I'm reading the datasheet. It does have some pluses, and also some minuses, including some of the minuses in common with that MCP chip, which is the tiny tiny tiny package, impossible to solder by below average humans (me).
One plus is it was meant to handle a load, but one minus for my application is its use of pulses. Although it may not be a switch mode design, it still does make use of pulses and I'm wondering how much of that can cause parasitic issues for nearby small signal circuits.
One other feature that I see as a minus is it's a shunt regulator, which isn't the best example of power saving design. Plus it's definitely not a constant-current/constant-voltage charging scheme, which I find a bit "violent" for what I have in mind. But it's worth considering, and since it's an LT chip, the model is right there in ltspice, so I will definitely give it a whirl and see how it fares.
No worries there, it was meant to be for a rechargeable type from the get go, so it's going to be a LIR2032, and I was only planning for a charge current in the 14-15mA range, nothing more.
The load is going to be very light, just a few microamps, and I will choose parts to minimize that load to keep the battery charge going for as long as possible, several months at the very least.
Both much too tiny for me, and I wouldn't call the DFN type a leaded package, it's much too tiny for me to tackle.
Sure they are, for the consumer market, manufactured in well equipped factories, but not for the common diyer.
Yes, those are more or less what I had in mind, but not the beefiest. Most are too much for my needs. I don't need to do any fast charge and the charge current needs to remain small.
One bad thing about the shunt regulator is when the battery is fully charged, the same amount of current continues being drawn, which is a serious power waste (and some heat needlessly generated).
It might. I will run some sims, since I don't have to wrestle some models to make it work, I should be able to get a sense of it fairly quickly.
The load that I have in mind for the battery is tiny, a small fraction of its charging current, much less than 1mA in the worst case, which is good to make the battery last long when there is no power to charge it.
Yes, interesting stuff, and it's pretty much what I've been gathering lately. I think I'm up to speed on the charging requirements for that type of cell. The constant-current/constant-voltage scheme with the preconditioning trickle is rather gentle on the batteries, if we choose a low enough charge current. And the charge termination is also a nice feature, which with a good charger chip, should bring down power usage to a bare minimum. We don't get that from a shunt regulator though.
I'll be simulating...
I have been going through all of the part retailers and manufacturer's device selectors, looking for that 5 legged sheep, to no avail yet.
Most of them only offer those tiny smt chips that I can't tackle, so by lack of DIP case available, I was entertaining the idea of having someone else solder the tiny chip on a little rig made of a DIP socket and pcb, to turn them back into a DIP package. I just can't touch those tiny things, not any more.
I came across something useful to understand more of the spice models.
I had not seen info like this posted before, so I'm sharing it. It may help many understand a little more how this works.
Although this is aimed at pspice specifically, I would think it also applies to ltspice models.
I've been looking for more reliable and confirmed good models for the 2N3055 and MJ2955, and what I found so far doesn't fare very well.
Some simulations deal fine with poor models, others completely crap out, and I have been dealing with a leach super amp simulation lately that I'm trying to get working with 2N3055/MJ2955, and none of the models I have allowed it to work properly.
I have simplified models from ST, which I put on a test rig and I can see it's way overly simplified. No wonder the sims can't deal with that.
I've used models from onsemi made by modpex, and those aren't very good either.
Plus there is always some confusion with the 2N3055 models, with some supposed to be for the 2N3055H, which did exist long ago, but now those models seem identical to the regular 3055. There is a 2N3055A, slightly different, and still not working out so well either.
Some are using the models for the TIP/MJE flavors, and they're somewhat similar, but not so good as well.
So I'm looking for more reliable models for the 2N3055 and MJ2955.
I had not seen info like this posted before, so I'm sharing it. It may help many understand a little more how this works.
Although this is aimed at pspice specifically, I would think it also applies to ltspice models.
I've been looking for more reliable and confirmed good models for the 2N3055 and MJ2955, and what I found so far doesn't fare very well.
Some simulations deal fine with poor models, others completely crap out, and I have been dealing with a leach super amp simulation lately that I'm trying to get working with 2N3055/MJ2955, and none of the models I have allowed it to work properly.
I have simplified models from ST, which I put on a test rig and I can see it's way overly simplified. No wonder the sims can't deal with that.
I've used models from onsemi made by modpex, and those aren't very good either.
Plus there is always some confusion with the 2N3055 models, with some supposed to be for the 2N3055H, which did exist long ago, but now those models seem identical to the regular 3055. There is a 2N3055A, slightly different, and still not working out so well either.
Some are using the models for the TIP/MJE flavors, and they're somewhat similar, but not so good as well.
So I'm looking for more reliable models for the 2N3055 and MJ2955.
Anything that gets closer to reality would be good.
