Wow, that's very cool Jon. It works! The big difference I see is the value of the Rbe in the darlington pair. Although it looked fishy, I plugged it in as shown in the Multicomp datasheet, http://www.farnell.com/datasheets/296716.pdf. Now I see that STMicro has a different opinion, one that aligns with yours, http://www.st.com/stonline/books/pdf/docs/4132.pdf.
Other than moving that pesky capacitor, did you do anything else?
Thanks tons!
Other than moving that pesky capacitor, did you do anything else?
Thanks tons!
Yeah, this corrected circuit works in all three simulators I was trying. Silly me. I built this amp and it worked perfectly. I just couldn't see my drawing error. Thank goodness I'm a little better with the soldering pencil. I am really grateful for your help.
Hey, this software simulation stuff is really cool! Someone ported Falstad's circuit simulator to a smart phone. How cool is that!?
Hey, this software simulation stuff is really cool! Someone ported Falstad's circuit simulator to a smart phone. How cool is that!?
The main problem with your simulations was the VAS connection.
Also the power transistors you'd chosen aren't a complimentary pair which resulted in asymmetry in the output so I just left them as generic npn and pnp. Some of the capacitors values appeared with some extended character set characters instead of u in the uf so I corrected these and changed the input filter to a non-polar 100pF. I connected input and output test points and change the simulation parameter to do a 1 second transient run instead of a frequency sweep. That's pretty much the extent of my changes.
I didn't change the topology of the output stage but I would recommend you use some different transistors for the VAS and the output stage and then run the simulation again. There is some crossover distortion at the moment because the biasing of those 3 diodes isn't correct.
Oh and if you need help with LTSpice and can't get it here there is a good group on Yahoo.
Also the power transistors you'd chosen aren't a complimentary pair which resulted in asymmetry in the output so I just left them as generic npn and pnp. Some of the capacitors values appeared with some extended character set characters instead of u in the uf so I corrected these and changed the input filter to a non-polar 100pF. I connected input and output test points and change the simulation parameter to do a 1 second transient run instead of a frequency sweep. That's pretty much the extent of my changes.
I didn't change the topology of the output stage but I would recommend you use some different transistors for the VAS and the output stage and then run the simulation again. There is some crossover distortion at the moment because the biasing of those 3 diodes isn't correct.
Oh and if you need help with LTSpice and can't get it here there is a good group on Yahoo.
Right, I knew the transistors weren't right. The standard libraries have so little. I joined the Yahoo group & learned how to add the SUBCKT darlngtn.lib library so now I can use the TIP142/147 pair. Also updated the standard.bjt & standard.dio.
Still, the transistor library is missing many power and most Japanese transistors. I probably don't totally understand the LTspice library structure, but it seems kludgy. It's a bit difficult adding entire libraries. When you do, you don't get to pick your part, you have to type it in, which means you have to memorize everything that's in the new library. To get the selection option, you have to add the parts one by one or risk duplication. Looks like library maintenance will be a big part of the effort with this simulator.
The character set problem is maybe because I'm running LTspice on Linux under a WINE environment. It does not appear to cause any problems so LTspice must be aware of the Linux character set.
As for topology, I'm not working on the amp, I'm working on short circuit protection schemes. So I wanted a simple amp to start with.
Anyway, thanks for getting me unstuck. Now I'm off & running, albeit at a snail's pace.
Still, the transistor library is missing many power and most Japanese transistors. I probably don't totally understand the LTspice library structure, but it seems kludgy. It's a bit difficult adding entire libraries. When you do, you don't get to pick your part, you have to type it in, which means you have to memorize everything that's in the new library. To get the selection option, you have to add the parts one by one or risk duplication. Looks like library maintenance will be a big part of the effort with this simulator.
The character set problem is maybe because I'm running LTspice on Linux under a WINE environment. It does not appear to cause any problems so LTspice must be aware of the Linux character set.
As for topology, I'm not working on the amp, I'm working on short circuit protection schemes. So I wanted a simple amp to start with.
Anyway, thanks for getting me unstuck. Now I'm off & running, albeit at a snail's pace.
The choice of transistor models in LTSpice may be limited but don't forget Linear produced the tool to assist with simulation of their opamps and regulators etc. not as a general purpose simulator (although this is how many of us it - and thanks to Mike for that) !
If I want to simulate a complex circuit with many new parts I'll create a sub file and include that on the schematic but more often than not I simply add a spice directive 'S' and put a model for a transistor I want to use directly onto the schematic. If you can only find subckt definitions than create a new assembly and use that to reference the part.
If I want to simulate a complex circuit with many new parts I'll create a sub file and include that on the schematic but more often than not I simply add a spice directive 'S' and put a model for a transistor I want to use directly onto the schematic. If you can only find subckt definitions than create a new assembly and use that to reference the part.
