I learned my electronics when there was no such things as computers, or they were in their infancy. I still have some parts from an old Univac that was being scrapped by a local surplus dealer. As far as audio amplifiers go, they're pretty simple stuff. I have a good feel for what will and won't work, so I just fly by logical deduction and the seat of my pants. Computer simulations are for sissies. 
I simply build it, plug it in, let it rip and then refine it.
Now this only applies to my DIY home audio stuff. Were I designing more complex circuits, computer simulations certainly have their place. Also since reading this forum and the knowledge gained here, and other places on the web, I've learned a few new tricks from others smarter then me.
Victor
I simply build it, plug it in, let it rip and then refine it. Now this only applies to my DIY home audio stuff. Were I designing more complex circuits, computer simulations certainly have their place. Also since reading this forum and the knowledge gained here, and other places on the web, I've learned a few new tricks from others smarter then me.
Victor
Spice can be a great tool.... How well the results compare to bench measurements depends on the quality and depth of the models you use...and how much parasistics you add in...
On my job we use Spice for all Worst-Case-Analysis and it "better" match the lab data pretty accurately or else you will be looking for a new job.... Some of the models can take weeks to months to model.... Some of the circuits are very extensive and can take up to 6 months to get the full schematics and models fully running and looking good...
When I did IC design, we would first design the circuits after many months, then the schematics go over to the IC layout group...after layout and metal layers are finished, we would use parasistic extraction to pull all the parasitics from the layout and these would be put into the original schematics for more accuracy when running the simulations that would take sometimes 4 days to complete useing 2 servers....
I believe there are some descent tube models out there....
Or you can start writting your own models....
Keep in mind that real tube characteristics are all over the place and nothing in the model is a constant...every aspect of a tube has a non-linear variables ...for example gm, rp and mu are variables that can greatly vary depending on operating conditions...some models treat these as constants...which in that case, you can only operate the model fairly close to it's measured operating points for OK results for small signal analysis....
Also in real life one tube to the next will vary, so a model can represent an average tube.... the data sheet models and curves are full of BS..they cherry-picked the cream of the crop.. top 5% in order to make there curves look better than the competitor...
Chris
On my job we use Spice for all Worst-Case-Analysis and it "better" match the lab data pretty accurately or else you will be looking for a new job.... Some of the models can take weeks to months to model.... Some of the circuits are very extensive and can take up to 6 months to get the full schematics and models fully running and looking good...
When I did IC design, we would first design the circuits after many months, then the schematics go over to the IC layout group...after layout and metal layers are finished, we would use parasistic extraction to pull all the parasitics from the layout and these would be put into the original schematics for more accuracy when running the simulations that would take sometimes 4 days to complete useing 2 servers....
I believe there are some descent tube models out there....
Or you can start writting your own models....
Keep in mind that real tube characteristics are all over the place and nothing in the model is a constant...every aspect of a tube has a non-linear variables ...for example gm, rp and mu are variables that can greatly vary depending on operating conditions...some models treat these as constants...which in that case, you can only operate the model fairly close to it's measured operating points for OK results for small signal analysis....
Also in real life one tube to the next will vary, so a model can represent an average tube.... the data sheet models and curves are full of BS..they cherry-picked the cream of the crop.. top 5% in order to make there curves look better than the competitor...
Chris
I prefer to run my "parallel analog computer", though I'm getting exhausted when can't get asleep calculating something, especially when it hangs up having no data to get a desired result... so, it needs to be feed by an experimental data so sometimes I have to wake up during the night to get some curves in question. This is the single way to create something new; however you may use digital computer with software to mimic what people already have done before... and repeat their mistakes... However, it is my own humble opinion; some orthodox engineers may disagree...
I use Beige Bag B2Spice - why? because tube guru John Broskie was co-developer/author of it and has models for most of the tube things I want to use built in.
Its not freeware like some other modelling programs but it costs about half of what I spent this morning for 10 off Genelex Gold Lion re-issue KT88.
Having said that the only "real" model for a vacuum tube is a vaccuum tube. A mathematical expression can only ever be an approximation.
The modelling will generally pickup the "massive stuff ups" and will in general give you a guide as to the value or otherwise of minor tweaks.
In the end you just have to build and listen to it.
Cheers,
Ian
Its not freeware like some other modelling programs but it costs about half of what I spent this morning for 10 off Genelex Gold Lion re-issue KT88.
Having said that the only "real" model for a vacuum tube is a vaccuum tube. A mathematical expression can only ever be an approximation.
The modelling will generally pickup the "massive stuff ups" and will in general give you a guide as to the value or otherwise of minor tweaks.
In the end you just have to build and listen to it.
Cheers,
Ian
Thanks for all the interesting replies. Yes I think modelling has it's place, but I tend to spend more time struggling with the model and software limitations than I should.
For example, I was looking at a model this morning, and an SRPP 12AU7 setup was acting like a high pass filter, not an amp. Until I realised I was using models that included heater modelling and ltspice was doing the ac sweep analysis BEFORE THE TUBE HAD HEATED UP !
Also, I suspect many models do not calculate miller capacitance and other important stuff like that.
For example, I was looking at a model this morning, and an SRPP 12AU7 setup was acting like a high pass filter, not an amp. Until I realised I was using models that included heater modelling and ltspice was doing the ac sweep analysis BEFORE THE TUBE HAD HEATED UP !
Also, I suspect many models do not calculate miller capacitance and other important stuff like that.
There is no substitute for real glowing tubes, but I use LTspice in order to keep the glowing from turning into a full scale fireworks show! If I am designing a straight forward vacuum tube circuit, I usually slap some tubes on the breadbord and experiment until it works. For some of my more unusual ideas, I use LT spice to weed out the stupid circuits that don't stand a chance of working. It also allows me to get a design closer to reality before building it. My augmented cathode follower amplifiers and other high efficiency designs would not have ever worked without a lot of simulation up front. I have 23 versions thet blew up on the simulator before I got close enough to start building. I still fried a few parts.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.