Spice simulation

john curl said:
Spice emulations? I don't need no stinking Spice emulations! (think bandito icon)
Count me in with Bob Pease on this one. ;-)
I just measure them in REAL-TIME with capacitors and a square wave generator, as I learned to do at my mentor's side, 40 years ago.


Hi John,

Either way works. I didn't use SPICE on my MOSFET power amplifier 25 years ago, and was able to get good results. It is also very important that people understand the limitations of SPICE, and don't just totally depend on it. Just as you have to LISTEN to an amplifier after you measure it, you have to MEASURE your built amplifier after you simulate it. I've rejected numerous JAES paper submissions because people claimed results based on simulation and never built the thing.

Having said all that, however, I find SPICE to be incredibly valuable in providing insights and in doing experiments and seeing what works and what doesn't have a chance. I have truly learned a lot when SPICING my own amplifiers, and have had several eye-opening experiences.

You should give it a try. It is never too late for an old dog to learn new tricks. I am also an old dog, and continue to learn new stuff every day.

Cheers,
Bob
 
Bob Cordell said:
.................... I find SPICE to be incredibly valuable in providing insights and in doing experiments and seeing what works and what doesn't have a chance. I have truly learned a lot when SPICING my own amplifiers, and have had several eye-opening experiences................

:yes:

And it saves a lot of time (and semiconductors!).

Cheers,
 
Bob Cordell said:



Having said all that, however, I find SPICE to be incredibly valuable in providing insights and in doing experiments and seeing what works and what doesn't have a chance. I have truly learned a lot when SPICING my own amplifiers, and have had several eye-opening experiences.


(italics mine)


Bob,
If you have time and are bored, read through the first part of the Aleph-X thread. (Yes, I know the thread's enormous; the first ten or twenty pages will do for current purposes.) I opened the thread and was immediately assailed by people whose first impulse was to simulate the circuit. Time after time I was told it wouldn't work. Why? Because their simulations said so! I knew differently because I already had a functioning prototype built with real, live parts. (IRF MOSFETs in case it matters.)
Now, the standard rebuttal is that "obviously" the person doing the simulation didn't know what he was doing. Once that argument runs its course, people begin pounding their chests claiming that their simulation software is better than the other guy's...etc. etc. etc. "Oh, bother!" said Winnie the Pooh, "Here we go again!"
My point is that simulations are capable of giving "false negatives."
I sometimes wonder how many marvelous ideas have been abandoned simply because someone didn't have the gumption to build a circuit after a simulation told him it wouldn't work.
If he did have the courage, why waste time doing the simulation? Go ahead and build the bloody thing and know absolutely, unequivocally, rather than guess.
Incidentally, the Aleph-X does not have an output network of any sort and somewhere in that thread I mention tossing a 4uF (I think...it's been a while) cap on the output. Nothing blew. Nothing got wonky. Film cap, I believe, for those who want something to gripe about.
Granted, the Aleph-X is not a "normal" topology, and I'm sure ten guys will post over the next hour pointing that out. However, I sometimes wonder just how much of the output network is necessity...and how much is "because it's always been done that way."

Grey

P.S.: Please don't ask me to try this load or that or any of umpteen other imaginable tests--at least not right now. I've got six month-old twin boys and a four year-old daughter that I watch over during the day and I work at night. What little time I have for electronics is currently aimed at conjuring an Aleph-X Ver. 2.0 out of thin air.
(And to the nay-sayers who might wish to point out that lightning ain't likely to strike twice, my response is simple: I know. But Nelson kinda tossed out the challenge and I try to accommodate the guy, 'cuz he's been good to me. So there.)
 
My first negative experience with computer modeling was in 1967 (yes, 40 years ago) with ECAP (Electronic Circuit Analysis Program) written by IBM, most likely for the military.
We (I was immediate supervisor) were running worst case simulations of many of our motor drive circuits. Sometimes they would show (FAIL) even though a simple slide rule calculation would show OK. The computer analysis was wrong, but what could we tell the client? We were caught short.
Computer analysis is just fine, for what it is and does, but don't think that everything that the computer gives you works, or visa-versa.
 