Turns out the library editing was much easier than I expected. I found a bunch of pspice libraries on the web and just added a huge number of 2SA..2SD devices and most of the TIP devices. There wasn't much in the way of conflicts to deal with. It's all good!
No haven't tried winetricks, but just downloaded it. There's a ton of stuff there. What are the essentials?
No haven't tried winetricks, but just downloaded it. There's a ton of stuff there. What are the essentials?
Linux doesn't have all the required fonts. Windows fonts aren't open-source so WINE can't distribute them. Winetricks will download them from the microsoft site.
Mike has tried to get LTSpice to work on WINE, so installing windows libraries won't necessarily get you anywhere. There may be some peripheral things like fonts that will make it easier. Winetricks can also enable font smoothing like on some versions of windows. Just look through the list, it'll come together.
- keantoken
Mike has tried to get LTSpice to work on WINE, so installing windows libraries won't necessarily get you anywhere. There may be some peripheral things like fonts that will make it easier. Winetricks can also enable font smoothing like on some versions of windows. Just look through the list, it'll come together.
- keantoken
Hi keantoken,
Thanks for the kind words about the library I put up on my site. I was originally inspired by Andy Connors' work on power transistors, where he looked at just how bad the vendors' models were, and set out to make some better ones. I got involved because I felt I needed better models than vendor models to do semi-decent simulations for my book. Figuring out how to create those needed models actually led to the chapter on how to make models.
I have to say the whole adventure was a humbling experience. After I completed the book, I set out to tweak some of the small-signal transistor models I had been using and was frustrated to learn that they were not that great, either. It was a lot of work to measure and get the models at least close in what seemed to be the more important areas. There were some eye-openers revealed by my measurements of real devices. In many cases I found the manufacturers' number for Early effect, VA, to be way off.
I'll soon be posting an update to the models and adding some. One area where I have been investing effort has been to get RB closer to correct. This is another one where vendor's usually punt. I learned how very difficult it is to infer RB. I originally tried to do it with DC measurements of excess Vbe. That works pretty well with power transistors, but not so well with small-signal transistors. Wrong RB values may give wrong amounts of excess phase in the transistor model and wrong noise simulation measurements. So I have actually been using sensitive noise measurements to better infer RB. So far the results are encouraging.
Cheers,
Bob
Hello Bob!
I didn't know there were any models there until recently when a friend pointed them out. I was glad to learn that the idea of DIY models wasn't totally lost.
I often find myself fixing crazy errors in models I use. For example, the KSC1845 model gives an RC of 180R. Because of this, VceSAT was broken and current mirrors wouldn't work. For a newcomer who expects a vendors' own model to be accurate, it is a put-off to 'discover' that the fast, low-noise transistor won't work for their project.
Good luck with the models. It is a service to the DIY community.
- keantoken
I didn't know there were any models there until recently when a friend pointed them out. I was glad to learn that the idea of DIY models wasn't totally lost.
I often find myself fixing crazy errors in models I use. For example, the KSC1845 model gives an RC of 180R. Because of this, VceSAT was broken and current mirrors wouldn't work. For a newcomer who expects a vendors' own model to be accurate, it is a put-off to 'discover' that the fast, low-noise transistor won't work for their project.
Good luck with the models. It is a service to the DIY community.
- keantoken
your problem here seems to be that the base of Q6 should connect to the collector of Q7, not the emitter. (that's the most obvious one, the implementation of the darlingtons is another one)
Thanks Jed, Jon got me pointed in the right direction.
I loaded Bob's models and look forward to using them.
Got sidetracked by winetricks. Loaded some wine components and kind of messed up my system. I'm not sure which of the tricks did this, but my LTspice toolbar icons are unreadable because of the colors. I've been trying to figure this out without luck. I have several WINE apps that are doing this now. I think it has to do with the use of transparent color which is rendered as black.
LTspice seems to be the only app I have that uses transparent color in the toolbar so it's the one that's a problem.
I'm guessing this is a transparency rendering issue.
Here's what my LTspice menu looks like in WINE.
I loaded Bob's models and look forward to using them.
Got sidetracked by winetricks. Loaded some wine components and kind of messed up my system. I'm not sure which of the tricks did this, but my LTspice toolbar icons are unreadable because of the colors. I've been trying to figure this out without luck. I have several WINE apps that are doing this now. I think it has to do with the use of transparent color which is rendered as black.
LTspice seems to be the only app I have that uses transparent color in the toolbar so it's the one that's a problem.
I'm guessing this is a transparency rendering issue.
Here's what my LTspice menu looks like in WINE.
winetricks trick: load only one download at a time. if a dll or something messes up wine, you know which one did it...
still, having the base of your negative side output effectively tied to the negative rail through a 33 ohm resistor is going to make the sim not work... it should be on the collector of the VAS, which is where the voltage swing is developed
still, having the base of your negative side output effectively tied to the negative rail through a 33 ohm resistor is going to make the sim not work... it should be on the collector of the VAS, which is where the voltage swing is developed