My Dream Job

[Shenandoah]

Hello all,

While I'm struggling and learning the hard way I'm very interested in getting a first time job as as an IC/chip designer. the problem is that my degree is not EE, it's in software. Is there any recommended references on this area so that I can pass an interview or a test?

Thanks.

[Bored@Work]

Just fake it, like you probably faked your way through university.

[dr.diesel]

IC/chip designer

There are many aspects to chip design, software being part of the process.  Any particular part that interests you?

[lgbeno]

Digital or Analog IC design?  I could see where a software background could dovetail with digital chip design.  I also like the suggestion on the design automation side as well.

Honestly though, analog IC is quite difficult let alone without any experience that you would receive by taking classes.  If that is what you are interested in, I'd suggest taking a class or two through continuing Ed somewhere.

[Shenandoah]

Just fake it, like you probably faked your way through university.

How did I fake my way through university? 

IC/chip designer

There are many aspects to chip design, software being part of the process.  Any particular part that interests you?

I'm more into microwave mixed digital/analog chips. But before I start learning the automation software I would love to learn the theory on paper first.
I was thinking of applying to Internships but they are only available to uni students, not self-taught persons, unfortunately.

Analog IC is tricky but I will make it...just need the opportunity.

Not really interested in going back to school and waste time and money, I can do the reading at home and teach myself properly instead of sleeping in classes...profs are so damn boring, talking to themselves reading 20 years old lec notes. 

[dr.diesel]

There are many aspects to the software side as well:

 - Production automation
 - Visual inspection/quality control AOI etc
 - Shop floor, robots, QC databases, SAP etc
 - Custom engineering software

What "kind" of software guy are you now?  Making stuff with mixed digital/analog chips can be a world of different from manufacturing them.  In any case, software has a HUGE role in making chips, or pretty much anything these days.

[gregariz]

http://www.mosis.com/

[free_electron]

Analog IC is tricky but I will make it...just need the opportunity

Not a chance... Analog IC design has very little to do with normal electronics. It's more material physics on one end, and black magic on the other end.

Start by reading 'the mos transistor' by h.m. Veendrick...

[EEVblog]

While I'm struggling and learning the hard way I'm very interested in getting a first time job as as an IC/chip designer. the problem is that my degree is not EE, it's in software. Is there any recommended references on this area so that I can pass an interview or a test?

If you don't have formal education in the field (some degrees have an IC design component, or it's optional) or practical experience in actually designing chips, then I think your odds of getting a job in the industry are practically zero.
IC design is not like regular component level electronics designs. Whilst you could self study it, actually applying it practically is going to cost a lot of money. It's not like you can just mock it up on a breadboard cheaply.

[Shenandoah]

Understood, but how the pioneers made it? To gain experience or at least start learning it practically I need the chance like any other industry.
What kind of talents/skills starters show to the employer assuming it's their first IC design job?

[dr.diesel]

Pioneers invent, ie you don't go to school for a degree in entrepreneurism, you just do it.

IC design, let's say on the fab side, have basic elements associated, these can be taught in school.  Employers look for these and hire those people!

[nctnico]

First you need to take some courses in chip design. My EE study was geared towards chip design because Phillips (now NXP) sponsored the equipment hoping some of the students didn't find IC design utterly boring (what is an EE without an oscilloscope?) which most of us did.

Anyway, back in the old days chips where build on breadboard using discrete components. Nowadays its just simulation, simulation and simulation and even more simulation. To make it worse: simulating a piece of electronics is an art in itself when it comes to temperature and process variations. And don't think digital isn't analog...

From your perspective I'd try to get a job in a company which designs chips as a tester for writing test benches (most likely Verilog). Then move on to logic verification (breadboarding using FPGA) and from there you may get to design actual chips.

[EEVblog]

Understood, but how the pioneers made it?
To gain experience or at least start learning it practically I need the chance like any other industry.
What kind of talents/skills starters show to the employer assuming it's their first IC design job?

All that matters if how you can get into the industry now.
And although I'm not into the IC design area, I can imagine that it is not unlike component level electronics design.
Yes, I can imagine it would be possible to get a job being self taught, the problem with IC design, like most of electronics, it's ultimately a practical applied field.
There usually two routes in:
1) Formal education and then a grunt graduate job.
2) Based on your practical and/or job experience.

