Author Topic: My Dream Job  (Read 56708 times)

0 Members and 1 Guest are viewing this topic.

Offline rbola35618

  • Frequent Contributor
  • **
  • Posts: 298
  • Country: us
Re: My Dream Job
« Reply #25 on: June 28, 2013, 04:51:51 pm »
If you want to learn about Integrated circuit design then you need to get the Jake Baker book titled "CMOS, circuit design, layout, and simulation".  It is the best book on analog IC design. It is very readable and lots of SPICE simulation.

On top of that, Dr. Baker has a website where he post his videos of the IC design classes that he teaches at the University of Nevada.

http://cmosedu.com/


http://cmosedu.com/jbaker/jbaker.htm

http://cmosedu.com/jbaker/courses/courses.htm


This is where is tutorial videos are. He is a great teacher. He was one of my professors at the University of Idaho.

http://cmosedu.com/videos/videos.htm

Hope this helps



« Last Edit: June 28, 2013, 05:50:26 pm by rbola35618 »
 

Offline lgbeno

  • Frequent Contributor
  • **
  • Posts: 349
  • Country: 00
My Dream Job
« Reply #26 on: June 28, 2013, 05:06:38 pm »
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?

Fab materials have a lot to so with it but its really a matter of hierarchical design all discrete components are constructed of base transistors, just abstracted to higher level elements like Opamps or an ADC by the time they are board level.

Yes hierarchical design and IP are used in the IC design tools as well but a big IP library for a specific process is needed.  This is really the bases for ViaDesigner.

We actually didn't use the book much, for this class mostly the professors notes. I took my classes at USC online. I'll check which one it was when I get home.
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 26757
  • Country: nl
    • NCT Developments
Re: My Dream Job
« Reply #27 on: June 28, 2013, 05:15:56 pm »
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.
AFAIK the chip manufacturer gives you a library to work with. Usually with standard transistors and even logic gates. From there its like building with Lego bricks. During my EE course I have designed two digital chips (never produced though) and those where build from standard library components. Some parts like and/or arrays to implement logic functions like in a CPLD where generated from truth tables and imported as a functional block. Those functional blocks (including the I/O macro cells) had to be placed onto the chip much like placing components on a PCB and connected using much like components on a PCB. All in all it was actually pretty simple once the simulation was finished. But this was 2um NMOS  ;D

I think there are some open source IC design packages like SPACE from the University of Delft ( http://www.space.tudelft.nl/ ). During my EE study I have used their Nelsis package which is an early version of the SPACE package.
« Last Edit: June 28, 2013, 05:17:54 pm by nctnico »
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline ShenandoahTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: 00
  • User is banned
Re: My Dream Job
« Reply #28 on: June 28, 2013, 05:23:30 pm »
Quote
AFAIK the chip manufacturer gives you a library to work with. Usually with standard transistors and even logic gates. From there its like building with Lego bricks. During my EE course I have designed two digital chips (never produced though) and those where build from standard library components. Some parts like and/or arrays to implement logic functions like in a CPLD where generated from truth tables and imported as a functional block. Those functional blocks (including the I/O macro cells) had to be placed onto the chip much like placing components on a PCB and connected using much like components on a PCB. All in all it was actually pretty simple once the simulation was finished. But this was 2um NMOS

I feel much better now. Lego bricks for IC creator.  :-+
 

Offline Stonent

  • Super Contributor
  • ***
  • Posts: 3824
  • Country: us
Re: My Dream Job
« Reply #29 on: June 29, 2013, 01:54:02 am »
Maybe one day you'll just write your chips up in Verilog or VHDL and print your silicon on a printer.  :-+
The larger the government, the smaller the citizen.
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37664
  • Country: au
    • EEVblog
Re: My Dream Job
« Reply #30 on: June 29, 2013, 05:01:05 am »
lgbeno, thanks a lot for clarification. So it's the fabrication technology and materials that determine what components have to be used.

