The same quest here, I am using the older generation Toshiba devices because it has proliferated the e-parts market in this part of the world [genuine or not I really can't tell but they seem to work ok].They can easily be had and comes cheaper but the challenge is putting them in simulation, I am totally lost in choosing which models out there speaks the truth...
...wish I had "brains" from the likes of Maestro keantoken so I can verify and or create my own models 
I am for sure not the type who is attached from the past
I also build Class D amps...oh well this audio thing...
Last edited:
The same quest here, I am using the older generation Toshiba devices because it has proliferated the e-parts market in this part of the world [genuine or not I really can't tell but they seem to work ok].They can easily be had and comes cheaper but the challenge is putting them in simulation, I am totally lost in choosing which models out there speaks the truth...
Well, hopefully with this ongoing work to get one model working properly, we can learn enough to get more of them made. Having as much "how-to" as possible gathered in the discussions, we should be able to figure this out.
After all, there are quite a few who already have been doing this, and for the rest of us who aren't as comfortable with it, the sharing should make this possible.
I too don't have the brains any more, and so many times I have to re-learn things, because I forget over time, so it's good to have places where to find the info again.
I am for sure not the type who is attached from the past
Same, here. I have other projects not involving any older stuff, but there is fun to have with that old stuff as well. Plus as mentioned, there are places on this blue planet where the older parts are cheaper and easier to get.
By the way here is the link for the digitizer app I was mentioning about.[first introduced by andy_c I think]...hope it may encourage some members to also do spice modelling Engauge Digitizer Member Than's feedback keeps reverberating in my head "work on models" ...quite true indeed Spice is all about the right directives and device models is all about the valid parameters.
Engauge Digitizer Member Than's feedback keeps reverberating in my head "work on models" ...quite true indeed Spice is all about the right directives and device models is all about the valid parameters. By the way here is the link for the digitizer app I was mentioning about.[first introduced by andy_c I think]...hope it may encourage some members to also do spice modelling
I had never heard of that one before, and it does have a mac version, so I'm definitely in and will try it out. I just grabbed it from the mac app store and will give it a whirl shortly.
Thanks for posting the info, when mentioning something like this, it's always much better with more info and links.
Datasheet charts are often shared between manufacturers and re-digitized, so it is best to try and find the original datasheets and use those, otherwise you may be trying to model distorted curves. Even then, the curves may be from an earlier version of the device which doesn't match the current version. That is why I hesitate to make models when I don't have direct measurements from a current device. Some charts are even averaged over multiple sample devices which prevents them from representing any single device or even being possible to model.
Last edited:
It's not always so easy, especially for older parts, to get a hold of the original datasheets, and what we find isn't always from a known valid source. We just have to be content with what we find. No choice there.
Measuring actual device is very good to do, when possible, but also not so easy and not always possible, especially when we don't even have the parts to measure to begin with. Or perhaps we have very few parts to measure, and too few isn't so representative to get a good idea of the range of characteristics.
The measuring equipment at hand may also be quite limited, or non existent.
I have that transistor meter that can check the second breakdown voltage, which was the main reason why I got that one, and it does Vcesat and Hfe, but that's just a plain digital meter giving numbers and not curves, so this is quite limited and I must say I always have doubts about those simple readings anyway.
We all have different means to do things, and unfortunately, some of us have far less than others.
I wish I had good lab equipment, like power supplies, meters, frequency meters, scopes, signal generators, distortion meters, etc... But can't afford any of that. And I'm obviously far from being alone in that respect.
So I do the best I can with the limited resources at hand.
Having the possibility to do so much virtually nowadays, really helps a lot, so we can do far more than before when computers didn't even exist or were too primitive.
I'm just trying to do as much as I can with what I have. This should suffice for most things though.
Making models would be better done with more stuff, but we can manage to get something with what we do have.
I must report that unfortunately that engauge software gizmo isn't stable enough on the mac.
I tried it on a chart and it crashes early in the process, at the same point apparently, because when it crashed, I tried again, and it crashed again at the same stage, so it's obviously buggy.
Maybe if it's actively maintained, the bugs may be fixed soon, I hope.
I tried it on a chart and it crashes early in the process, at the same point apparently, because when it crashed, I tried again, and it crashed again at the same stage, so it's obviously buggy.