My experience with simulations, that keep me chuckling when they are put forward as hard facts, is that the simulations don't take into account the physical layout from input, to output, to power supply, not to mention the PCB layout. I've spent 30 years stumbilng through this maze and there are quite a number of the instabilities that result from the thousands (?) of ways to physically implement a circuit that do occur, that aren't really covered in the textbooks.

If a zobel network or coil is necessary in some cases but not in others, what is the variable? If two (or 30) people build a chip amp and they all get different results, what might be going on? What is actually happening when you place a ground plane around your circuit? If you hook a 8ga litz wire ground (that's a joke, son) to your heatsink, is it really grounded? Why are some implementations plagued by hum and RFI and others quiet as a they should be? What is a ground?

I won't go on, but the concept deserves some thought.

Mike.
 
john curl said:
My first negative experience with computer modeling was in 1967 (yes, 40 years ago) with ECAP (Electronic Circuit Analysis Program) written by IBM, most likely for the military.

Honestly John, by similar resoning you and many other EEs would probably still be using valves because the transistors were so bad 40 years ago. :)

Since computer science is a younger discipline than electronics, I think it is fair to assume that it has advanced much more than (at least analogue) electronics has in the same time. Furthermore, with the computing power we have today in an ordinary PC, we can easily do simulations of a kind that nobody would have dreamed of then, because the models we use today would have been too complex to be used on 1960s computers. Most probably the models used in that ECAP system were extremely crude and inaccurate to make computer simulation possible at all.


We (I was immediate supervisor) were running worst case simulations of many of our motor drive circuits. Sometimes they would show (FAIL) even though a simple slide rule calculation would show OK. The computer analysis was wrong, but what could we tell the client? We were caught short.
Computer analysis is just fine, for what it is and does, but don't think that everything that the computer gives you works, or visa-versa.

This I totally agree with. How long didn't some aerodynamics experts tell us that bumble bees can't fly? (Fortunately for the bumble bees, recent research has shown that they were right all the time, so they can go on flying happily).

One should never trust computer results as a truth. First, you must have a fair understanding of how the circuit is supposed to work, and what the component models look like. Then you must consider if the answers are reasonable, just as with any calculation. Obviously Spice simulations can be very accurate and reliable if you know what you are doing. ICs are regularly designed using Spice as the verification tool, since a breadboarded model would be both impossible and too inaccurate. But ICs are simpler in the sense that it is a more controlled environment. For discrete circuits we do get a lot of added error sources, like parasitic, mutual induction between PCB wires etc. etc. So a perfectly accurate simulation of how a finished circuit will behave is hardly practically feasible.

It is probably true as you said, that occasionally someone has decided not to build a working circuit because it didn't seem to work in the simulator. However, then it is most certainly mishandling of the simulator, just as you can mishandle an oscilloscope or any tool you use. I think on many more occasions, people have been spared from building a prototype that never had a chance of working, because they detected it already in simulation. In fact. I once found a problem when simulating a circuit that I most probably never would have found if building a prototype. I don't think I have the instruments i would have needed and even if, I wouldn't have found the problem unless I already knew I should look for it. Since I didn't think of it when inventing the circuit, why should I have thought of trying to look for it?

That said, I see no need to convince experineced EEs like you to start using Spice if you are happy without it. You have found your ways to think and design long ago and are comfortable with them. I am sure Spice could help you to get new insights, like Bob suggested, but of course only if you are curious about it yourself. Spice won't do it for you.

For less experienced people, I think Spice can be a great tool to experiment with circuit behaviour in general, to check if one has understood the theory correctly. It should not replace the theory, though. It would be utterly stupid to use trial and error in Spice instead of learning Ohms law, to take an overly stupid example. Spice can also be useful to test concepts and compare topologies with ideal components. The latter may sound useless to some, but I think otherwise. By using ideal models, or models which are elaborate but not corresponding to a real component, one can get more general insight than if always only using specific component models or breadboarding with real components. It can even be enlightening sometimes to simulate non realistic circuits to understand things. But of course, if the goal is to build a working prototype, one eventually has to build it, measure it, and most probably redesign it. There might be less of redesigning though, if one makes the right simulations first.
 