It's much deeper than that. The fabrication technology, material and physics determine how the "components" are realised on the final chip.
When you design a chip you are making the components and interconnects with basic physics and very complex process.
It is totally different to regular electronics where you use each known component and connect them up on the PCB or breadboard.

 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37664
  • Country: au
    • EEVblog
Re: My Dream Job
« Reply #31 on: June 29, 2013, 05:03:37 am »
I feel much better now. Lego bricks for IC creator.  :-+

Yes, you can do that, but this doesn't make you a chip designer any more than putting a design into an FPGA makes you a chip designer.
Only in this case it costs more, a lot more, and you have some more flexibility.
 

Offline gregariz

  • Frequent Contributor
  • **
  • Posts: 545
  • Country: us
Re: My Dream Job
« Reply #32 on: June 29, 2013, 05:25:27 am »
AFAIK the chip manufacturer gives you a library to work with.
Some colleges actually have their own process (usually an old one) and train their students to create their own devices on silicon or some other exotic wafer material. They then train them to characterize them. Someone has to be able to create processes rather than just the chips.
 

Offline gregariz

  • Frequent Contributor
  • **
  • Posts: 545
  • Country: us
Re: My Dream Job
« Reply #33 on: June 29, 2013, 05:27:21 am »
Maybe one day you'll just write your chips up in Verilog or VHDL and print your silicon on a printer.  :-+
It was called Verilog-A
 

Offline jancumps

  • Supporter
  • ****
  • Posts: 1272
  • Country: be
  • New Low
Re: My Dream Job
« Reply #34 on: June 29, 2013, 07:13:30 am »
Maybe one day you'll just write your chips up in Verilog or VHDL and print your silicon on a printer.  :-+
Chris Gammell wakes up now...
 

Offline nctnico

  • Super Contributor
  • ***
  • Posts: 26757
  • Country: nl
    • NCT Developments
Re: My Dream Job
« Reply #35 on: June 29, 2013, 01:46:11 pm »
lgbeno, thanks a lot for clarification. So it's the fabrication technology and materials that determine what components have to be used.
It's much deeper than that. The fabrication technology, material and physics determine how the "components" are realised on the final chip.
When you design a chip you are making the components and interconnects with basic physics and very complex process.
It is totally different to regular electronics where you use each known component and connect them up on the PCB or breadboard.
AFAIK it doesn't work that way. If this where true then each chip designer would need deep knowledge about the proces. That would be very inefficient. For analog circuits you'd typically specify the geometry which are fed into the spice model. The underlying Spice model is created by the people who have a deep understanding of the actuall process parameters through calculation and characterisation (a nice word for trial & error). For that job you'd probably need a physics degree and a good brain.

And there are specialist jobs as well. In a newsgroup I follow one of the regulars is an old school chip designer specialised in analog and mixed signal chips (everyone of us probably have used or have a chip he helped design); at some point he said he wasn't involved in designing I/O pads because that is an art in itself.

The more I think about it the more I get convinced there are many similarities to designing a PCB with standard components. The biggest difference is the cost when something goes wrong so designing a chip becomes a largely academic excerise using a simulator.
There are small lies, big lies and then there is what is on the screen of your oscilloscope.
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37664
  • Country: au
    • EEVblog
Re: My Dream Job
« Reply #36 on: June 29, 2013, 02:12:46 pm »
Maybe one day you'll just write your chips up in Verilog or VHDL and print your silicon on a printer.  :-+
Chris Gammell wakes up now...

and realises it was all a dream  :P
 

Offline AlfBaz

  • Super Contributor
  • ***
  • Posts: 2183
  • Country: au
Re: My Dream Job
« Reply #37 on: June 29, 2013, 03:55:46 pm »
I cant help but think there must be levels of abstraction when it comes to chip design and amongst them they probably discuss things parallel to do you need to know asm to program c++.

I wonder what the process flow for chip design is.