Would you share your improved models please?
Since I had so much trouble with those models when trying to simulate a super leach using 3055/2955 sets, I switched to using the MJ15015/16 models instead of the 3055/2955, but using AKO: and keeping them called 3055/2955 in the sims. Hoping to find better models to get this working.
That super leach won't simulate right even using those MJ models, but those are also from onsemi modpex, so I guess they can't be trusted either.
Anyway, the super leach wouldn't fully simulate right even with the MJ15003/4 models, but there again, they came from onsemi modpex, so it figures...
This model is the one from modpex as is, and it is identical to the one I have, so it's un-tweaked.
Have you only tweaked BF ? By how much?
But I suspect there is far more to tweak than just hfe.
Last lot tested (2N3055G) had BF=78...110. Was a lot bought from China, very cheap but also low quality.
You are right, probably we should redraw the operational curves of each transistor in order to build real models but I didn't find any reason to do this. I use Multisim only to check how the circuit is working, as a preliminary test before build.
You are right, probably we should redraw the operational curves of each transistor in order to build real models but I didn't find any reason to do this. I use Multisim only to check how the circuit is working, as a preliminary test before build.
But this info can help, along with the datasheet data, to come up with typical/average values that can be expected from parts in the wild, and we can plug this into the models to tweak them to better reflect the real thing.
There are many parameters though, so it's not that easy. The gain is an easy one to tweak.
I was just comparing all my models, and found many are the same, so more to be eliminated and the few left have often very different parameters, which could either be fine or way off. This is the kind of things we could focus on tweaking, and aim for a better, more reliable model, typical of average parts, working properly and share all this here.
The 3055/2955 models could of course also be shared on the dedicated thread:
http://www.diyaudio.com/forums/solid-state/5493-pspice-models-2955-3055-a.html
No seriously good model was unfortunately posted there yet, but we could remedy that.
I have several 3055/2955 sims that need good models. I'd like to get those working well.
I never used multisim, and I guess couldn't because it's not available for mac, but whatever models are used there should also be usable with ltspice.
If your models aren't good enough, then your sims could be way off and negate the advantage of simulating before building. Good models are a must.
If enough people collaborate on tweaking models, we can build up a library of known reliable models. Many people can corroborate the models' validity and confirm them, which can reinforce their validity status.
I made a few test rigs, which I could post, along with a few results, and perhaps some would have suggestions to make those rigs better and make more to test other parameters.
It's a long painstaking process, but a little at a time it can be done.
My test rigs have already revealed how some models are totally inadequate, at least on some parameters, and those models can be either fixed or avoided in sims.
With data from many sources, we can come up with a consensus on what each parameter should be.
Some users may even want more than one model for their sims, with best/typical/worst values...
I would settle for typical myself and keep it simple (K.I.S.S)
Is there a particular reason you are using the 2N3055 and MJ2955? Even the MJL2093/4 would be far better.
Cheers,
Bob
I cut down my 2n3055 models to only 4 now. 2 of them being from ST, the other 2 are modpex 2004 from onsemi, and one of the 2 from ST is really bad, with too few parameters, overly simplified.
As an example of the disparities, I'll compare what they have for early voltage parameter, VAF:
The ST model that is overly simplified doesn't even have that parameter, so that's quite bad already, and the other has VAF=90.
The other 2 from modpex have VAF=49.3238 for a 2n3055a, and the other has VAF=31.1252, for what was supposed to be a 2n3055h, but it's actually just their current 2n3055g, because the models are the same.
That's already quite a spread, for what's supposed to be the same part.
And it seems none of those values are right to begin with anyway.
As an example of the disparities, I'll compare what they have for early voltage parameter, VAF:
The ST model that is overly simplified doesn't even have that parameter, so that's quite bad already, and the other has VAF=90.
The other 2 from modpex have VAF=49.3238 for a 2n3055a, and the other has VAF=31.1252, for what was supposed to be a 2n3055h, but it's actually just their current 2n3055g, because the models are the same.
That's already quite a spread, for what's supposed to be the same part.
And it seems none of those values are right to begin with anyway.
Hi Bob, thanks for droping in
This is for 3055 based amps, or 3055/2955. There are active threads and many hobbyists interested in building such amps. Myself included.
There are several projects that are a little different from the ordinary, such as grounded bridge (crown like), leach super amps, qsc grounded collector, and even bryston type output stage config to test out.
I'm not the only one having an old stock of good parts, and some people in some parts of the world don't have easy access to certain parts, but have lots of the old venerable 3055s.
So there is plenty of interest and diy projects to build.
We need good models to get the projects as close to buildable as possible, to minimize unfortunate results and speed things up with proto and builds.