Maybe if it's actively maintained, the bugs may be fixed soon, I hope.
In my tweaking efforts to fit the curves for Hfe/Ic to better resemble the chart on the datasheet, I find it difficult to come closer to it.
For example, I don't see any parameter that would allow control over the rolloff speed.
I brought the curve at one point fairly close to matching the datasheet, and at one end (lowest X value) it's somewhat in the ballpark, but at the other end, it's just too high and the curve isn't peaking where I need it to.
Looking at the 25C Hfe curve on the datasheet chart, that curve peaks at somewhere near 0.2-0.3amps, but I was only able to let it peak at somewhere a little below 0.1, which is outside of the range shown on the datasheet chart.
I make it trace further down on the X axis, but I'm looking at where the curve sits at 0.1amps, which is the axis on the datasheet chart.
What would be good to do, is have the lowest X axis value be in the ballpark, then the peaking maximum at roughly the same amount (Y axis) and position on the X axis, with possibly a reference point somewhere along the way (at 1A on this curve for example) to be close enough, and the other end on the X axis end up somewhere near it needs to be.
But this seems nearly impossible.
Q quick glance at the curves I was able to get make them seem like they kind of look like the ones on the datasheet chart, but it's really not quite so close.
For example, I don't see any parameter that would allow control over the rolloff speed.
I brought the curve at one point fairly close to matching the datasheet, and at one end (lowest X value) it's somewhat in the ballpark, but at the other end, it's just too high and the curve isn't peaking where I need it to.
Looking at the 25C Hfe curve on the datasheet chart, that curve peaks at somewhere near 0.2-0.3amps, but I was only able to let it peak at somewhere a little below 0.1, which is outside of the range shown on the datasheet chart.
I make it trace further down on the X axis, but I'm looking at where the curve sits at 0.1amps, which is the axis on the datasheet chart.
What would be good to do, is have the lowest X axis value be in the ballpark, then the peaking maximum at roughly the same amount (Y axis) and position on the X axis, with possibly a reference point somewhere along the way (at 1A on this curve for example) to be close enough, and the other end on the X axis end up somewhere near it needs to be.
But this seems nearly impossible.
Q quick glance at the curves I was able to get make them seem like they kind of look like the ones on the datasheet chart, but it's really not quite so close.
The power of sharing!!!
We're better off pooling our info and efforts.
But you do have one advantage compared to many of us diyers: more know how!
Diyers learn from each other and there are those, like you, who know more. Plus we have our revered gurus (Bob, Doug, Jan, Nelson, etc...) who really help us all a lot. Without all this, we wouldn't have much of a diy community. This community endures, despite some erosion over time. I remember when I started in the 70s, that I felt a little isolated, because in those days we had no internet and all that, so the sharing was local, and there were a few others with whom to share with. In those days, the biggest medium for the community to claim its existence was through magazines, but the real sharing was quite limited.
Over the years, to my dismay, I've seen that part of the community dwindle a lot, with disappearing magazines and far fewer local diyers. Now I don't even know anyone local to where I was during the 70-80s who does any electronics, and of all the magazines very few are left now.
I'm glad there is a vibrant community on the net, and I think it will keep going, but this is due mostly to the globalization via the net, which allows us, isolated diyers, to stay in touch and exchange. Otherwise in areas like where I am at the moment, there are so few left who are still into it.
Have you checked the Installing versions.txt that comes with it? It contains a few troubleshooting tips. For example in a windows OS it won't run without the 32x Visual C++ 2015 installed. I'm on x64 so no go for me...
andy_c seemed to be the only one who has the knowledge to use it for spice modelling. He posted some tips on its usage long time ago in this thread.Have you checked the Installing versions.txt that comes with it? It contains a few troubleshooting tips. For example in a windows OS it won't run without the 32x Visual C++ 2015 installed. I'm on x64 so no go for me...
I opened an issue ticket on the github site and got a response from Mark who maintains it.
He's thinking there is a build issue because he built it on osx 10.11.5 and I'm running it on 10.8.5.
It's supposed to be able to run on 10.6.6 and after, but it's hard to keep things running over so many revs over so much time.
It may be a very small issue, this type of thing usually is, but the cause needs to be identified first.