Having said that, i have a book that covers low level semiconducter theory, lythography, and the workings of basic electronic semiconductor devices from resistors to mosfets.
I've had it for over 10yrs and still haven't read it

 "Understanding Semiconductor Devices" by Sima Dimitrijev of Griffith university Australia. Published by Oxford university press
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8515
  • Country: us
    • SiliconValleyGarage
Re: My Dream Job
« Reply #38 on: June 29, 2013, 04:17:14 pm »
It depends what you define as chip design.
Coding vhdl or verilog is not chip design. It is system design.

Designing the actual and, or , not , doing the layout , extract parasitics and make a library of standard cells is chip design. You need to know how to create a transistor, how to define the wells, where to put the bulk connection, isolation rings and do it all in such a fashion that it yields the smallest surface area, fastest speed and lowest powerconsumption without blowing up in your face. The circuit needs to work reliably over process variations (layer thickness etc ) and temperature span.

For analog there is even more involved. You cannot simply say : i want 1 kiloohm here.. You'll get something that will be between 800 ohm and 4k...
Only the ratios will work. Absolute values are unknown. Cant make capacitors either on chip except for extremely small ones... So everything is dc coupled in your design....

There is a linrary with standard blocks but most commonly these are taken, modified a bit and then used. You cant simply slap down 4 opamps and wire them up with 5 resistors.... The design would be waaay too large. You'd strip the output stage for the opamps used internally. Youll also strip,the bias generators as you can share the bias generator between opamps.
So that stuff invariably ends up being full custom starting from a few base blocks with a full custom layout. You will need to interact with the ic layouters as they will extract parasitics and you will have to see if there is impact and provide feedback. Select carefully as any square micron you add increase cost of the chip thus reducing profit....

What block to place where ? How do we bond-out. How do we thermally distribute ? What about gradient compensation ?

There is a lot of factors to account for that do not exist in traditional design.

Then there is the whole testability problem. You will need to inject test multiplexers to be able to probe internal nodes during ic testing. Every singgle chip produced goes through that.

A circuit you breadboard with an lm324 and some resistors in half an hour takes about a week to design on silicon... Design as in : here is the GDS tape , please make the masks and fab it.
Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline ShenandoahTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: 00
  • User is banned
Re: My Dream Job
« Reply #39 on: June 29, 2013, 10:41:59 pm »
Quote
Maybe one day you'll just write your chips up in Verilog or VHDL and print your silicon on a printer.

I can see this in the near future, my grandchildren will be laughing at me when I tell them about these big ugly retarded fabrication facilities.  :-DD
 

Offline ShenandoahTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: 00
  • User is banned
Re: My Dream Job
« Reply #40 on: June 29, 2013, 10:44:10 pm »
Quote
Yes, you can do that, but this doesn't make you a chip designer any more than putting a design into an FPGA makes you a chip designer.
Only in this case it costs more, a lot more, and you have some more flexibility.

Yes I got it Dave, chip designers are contracted scientists from Mars.
 

Offline DaveW

  • Frequent Contributor
  • **
  • Posts: 284
  • Country: gb
    • WattCircuit
Re: My Dream Job
« Reply #41 on: June 30, 2013, 12:09:23 am »
As you brought up Bob Pease, he gave advise to someone on how to become a chip designer. Just start adding up all of the years of experience that he advises; and that's for someone with EE qualifications, although like you, they're working in software

http://electronicdesign.com/analog-amp-mixed-signal/bob-s-mailbox-advice-young-engineer-julie-resistors-key-less-acceleration-an
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37664
  • Country: au
    • EEVblog
Re: My Dream Job
« Reply #42 on: June 30, 2013, 12:20:24 am »
I wonder what the process flow for chip design is.

I imagine it depends entirely what type of chip you are talking about.
Jeri for example is currently doing a video compression/whatever chip for your CastAR glasses. That would be as high level abstraction as is possible. She just wants to do some digital stuff and get the power consumption and cost down from the existing FPGA solution.

Contrast that with someone at Analog Devices designing the latest ultra low noise, ultra low power, ultra low drift opamp.

They are chalk and cheese.
 

