EMRFD Message Archive 13146
Message Date From Subject 13146 2016-09-15 07:11:50 Alan Ibbetson How far do you go with circuit simulations before soldering? I have a question more for those on here who put bread on the table from
electronics rather than just dabble in it as a hobby. How much use do
you make of simulators such as LTspice before committing to a prototype?
How many prototype iterations do you suffer before reaching the final
production PCB? And, crucially, how different are the component values
between what you had on the computer and what you finally put onto the
Yes, I realise this is a "how long is a piece of string" question, so
let's put down some SWAG parameters. The project should have more than
one stage, so not just (say) a preamp. Maybe an IF strip? Not as complex
as a complete receiver. Something in between. Oh, and probably not above
30MHz, so we are not into the realms of component lead lengths and suchlike.
The reason I ask is I am building my first attempt at a switchmode PSU.
I have a 2KW PFC pre-regulator regulator working, based on the LT1248
chip. I went straight from LTspice to KiCAD layout to mail-order PCB fab
to soldering iron and it "all just worked". My more experienced friend
is saying that such an approach for the DC/DC converter stage, again
"just" a single chip but with more complex magnetics than a PFC, is
certain to end in tears because things like feedback compensation never
work out in practice like they did in the computer due to circuit strays
and imperfect component models.
Do you use simulations or go straight from datasheets to prototypes?
73, Alan G3XAQ
13147 2016-09-15 08:10:39 Russell Shaw Re: How far do you go with circuit simulations before soldering? 13148 2016-09-15 09:31:02 kb1gmx Re: How far do you go with circuit simulations before soldering? the parameter used here in the office is time and material. The more time and material costthe more sims/models/EDA will be used to get validation of a design.For something simple like a RF amp it may be a few minutes with paper and a calculator.For a complete system it may take several tools and weeks of time with genesis, XFTD,and NEC. It also has to be "big enough" to justify time learning and becoming skilledat using the tool. It also requires the knowledge that physics or theory says what can't orcan if done right. Usually that is when time to go back and redesign or material lead timeare critical path items. An example is a 24 week project with key parts having 16-20weeklead times, you don't have 8weeks to design and simulate to a fine detail.Your friend hit it. If the sim has what I call fidelity, that is it captures the relevant dataand produces faithful result with correlation the eventual real thing great. Usually itrequires experience and skills to know when you have all the relevant data to get itright. Sometimes you have to breadboard/brassboard the tricky parts and do a bitof testing and use the resulting data to tune up the sim for a closer to real result.Generally the higher the power involved the more steps to insure the costly parts arenot destroyed on the first pass. Even then its to a point then move on.Off the office bench I use many of the same tools but since most deigns are modularits easier. Design is again paper and calculator and then build test as model to confirm.If not "in spec" find out why and fix. continue to next module. If the design uses what Ilike to call library elements that are plug and go then build and confirm function andmove on. After 4 decades in electronics that is a large library to pull from.Generally Sims/Models are useful but they can get on the way or be more the projectthan the project. If the parts are cheap and easy its sometimes better to blow ups fewand learn than ponder a dimes prediction it will work, or not. Along the same pathdata sheets can be misleading or miss critical detail for a given design. The morecomplex the part used the more likely the latter.Allison 13149 2016-09-15 09:33:42 g3xaq Re: How far do you go with circuit simulations before soldering? Hello Russell. I wonder, in your commercial environment, when given a new project do development engineers create a circuit simulation first or a hardware prototype?
I've read that some aerospace customers for example demand a sim model as part of a project documentation. I'm wondering if in this kind of environment the development process is simulation -> breadboard(s) -> production PCB -> documentation, or is it breadboards -> PCB -> simulation model (tweaked to reflect the PCB implementation) -> documentation?
73, Alan G3XAQ
13150 2016-09-15 16:35:11 Russell Shaw Re: How far do you go with circuit simulations before soldering? 13151 2016-09-15 18:43:20 Tayloe, Dan (Noki... Re: How far do you go with circuit simulations before soldering?
I like the simulation to “tweak” active filters values. The values given out of the calculators are odd values. At least for the low Q section, the R/C value can be moved around a bit and the simulation can show how much the filter shape “moves”. You quickly find out that shifting a higher Q (like “3”) section is not very practical, and that those stages need more precise values. The lower Q filter stages, not so much.