You'll have none of these in the IC field.
Yes, you can learn the theory and use some tools until the cows come home, but unless you fork out big money to get your designs fabricated and have experience going the whole cycle of getting a chip made tested etc, your odds of getting a job are very low I suspect.
You might get lucky and get a job using one of the IC design tools as part of big team, maybe.
If you do have the money to fork out on actually getting chips made, then your odds will be better, but still probably rather low.

Now, someone will no doubt bring up Jeri Ellsworth. As I recall, she got her first IC design job not because she self studied IC design and went looking for a job, it's because she had the experience required in the FPGA side of reverse engineering a particular product that a company happened to want. She had the ready to go IP. Basically, she got lucky that someone found her. She might not have had nearly the same luck if she self studied IC design and then actively went looking for a job.

[EEVblog]

From your perspective I'd try to get a job in a company which designs chips as a tester for writing test benches (most likely Verilog). Then move on to logic verification (breadboarding using FPGA) and from there you may get to design actual chips.

Yes, that's a third way in, via the side door.
The classic "working your way up" within a company.
I hired a guy one time that literally started as Dick Smith's janitor.

[gregariz]

The mosis service I linked above is what alot of universities use as a small quantity prototyping service to help train their students in analog/cmos design. If memory serves it used to cost 2-5K for a small quantity of chips. Just to give you some cost perspective from recent experience;

Digital ASIC: 100K-250K <--- you do the Verilog/VHDL design and simulation and make sure it works on an fpga.. then you hand it off to custom vendor ie someone like gigaoptix ( http://www.gigoptix.com/products/20-asic )

Analog ASIC: 500K-1MIL <--- you write a detailed specification for all of the performance parameters that the chip needs to conform to. You then hand it off to a custom chip vendor. The custom chip vendor usually has a library of pre-characterized circuit blocks that are tied to a particular fabrication process ie something like SiGe. They try to put these pre-characterized blocks through their software simulators (cost about 50-150K) and put everything together and fill in any blanks. Its very process dependent which is why you dont usually do it yourself. Check out the design packages at mosis for the various fab processes. They give you characterized (for the fab process) individual devices like transistors that you can use in Spice.

Mixed Signal ASIC: 800K and up <--- most companies avoid this due to cost. Plus it ends up being alot of voodoo. Vendors IP extends to knowing what processes can handle both logic and analog reasonably ie how do you put an ARM core and a RF amplifier together - alot of processes wont like it - the arm will be great but the noise in the RF amp will be crap - stuff like that. Lots of people get burnt here and so there a fewer custom vendors out there to do these jobs.

Its highly possible that you could use Spice and the fab packages available from Mosis to design basic analog chips. Its highly improbably that you would be able to get microwave or mixed signal devices happening without spending a huge about of money optimizing design by doing alot of prototypes. Some of the circuit and system software simulators that all the major analog  vendors use to help design these chips easily cost over 100K. You will need excellent analog discrete design skill to do that so if you are interested in pursuing it you should probably concentrate on getting that skillset ie rockstar transistor designer. Now one word of warning... I'm sure that there are a few small contract chip design houses out there... but if you look at the major chip vendors ie people like Maxim, Freescale etc... go to there career section and see what they want. Typically they'll want a PhD. You can always try to get your foot in the door with the independent contract firms... but I only vaguely know who they are so..

Just to give some perspective... when I was a student I only did one very simple transistor circuit on a cmos process. If memory serves the only one that worked .. or at least worked well was the one where the guy flooded the wafer surface with aluminium to keep all the track inductances low. Even though everyone had working Spice models.. it didnt matter. That sort of stuff is ... ie circuit blocks that work well for a process is what IP is all about and why its almost impossible to do this stuff as a hobbyist. Prototyping cost will kill you.

[Stonent]

Yeah because when you go to chips themselves, you're also dealing with the physics of the doping compounds and other materials. You'd have to know a lot of the characteristics of things like Gallium Arsenide and the meaning of things like "dopant-site bonding energy"

Or in more relative terms, I work in IT and have done so for about 17 years now, and for a period of time took an evening job at a store doing the fix customer computers thing for some extra money.  This while having my regular IT job. And most of the people I worked with were about 18 to early 20's and they'd say "Hey man, can you get me a job where you work full time doing IT stuff?"  And honestly for most of them I had to say no, because they've only worked on the home level on computers and didn't have experience with things like Active Directory, setting up network printers, tracing out network issues, working with AS/400 or Mainframe connectivity etc.

