Work from home as PCB designer?

[shoppy]

Hello,

Great work for the youtube channels, I really enjoy watching them.

I have electronics knowledge and I played with PCB design a couple of times, do you think that it is feasable to work full / part time as a PCB designer from home?
Are there sites lyke "Rent-A-Coder / Freelancer" that specialise in outsourcing PCB design projects?
(Is there a standard as how such a project should be outsourced(electrical schematic, fooprints, etc)?)

[AndyC_772]

Yes, it's certainly possible to work as a freelance PCB designer, and there are plenty of design bureaux out there which do precisely that.

The difficulty, though, is that if you've "played with PCB design a couple of times", you're a long way from being able to take on other peoples' jobs for money. You need to first find yourself somewhere to train, to get experience, to learn the skills you'll need before you eventually set yourself up as a business.

Have a look for PCB design services in your area; good ones may be expanding and looking to take on junior design engineers. Failing that, plenty of electronics companies design their own PCBs and may be able to give you a job.

I don't see any way to get the experience working from home, though.

[hagster]

Do you count PCB design as seperate to circuit design?

I would guess that most companies that have already designed the circuit are willing to put in the extra couple of hours to layout the PCB.

[AndyC_772]

I certainly would. I regularly design boards that take a week or more to lay out, and they're small and simple compared to the jobs I was doing ten years ago.

[tszaboo]

PCB design is more than just doing schematic and layout.
PCB is usually not the end product, it is a mean to make something work. Which usually means it also needs mechanical design, firmware, software.
And that means meetings and teamwork. Also, if you work as an EE the board has to be brought to life, tested, you usually need a lot of equipment to test everything.
That being said, if you specialize in something like layout, find a lovely company, maybe you could do 50% teleworking.
But for this, I would say the first step is to become that specialist. That takes time. Thousands of hours.

[mikeselectricstuff]

The sort of companies that are big enough to want someone to do just PCB design are probably big enough to do it in-house.
A lot of PCB design is experience and stuff that is hard or impossible to teach - sounds like you don't yet have the experience.
It's also rather too important to trust to some random rent-a-coder site

[shoppy]

Hello guys, thank you for all your kind answers.

What I understand by "PCB design" is that you do the component placing and trace routing starting from the electric wiring diagram and list of component packages.

Yes, it's certainly possible to work as a freelance PCB designer, and there are plenty of design bureaux out there which do precisely that....

Do you have some website or just point out what keywords I should use to search in Google?

[AndyC_772]

Imagine you're an engineer with a schematic you need laid out. What search terms would YOU put into Google? Do they yield the result you'd hope for?

If so, there's your answer. If not, there's a clue to help you design your own web site and get hits from potential customers.

[T3sl4co1l]

I've worked on a lot of jobs where the client has a basic schematic, perhaps pasted together out of odds and ends from prior products.  They want the design checked out, finalized, PCBified and manufactured (at least for protos).

It seems to me, the design checkout step may often be omitted from the RFQ, but it would be in poor conscience of me to ignore a circuit that has obvious problems.

A recent case, the client was inexperienced with Altium, and made numerous connectivity errors by assuming Port and Net Name objects are equivalent.  Easy enough fix (convert ports to net names and off-sheet connectors).

Size of the organization doesn't really matter, it's if their engineering staff are experienced and committed to quality.  A company that is normally dedicated to manufacturing a given product, has little need for design engineering -- one or two guys to handle maintenance activities is enough.  A company that's constantly innovating will have design engineers working all the time, and is more likely to have good quality.

Keep in mind, both cases are a business opportunity (give or take your quality of work).  One needs engineering, when they infrequently design a product; the other needs engineering, when they infrequently need more hands on a job.  (Engaging a contractor to fill in missing labor is a great way to keep tight project timelines.)

As for PCB freelancing, I've seen plenty of "professional" designs more goofy than some amateurs.  Not nearly as bad as beginners to PCB design, mind, but there is plenty of opportunity to grow there.