I can spend a lot of time on a filter design to tweak the cut-off frequency and/or the allowed ripple, and/or the stage gain so that the highest Q section comes out using standard values. IF the lower Q filters look close to a standard value (or maybe two standard values in series or parallel), that is when I go check it out in simulation.
13152 2016-09-16 04:05:07 Hans Summers Re: How far do you go with circuit simulations before soldering? HiPersonally it is rare for me to do much simulating, I just don't personally like it much, so far! I prefer to iteratively solder and measure. This applies to QRP Labs kits http://qrp-labs.com as well as other commercial projects I have done and am doing.The only exception is filters - LC and active - where I do find simulation is useful! The application circuits in most IC data sheets usually seem good enough as a guide.Mostly I manage with a single prototype build. Low volume prototype PCB's are very cheap now. Usually they either work perfectly first time or require small enough modifications that I am confident to go straight to the production batch.Sometimes I have even gone straight to manufacture without ANY prototype. If the circuit is simple and I controversial enough, and you trust the PCB CAD checks etc., and are brave (without crossing the very fine line to stupidity) - it can work! An example is the 50-ohm 20W dummy load kit http://qrp-labs.com/dummy . I got away with NO prototype build and it does lower the overall production cost.It should be mentioned that the same goes for firmware too... Development environments such as AVE Studio do include simulators but personally I have never used it. I prefer to iteratively write some code and test it, observe the result etc right there on the actual target hardware.I think everyone has a different development style they're working with. Lack of simulation works for me. But I know many people who spend 10x more time in front of a screen than melting solder. Undoubtedly the cost of the components, time pressure on the project, complexity of the product, etc etc are all factors that weigh in.73 Hans G0UPL AF7BF 13153 2016-09-16 06:48:34 AncelB Re: How far do you go with circuit simulations before soldering? Your skill with the simulator is a factor. If u include parasitics and
real SPICE models with consideration for environmental conditions you're
often good to go.
Ideal components don't consider trace routing, cross talk, EMI, voltage
drop, etc and depending on how close you are to the operating limits of
your components you can have a working bench prototype circuit that
'breaks' under environmental variations, temperature or noise coming
into the PSU from electrical tools nearby.
For commercial applications you must test for durability and sometimes
build in additional protections for foolish users.
13154 2016-09-16 07:19:34 Ashhar Farhan Re: How far do you go with circuit simulations before soldering?
As Wes says : Simulation is the greater experiment.
I have a competency gap in using LTSpice. Hence, except for mixers, I use simulations for every bulding block and frequntly, even simulation of an entire signal chain to evaluate noise figure and IMD.
If I have to choose just one reason for simulations, if cuts short a whole experiment set to answer a What-If question.
More importantly, even the simplest simulations show you in far excess what various components are doing without disturbing it. like measuring the feedback current in norton amplifier.
13155 2016-09-16 07:19:45 John Levreault Re: How far do you go with circuit simulations before soldering? 13156 2016-09-16 07:23:30 Chris Trask Re: How far do you go with circuit simulations before soldering? Myself, I'll use PSpice simulations to determine if a newly conceived topology is worth pursuing. If the initial simulation looks promising, I'll do a series of additional simulations to determine the robustness of the topology by changing devices (active and passive), bias currents, and anything else that would influence performance. If the topology remains valid, then I'll go to the prototyping stage.
>Your skill with the simulator is a factor. If u include parasitics and
>real SPICE models with consideration for environmental conditions you're
>often good to go.
>Ideal components don't consider trace routing, cross talk, EMI, voltage
>drop, etc and depending on how close you are to the operating limits of
>your components you can have a working bench prototype circuit that
>'breaks' under environmental variations, temperature or noise coming
>into the PSU from electrical tools nearby.
>For commercial applications you must test for durability and sometimes
>build in additional protections for foolish users.
N7ZWY / WDX3HLB
Senior Member IEEE
13157 2016-09-16 07:52:58 Dave Daniel Re: How far do you go with circuit simulations before soldering? Linear Technology (now Analog Devices) also has a simulation tool specifically intended for simulating SMPSs. It is called LTPowerCAD.