on the op amp, its gain and will possibly add some hysteresis to the comparator - it works fine in simulation but I doubt it will in real life without hysteresis.

There's a classic theory <=> practice isssue here - the simulation has no power supply bypassing. The op amp and the comparator will both need nearby supply bypassing, especially the comparator. Fast comparators tend to chuck at lot of current pulses into the rails and can become horribly unstable without adequate bypassing.

I chose 3mA nominal as the integrator charge current. That's enough to be quick, not so much that the current source and current switch will suffer much from drift from self-heating. I chose a nominal 1uA for the discharge current, and ended up with around 1.2uA. This was picked on the basis that I didn't think I could get away with a lower current. I also chose to simplify the design by not switching the discharge current. Logically one would, but with the high charge/discharge ratio it doesn't stop the integrator from charging over a useful range. Adding the necessary level shifting and switching would have complicated the design to no practical end. The small constant discharge current just drops out as a constant when you do the calibration described below.

The diode clamp that holds the integrating capacitor at its quiescent, between measurements, voltage is constructed with a diode wired JFET. With the low 1.2uA discharge current it was important that the contribution from leakage current from the clamping diode was minimal. A diode wired 2N3904 would probably work just as well in a low voltage circuit like this (the possible reverse bias voltage isn't enough to cause breakdown problems) but I like JFETs.

I expect both current sources to drift. Dual slope integrators automatically deal with many of the non-idealities of the real world, their biggest contribution in this regard is that although the integrating capacitor may drift, it's the raio between charge and discharge times that matters, not absolute voltage reached, so any drift automatically cancels out. Drift in the current sources will be dealt with by measuring a 100ns and a 200ns calibration pulses immediately after taking the actual measurement - a circuit isn't going to drift much over a few milliseconds. Having two known input pulse widths and two measured output pulse widths one can easily calculate the constants for a y = mx + c style equation where x is input pulse length and y is output pulse length.

Many decisions were quite arbitrary. The op amp and comparator choosen where picked on the basis of working with ±3.3V rails, being fast enough and not too expensive (there are no really cheap, fast comparators). They could easily change with little effect on the basic design.

Taking the simulated timings, sticking them into a spreadsheet, quantizing the results to a 100ns clock (including the 100ns/200ns calibration) gives a worst case error of 0.12% against simulated input pulses at 100ns, 101ns, 102ns,..., 200ns. How that holds up in the real world remains to be seen and will probably stretch the limits of my available signal sources - it might even be an excuse to buy some more test gear. 

So, that's an isolated part of a larger system with some of the thinking that went into the design. We've covered where the basic circuits ideas were borrowed from. We've covered some design decisions forced on us by real world component/circuit behaviour. I'm aware that I've only scratched the surface and I've almost certainly glossed over some decisions I made. But as a snapshot of the design process from "here's the problem" to "here we are ready to do some physical prototyping" I hope it's useful.

[ElbowPics]

Thanks for all the help everyone, it's really cleared up how to go about getting better at circuit design.

Cerebus, that breakdown is exactly the kind of thing I wish I could have a whole book of. Seeing how people choose between all their different options, what they consider to be feasible or infeasible, and why they consider some potential problems to be more important than others is really helpful.

Slightly unrelated but my school has been reworking their classes to make design a much bigger focus. Lectures for the students a year or two below me now cover actually making circuits and labs and assignments involve them designing things rather than only analysing them now. It's really good to see.

[Cerebus]

Cerebus, that breakdown is exactly the kind of thing I wish I could have a whole book of. Seeing how people choose between all their different options, what they consider to be feasible or infeasible, and why they consider some potential problems to be more important than others is really helpful.

Good, glad I hit the right note. I thought that was the kind of thing you were looking for.

Bob Dobkin, Bob Pease and Jim Williams are all notable analogue designers, as I presume you know, and all did good write-ups of designs with a lot of their reasoning for, and justification of, choices taken or not taken.

There's a book "Analog Circuit Design - A Tutorial Guide to Applications and Solutions" by Jim Williams and Bob Dobkin (and a sequel "Volume 2 Immersion in the Black Art of Analog Design") that is basically a collection of their application notes written for Linear Technology. It contains a lot of this type of stuff. The application notes are available for download individually from Analog's website, the book is just a convenient way of finding the same material concentrated in one place rather than scattered among a few hundred application notes. As the individual app notes are freely available I don't feel too bad about pointing out that copies of the books are available online in the usual shady places.

You can find collections and individual "lab notes" from Bob Pease scattered around. They tend to be more thematic, but a lot of them walk through various designs. Many are titled "What's all this \$x\$ stuff anyway?" where \$x\$ is the theme, like "Femptoampere" when he discusses low current measurement circuits and so on. Much of Bob's writing was published in EDN magazine and some of those articles can be found in their online archives. Bob Pease also put together a book "Analog Circuits - World Class Designs" that is worth a read, but it is more textbook than a journey through designs like the Williams/Dobkin book or Bob's lab notes. National Semiconductor made a series of videos with Bob that can be found on YouTube. Here's one, the video version of "What's all this Femtoampere stuff anyway?":



For some reason the digital guys aren't as eloquent as the analogue guys, so I find it hard to think of any resources from the digital side of electronics that are of the same quality as the Bobs and the Jims produced for the analogue side of things. On that point I'll leave the rest of the commentary on that topic to another famed analogue designer, Bob Widlar:

[schmitt trigger]

My two Yen;
As others have already mentioned: design, design, design. 

To which I would add; and don’t be discouraged if some of your design don’t work, or even if they fail spectacularly.
Failure is something that every engineer worth its salt has experienced.

A failure is an excellent opportunity to learn. Even if you destroyed some expensive components, if you learned something from the experience, it is not a complete loss.

[wizard69]

I rambled a bit in another post but I and others probably missed the most important element to development in any profession be it a plumber or a NASA engineer.    That is work with others !   Even better vary who you work with when getting started, that may very well mean switching jobs.   I guess the key word these days is a mentor, somebody to learn from and that you respect for their knowledge.

Now these days you have the advantage of the net which leads to world wide access to knowledge.   I'm not sure that the web is a complete substitute to professional development though.  For one there is not that immediate exchange of information be it technical or the look of puzzlement on somebodies face.

[perieanuo]

My two Yen;
As others have already mentioned: design, design, design. 

To which I would add; and don’t be discouraged if some of your design don’t work, or even if they fail spectacularly.
Failure is something that every engineer worth its salt has experienced.

A failure is an excellent opportunity to learn. Even if you destroyed some expensive components, if you learned something from the experience, it is not a complete loss.

mostly right, BUT:
design>prototype>fab>test really thorough; re-design>prototype>fab>test really thorough; re-design>prototype>fab>test really thorough; re-design>prototype>fab>test really thorough;
designing without proto/fab is zero experience, ltspice sim can and is tricky. so prototyping and test is essential in understanding what you miss in your pcb/sch
i can add, those days fab in china is really affordable to everyone who expects to do pcb design, when i started electronics in 90's it was prohibitive to do the level of 2 or 4 layer pro pcb like nowadays so apreciate your high oportunities in this domain

[Microdoser]

Read datasheets.

Get a scope.

Design and build a circuit, ideally one that means you learn something new. Do not rely on 'In theory that should work'

Use the scope to see if the circuit does what you expected. Often it will not.

Read the datasheets again and spot the thing you missed first time which is the reason the circuit does not do what you expected then redesign the circuit and build it again. Repeat this step until the circuit does what you expect. Ask other people for help if you really cannot figure out what you have missed but only as a last resort, ideally you should be learning the skills to figure this out by yourself.

Repeat this process until circuits do what you expect when you build them after reading the datasheets.

This may take a number of years. It may be wise to be paid by someone to go through this process.

There are no shortcuts.

[larsdenmark]

Here are some very general advise:

Figure out if you want to be a generalist or a specialist and plan your learning accordingly.

Learn to question data sheets and designs. Read the data sheet as an advertisement. What are the shortcomings? What it the data sheet NOT telling you?

Practice different types of design in different fields. What are the choices you would make if only one of these topics were important: development time, cost of finished product, (very high) accuracy, stability, longevity, maintainability. Your entire workflow from gathering requirements, performing the design and deciding what parts to use may change a lot depending on what your focus is.