Tim

[Old Printer]

There is also the interaction of components and traces on the board. As a hobbyist I don't have any formal training, and as such over the years I have been surprised to read about how adjacent components can negatively effect the circuit. PCB layout goes far beyond the convenient and pleasing placement of parts. Even after a formal education in electronics there is a lot of experience required to know what to, and what not to, place near what on a board.

[shoppy]

I'm not saying that it's easy or that I don't need any electrical engineering background, that is not the problem, I totally agree with you.

I was hoping for a simple answer lyke "that's a bad ideea, don't do that" or "it's a good ideea, here is a good place to start, see this lynk / google this out".
(hope this is not bad etiquete)

[Old Printer]

I'm not saying that it's easy or that I don't need any electrical engineering background, that is not the problem, I totally agree with you.

I was hoping for a simple answer lyke "that's a bad ideea, don't do that" or "it's a good ideea, here is a good place to start, see this lynk / google this out".
(hope this is not bad etiquete)

EE is the wrong forum for simple answers, it's not in our genes     but never let that stop you from asking.

[Bud]

My friend, if you think of PCB design is as a simple "placing component and connecting them together", you are very very far from the reality.

[free_electron]

The big thing is : what CAD packages will you support.
Anyone looking for outsourcing layout is going to run on one of the big 3 : Altium ,Cadence or Mentor. They will send you the database  : you run the routing and they will expect the routed file back. And for complex boards they may want you to have specctra as well ...

There are plenty of companies that do contract work for PCB layout.

Typically outsourced falls in two categories
- stuff that is simple enough it doesn't need babysitting and our internal people are better tasked at more complex stuff
- stuff that requires very specialized know-how and tools ( like Specctra)

[tszaboo]

My friend, if you think of PCB design is as a simple "placing component and connecting them together", you are very very far from the reality.

Yes, I always say: that part of the job, a trained monkey can do it. It's just connecting two points together with a line, without the lines crossing. Right?

[cdev]

Eventually, AIs, given enough contextual information, should be able to grok more and more of those rules and implement them.

[james_s]

Maybe some day, but I'm not holding my breath. I've tried out autorouters now and then and every time I found I spent more time screwing around and got inferior results to if I had just laid it out manually to begin with.

[marcuswilson007]

I would not trust anyone without years of audio experience to layout audio boards for me.
I have seen many audio kitset boards (also many commercial products) with serious layout errors.
Modern audio devices are such that a designer needs to understand DC, low frequency and RF engineering principles to layout good PCBs.

[Alex Eisenhut]

Do you count PCB design as seperate to circuit design?

I would guess that most companies that have already designed the circuit are willing to put in the extra couple of hours to layout the PCB.

"couple hours" 

 

What kind of companies are these???

[james_s]

I would not trust anyone without years of audio experience to layout audio boards for me.
I have seen many audio kitset boards (also many commercial products) with serious layout errors.
Modern audio devices are such that a designer needs to understand DC, low frequency and RF engineering principles to layout good PCBs.

Gotta start somewhere, nobody is born with years of experience. I've seen terrible boards laid out by engineers who had years of experience, and some really nicely done boards by people who were relatively new at it. It's something that needs to be judged on a case by case basis.

[T3sl4co1l]

Eventually, AIs, given enough contextual information, should be able to grok more and more of those rules and implement them.

Thing is, will the incremental cost of AI development actually go to zero, or asymptotically settle out?

After all, there is ultimately some cost of energy to computing such a solution, even for ideal quantum computers.  (We are probably centuries away from reaching the information-theoretic limit.)

The question is, can that cost be low enough that it is worth doing, relative to the cost of the conventional approach?

It may very well be that analog circuit design is not automated, in my lifetime, simply because it's not a big enough problem to justify its expense.  And its customers are disparate; there is currently no market where customers can pool their purchasing power that designers must compete for.

