Hi, How about your design method?

[hardings]

Hi,

I always use equation to get parameter of the circuit, calculation is very

heavy. My director said we should use "intuitive" method .

How do you think of that ?

 

[electronrancher]

it's the best idea - start using rules like: precalculate mosfet to get an idea of Vdsat..

Now you double the size of mosfet to get 2x the current, but Vdsat stays the same - excellent when you have to create low-voltage circuits, so running into the vdsat at top or bottom rail is a big problem.

Another good one is counting diodes in bipolar. If Vb is 2v, Ve has to be about 1.3. Or if emitter is connected to ground, you KNOW base can only go about 0.7v above gnd. Now you trace thru other branches starting from your known point, and can get a good idea of a novel circuit in only a few minutes.

Start rough, and increase complexity as you work to tighter and tighter tolerances. Even guessing current in one branch is about 2x current in another branch is much more efficent than calculating for 20 minutes to find out the answer is really 1.8903x.

I hated this method when I first had to learn, and even lost points on quiz due to "too many calculations" when the answer was perfect! But it turns out that starting with rough estimate is really the trick of all the good analog guys i have met - then they can modify very quickly to their desired spec.

 

[hardings]

Hi,electronrancher

That is wonderful way to design. But I hate it too!

"rough estimate" can have many start point and maybe let designer be

unsure of the circuit!

Maybe I need to learn more about that.

 

[electronrancher]

I agree, but starting points are always given by the planners of the IC - maybe we want hi gain, but we can sacrifice some output swing. Smaller mosfets here toggle fast in an amp, but swing is limited. Opposite case gives near-"rail to rail" swing but lower gain on your stage.

Actually, I had to debug a MAX-232 replica one time that showed a big glitch on cold startup. Probing the metal lines of the IC showed me an emitter going up to 2v when I expected only 200mV normally.

Counting diodes backards found a circuit that was collapsing, and putting all the charge-pump voltage into the wrong path. Six transistors later, it caused the glitch that I was called in to find. Even rough estimates were able to find the problem in less than 2 days probing, and the IC was fixed within the week.

Heh then I got promoted to senior engineer and was only 25! Rough estimates are OK at first - make it perfect once you know it works!

 

[sutapanaki]

Hi Hardings,

Yes, it is all about intuition. Rough etimates are not always equivelent to intuition, though. However they are more often helpful than not.
Also, go have a look at www.ardem.com.

 

[hardings]

Hi sutapanaki

Look like a good material. But there is no pdf file there and the 'response form' does not woking.

Can u get that and how do u comment on that?

 

[OpAmp]

Hi,
More than 90 percent of analog designers use only intutive approach to design thei circuit. But the remained 10 percent combine intuitive approach along with hand calculation. for example you can not use SPICE level 1 equations for submicron CMOS IC design, because those equations are no longer valid. I use some curve fitting methods to find some new and simple equation and then use hand calculation to find the optimum design.

 

[sutapanaki]

Hardings,

Yes, there are no pdf files there. Now I see that the responce form is not working again. It happened to me as well, but one time I got it working. I just tried now and for me it is working. I think if that dvd ever becomes reality it will be a good source of knowledge. However, the reason I gave you this url was to have a look at the book that was mentioned there and now is no longer on the web-site (I don't know why). The book is:

Fast analytical techniques for electrical and electronic circuits
Vatche Vorperian
isbn 0521624428

Unfortunately I haven't seen an electronic version of it. You can also try to find something useful at Intusoft's website.

 

[hardings]

Hi,

I use google to find this site:
https://assets.cambridge.org/0521624428/sample/0521624428WS.pdf
But it is incomplete.

OpAmp

I only use curve fitting method to get the capacitance charge or discharge voltage. And u?

 

[nickwang]

Gets some information from all your posts.
I got a question about analog design, would you please talk about how do
you get the layout for your design? I mean which tool you are using.
or the steps to the layout.
Thanks in advance

 

[bastos4321]

I use cadence virtuoso for IC layout and calibre for verification.

bastos

 

[sutapanaki]

well, the steps to the layout. I think you can hardly get a prescribed approach concerning the analog layout. It is not automatically done, so it is more or less an individual approach. But, group devices that should match together - either digitized or common centroid or combination of both. Don't route across transistors that should match. Use dummy devices around devices that should match. Floorplan so that you avoid crosstalk, noise and termal coupling. Distribute currents, rather than voltages (where possible). Don't match a transistor with another one that is miles away. So on and so on. I'm sure many people can add more of these DOs and DON'Ts.
As for the software, well it is just a software. It doesn't do the layout (or design) for you.

 

[nickwang]

Thanks,bastos4321 and sutapanaki.
basto, what do you use for schematic ? Cadence analog artist?

sutapanaki, thanks for your enlightenment. Though I know it is hard for analog design to be automatic, there is still ways of using software to help, as basto uses. But from your statements, I guess you depends heavily on your experience rather than tools, right?
Do you have to modify a lot on the software-output products(layout?) and modify before the software handles your design?, or both?
Thanks again

 

[linxf2003]

I think intuition is very important to an best enginner. I read a paper talking about the phD qualify exam in Berkerley, the professors are highly emphasized to check whether you can answer the questions by some simple way not using a complicate equation even it is correct.

 

[hung_wai_ming@hotmail.com]

almost 4 years from now on, anyone have new ideas?

 

[Vabzter]

Hi all,
Excellent post from all. but i have one question.lets take a simple 2 stage opamp for design..in my univ they gave us a datasheet of 0.35um process(un, cox, vth e.t.c)
Buts lets say in a company where u are designing for 90nm for less, then do u still find these process parameters or there is any other way around it..I mean in doing hand calculations u require these process parameters for Id e.t.c..
So how do u experienced designers designs..
Thanks
BR
Vabzter

 

[sutapanaki]

Well, you get process information which can include un, cox,vth etc. However for nm transistors those parameters increasingly bear no meaning anymore. Vth for example varies a lot with L UnCox is also not constant. In short your square-law transistors do not exist. This is more true for short L. But also for long L you have deviation from it caused by the vertical fields of the gate since the gate oxide thickness is so small now. I would recomend you to search the net for gm/Id based design. There are articles and other publications. Good luck

Hi all,
Excellent post from all. but i have one question.lets take a simple 2 stage opamp for design..in my univ they gave us a datasheet of 0.35um process(un, cox, vth e.t.c)
Buts lets say in a company where u are designing for 90nm for less, then do u still find these process parameters or there is any other way around it..I mean in doing hand calculations u require these process parameters for Id e.t.c..
So how do u experienced designers designs..
Thanks
BR
Vabzter