ANALOG DEVICES Signal Chain Designer TOOL USER LICENSE AGREEMENT

This is my personal take on things. I speak for myself, I am not a mouthpiece for ADI even though I work there and have done for quite a few years now.

The System Designer software needs to know the circuit in order to run

- that should be obvious. Some people view that as IP. ADI doesn't care, they just want to sell you parts. They would be interested if you came up with a configuration that crashed the software, or if you reported it gave bogus results, but beyond that they don't bother to look at what people submit. I'm roughly aware of the number of users as measured by EULA acceptances, and it's so large that they couldn't possibly look at everything that comes in. And to be frank, I'd hazard that most of them are pretty mundane, the product of the huge falloff in the teaching of analog circuit design skills in recent decades.

LT and ADI have different views with regards to simulation. LT puts their PC-based tools out there for people to use. I've played around with their simulator at home and it's quite good considering what I paid for it. It's certainly an improvement over other low-end SPICE-based packages, and is fine for the relatively simple signal chains that many customers want to design, but it would gag on the stuff I do at work every day.

ADI's internal simulator is the result of many critter-decades of investment, and is regarded as a significant competitive advantage. IBM, TI, ST, and the other mixed-signal biggies all have their own internal simulation capabilities, carefully guarded for the same reason. Accurate design of complex high-performance analog and mixed-signal chips requires a lot more than standard SPICE offers, and it's a lot of effort to get there - why would any of the big guys want to show their methods to a competitor?

LT has concluded that there's little competitive risk in them disclosing their simulator. ADI, TI, and others have a different opinion. Given that, the non-LT contingent are attempting to do what they can to support customers. You can like it or not, but that's the way it is despite what the conspiracy-theorists want to believe. If ADI or any other semiconductor company wants to know its competitors' tricks it's a simple enough matter to buy parts from Digikey or Mouser and reverse-engineer them; this has been going on in the industry since the beginning. As for "stealing" IP from customers, I've never seen it and doubt it would happen, the costs of getting caught would be too high and simply not worth the risk. We work *with* customers on proprietary system solutions every day, with both sides protected by appropriate non-disclosure agreements.

Use ADI's simulator, or TI's, or LT's, or one you've paid for, I don't really care. Every company offers some parts with performance unmatched by the competition, and when you need one you'll use it, regardless of who makes it or what kind of simulation support they provide. But recognize that simulation isn't just a commodity. Companies do what they can to support customers in designing in their products, but they operate in a competitive environment and need to balance what they give to customers against what they might be giving away to their competition.

I think I've said enough on this topic.

Steve

Reply to
Steve Goldstein
Loading thread data ...

OK sorry for making you clarify that. (I guess I was feeling like being the contrarian.)

Grin, half the time I use spice it's as a check of my algebra.

Oh and Steve G. ... Thanks for the nice response to Phil and SED.

George H.

Reply to
George Herold

I've been talking to Analog Devices about Spice modeling for more than four years... monthly or more with John Ardizzoni in Massachusetts... quoted several parts, with no results.

(I even had an inquiry in 2011 from Steve Goldstein asking about IBIS modeling, and what I needed to perform such a service :-)

Finally, in August, 2012, John Ardizzoni arranged an interview at Analog Devices, Santa Clara, ostensibly a job interview, but actually for a "business relationship".

There I met Paul Rako, and also talked with Paul Grohe, and several others I cannot now remember. The engineers and I got on famously... they _all_ knew of me from SED ;-)

Then I went out to lunch with Rako and Steve Sockolov, a VP at Analog Devices.

And we talked about modeling, and how I felt that the best approach was to provide me with schematics and netlists and device libraries, so that I could generate a near-exact behavioral fit to their real device.

Everything went quite cordially although, in retrospect, my actually wanting to know the guts of their chips was probably the killer.

After I departed I heard no more.... silence. When I finally inquired... "We've decided no."

BTW: The manager of Spice modeling, Steve Sockolov, is VP of MARKETING. Need I say more >:-}

I think Analog Devices management are paranoid... for that matter, so are LT's... and TI's... just look at the wave of encrypted models that only simulate on their proprietary simulator.

I have just concluded my first contract for a fully behavioral Spice model. This customer was not paranoid, and has provided me with full schematics, netlists and device libraries.

It would be foolish of me to divulge any proprietary information. I'd never get another job. Why are people so paranoid?

Troubling times :-( ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

After a private meeting, you just named a bunch of names, and trashed a manager, all in the public record, online forever. That won't encourage future companies to meet privately with you, or to hire you.

Bet they had other reasons for not engaging you. They were right.

Of course I'm killfiled, and followups are killfiled, so you can't see this.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation
Reply to
John Larkin

Amazing!

I told you about the Parker Bros, twins, who started a Power Supply company in LA area, and would tour their competitors though their facility showing in complete detail the reasons for their success.? How can you DO that? received the answer, "they'll never use what they see, they think they're right about the way they're doing things now, and will never change." The twins actually gleaned information FROM their competitors during these tours, too.

And thus the conclusion: It's not the idea that succeeds, but who carries out that idea that succeeds.

Reply to
RobertMacy

There is too much paranoia. Especially about pricing. Some people make it really hard to get pricing. It's probably to their advantage that competitors let me know their pricing; if nothing else, it keeps me from grossly underbidding them, and it may keep me from competing entirely.

One company once did copy one of our products. My biggest cuctomer told me, and blackballed the copier.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation
Reply to
John Larkin

You don't make sense.

Reply to
j.mercredi

However, as I have mentioned a few times recently, this simulating speed of switchers is a complete dead issue today.