There's also the matter of trickle-down economics.  Some technologies trickle.  Three decades ago, no one could imagine affording a computer and CAD system, out of pocket.  Now we have free tools (like KiCAD) that run on absurdly affordable platforms (like a $100 used PC).  That worked out nicely.  On the other hand, it's been over half a century, and IC manufacture has still not made such a transition -- the industry secret sausage is still carefully maintained behind NDAs, and manufactured behind $10k+ NREs.  (Yes, there are some cheaper fabs, but nothing hobbyist level, not by a long shot.  In contrast, small PCBs are cheaper than a meal.)

Which history will apply to these technologies, when and if they are developed?  Only time will tell.

Tim

[tszaboo]

If you think about, how much companies are paying for the layout designers, I can see it being automated within a decade or so. Autorouters are getting more and more intelligent. You can fan-out a BGA with a click in newer versions of Altium. The scary thing about machine learning is, that you dont need that much human interaction, after the setup is made.

Now, this is the situation now:
Auto routers can route a board.
Length tuning and other optimization is not that difficult to do.
Component placement can be automated.
Digital simulation is a solved problem, high end solutions are capable of providing useful feedback.

So it is not unconceivable to set up a machine learning algorithm to learn placement and routing.

On the other hand, PCB routing is an NP complete problem. Which means, that increasing the connections to be routed, the time doesn't increase linearly, more like exponentially (kinda, sorta). Throwing more hardware at the problem is definitely solving it though, since stuff on the PCB costs money, thus it will not get infinitely complex (like IC design is going now).
You would still need the engineer though. To set up the board size, connectors, schematic, footprints, general guides to steer the AI.

But imagine a company, with access to a small supercomputer (or cloud provider), doing overnight PCB layout consulting...

[T3sl4co1l]

But imagine a company, with access to a small supercomputer (or cloud provider), doing overnight PCB layout consulting...

Yech, who knows if it's even correct?  Just sounds like an opportunity to waste more money.  How many prototypes will you spin (minding that an article of the above imagined situation will run >$10k to fabricate and assemble) before realizing you should've had an engineer sit and look at it for a few days, first?

I laugh at "machine learning".  Possible to improve it?  Certainly.  Possible to do better than a squishy meat-brain?  Without a doubt!  Is anyone even trying that?  Hell no.  Current tech is only statistical pattern matching.  Which one of those thousands of nets is wrong?  How will you know?

For sure, combining algorithmic assurance with machine learning heuristics is the next frontier.  But how long will that take?  A decade, or what?

Tim

[tszaboo]

Which one of those thousands of nets is wrong?  How will you know?

By wrong, I assume you dont mean wrong connection. I will assume you mean excess EMC or reflections, or other signal integrity problems.
The thing is, who knows? I dont simulate the connections that I route, because software costs $$$. I know it is possible to avoid a lot of problems, if you simulate it, so the  computer already has the advantage. And I saw so many bad designs throughout my career, I wouldn't be surprised, if a computer can make better results than that.
I mean I saw boards, where an RNG generator can probably route it better.

And the computer can do route->simulate->route->...->route->simulate-> "ah, that is how it works" ->  route -> acceptable result. After it learns it.

[cdev]

They will figure out a way to have the users do the refinement of the software, unknowingly. I'm sure this is already happening somewhere. Probably in multiple contexts. When we consider the amount of boring tasks which machines are exceedingly well at doing, or are passable at doing but getting better quickly, our entire way of looking at society and work is not long for this world. Virtually every profession, skilled or unskilled, will be impacted by automation within a few years, some more than others. I think this will be one of the ones that is heavily impacted. Not overnight, but within a decade from today.

I say this having seen a similar process impact architecture and construction via a friend who works in that field. Twenty years ago, they were already almost there with buildings. (Of course, lots more money (per project) was involved there.)

IMHO, this problem, technically, is similar enough to that one to seem likely to be driven by a similar dynamic. Very high end tools that do these kinds of tasks well probably already exist, but are not widely used because of cost. possibly cost of the hardware as well as software. But just as with everything else, due to rapid improvements in technologies like cluster computing, the price of solving very complex problems will come down.