With electronics, I'm doing this as partly a hobby and partly a challenge. I'm playing with microcontrollers now, have a cheap logic analyzer on the way, playing around with KiCAD and maybe might even get into FPGAs at some point. But also I need to finish my college degree that I've stopped and started numerous times, because I've just about topped what I can do with my career without my degree being finished.

It's not that they were dumb in any way, just that they couldn't just say "well I don't know what's wrong, we're just going to have to format and start over" when the assembly line has stopped and the company is losing 1000s of dollars a minute.

[lgbeno]

I would say that Jeri has more of a digital IC background than anything.  I think a software developer with the right mindset could morph into a digital IC designer without much of a stretch.  Especially if he/she is doing some FPGA work or prototyping first.  In fact a software person who understands digital design is an indispensable resource IMO.  Analog is quite a bit different simply because understanding device physics is so important.

All of that said I wouldn't want Shenandoah to be too discouraged.  Just friendly advice that it will take a lot of perseverance.

Mixed Signal ASIC: 800K and up <--- most companies avoid this due to cost. Plus it ends up being alot of voodoo. Vendors IP extends to knowing what processes can handle both logic and analog reasonably ie how do you put an ARM core and a RF amplifier together - alot of processes wont like it - the arm will be great but the noise in the RF amp will be crap - stuff like that. Lots of people get burnt here and so there a fewer custom vendors out there to do these jobs.

gregariz, great breakdown and for most vendors in the industry, this is very true. 

So full disclosure, I'm a FAE for Triad Semiconductor and Mixed Signal ASIC design is our business.  We do things differently and as a result, we're able to do a mixed signal ASIC design starting around 100K NRE.  800K would be on the far upper end of our spectrum for a custom IC design.  We have IP and proven silicon for ARM cores, precision ADC's and all kinds of analog including power management.  In the future we think that using our via only approach and our ViaDesigner software that we could get this NRE as low as 10K or 0 if the volume makes sense.  But anyway, if interested, send me a PM and we can chat more about it.

So now that cat is out of the bag, my best recommendation for the question at hand is that it sounds like you are really dedicated to making this happen.  Working your way up is the best thing to do, especially since the world is in need of good IC designers.  My advice would be to use your current talents as a segway to do IC design.  Maybe that means finding a company that does both (which there are many) get in with your current credentials and start talking to your manager about where you would like to see your career in the next few years.  If your manager is good he/she will listen and do whatever to help you fulfill your goals.

[chickenHeadKnob]

Besides Jeri Ellsworth, the one non-typical background IC designer I can think of is Charles H Moore, the inventor of the FORTH language.
In the 90's he went on a bit of mission to create a special visual Forth, OKAD that was intended as an ultra minimalist chip layout tool. I thought he was going a bit kooky myself. If I recall correctly his IC design house did have some success in selling stack/forth machine cores.
 

[Shenandoah]

Yeah I heard that Jeri Ellsworth is a chip designer, and she is self-taught. This is very encouraging.

I thought that IC design could be simpler. First prototyping the circuit using discrete components, and then layout design and fabrication is done using software, which automates the process. I'm still interested to read some good literature on this subject. I found these books:

http://www.amazon.com/Analysis-Design-Analog-Integrated-Circuits/dp/0470245999/ref=sr_1_2?s=books&ie=UTF8&qid=1343870038&sr=1-2

http://www.amazon.com/CMOS-VLSI-Design-Circuits-Perspective/dp/0321547748/ref=sr_1_1?s=books&ie=UTF8&qid=1372404554&sr=1-1&keywords=cmos+vlsi+design

http://www.amazon.com/Fundamental-Principles-Optical-Lithography-Microfabrication/dp/0470727306/ref=pd_sim_b_1

http://www.amazon.com/Principles-Electronic-Materials-Devices-Kasap/dp/0073104647/ref=sr_1_1?s=books&ie=UTF8&qid=1372404728&sr=1-1&keywords=electronic+materials+and+devices

Makes sense and it's like in mechanics, Newtonian physics does not work for atomic objects, that's where quantum mechanics is applied. The same can be said for IC where I think KVL/KCL/Ohm's laws are useless, we need to apply Maxwel's and semiconductor physics instead. 