John,
I call the phenomenon Science as Religion (I'll get the Nobel Prize for the concept someday). The irony is intentional. People who treat science as Revealed Truth refuse to question things in a manner precisely analogous to the most fervent religious zealot refusing to face the problems inherent in his faith. And, exactly as one would expect, they condemn most bitterly those who question the underpinnings of their faith as heretics, unbelievers, and worshipers of false gods.
Not to mention purveyors of snake oil.
(Speaking of snakes, it's interesting to note the reputation of the Serpent in Christianity. The analogy goes deeper than one might first think.)
Science, in order to be valid, must be tested against reality or it literally becomes another religion. In our case, reality means the listening room, not the test bench. This is not some abstract theoretical exercise--it's an attempt to build a real system that recreates music. But, oh Gussie, the cries of anguish from those whose Revealed Truth is found wanting. Just like all whose deity is mysteriously AWOL when put to the test, they blame everything except their belief system.
Don't get me wrong--although I gave up on simulation programs, I still think THD is nice. In moderation. I've got an HP 339. I even use it on alternate leap years. (Gasp!) But I've had far too many experiences that flew in the face of the Theory Uber Alles religion to show up on Sundays.
Reality dictates Theory, not the other way around. When people put theory above reality, it's no longer science...it's religion.

Grey
 
Jeez, I should think it should be obvious--it's right there in the spec sheets. Devices behave differently when they get hot. But you can't simulate that because you can't know the ambient temperature in the room...
hence the efficiency of the heat transfer from the heatsink to the room...
hence the actual temperature of the heatsink...
hence the actual temperature of the output device...
hence the actual temperature of the semiconductor chip itself...
hence the actual, literal, real world behavior of the MOSFET's bias and transconductance in the practice of delivering music to the listener's ear.
For want of a nail, the kingdom fell.

Grey
 
andy_c said:


A quick browse indicated that the models are for low power. However, it is, no doubt, an incremental step forward.
I work with mainframe computers. At this point, meaning sixty or seventy years into the mainframe era, the technology is pretty mature. Your average mainframe works well into multiple "nines" of reliability. The system gets IPLed (more or less equivalent to rebooting a PC) every few months or so, whether it needs it or not.
Ah, but the network stuff is another matter entirely. I have a screen here above my head that lists problems on the network. At this moment, 21:52 on 5/9/07, there are 18 different things pending on the network here on campus. Reliable? Uh unh. Not even in the ballpark.
Yet the network guys are eternally patting themselves on the back about how great they are and how marvelous the network is.
Us mainframe guys? We're like the proverbial grandfather clock in a thunderstorm. We just keep on ticking. Reliable. Dependable. Always there.
No wonder anything important--student grades, accounting, dorm assignments, class roles, and so forth, gets run on the mainframe. Running that stuff on the network would be a recipe for disaster, no matter how cute the flashy graphics look on a PC.
There is no doubt in my mind that electronic simulation programs will someday be useful. At this point, though, it's a woefully immature technology and I question its utility for audio work. Like the network, they look good and give (false) confidence. Also like the network, they crash inexplicably and without notice.
Meanwhile the old-fashioned mainframe just keeps chugging along, doing the heavy lifting. Ain't sexy, ain't cool, ain't much loved, except by those who want to get the job done.
Kinda like using a soldering iron and real parts. It may not be the modern thing to do, but it works. And that's pretty cool in my book. I'll take results over flash any day.

Grey
 
Christer said:


....Spice can also be useful to test concepts and compare topologies with ideal components. The latter may sound useless to some, but I think otherwise. By using ideal models, or models which are elaborate but not corresponding to a real component, one can get more general insight than if always only using specific component models or breadboarding with real components. It can even be enlightening sometimes to simulate non realistic circuits to understand things. .......