Eventually, AIs, given enough contextual information, should be able to grok more and more of those rules and implement them.

Thing is, will the incremental cost of AI development actually go to zero, or asymptotically settle out?

After all, there is ultimately some cost of energy to computing such a solution, even for ideal quantum computers.  (We are probably centuries away from reaching the information-theoretic limit.)

The question is, can that cost be low enough that it is worth doing, relative to the cost of the conventional approach?

It may very well be that analog circuit design is not automated, in my lifetime, simply because it's not a big enough problem to justify its expense.  And its customers are disparate; there is currently no market where customers can pool their purchasing power that designers must compete for.

There's also the matter of trickle-down economics.  Some technologies trickle.  Three decades ago, no one could imagine affording a computer and CAD system, out of pocket.  Now we have free tools (like KiCAD) that run on absurdly affordable platforms (like a $100 used PC).  That worked out nicely.  On the other hand, it's been over half a century, and IC manufacture has still not made such a transition -- the industry secret sausage is still carefully maintained behind NDAs, and manufactured behind $10k+ NREs.  (Yes, there are some cheaper fabs, but nothing hobbyist level, not by a long shot.  In contrast, small PCBs are cheaper than a meal.)

Which history will apply to these technologies, when and if they are developed?  Only time will tell.

Tim

[hammy]

I was hoping for a simple answer lyke "that's a bad ideea, don't do that" or "it's a good ideea, here is a good place to start, see this lynk / google this out".

@shoppy:

There are several free pdf design guides out there, you can use for studies. A search engine and these keywords should help to find these files:
"Optimum Design Handbook"
"Right The First Time"
"PCB Design Tutorial"
"Toradex - Layout Design Guide"
"PCB and DFM" from Seed Studions

The first level is to learn everything about "normal" pcb design. "Normal" pcb design means: breakout boards, small circuits with just a voltage regulator and some components, maybe a small microcontroller (PIC, Atmel). PCBs with no more complexity like an arduino shield.
After mastering this level, after some time, you can start with these keywords: High Speed Design, EMC , EMI and HDI.

Working from home, designing PCBs for companies, requires usually a deeper knowledge. A part of this required knowledge is nowhere written down. You can only learn this stuff from another experienced pcb designer. Usually inside a company, sometimes you can learn from a retired designer who is happy to share his knowledge. And through experience.
You also need some knowledge around the design itself: DFM, Layer StackUp, Gerber, BOM, IPC standards ... a lot of stuff!

If you work full time on this, you should be able to design quite good PCBs after two or three months. To master complex boards you may need between a half or a full year. During this you have to get used to your selected EDA software.
The EDA software you choose should be the same used in the industrie you want to work for. Normally this software is quite expensive.

Maybe you are talented and you can get the knack of it faster, it is still a lot of work and a lot to learn. If you are a self-taught person, it is possible. The learning speed also depends on how much money you can invest in front, it is maybe wise to take one step after another. You could design some PCBs for other guys in your local makerspace and get some experience. This way you also can build up some reputation, very important for your later work.

There are online marketplaces you can look for pcb design projects: guru.com and upwork.com

Commercial training is also a possibility to learn special knowledge. And there are also certification courses you can attend. A lot of money is needed for these trainings and courses.

EDIT: I forgot to mention the "Fedevel Academy". Robert offers online courses for PCB Design. These are excellent courses! I highly recommendend them.

HTH! Have fun!
Cheers
hammy

[rx8pilot]

On my most recent project, I spent a single day designing the circuit and creating the schematic. Then I spent 5 days laying out the PCB - which is vastly more sophisticated than connecting the parts. It is a mix of high power, low voltage analog, and high-speed digital. The level of detail required for success is mind-numbing.

In our modern world, most PCB layouts are way harder than connect the dots. It is a critical part of the engineering. As others have said, experience is key and you cannot rush experience. You cannot teach experience.

[james_s]

I've done boards that I laid out in less than an hour, I've spent tens of hours on others, it really depends. Sometimes the complexity of the schematic has little relation to how time consuming the board layout portion is.