Learn to simulate. Learn what shortcomings your simulation tools have. Learn about new tools to see if you can improve results by having newer or better tools. Make sure that you still understand what a circuit can do and do not rely on only being able to conclude something from a full simulation. Make sure you still can do a back-of-the-envelope calculation for fast calculations.

In a professional setting the two most important things are time and money. Practice to estimate the time and budget required to make a design (what choices do you make if your design is produced 5 times as opposed to 1 million times). Document your work. Use revision control for your designs, simulations and code.

Lastly, the most important way to learn how to be a professional in any profession is to get a job. Make sure you have skills that are wanted in your area. This also covers personal skills such as communication skills, being likeable and having interests outside of being an electronic engineer. Write your resume now. Practice writing job applications. Build a network.

[tggzzz]

Figure out if you want to be a generalist or a specialist and plan your learning accordingly.

Learn to question data sheets and designs. Read the data sheet as an advertisement. What are the shortcomings? What it the data sheet NOT telling you?

All sensible advice, but I think it is worth emphasising the above two points.

Additional questions...

Decide whether you want to work in academia or industry; their objectives are different.

If you want to be a specialist, how are you going to decide which specialty? After my first job I decided I didn't know whether I wanted to remain technical or move into project management or sales/marketing. So I found myself a job where I could try all of those at different phases of a project.

Join your national professional engineering organisation, and go to local events.

[Siwastaja]

Remember, only a tiny portion of value goes into so-called "industrial" sector. Despite smallest margins, most money is in low-cost consumer market. Both are designed by very much "professionals" yet the design strategies are very different.

Forums such as this tend to create a false dichotomy "hobbyist" vs "professional" where "professional" tends to mean high-reliability industrial control designer.

In reality, professionals vary from garage audio circuit designers who have no clue how to design robust electronic devices, to really experienced but very constrained designers at massive multi-national consumer gadget manufacturers, and finally those who design industrial products that sell in small numbers like thousands to tens of thousands at 100x the BOM cost, in USA or Europe.

Plus obviously a dozen of different kinds of professionals these examples do not cover.

Like Cerebus explains, you can add connotations to "professional" but you need to be careful otherwise your "professional" view ends up very limited. At least "getting a living doing it" is a robust definition if not perfect.

[AndyC_772]

I was fortunate in that I had a good mentor in my first job post graduation.

He taught me a valuable lesson which has stuck with me throughout my professional career, which is: when you have completed the first draft of your design, think of all the possible ways in which it could fail, and in each case, make sure it won't fail for that reason.

This encompasses many different disciplines. It could be, for example, that a design will fail because the power supply can't provide enough current, in which case, review your power system design. It might break under vibration, in which case the electrical design may not need changing but you'll need to reconsider how components are physically mounted.

Other reasons you might want to change a product could be to do with solderability, timing margins, EMC, race conditions in software, running out of memory, lack of debug points, component availability, noise, cost, or some change in the market that means your product just isn't viable any more. A good designer is multi-disciplinary, and needs to learn from people who have expertise in many different areas.

One of the hardest things can be to find a way to say to someone "yes, I understand the demands of your job, but I have to balance that against other conflicting demands". You can't please everyone all the time, and don't need to - but it's a good idea to listen to everyone.

[Cerebus]

A reminder, the OP asked "How do you learn to design [circuits] like a professional?", not how to find a job or pick a career path. Several people seem to be repeated veering off in that direction.

Don't be led off on the wrong track just because the word "professional" got used, if he'd used "How do you learn to design [circuits] like a expert?" I suspect people wouldn't be so quick to head off on the wrong track.

Also I suspect that a lot of folks are unintentionally gravitating toward the easy path because they can explain how they found a job, but can't explain how they bridged the gap between learning a lot of theory at university/from books and how they got to the point where they can design practical, working circuit designs that meet specifications and that don't turn into puffs of smoke or radio transmitters at the first excuse. That kind of historical introspection is hard to do, if it wasn't we'd all be writing world class tutorials all the time.