Offline ftransform

  • Frequent Contributor
  • **
  • Posts: 728
  • Country: 00
Re: My Dream Job
« Reply #43 on: June 30, 2013, 03:16:19 am »
in a new process analog devices just hired drunken hicks to yell "shut the fuck up and hold still" at this chips when they are carried on by on conveyor belts.


No magic there... thats how chopper stabilized is really made, they are regular op amps threatened with a butchers knife.
 

Offline c4757p

  • Super Contributor
  • ***
  • Posts: 7799
  • Country: us
  • adieu
Re: My Dream Job
« Reply #44 on: June 30, 2013, 03:19:13 am »
That's why I don't like AD - it took them 48 years to figure out that shouting and cussing at silicon makes it behave better. Every computer user in the history of computer users could have told them that!
No longer active here - try the IRC channel if you just can't be without me :)
 

Offline ftransform

  • Frequent Contributor
  • **
  • Posts: 728
  • Country: 00
Re: My Dream Job
« Reply #45 on: June 30, 2013, 03:22:44 am »
AD actually copied the process from texas instruments, which discovered it due to the nature of texas
 

Offline Stonent

  • Super Contributor
  • ***
  • Posts: 3824
  • Country: us
Re: My Dream Job
« Reply #46 on: June 30, 2013, 03:32:11 am »
AD actually copied the process from texas instruments, which discovered it due to the nature of texas

Texas: It's like a whole other country (tm) !
The larger the government, the smaller the citizen.
 

Offline EEVblog

  • Administrator
  • *****
  • Posts: 37664
  • Country: au
    • EEVblog
Re: My Dream Job
« Reply #47 on: June 30, 2013, 06:28:50 am »
Texas: It's like a whole other country (tm) !

Just like Africa!  ;D
 

Offline Stonent

  • Super Contributor
  • ***
  • Posts: 3824
  • Country: us
Re: My Dream Job
« Reply #48 on: June 30, 2013, 07:50:11 am »
Texas: It's like a whole other country (tm) !

Just like Africa!  ;D

 ;)

But if you do want to get technical...

Quote from: Wikipedia
The Republic of Texas (Spanish: República de Texas) was an independent sovereign nation in North America which existed from March 2, 1836, to February 19, 1846. It was bordered by the nation of Mexico to the southwest, the Gulf of Mexico to the southeast, the two US states of Louisiana and Arkansas to the east and northeast, and the United States territories encompassing the current US states of Oklahoma, Kansas, Colorado, Wyoming, and New Mexico to the north and west.
Formed as a separate nation after gaining independence from Mexico in 1836, the republic claimed borders that included all of the present US state of Texas as well as parts of present-day Oklahoma, Kansas, Colorado, Wyoming, and New Mexico based upon the Treaties of Velasco between the newly created Texas Republic and Mexico.
The larger the government, the smaller the citizen.
 

Offline ShenandoahTopic starter

  • Regular Contributor
  • *
  • Posts: 53
  • Country: 00
  • User is banned
Re: My Dream Job
« Reply #49 on: June 30, 2013, 04:59:56 pm »
Quote
As you brought up Bob Pease, he gave advise to someone on how to become a chip designer. Just start adding up all of the years of experience that he advises; and that's for someone with EE qualifications, although like you, they're working in software

http://electronicdesign.com/analog-amp-mixed-signal/bob-s-mailbox-advice-young-engineer-julie-resistors-key-less-acceleration-an

Why should someone's experience be considered an applicable advise to anyone? People different in their learning curve, abilities, motivations, life style and circumstances, accessible resources, and of course luck. I can see his advise is true for average EE graduates IC-designer wannabe vs. someone with true passion and endless motivation. What takes you 20 years to achieve it may take others much more, and some others much less.

Working at the very low level and being closer to the hardware physics does not necessarily make it much harder than working at the higher level.
When designing using FPGA and HDL you may face problems much much harder than IC chip layout and physics. It all depends on the problem domain.

Have you checked TI's ultra fast ADCs?  :clap:
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf