Hi ! sorry for the very generic question in the title. I wonder if The Lounge or Everything Else coud be a better place to ask.
When you design a circuit do you simulate it first or not ?
and if so, is there a fair correlation between simulation performance and then actual performance measured on a prototype ?
Thank you all and have a nice day, gino
When you design a circuit do you simulate it first or not ?
and if so, is there a fair correlation between simulation performance and then actual performance measured on a prototype ?
Thank you all and have a nice day, gino
Yes and no.
Yes: I try hard to understand the mechanisms that make a given design work, both on a qualitative and quantitative level. I do back-of-the envelope calculations. I sometimes write computer code that models (parts) of a system. I try to get a feeling for the uncertainties in these models. I sometimes test my models by making real-world implementations of sub-circuits.
No: I don't use ready-made simulation tools like SPICE. They are black-boxes to me, and it kills the thinking part for me, which is equivalent to taking the fun away. I don't have a lot of experience with SPICE and the like, but from what I read and hear these tools have a reputation for not giving accurate results. I don't see how/why I would like or need SPICE et al for what I do.
Research/Understand > Simulate > Understand it's back to the drawing board > Repeat
Then eventually > Works in simulator > build.
It's cheaper and safer to make an oops in the simulator however the simulator isn't entirely accurate at the end of the day due to variances.
From the day job two things have always stood out:
* Paralysis by analysis
* 5Ps - Planning prevents pish-poor performance. Planning, but knowing planning is going to change whilst you're doing and planning for that too.
Then eventually > Works in simulator > build.
It's cheaper and safer to make an oops in the simulator however the simulator isn't entirely accurate at the end of the day due to variances.
From the day job two things have always stood out:
* Paralysis by analysis
* 5Ps - Planning prevents pish-poor performance. Planning, but knowing planning is going to change whilst you're doing and planning for that too.
Last edited:
Simulation (I use LTSpice) is absolutely essential to me when designing DAC passive filters. One reason being - online filter calculators - like this one - assume inductors and capacitors have infinite Q. LTSpice allows me to put a real-world ESR into the inductors and see the effect of finite Q.
I grew up with breadboards and TDEC's and so experienced the world of magic smoke and what would stand a chance of working and what wouldn't. Designs were refined for real on the fly.
Simulation came along and is a wonderful tool but it doesn't teach the above... you make a mistake and the amplifier works with 40kW dissipation in the output stage... and you don't experience the error for real.
Somethings unstable... but nothing goes up in smoke.
I'm amused when simulations are refined to the point of seeing folk stressing over whether the 0.05pF cap should be here or there... get real
Same for wanting specific models of opamps and transistors... it's not going to make one jot of difference to what you actually build for 99.9% of what are in reality (when it comes to audio anyway) very low frequency circuits.
Simulation came along and is a wonderful tool but it doesn't teach the above... you make a mistake and the amplifier works with 40kW dissipation in the output stage... and you don't experience the error for real.
Somethings unstable... but nothing goes up in smoke.
I'm amused when simulations are refined to the point of seeing folk stressing over whether the 0.05pF cap should be here or there... get real
Same for wanting specific models of opamps and transistors... it's not going to make one jot of difference to what you actually build for 99.9% of what are in reality (when it comes to audio anyway) very low frequency circuits.
Hi,
FWIW: I'm working on my first project from scratch and never had been more happier than sim with Xsim or Vitruix. I like a lot Xsim for its casual use. helped with some other softs like Jeff Bagbys's (RIP) .
Thanks to the good advices of experienced people here I followed, I found it very simple to play with. You can see immediately what are the changes in a crossover for a learning purpose as a real project if you import or find the impedance and spl curves (.zma & .frd). Of course, it will always be better to sim the real driver in the real cabinet for the xover : good news, you can do it with a lot of good soft.
So as far I'm concerned whoever I'm a beginner : yes it will help you and avoid some huge errors and wins you a lot of time. imho it helps to make the link between the theory and the practice.
FWIW: I'm working on my first project from scratch and never had been more happier than sim with Xsim or Vitruix. I like a lot Xsim for its casual use. helped with some other softs like Jeff Bagbys's (RIP) .
Thanks to the good advices of experienced people here I followed, I found it very simple to play with. You can see immediately what are the changes in a crossover for a learning purpose as a real project if you import or find the impedance and spl curves (.zma & .frd). Of course, it will always be better to sim the real driver in the real cabinet for the xover : good news, you can do it with a lot of good soft.
So as far I'm concerned whoever I'm a beginner : yes it will help you and avoid some huge errors and wins you a lot of time. imho it helps to make the link between the theory and the practice.
Last edited:
Hi, Gino!
Excellent example with simply stupid question and very longread answer.