[tggzzz]

A reminder, the OP asked "How do you learn to design [circuits] like a professional?", not how to find a job or pick a career path. Several people seem to be repeated veering off in that direction.

Answering questions is relatively easy. It is much more difficult to find the "best" question(s) to ask.

The OP is a beginner, and may decide there are other questions that turn out to be more important for him than the specific question they asked.

[Cerebus]

The OP is a beginner, and may decide there are other questions that turn out to be more important for him than the specific question they asked.

It would be courteous to wait for him to decide that, rather than substitute the question that one thinks he should have asked and that one likes to habitually answer.  And, no, I don't think that there's a way you're going to legitimately squeeze "XMOS xCORE processors", "Russell Group universities", or "Original Rocky Horror Show T-Shirt that still fits me" into the thread either. 

Edit: You probably can squeeze parallels with "Gliding, learning how to" into the thread however. 

[tggzzz]

The OP is a beginner, and may decide there are other questions that turn out to be more important for him than the specific question they asked.

It would be courteous to wait for him to decide that, rather than substitute the question that one thinks he should have asked and that one likes to habitually answer.  And, no, I don't think that there's a way you're going to legitimately squeeze "XMOS xCORE processors", "Russell Group universities", or "Original Rocky Horror Show T-Shirt that still fits me" into the thread either. 

Edit: You probably can squeeze parallels with "Gliding, learning how to" into the thread however. 

Another thing the OP might like to consider: people are rarely impressed by the introduction of irrelevant strawman points.

[Cerebus]

The OP is a beginner, and may decide there are other questions that turn out to be more important for him than the specific question they asked.

It would be courteous to wait for him to decide that, rather than substitute the question that one thinks he should have asked and that one likes to habitually answer.  And, no, I don't think that there's a way you're going to legitimately squeeze "XMOS xCORE processors", "Russell Group universities", or "Original Rocky Horror Show T-Shirt that still fits me" into the thread either. 

Edit: You probably can squeeze parallels with "Gliding, learning how to" into the thread however. 

Another thing the OP might like to consider: people are rarely impressed by the introduction of irrelevant strawman points.

I'm just joking about your hobby-horses, but I'm serious about not veering off topic for two reasons (1) I don't think we've accumulated all the pieces of the answer to the question posed yet, (2) this seems to be a recurring question recently from current undergraduates and new graduates that have been let down by courses that lean almost entirely or mostly on pure theory and it would be nice to have a thread to point them at when the question recurs (which I suspect it will). There are dozens of threads where peopled have handed out, often good, career advice for new/soon-to-be graduates, there aren't many (any?) about bridging the skills gap between taught theory and practical circuit design.

[JustMeHere]

The same way you get to Carnegie Hall.

[Cerebus]

The same way you get to Carnegie Hall.

On the D, E, F, N, Q, R, or W subway trains? 

[PixieDust]

Maybe one step at a time?

Ah found the answer:

https://www.carnegiehall.org/Explore/Articles/2020/04/10/The-Joke

[wizard69]

Read datasheets.

This is huge and I might add read everything your can that is even remotely associated with your interests.    That includes the trade magazines.

Get a scope.

A good DMM would be very high on my list.   Growing up, when I first got interested in electronics any meter would have been like a gift from heaven.   The two together (scope and DMM) would have caused me to end up permanently giddy.

Design and build a circuit, ideally one that means you learn something new. Do not rely on 'In theory that should work'

Use the scope to see if the circuit does what you expected. Often it will not.

Read the datasheets again and spot the thing you missed first time which is the reason the circuit does not do what you expected then redesign the circuit and build it again. Repeat this step until the circuit does what you expect. Ask other people for help if you really cannot figure out what you have missed but only as a last resort, ideally you should be learning the skills to figure this out by yourself.

In my mind this is what separates a "professional" from an amateur in any field.   Professionals solve problems using their knowledge and experience.    Unfortunately I see this daily with engineers and technicians that can't "figure things out".    Cranking the numbers is a valid skill but that doesn't always solve a problem, especially if there is no canned solution to follow.

Repeat this process until circuits do what you expect when you build them after reading the datasheets.

This may take a number of years. It may be wise to be paid by someone to go through this process.

There are no shortcuts.

This is an excellent post!