If you have good 3055/2955 models, that would be a good addition to your shared model lib.
Centralizing a collection of shared and known good models would really be great for diyers.
Did you mean the MJL21193/4 ones?
Much newer and beefy, and more expensive, not to mention possibly more difficult to procure for some people in some parts of the world.
We diyers are all over this blue planet, and unfortunately some have more difficulty to get certain things.
It's also cool to be able to design something with such old parts that can perform well in comparison with newer parts and attempt to use topologies not visited in the past when those old parts were all that was main stream.
I dug out one more model that was posted by Jan Didden some time ago for the 2n3055.
Comparing it to the others, it departs even more from a consensus.
That model has the highest VAF at 100, which may likely be nearer to the truth.
However, it has a BF of 360, which seems to be very excessive for the 3055. Compared to the others, 73 for the ST models, and 199.697 for the modpex 2n3055a, and 129.119 for the other modpex 2n3055g (h).
According to what I found about what the parameters are, BF being called ideal maximum forward beta, I assume this would be the maximum possible hfe, which according to the ST datasheet, could be anything from 50 to 250 (depending on group), and a datasheet from onsemi states it at 70.
So among those 5 models, the closest to a fairly representative, as far as max hfe is concerned, would be the ones from ST, with the BF=73.
One other parameter that varies hugely is IS:
IS=2.37e-8 for the ST models
IS=7.56605e-11 for the modpex 2n3055a
IS=2.37426e-14 for the other modpex 2n3055g
IS=4.66e-12 for the model posted by Jan Didden
So different. And that's just a few important parameters. The same thing is true on most others, not to mention some parameters not being present in some models, and the ST model having far too few of them to be any good.
Can we fix this?
I was trying to look up what IS would need to be. It depends on what they mean by saturation current, but I would assume it's the base current required to saturate the device, and this depends on voltage on vce, so that couldn't be a simple number, so I assume this parameter must have at least one other sibling that defines something else like a slope, and perhaps one for a knee or some roll off.
The datasheets really aren't too detailed, and we need more that what they provide to come up with all those parameters.
Comparing it to the others, it departs even more from a consensus.
That model has the highest VAF at 100, which may likely be nearer to the truth.
However, it has a BF of 360, which seems to be very excessive for the 3055. Compared to the others, 73 for the ST models, and 199.697 for the modpex 2n3055a, and 129.119 for the other modpex 2n3055g (h).
According to what I found about what the parameters are, BF being called ideal maximum forward beta, I assume this would be the maximum possible hfe, which according to the ST datasheet, could be anything from 50 to 250 (depending on group), and a datasheet from onsemi states it at 70.
So among those 5 models, the closest to a fairly representative, as far as max hfe is concerned, would be the ones from ST, with the BF=73.
One other parameter that varies hugely is IS:
IS=2.37e-8 for the ST models
IS=7.56605e-11 for the modpex 2n3055a
IS=2.37426e-14 for the other modpex 2n3055g
IS=4.66e-12 for the model posted by Jan Didden
So different. And that's just a few important parameters. The same thing is true on most others, not to mention some parameters not being present in some models, and the ST model having far too few of them to be any good.
Can we fix this?
I was trying to look up what IS would need to be. It depends on what they mean by saturation current, but I would assume it's the base current required to saturate the device, and this depends on voltage on vce, so that couldn't be a simple number, so I assume this parameter must have at least one other sibling that defines something else like a slope, and perhaps one for a knee or some roll off.
The datasheets really aren't too detailed, and we need more that what they provide to come up with all those parameters.
Hi Bob, thanks for droping in
This is for 3055 based amps, or 3055/2955. There are active threads and many hobbyists interested in building such amps. Myself included.
There are several projects that are a little different from the ordinary, such as grounded bridge (crown like), leach super amps, qsc grounded collector, and even bryston type output stage config to test out.
I'm not the only one having an old stock of good parts, and some people in some parts of the world don't have easy access to certain parts, but have lots of the old venerable 3055s.
So there is plenty of interest and diy projects to build.
We need good models to get the projects as close to buildable as possible, to minimize unfortunate results and speed things up with proto and builds.
If you have good 3055/2955 models, that would be a good addition to your shared model lib.
Centralizing a collection of shared and known good models would really be great for diyers.
Thanks for the response - these are all good points. I'm a little surprized that even a crappy model gives LTspice trouble, if I understood you properly.
Cheers,
Bob
Yes, I'm sorry, you are correct; I did mean MJL21193/4 devices.
BTW, I'm guessing that it might not be too hard to dumb down models for those devices to sort of look like at least an approximation to the 3055 and 2955 devices that would at least behave in LTspice. The first thing to dumb down would be ft, via increasing Tf.
Cheers,
Bob