First of - yes, we do simulate. Modern simulators are very precise in terms of prediction performance of the circuit in the audioband.
We do simulate of all bias points and stage working mode, open loop gain simulation aka stability and margins estimation, noise simulation and output thermal dissipation simulation.
But there are a huge trap.
While we reach more than 120 dB feedback depth through all audioband, THD more than 130 dB, CMRR at least of 100 dB, lowest possible noise and concrete wall stability it's all seems that we reach the sky in terms of audio reproduction quality.
But this is a real mistake.
Such an amplifier still have a huge dependance of sound reproduction with components used. Even change of main capacitors have easily audible influence to the sounding. More warmth or more fatless, more precise or more engaging, more audible bass or more audible hiss.
So yes, we do simulate technical aspects of the amplifier, but then we do listen obtainable different parts and really consult our customers with deep tuning for their expextations in the real systems in real listening rooms.
Excellent example with simply stupid question and very longread answer.
First of - yes, we do simulate. Modern simulators are very precise in terms of prediction performance of the circuit in the audioband.
We do simulate of all bias points and stage working mode, open loop gain simulation aka stability and margins estimation, noise simulation and output thermal dissipation simulation.
But there are a huge trap.
While we reach more than 120 dB feedback depth through all audioband, THD more than 130 dB, CMRR at least of 100 dB, lowest possible noise and concrete wall stability it's all seems that we reach the sky in terms of audio reproduction quality.
But this is a real mistake.
Such an amplifier still have a huge dependance of sound reproduction with components used. Even change of main capacitors have easily audible influence to the sounding. More warmth or more fatless, more precise or more engaging, more audible bass or more audible hiss.
So yes, we do simulate technical aspects of the amplifier, but then we do listen obtainable different parts and really consult our customers with deep tuning for their expextations in the real systems in real listening rooms.
Hi ! thanks for the advice and i understand now that maybe this is a question for the Lounge ... this is more technical section.
I would like to migrate there ... but i have got some very valuable advice already here.
I will answer there
Please close this thread and excuse me again
Kind regards,
gino
Hi ! i thought a lot and i think that such a generic question is better posted here
When you design a circuit do you simulate it first or not ?
and if so, is there a fair correlation between simulation performance and then actual performance measured on a prototype ?
Thank you all and have a nice day, gino
When you design a circuit do you simulate it first or not ?
and if so, is there a fair correlation between simulation performance and then actual performance measured on a prototype ?
Thank you all and have a nice day, gino
When I started "designing" electronic circuits I took whatever I could scavenge from discarded 50's vintage TV's, radios and HiFi sets, and strung them together on a pine board soldered to brass upholstery tacks. The probability of success was very low, but I kept at it.
Today some of my dumb ideas go to a perf board prototype, or other kinds of breadboard, but more often than not the idea gets "built" in LT spice. I come from a 41 year electronics engineering career. Simulation came to low level engineers in the 80's, but became really useful in the 2000's.
Simulating a circuit will only work out great if you have good models for all of your parts. Modern parts generally have good models, often provided by the manufacturer. Vacuum tube models however, vary from excellent, to useless. In this case a simulation will usually tell you if your idea will NOT work, but it may not accurately predict success, and may be less likely to predict things like the distortion percentage.
If a tube circuit doesn't work, try some different models for your tubes, or try some different tubes. This is where the biggest errors lie in vacuum tube circuits.
Today some of my dumb ideas go to a perf board prototype, or other kinds of breadboard, but more often than not the idea gets "built" in LT spice. I come from a 41 year electronics engineering career. Simulation came to low level engineers in the 80's, but became really useful in the 2000's.
Simulating a circuit will only work out great if you have good models for all of your parts. Modern parts generally have good models, often provided by the manufacturer. Vacuum tube models however, vary from excellent, to useless. In this case a simulation will usually tell you if your idea will NOT work, but it may not accurately predict success, and may be less likely to predict things like the distortion percentage.
If a tube circuit doesn't work, try some different models for your tubes, or try some different tubes. This is where the biggest errors lie in vacuum tube circuits.
Hi ! i understand my question is quite trivial but any design is based on calculation.
To have a SW that does the calculation for us is too handy not to use it.
My guess is that main issues with simulation could be:
1) availability of models that reproduces well the behaviour of the part used
2) lay-out variable that is not considered by the sim software
I want to be honest ... i am awaiting for the first negative answer ... like ... no i do not sim. Then what ?
Hi ! so if parts are available you calculate your circuits in which way ?
Hi thank you for the confirmation. Actually i am starting to try some very basic circuits ... just to see how different is real from ideal.
Hi ! you say understand the mechanism. If i look at let's say a tube datasheet i see figures and plots describing the behaviour of the device. A device is defined by its electrical behaviour under different working conditions.
I think that simulation is a god's gift. Unfortunately the models used maybe do not match exactly the device behaviour. And there are other variables like for instance lay-out but not for all devices.
I would use a sim to rough-cut the design and then try some adjustement by hardly by ear. I mean starting with sim and ending with scopes.
Hi ! you confirm my thinking. They are indeed great tools for the designers. Of course models are not always perfect or even available. But they are almost essential to me.
Hi sorry i have moved the question to a more proper section
https://www.diyaudio.com/forums/the-lounge/363690-simulation-circuit-design.html#post6429335
Please excuse me again
i will answer You there
Thanks a lot, gino
Moderation edit
Threads have been merged and moved back to 'Software'.
https://www.diyaudio.com/forums/the-lounge/363690-simulation-circuit-design.html#post6429335
Please excuse me again
i will answer You there
Thanks a lot, gino
Moderation edit Threads have been merged and moved back to 'Software'.
Last edited by a moderator:
Hi ! may i ask what TDEC is ? my question was not provocative at all. I just would like to have some confirmations to my opinion. Let's take the process of select the working points for a tube. A sw that can help us to find an optimal one is very handy. I am interested in very simple filters for psu ... passive and active. LTSpice can provide very useful results and makes very easy to check the impact of different parts values of the psrr of the circuits. Very very convenient. I have to try a prototype now ... but i am confident that i will get very nice results.I grew up with breadboards and TDEC's and so experienced the world of magic smoke and what would stand a chance of working and what wouldn't. Designs were refined for real on the fly. Simulation came along and is a wonderful tool but it doesn't teach the above... you make a mistake and the amplifier works with 40kW dissipation in the output stage... and you don't experience the error for real.
Somethings unstable... but nothing goes up in smoke.
I'm amused when simulations are refined to the point of seeing folk stressing over whether the 0.05pF cap should be here or there... get real
Same for wanting specific models of opamps and transistors ... it's not going to make one jot of difference to what you actually build for 99.9% of what are in reality (when it comes to audio anyway) very low frequency circuits
I completely agree that any tool must be used with a sane approach. So you mean that is possible to have a good result from a sim and yet end up with smoke and flames ? i mean an unstable design ? i am asking because i have already ended with burning my fingers on a heatsink for modding a unit without really knowing what i was doing.
Hi ! thanks a lot for the very valuable advice. So another
for sim sw. I am about to start myself and i would like to start with the right foot. Hi ! thank you very much for the very valuable advice. I can understand ... the better the model the better the simulation. Anyway i think that speaking of tube measuring it can lead to using a model of another tube with similar parameters ? more or less of course. My point is that sim sw when used rightly are a godsend They help a lot indeed.
And yes the availabilty of a reliable model is of paramount importance for the final result.
Hi ! may i ask you which way has provided the best outcome ?
Last edited:
I find simulation handy for a circuit I am not sure about.
I designed a transistor curve tracer that had loads of differential amps in it.
I used simulation to check they functioned as expected.
I also did a simulation of a 12 volt valve pre amp to make sure valve worked at 12 volts B+. The simulation showed it working.
When I built up my pcb I found with some valves it didn't work. I had to put 470K resistor from grid to B+ to get it to conduct.
So sims are good for some things but do sometimes get caught out.
I designed a transistor curve tracer that had loads of differential amps in it.
I used simulation to check they functioned as expected.
I also did a simulation of a 12 volt valve pre amp to make sure valve worked at 12 volts B+. The simulation showed it working.
When I built up my pcb I found with some valves it didn't work. I had to put 470K resistor from grid to B+ to get it to conduct.
So sims are good for some things but do sometimes get caught out.
Hi ! thank you very much indeed for the very valuable reply. Do you have any opinion about why the sim gave a wrong outcome ?
Speaking of tubes if i understand well a model contains just the tube parameters and the curves describing its behaviour under different working conditions ?
The sim will just do automatically what otherwise must be done with a calculator and a ruler by the designer ?
Of course the availability of a reliable model of the tube is of paramount importance to get reliable results.
If i wanted to use let's say a nos part of which a model is not available i would try to make a model of that tube I guess is a long and exhausting process of measurement ... but after that designing could be a breeze ?
Last edited:
Thank you very much again. You have pushed the envelope too much ?
Sorry for my silly question ... is it a little like with engines when you increase too much for instance the pressure of the turbo ... you get indeed a huge power but the engine will get its life shortened and it could also break ?
So in the end it seems to me that the sim itself was just ok ... the input was not ok. Great. I see.
I definitely have to find a good book on LTSpice
... for dummies of course. - Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.