[janekm]

BTW several of the Chinese PCB factories are starting to offer PCB layout service (my suspicion is it's a way to use the CAM operator's time between jobs). I haven't used one affiliated with a PCB factory but I did use an independent Chinese layout bureau once when I didn't have enough time to do finish it myself, they actually did a really good and clean job for a ridiculously low fee (perhaps too low... when I asked them to do more work later they said they were too busy  ).

Anyway tl/dr, between autorouters and "layout factories" I wouldn't bet my income on just layout.

[Alex Eisenhut]

between autorouters and "layout factories" I wouldn't bet my income on just layout.

I've made money fixing the "cheaper" alternative for other people.

The old adage of "there's never money to do it right, but there's always time to do it twice" applies here.

[rx8pilot]

The old adage of "there's never money to do it right, but there's always time to do it twice" applies here.

Yes. That.

Modern PCB layout needs an engineer familiar with the circuit to oversee regularly and the layout person needs to have a reasonable understanding of the engineering principles at play. Poorly considered layouts can break even a robust schematic design.

I wonder how these services are executed practically. If I send over a schematic and they layout a dumb PCB - do I still pay? How much will they tolerate my demands to fix stuff and improve things before the price quadruples?

[free_electron]

I can see it being automated within a decade or so.

i'm getting a bowl of popcorn. this will get interesting. especially since exactly THAT has been said the last 40 years ... I'm wondering when that 10 years span exactly starts ....

Autorouters are getting more and more intelligent. You can fan-out a BGA with a click in newer versions of Altium.

That is NOT autorouting. That is just fan-out. Fan-out is an easy problem to tackle. Even then ... in lots of cases it doesn't work. Try placing and routing all the decoupling caps on let's say a 688 pin BGA processor ( not that big of a beast these days ... ) and then click that fan-out button and see where you end up. Plenty of pins unreachable...
Back to the 'hooman' to make pathways for the 'clickety thing' to complet the fan out... no 'hooman', no solution...

Auto routers can route a board.

Autorouters are nothing more than a bunch of algorithms. It take a 'hooman' in the drivers seat , setting up all the criteria , selecting the algorithms , in which order to launch them , and to clean up the initial placement and preroutes ( like power , decoupling , series and parallel termination , local power switchers and regulators etc ... ) before you can launch the router.

Length tuning and other optimization is not that difficult to do.

  obviously you have never done a 'serious' board like a graphics card , computer motherboard, or even something smaller like let's say a DDR4 memory stick ... Cramming all that stuff in the available space is a nightmare. you assume it is easy because you are thinking in terms of 'unlimited' space. Wait till they tell you : you have the size of a 2 postage stamps. cram it in ...

Component placement can be automated.

Really sherlock ? You find me a placement tool that can optimally place all the parts around a switching regulator. Bulk and output caps , coil , and power pathways. Let's just start there. PLace the parts and properly route them with the proper widths. Collapse all current loops and optimize for density. Wake me when you found such a tool. Then we'll tackle something slightly more complex.

Digital simulation is a solved problem

Quote the Late Widlar. Digital ? any idiot can count to one ...
It is so easy to say : it's digital. one and zero. except on the board these things are voltages , that switch , that become waves and fields that interact with the routing and all the 'analog cruft' on the board. Good luck finding that 'one' in all the noise around the trace that needs to carry those electrons. Ever tried to properly run USB3 , Sata , or other 'common' buses these days ? you cant do that on a single layer with 8 mil traces ... and no the computers can;t figure it out by them selves either. It takes a 'hooman' to figure out what layer stack will be used , calculated the required widths and clearances per layer and feed all that information in to the router. and then the computer can place a trace of proper width and length between point a and b. There is not a single PCB layout tool where you can simply feed it the layer thicknesses and the required impedances (through the schematic) and it will do the calculations for you. Needs 'hooman'. And if the layer stack needs adapting for whatever reason ( manufacturing,  materials shortage, other busses with other requirements on same board ) guess who gets to do all the 'fun work again' : the trained descendant-of-ape ...
A a matter of fact : coming up with a layer stack itself is a very complex issue. you need thick copper for all the massive currents modern cpu's eat , and at the same time you want thin copper to be able to route all the high speed stuff...

high end solutions are capable of providing useful feedback.