When I started SuperSpice (XSpice). before 2000, I had a Current Mode Controller model, all in a Spice3 behavioural model that took several minutes to run. Today, this model takes 7 seconds and My Dell machine is 4 years old, and runs at 30 GFLOPS

I have the analog bit of a 2nd Order Delta-Sigma converter taking 20 seconds, for 2ms tran time at a 1MHz clock.

Kevin Aylward B.Sc.

formatting link
formatting link
-SuperSpice

Reply to
Kevin Aylward

Ahmmmm... well you walked into that one, baring in mind I cant find anyone to pay me for model generation...

So... Oh.... I had a look at your analog versions of digital logic. Around

10 net lines a piece...ahmmm....indeed

These are my behaviour digital logic from my SS lib:

All 0V to 1V logic, but analog inputs can span any voltage, cleanly. Low gain, but enough, and minimizes convergence issues.

.SUBCKT inv_XN A O

  • _SS_Symbol [Behavioural.ssm] [invertor] B1 Ox 0 v=1 - tanh( 2 * tanh(5 * v(A) - 3.4) + 2) C1 O 0 10f R1 Ox O 100 .ENDS

.SUBCKT nand_XN A B O

  • _SS_Symbol [Behavioural.ssm] [nand]
  • B1 Ox 0 v=1 - tanh( 2 * tanh(5 * v(A) - 3.4) + 2)*tanh( 2 * tanh(5 * v(B) - 3.4) + 2) C1 O 0 10f R1 Ox O 100 .ENDS

.SUBCKT nor_XN A B O

  • _SS_Symbol [Behavioural.ssm] [nor]
  • B1 Ox 0 v= 1 - tanh(2.5 * (tanh( 2 * tanh(5 * v(A) - 3.9) + 2) + tanh( 2 * tanh(5 * v(B) - 3.9) + 2)) ) C1 O 0 10f R1 Ox O 100 .ENDS

er... note the one line core code bit.

The d-type is in the SuperSpice download :-)

Kevin Aylward B.Sc.

formatting link
formatting link
-SuperSpice

Reply to
Kevin Aylward

My aim is for a model where data sheet info can be entered directly. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Can you be more specific? What doesn't make sense to you?

--

John Larkin         Highland Technology, Inc 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com
Reply to
John Larkin

I had a go at running the inverter. Spice3/XSpice conversion is a bit of a hassle.

The conversion of

G1 1 2 VALUE {expression()}

to

B1 1 2 i=expression()

Is not too bad. However... Spice3/XSpice don't take K, m etc in the expression() bit. It must be all in scientific notation, despite normal netlist being ok with multipliers. This gets to be a pain in manually converting.

I did do a DC sweep, and yeah, it wants to switch in uV. Not a great idea Jim. My versions have about 12db gain, which is all that's needed.

So, how often do people come up to you and say, "I'm an engineer Jim, not a doctor !",

Kevin Aylward B.Sc.

formatting link
formatting link
- SuperSpice

Reply to
Kevin Aylward

Kevin,

I saw your post saying a DC sweep of my inverter gave you a transition width of 1uV, but now I can't find my way back to it to reply. It should be 1mV, which is similar to a _real_ CMOS inverter. How'd you get 1uV? ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Hi Kevin,

Now I see how you arrived at 1uV. I may indeed have over-done it. I emulated the behavior of real CMOS _except_ my currents controlling delay time have no equivalent to channel-length modulation, just end-stops.

So, Kevin, THANKS for the prod! I'll fix that, which should bring the "gain" down dramatically. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Another way of bounding solves all issues at once. Kevin, thanks for the prod/nudge! I aim for perfection >:-} ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

I am using an engine that is not as good as others in convergence, so I am really paying attention to certain aspects of modelling.

Key points.

1 Minimise the number of nodes for speed. 2 Make everything smooth for convergence. 3 Limit everything for convergence.

I did note that you are on the same lines as me with your own clamper.

I made this one recently. It acts like 2 parallel diodes (not series), but the on voltage is changeable to vmax instead of ~0.6V. However, it has a max current to stop it clamping to infinite current.

.SUBCKT clamp_XN 1 2 vmax=1 imax=1

  • .param i0=1e-18
  • b1 1 2 i={imax}*tanh({i0}*sinh(100*tanh({40/vmax/100}*v(1,2) ) )) .ends

Its a one liner, avoiding an extra resister node to limit the clamping. The divide and X 100 is to stop sinh going out of bounds.

Anyone can use any model in SS, just include a reference to SS

Kevin Aylward B.Sc.

formatting link
formatting link
-SuperSpice

Reply to
Kevin Aylward

Hi Kevin, I like your "double TANH" scheme, but you need more gain to match the transfer curve of a real device...

I have too much gain, my model is more like an OpAmp, very high gain at DC, then roll-off; so I'll have to remedy that. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Yep. That's what it takes.

And I'll have to watch the use of "u", "m", "k"... since I didn't know that some Spice's can't handle that :-(

...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Hey Kevin, My 3F3 manual says that "m", "u", etc, are Kosher. Which Spice variants can't handle that? ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Depends on what device :-)

45nm devices have Va < 1V, so they actually don't have much gain. 12db is all you need. I have an LNA example
formatting link
that illustrates this.

Models are a compromise. I don't necessarily try and match a real device where such matching is not going to significantly effect the final result. For Analog models of Digital logic, other than final input and outputs, the gain between logic really don't matter at all. It can help though for speed and convergence though to keep it low.

I have a Behaviourial PLL with divide 8 counter all done with this tanh() logic in the current SS download (PLLBehavioural.sss)

Kevin Aylward B.Sc.

formatting link
formatting link
- SuperSpice

Reply to
Kevin Aylward

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.