[EEVblog]

Yeah I heard that Jeri Ellsworth is a chip designer, and she is self-taught. This is very encouraging.

Just remember that she didn't go apply for an IC design job.

I thought that IC design could be simpler. First prototyping the circuit using discrete components

No, it doesn't really work like that. As has been pointed out, IC design has little to do with practical component level electronics design.
It's all to do with manufacturing processes and physics.
Higher level stuff like ViaDesigner as has been pointed out, takes a lot of the mystery out doing custom chip design.

[Shenandoah]

Why is a discrete circuit any different than IC circuit? Is it noise and other physical requirements that change the circuit implementation? Do we have different schematic for the same functionality when it's implemented in discrete components and when it's IC?

[Codemonkey]

Do we have different schematic for the same functionality when it's implemented in discrete components and when it's IC?

Yes. You can't build a mixed signal chip on veroboard then expect the same thing to work on a bit of silicon.

Seriously, as others have said, you're not going to get anyone letting you play with their $$$$$$ tools designing a chip for them unless you've either got previous proven experience, and a suitable qualification in the discipline. No company in their right mind is going to risk binning $100K+ worth of mask set cos some random bloke fancied himself as an IC designer but had no formal training.

Its like me saying I've got no knowledge of brain surgery but hey, I'm pretty handy at cutting PCB tracks with a scalpel, how hard can it be ? Its just cutting stuff right ?

You're best bet is learning something like Verilog and becoming a digital designer, but even then, implementing something in a CPLD or FPGA is NOT the same as having an IC produced from the same code.

I happen to work for a company that has produced complex mixed signal devices, some of the software guys I work with have worked on some of the digital design in the past, but the mixed signal team are very specialised and spend almost all of the time sat in front of a pc running simulations, and very occasionally can be found in the lab when a new design of device returns from the fab. Its not even remotely the same as your average electronics engineer sat in a lab breadboarding stuff.

[Shenandoah]

Yes. You can't build a mixed signal chip on veroboard then expect the same thing to work on a bit of silicon.

Please I'm very curious to know why, can you give me a simple example why the schematic is different?

Seriously, as others have said, you're not going to get anyone letting you play with their $$$$$$ tools designing a chip for them unless you've either got previous proven experience, and a suitable qualification in the discipline. No company in their right mind is going to risk binning $100K+ worth of mask set cos some random bloke fancied himself as an IC designer but had no formal training.

You just shattered my dreams.

Its not even remotely the same as your average electronics engineer sat in a lab breadboarding stuff.

Got it! It's a black magic.

I remember seeing a picture of very famous IC designer Bob Pease bread boarding and heard that he dislikes SPICE?

[lgbeno]

Yes. You can't build a mixed signal chip on veroboard then expect the same thing to work on a bit of silicon.

Please I'm very curious to know why, can you give me a simple example why the schematic is different?

I admire the strong drive to do this, if there is a will then there is a way...

For traditional IC design, attached is a report for a project that I did for a 400 level masters course.  We had to design an opamp which conformed to the specs on the last page, the first two pages is the schematic for just this opamp!  I barely got a B- on this design and this is no where near a state of the art or something that would be worth tapeing out. 

Lets say that you are designing in CMOS, really your only components to use are NMOS and PMOS fets.  The designer is responsible for adjusting the length, width and multiplier of each transistor to get the sort of performance needed for the circuit at hand.  Start talking about a high level integrated chip and it goes up from there.

Schematics sure give you an appreciation for that humble Triangular symbol that we use for an opamp everyday

This is one area where I think formal education is required, not only to listen/sleep through "boring lectures" but also to do these types of sample projects and learn the hard way.  I poured hundreds of hours into that simulation to get a B-...  It can be done though!

[Shenandoah]

lgbeno, thanks a lot for clarification. So it's the fabrication technology and materials that determine what components have to be used. So choices are limited when it's on one chip. I'm getting the idea...but need to research more.
What book did you use at this course?