Ah , the good old 'let's run it. look at it  ,tweak it some more and re-run it.. approach. Typical manager (aka poo-flinging baboon type) speak.
How many run cycles will be needed before it works ? 10 ? 20 ? 100 ? 1000 ? Sure the computer is fast. But then the 'hooman' needs to check the end rsult of the run , tweak the setup. That is where most time is lost. If you approach it without know-how and rely on this mechanism you end up with total time ( computer + hooman ) orders of magnitude longer than an experienced 'hooman'.

So it is not unconceivable to set up a machine learning algorithm to learn placement and routing.

Quick , someone throw oodles of money at this and get very rich .. or the laughing stock of the industry ...

You would still need the engineer though. To set up the board size, connectors, schematic, footprints, general guides to steer the AI.

First sensible thing you said so far... And this is where most of the time is already spent in the first place.

One of the complex boards i did was going to attempt the 'specctra' autorouter. Had a hired specialist in the field to do the runs. Everything was set up: worked with the GPU maker to do the initial power distribution and other stuff. Everything was ready to go. Only the memories and gigabit lanes between the GPU's were going to be tackled by Specctra. All rules were loaded. Everything else was placed and routed.

First run : aborted after 30 minutes. Not enough routing channels. Specctra could not complete because it could not find enough space to throw down all the tracks.
So the hoomans (two) spend 2 days following the advice of the expert and shuffle via's on a grid so that there would be plenty of horizontal and vertical pathways.
Second run : still no enough pathways. Congestion in area's around memories. Can we use different pinout ?
Hooman : are you effing insane ? I can't change the memory pinout. The chip is like that . Standard Samsung part. GPU vendor uses same on their boards. They already tuned their GPU pinout for easy layout to the memories.
So we studied the congestion map , moved decoupling , moved via's created more lanes in the congested area's.
third run. now congestion in other area's.

15 or 16 runs with every time 2 to 3 days to move stuff around ( sometimes by 1 or 2 mils , converting a via here and there to a laser via to shave another mil or 2 so there was room for one more trace )
Spent 2 months playing this tag game. 2 layout engineers and another expert driving Specctra. It finally completed. The layout looked like a plate of spaghetti, projectile vomited into a pcb. Sure the traces all met impedance , length and diff length criteria, but half busses plowed straight underneath the switching inductors, and so many via's were shot that the ground plane was swiss cheese. Yes there was coupled copper underneath the traces, but everywhere else it was snot. A power analysis showed the copper was so shredded it couldn't even handle the return currents for the power of the GPU... we had to add 2 more planes just for that.
So 2 more power planes and 2 more routing layers later , bringing the board to a 16 layer now, we Specctra again. Marginal improvement but it completed this time.

As a side project we had 2 other engineers do the routing by hand (only the specctra portion). They finished in 4 days , length tuning and all. And got away with a 12 layer board.

We talked to the GPU maker about this. They looked at us and laughed.. Could have told you that....
Their layout engineers did graphics boards on 8 layers in 2 weeks. By hand. They had experience doing that stuff for years.
We did it on 12 layers in 4 weeks ( we had 4 gpu's interconnected with a bunch of other stuff , double complexity ). With less experience in such demanding interconnects.

The autorouter took 3 people during 2 months pushing and shoving a 80% routed board to make enough room for the router to 'vomit' its traces everywhere, without even understanding that it can't route a PCI-x underneath the magnetics of a switcher !

SPECCTRA : Stupid Program, Expert in Creating Crappy Traces and Routing Aborts.

We didn't even bother to produce that board ... the hand routed one worked first time right.