One of the most important uses I find for simulators, apart from sparing charred silicon after some silly mistake. :up:

Rodolfo
 
AX tech editor
Joined 2002
Paid Member
GRollins said:
Jeez, I should think it should be obvious--it's right there in the spec sheets. Devices behave differently when they get hot. But you can't simulate that because you can't know the ambient temperature in the room...
hence the efficiency of the heat transfer from the heatsink to the room...
hence the actual temperature of the heatsink...
hence the actual temperature of the output device...
hence the actual temperature of the semiconductor chip itself...
hence the actual, literal, real world behavior of the MOSFET's bias and transconductance in the practice of delivering music to the listener's ear.
For want of a nail, the kingdom fell.

Grey

Actually, Grey, all of these things can and have been simulated. IsSpice has published several newsletters in the past addressing these kinds of issues. Guess you'll have your kingdom after all ;) .

Jan Didden
 
Sorry for yet another Spice post, but I cannot silently stand by when people post the most surprising nonsense about simulations and computers.


Greg, I don't know if your lengthy discussion of mainframes vs. networks was intended to say anything about simulations. But if so, I cannot see the relevance of network reliability issues for a discussion on simulations, unless you are doing distributed simulations over a network, which probably none of us are doing.


GRollins said:

There is no doubt in my mind that electronic simulation programs will someday be useful. At this point, though, it's a woefully immature technology and I question its utility for audio work.

There is no doubt in my mind that that that day did come long ago, and you just missed it, since Spice has been used regularly by the industry for a very long time. Except for IC design maybe, simulations do usually not tell the whole story, but they are undoubtedly ver useful. Do note that Spice is an electronic circuits simulation program. It is not intended for modelling spatial issues like component placement or heatsinks. However, with a lot of effort you could probably make an electrical model of your PCB and a model for the electrical effects of the thermal arrangement (heatsinks, component placement etc). However, it is most likely not worth the trouble to do so for a specifik design. Neither is Spice intended to simulate psychoacoustics, which would be necessary to giva an accurate simulation of what you will think an amplifier sounds like. To simulate what an amplifier sounds like would also require simulating the speakers and the room acoustics, also something Spice was never intended for. So please, do not call Spice useless just because it doesn't wash up your dishes or move your lawn or any other thing it never was intended to do. Also do not accuse Spice for the errors that the users are responsible for.


Like the network, they look good and give (false) confidence. Also like the network, they crash inexplicably and without notice.
Meanwhile the old-fashioned mainframe just keeps chugging along, doing the heavy lifting.


Sorry, but I am afraid you do sound quite confused. Trying to compare network reliability with simulation results seem like a very strange and useless analogy. And what is it that crashes like the network and in contrast to the mainframe? The simulation program? Not the one I am using at least.

Is it that you are trying to compare network technology, simulation programs and computers (but only mainframes)? A very strange comparison indeed, whichs leads me fear you don't have a clue what you are actually talking about.
 

GK

Disabled Account
Joined 2006
Christer said:
Sorry for yet another Spice post, but I cannot silently stand by when people post the most surprising nonsense about simulations and computers.

:rofl:
I had a similar reaction, but I only broke from silence when my jaw hit the table.

Your response is most reassuring. Some times I question wether it's me that's barking mad, or if it’s everyone else around me.
 
Bark, Bark! SPICE is OK, but you should know its limitations and that it can sometimes be misleading.
When I was young, SPICE was only a concept, but we had specific electronic analysis programs that worked on mainframe computers, mechanical calculators, and slide rules. When we wanted to evaluate a design, we built it. I used to build them for others, at first, then I finally got to build my own designs. This was the way of the engineer of 35 years ago and more. Like the proverbial: "When I was young, I had to walk to school." often said by parents to childern, when I was young we had to actually build our own circuits in order to see how well they work, and we got used to it.
You, newbe engineers, are taught in school how to use Spice, and you can't see a world without it. You also tend to rely on it, perhaps more than is completely wise in all situations, but there is nothing wrong with using it, if you wish.
This question came up when Bob Cordell apparently was surprised that I had not made a SPICE emulation of the JC-1 power amp. I told him (in so many words) that it was not really necessary, since Parasound did it the hard way, we built it and tested it.
The SPICE scare stories are to point out that no emulation is perfect, and to keep that in mind when designing.