He'll fix it somehow, he seems serious about it. That's a good thing.
If I can get it working, I'll do the 3055 SOA and Hfe/Ic curves at first, and try tweaking some more.
Maybe you have some devices in mind for which you would like models, maybe we can handle those at the same time.
Keeping our efforts logged here should be good and help many others do the same thing.
That is good news, hope he will also create a version for x64 bit Win OS.
For the models, I am after the older Toshiba [TO-92M] and Sanyo devices [D1047/B817 and the likes]. I am actually using them I only guesstimate a model to emulate these devices during simulation.
Rght now I need to do revision on two pcb lay-out and on a new CADware [back to beginner mode
] another two design that needs modification in simulation, not much time ehh I did made a few sim runs following Maestro keantoken asc file. I could not plot a usable graph though I guess spice modelling needs more focus and more time. [at least for me ] This thread has started long years ago, but it seems only andy_c keantoken syn08 and Harry Dymond has made spice model contribution. [not to mention Bob Cordell's creation] I just wish a sticky reference to this thread should still be visible in the Solid State forum so It can attract curious minds and beginners and make the thread fast moving again.Sa muli,
Albert
I think I saw some mention of this somewhere. Not sure exactly, as this wasn't my focus at all, but I'm sure I did see a 64 bit mention.
First thing is a datasheet. If you have that, why not posting it here for starters?
If you have anything handy to make some measurements, that can be additional data or confirmation for the datasheet's data.
Then we can take the parameters one at a time and go through them. Hopefully learning something along the way, which is useful for any model and keep tweaking.
With a few test rigs like the one posted by keantoken, and others, we can see what we're doing and how close we get to get it right.
I've been looking at more model parameters, and I really don't see how we can get the needed data for those in the datasheets. More learning to do...
There are 3 similar parameters right now that I'm wondering about:
RB > ohmic base resistance
RE > ohmic emitter resistance
RC > ohmic collector resistance
They are basically the same thing. Just for each device leg.
They default to 0, and few models sometimes don't specify them leaving them at default.
For the 3055 model, looking at the ones I have, there is a good amount of disparity, sometimes several orders of magnitude. So how could all this be right?
Beats me!
Example: RE can go from some 350u to over 8.5m, including some at 0 with that parameter missing.
I don't see how that can be all for the same device. Makes no sense. There must be a correct value, even if a little fuzzy. It should be in the ballpark, a small value of course, but all should reflect this in about the same range.
And when we see parameter values specified with some 10 or more digits, that has to be something computer made and no human made. They used some software to generate that automatically, and no human double checked the data.
Last edited:
That's cos 'real life' 2n3055 vary like that. eg you will find ft specified from 800kHz to 3MHz from different reputable makers ... let alone the cowboys.
Those of us who were involved with commercial amps in da 70s used to believe Motorola had the fastest 3055s and RCA the most robust ... but even that was 'finger in the air' stuff.
If you are making an amp with 2n3055 today, your circuit better be happy with important parameters varying by at least an order of magnitude if not more.
8.5m sounds more realistic to be, but is small enough it might not matter. You can look at the Vbe plot and look at where Vbe becomes resistive, because the emitter resistance takes over and it becomes a straight line. You can calculate the resistance from that line. But this can be due to both Re and Rb, so more information is helpful. Rc can be adjusted when doing Vcesat curves.
Some of that big variance in parameters is also just due to crappy modelling by manufacturers.
Cheers,
Bob
We need at least 2 models for the 2N3055 as I have stated before, there was the
very early single diffused RCA (<1MHz Ft) device and the later Motorola (2.5MHz Ft)
triple diffused. I don't know if RCA ever made a higher Ft part or not.
Perhaps a few more are needed to cover all the years of production
from all the different manufacturers. People need to keep in mind that even within
a particular type there are wide process variations from batch to batch as indicated
by the tolerances on all the data sheet specs.
If you measure Vbe vs. Ib say with no collector supply so that Ic =0 and then with a
collector supply my guess is that adjusting RE for a match will get you the correct value.
There could be other effects but it might be close enough.
Once RE and RB are known RC could be adjusted to match Vce vs. Ic at high Ic. Not
sure if this is the best way - anyone?
It is possible that for Ic less than an amp or two RE and RC are not very important.