[tszaboo]

I can see it being automated within a decade or so.

i'm getting a bowl of popcorn. this will get interesting. especially since exactly THAT has been said the last 40 years ... I'm wondering when that 10 years span exactly starts ....

We dont need to agree on this. A lot can happen in 10 years. First capacitive touchscreen smartphones came out 10 years ago, now there is 1 billion sold every year.

Length tuning and other optimization is not that difficult to do.

  obviously you have never done a 'serious' board like a graphics card , computer motherboard, or even something smaller like let's say a DDR4 memory stick ... Cramming all that stuff in the available space is a nightmare. you assume it is easy because you are thinking in terms of 'unlimited' space. Wait till they tell you : you have the size of a 2 postage stamps. cram it in ...

No shit, I also haven't designed millimeter wave radars, or nuclear submarine control panels. I guess most of us don't do that. The amount of length tuning is enough for the other 95% of jobs. Besides, with the HBM, all those 256bit+ memory interfaces could go away in the future. Or just look around a bit, look at for example the PS4 PCB. They sold 75 million of it, and the PCB is basically infinite size compared to the components on it.

And there are other type of engineering problems, which are hard.

Really sherlock ? You find me a placement tool that can optimally place all the parts around a switching regulator. Bulk and output caps , coil , and power pathways. Let's just start there. PLace the parts and properly route them with the proper widths. Collapse all current loops and optimize for density. Wake me when you found such a tool. Then we'll tackle something slightly more complex.

A typical buck converter with 12-60V input and few amps output can be placed and routed probably in an hour, provided you have the space.
But really, why would it be difficult to specify the current loops and define to make this as small as possible?

high end solutions are capable of providing useful feedback.

Ah , the good old 'let's run it. look at it  ,tweak it some more and re-run it.. approach. Typical manager (aka poo-flinging baboon type) speak.
How many run cycles will be needed before it works ? 10 ? 20 ? 100 ? 1000 ? Sure the computer is fast. But then the 'hooman' needs to check the end rsult of the run , tweak the setup. That is where most time is lost. If you approach it without know-how and rely on this mechanism you end up with total time ( computer + hooman ) orders of magnitude longer than an experienced 'hooman'.

So it is not unconceivable to set up a machine learning algorithm to learn placement and routing.

Quick , someone throw oodles of money at this and get very rich .. or the laughing stock of the industry ...

That is how machine learning works. Time is irrelevant, also the number of iterations.  If it takes a machine a year, and a small supercomputer to learn the basics of routing, would that matter? 

Digital simulation is a solved problem

Quote the Late Widlar. Digital ? any idiot can count to one ...

We both know, that modern simulators will produce feedback on reflections, voltage drops, power plane analysis, GND bouncing and almost every relevant aspect of a digital design. Its not just the 1-0-1 that is simulated.

All you are saying, is that programmers are bad at their job, because they cannot set proper rules and proper priorities for routing.
I'm quite sure you know the difference between a deep learning algorithm and a "bunch of script thrown together by hoomans".

[hammy]

That is how machine learning works. Time is irrelevant, also the number of iterations.  If it takes a machine a year, and a small supercomputer to learn the basics of routing, would that matter? 

I'm quite sure this is the reason for all these "store your circuit and design in our cloud"-Software. They can train their AI against these files, until it learned all basics and is able to improve itself. 

[AndyC_772]

We dont need to agree on this. A lot can happen in 10 years. First capacitive touchscreen smartphones came out 10 years ago, now there is 1 billion sold every year.

This is a ridiculously easy debate to settle. Just present even one example of a non-trivial PCB that's been placed and routed well by an AI.

It need not be something as complex as a PC motherboard to prove the point; something along the lines of a piece of consumer A/V equipment would do. Anything with a few different functional blocks: power, some analogue circuits, a microcontroller, that kind of thing. The sort of PCB that even a half way capable layout engineer would do without having to think too hard.

I'd be impressed by any algorithm that works out for itself that there are right ways and wrong ways to lay out a simple SMPS without being explicitly told. Or one that can work out that a design has an isolation barrier which needs a sane physical layout to keep all the hazardous voltages in one area of the board. Double points if it can work out that of course you keep the microphone inputs away from the mains supply, or that the SPI interface needs to be simulated with IBIS models but the keypad itself probably doesn't.

[tszaboo]

http://www.icd.com.au/articles/AI_PCBD_May2016.pdf
Google X
https://pdfs.semanticscholar.org/ad48/d28bb04c196036f743c89f5e519ff7b74924.pdf
Fujisu
https://semiengineering.com/machine-learning-meets-ic-design/
Synopsys

Sure not one shows a product designed by AI.

I'd be impressed by any algorithm that works out for itself that there are right ways and wrong ways to lay out a simple SMPS without being explicitly told. Or one that can work out that a design has an isolation barrier which needs a sane physical layout to keep all the hazardous voltages in one area of the board. Double points if it can work out that of course you keep the microphone inputs away from the mains supply, or that the SPI interface needs to be simulated with IBIS models but the keypad itself probably doesn't.

You are explicitly told how to lay out an SMPS. They told you that the current loops should be small, and the isolation barieres between these nodes should be this much, according to IPC recommendation. So why should an AI work with less info than an engineer?
It would be the case of asking the AI to draw a flower, without telling it what a flower is.

[rx8pilot]

I just spent a few days laying out an SMPS. There were so many judgment calls that were settled by weighing a variety of issues that are not part of a spec. The end result needs to accommodate a variety of mechanical restrictions, thermal issues need to be matched with an existing design, test points needs to be placed in a way that would make it easy to test.

Density is critical, performance is critical, etc....my guess is that I would like to have some automation tools to help in certain areas but the rules of this design are just too much to even explain to a co-worker - much less an algorithm.

[cdev]

How much more would you pay for a power supply designed with the utmost care by a human being, with a signed calibration certificate, rather than optimized by a program?

I just spent a few days laying out an SMPS. There were so many judgment calls that were settled by weighing a variety of issues that are not part of a spec. The end result needs to accommodate a variety of mechanical restrictions, thermal issues need to be matched with an existing design, test points needs to be placed in a way that would make it easy to test.

Density is critical, performance is critical, etc....my guess is that I would like to have some automation tools to help in certain areas but the rules of this design are just too much to even explain to a co-worker - much less an algorithm.

[rx8pilot]

In the end, I only care about the design meeting the needs. When the needs are very difficult to define, it just makes it less likely to be algorithmically practical.

I would pay the exact same amount of money regardless of how it happened.

Short and misplld from my mobile......

[T3sl4co1l]

Comparing primitive AI to a competent designer seems to be a strawman.

99% of circuits don't need that care, or can be refined through review, or iteration.

The average board design seems to be an MCU, some buttons and LEDs, sensors, maybe some interface hardware, maybe some power management, and that's about it.  Very simple, nothing that needs human attention.  Many that I have seen, look that way, though they probably were done by humans.

Comparing AI to the average human is a much more economically relevant comparison, the one that matters.

Tim

[TiN]

i'm getting a bowl of popcorn.

I got a big one.

By the time you specify and rules and settings for autorouter thingy, you'd be done with job already by just using "hooman" experience. Better computer tools make job easier, but it's not meaning it will do it FOR you There is reason why CAD means computer-AIDED-design

Say automatic algorithms getting better? Sure. As do speeds on common hardware gets faster. 16GHz lanes on PCIe Gen4 and 10GHz on existing TBT/USB lanes have much more in analog domain, than many expect from "digital design" crowd.

Take this as just 5c from person, designed some latest-gen PCIe graphics cards/motherboards .

http://www.icd.com.au/articles/AI_PCBD_May2016.pdf
Google X
https://pdfs.semanticscholar.org/ad48/d28bb04c196036f743c89f5e519ff7b74924.pdf
Fujisu
https://semiengineering.com/machine-learning-meets-ic-design/
Synopsys

Sure not one shows a product designed by AI.

There is none shown here 

[shoppy]

@ hammy, Thank you!