General > General Technical Chat

Is there a divide between board level electronics engineering and ASIC design?

<< < (2/2)

tooki:

--- Quote from: karpouzi9 on April 09, 2023, 06:30:31 pm ---That is a lot of letters to spell out the sentence "it's not for me"!

--- End quote ---
Well ASIC design isn’t for me, that’s true, but dismissing my entire reply like that is rather rude and ungrateful of you.

Your entire original question is predicated on the assumption that ASIC and board level EEs are interchangeable, and that the only reason we’d want to make them appear distinct is to manipulate wages. I don’t think you’ve considered the possibility that they actually aren’t interchangeable. That’s why I wrote “letters” to you explaining my position with why I think what I think, and what characteristics I would look for in an employee and why.

Also, I asked you the very real question of why you think merging the two disciplines is actually desirable to begin with. I’d be curious for your answer.

SiliconWizard:
"ASIC design" is microelectronics, it's of course part of EE but as a specialization as others have said.

The "divide" is that it's two different activities. Analogies have their limits, but it's a bit like comparing a web front-end software developer with someone developing firwmare in assembly.
Is it software in both cases? Sure. Do you expect one to be able to do the job of the other? Usually not, unless they have managed to master both. Which is not as common.

So it's largely a high-level/low-level kind of divide. Not sure what kind of opinion you are after though?
Yes these are two pretty different activities with a different mindset and different tools, even though it's electronics in both cases.

nctnico:

--- Quote from: tooki on April 09, 2023, 05:57:18 pm ---
--- Quote from: karpouzi9 on April 09, 2023, 04:53:21 pm ---This thread seems to have sunk beneath the tide. Any new ideas for increasing overlap?

Or is it better for the populations of [board-level] and [asic-level] EEs to stay separated so as to reduce their perceived interchangeability and increase wages within each group?  :-//

--- End quote ---
Why would you want to increase overlap? Specialization is unavoidable given the sheer complexity of the things we do. It’s impossible for an individual to be an expert in everything. I’d rather leave chip design to chip experts and board-level engineering to board-level experts, rather than having everyone do both of them poorly.

--- End quote ---
It also depends on what part of chip design you are doing. Likely most of the people involved in designing chips don't have to care about the physics parts of transistors but 'only' create circuits from existing parts. My EE study was very biased towards analog and digital ASIC design and I had to design 2 chips during my study. The chip design workflow was based on using parts from libraries. From what I was taught, you are not going to develop your own transistors but you use standard parts that where created for the particular chip process (size & technology). From the standard transistors and/or logic gates you create blocks of components (where digital parts like lookup-tables, etc can be auto-generated). Once you have simulated the functionality to an extend that you have good test coverage, you put the blocks on a chip and wire them up. Much like you'd do with putting parts on a PCB and making the connections. Once that is done, you extract the circuit from the chip which now includes all the parasitic capacitances and resistance from the 'wiring' and verify the design.

This workflow is pretty similar to creating a PCB using off the shelve parts. However, due to the high mask costs and long turnaround times you can't create so many prototypes so circuit simulation is the biggest part of the job in order to make sure the chip works the first time. Too tedious and way too little hands-on for my taste though.

CatalinaWOW:
There is a difference.  But not a divide.  There are skills that span across both areas.  And others that are very specific to each area.  Just like other areas that are different but the same.  Power electronics.  Audio.  So the divide is an arbitrary selection of a point on a gradient.

The divide is changing over time.  When I started my career a flip-flop or ring counter were board (or chassis) level designs.  Now they hardly rate as a component and things that used to a take a cabinet full of boards are put on a single chip.

One skill that applies across the entire spectrum is the whole specification, application, interface and cost integration job - often called system engineering.  It is also a skill that is usually in high demand.  Note that there are many titled and trained as systems engineers, but relatively few that are actually good at it.  This skill is not necessarily a good key to getting a board level assignment.  The perception is that the total value proposition of a board doesn't demand much of this tasking, which is correct.  But rounding this to zero is often the action which results from the perception and is wrong.  The skill would be just part of what a good board designer uses, and would be among the things that differentiate a good one from an average or bad one.

TheUnnamedNewbie:

--- Quote from: karpouzi9 on April 10, 2023, 07:06:52 am ---
In some unreal future, everyday engineers could design or extend their own ASICs from open-source blocks that would do the same things at a fraction of the cost, lead time, pin count, and environmental impact. These blocks would be anything from a MOSFET, to an opto-coupler, to an opamp, to a mid-range CPU or even flash memory.


--- End quote ---

I mean part of that is already a thing. A lot of companies don't design their own in-house I2C, HDMI, ethernet, ESD, whatever. You buy/license them from companies like Rambus, Synopsis, Cadence, etc...
 The problem ends up being the high up-front cost to tapeout. Even if the design was cheap (sometimes these IPs are paid in a per-die-manufactuerd way so little upfront cost in the design), taping out a chip in a <100 nm technology is going to have up-front costs ranging from 1 million USD to many million USD, just to have your maskset made. You need a huge volume to make up for that.

Ofcouse, you can make chips in older technologies too, and a lot of times this is what they do for true ASICs (I mean, a customer goes to an ASIC designhouse and asks for a very specific chip for a single product). But unless you have huge volume, it is just not feasible (and it is high risk!)



To come back to the original question/topic:

--- Quote from: SiliconWizard on April 09, 2023, 09:27:06 pm ---"ASIC design" is microelectronics, it's of course part of EE but as a specialization as others have said.

The "divide" is that it's two different activities. Analogies have their limits, but it's a bit like comparing a web front-end software developer with someone developing firwmare in assembly.
Is it software in both cases? Sure. Do you expect one to be able to do the job of the other? Usually not, unless they have managed to master both. Which is not as common.

So it's largely a high-level/low-level kind of divide. Not sure what kind of opinion you are after though?
Yes these are two pretty different activities with a different mindset and different tools, even though it's electronics in both cases.


--- End quote ---

I think this is a good analogy to use.

A place where I see more and more overlap is when you get to anything high-speed/high-frequency, because in a lot of those situations you are constrained by pretty much every single thing in the signal path. So the people working on the 112 Gbit/s serdes need to understand what the limitations are for each part of the chain. At least the system level guys do, those who come up with the standards and architectures. Though now that I think about it, once those (usually research) engineers paved the way, the guys following just have a standard to follow and I guess don't really need to know what is going on anymore.

Navigation

[0] Message Index

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod