Unusual SPICE model parameters?
Hi Bob
Just about to do some detailed simulations so I checked my SPICE models before I wasted any time.
When I compared with CORDELL-MODEL.TXT I noticed a few values that made me curious. I hope this is not too pedantic for you.
In the 4302 model the XTI (Exponent of saturation current with temperature) =1 rather than the default =3. This is determined by fairly fundamental semiconductor physics and seems it should not vary so far from ideal?
Similarly the XTB is =.12 : isn't this usually about 1, even if the default is zero?
There is an ISC value but this parameter is not discussed in your book. If you derived this value can you explain a bit? Or was it inherited from a supplied model? Also the ISC exponent NC=4. This is even more obscure! The default is =2 and it seems odd to alter so much without solid data.
Finally NR=1.1 This is close to the default. Why has it been altered?
Some of these parameters have similarly mysterious values in other models, the 4281, 3503, 1381 for instance, and I am not sure if this shows they are deliberate choices or just inherited from factory models that seem to be produced by random monkeys.
Best wishes
David
Hi Bob
Just about to do some detailed simulations so I checked my SPICE models before I wasted any time.
When I compared with CORDELL-MODEL.TXT I noticed a few values that made me curious. I hope this is not too pedantic for you.
In the 4302 model the XTI (Exponent of saturation current with temperature) =1 rather than the default =3. This is determined by fairly fundamental semiconductor physics and seems it should not vary so far from ideal?
Similarly the XTB is =.12 : isn't this usually about 1, even if the default is zero?
There is an ISC value but this parameter is not discussed in your book. If you derived this value can you explain a bit? Or was it inherited from a supplied model? Also the ISC exponent NC=4. This is even more obscure! The default is =2 and it seems odd to alter so much without solid data.
Finally NR=1.1 This is close to the default. Why has it been altered?
Some of these parameters have similarly mysterious values in other models, the 4281, 3503, 1381 for instance, and I am not sure if this shows they are deliberate choices or just inherited from factory models that seem to be produced by random monkeys.
Best wishes
David
Last edited:
Not any more than I could glean from an internet search or the LTSpice manual at the time. I would look up the function of the parameters and try to use it for its intended use. If you just leave out a parameter LTSpice uses the default, so I simply would not enter parameters I didn't understand or didn't need.
To be honest, many of my models were modified from vendor's models because it takes time and there are diminishing returns from getting a model more than close enough for horseshoes to the original device in most cases. Simulation can't be a substitute for measurement because even the model algorithms are still undergoing improvements and changes. Even when the simulation and models are accurate, user error tends to make a simulation inaccurate. In the end it is usually faster to build the prototype and adjust it rather than do most of the work in SPICE.
To be honest, many of my models were modified from vendor's models because it takes time and there are diminishing returns from getting a model more than close enough for horseshoes to the original device in most cases. Simulation can't be a substitute for measurement because even the model algorithms are still undergoing improvements and changes. Even when the simulation and models are accurate, user error tends to make a simulation inaccurate. In the end it is usually faster to build the prototype and adjust it rather than do most of the work in SPICE.
Accurate predictions from theory prove that you really do understand a subject.
When I research a Spice parameter I inevitably learn about what influence the modelled variable has in a real circuit.
My new(ish) soft start thread has a lack of simulation that reflects the lack of real data. Happen to know the primary inductance of a toroid power transformer of around 600VA? I don't have one handy, a test-it-and-see person like yourself should have one to measure
Best wishes
David
When I research a Spice parameter I inevitably learn about what influence the modelled variable has in a real circuit.
My new(ish) soft start thread has a lack of simulation that reflects the lack of real data. Happen to know the primary inductance of a toroid power transformer of around 600VA? I don't have one handy, a test-it-and-see person like yourself should have one to measure
Best wishes
David
Primary inductance is what you measure if the secondary is open circuit, so like a choke.
Any reasonable VA toroid? It would be of interest to see how this scales.
Bob's book does discuss this obliquely, just to stay on thread
Best wishes
David
Last edited:
Hi Dave,
Sorry for the late reply. My wife has me doing the bathrooms. I'll have to look into this one and get back to you. I've still got lots of notes from doing those models.
In some cases I did use manufacturer's numbers, at least as a starting point, where I had no compelling reason to do otherwise based on my measurements. Although I directly measured many things in my lab, I also used data sheet information where I could not determine otherwise. The resulting models were simulated for many different behaviors against the lab data and the datasheets. In some cases beta droop was difficult to model with the available SPICE parameters and I had to wiggle some to get a best fit. Often the compromise between beta droop at high current and beta at low currents is a difficult one.
I'll have to get back to you with more detail on your specific questions.
Cheers,
Bob
Thanks.
I assumed so, but the absence or presence of some parameters is notably erratic. I look forward to learn more.
The Spice parameter NK should help here. This is undocumented in LTSpice but seems to work.
Best wishes
David
I stand corrected. Seems it was undocumented in the earlier release I tried but is now included. It is not in Bob's book AFAIK and I assume this was the reason. Thank you.
Last edited:
Combining PDF's
Perhaps Jan, Scott or others could suggest something for the following.
My Schematic capture software generates multiple sheets. When I go to print the sheets to PDF using Bullzip, Bullzip generates individual PDF files. Is there a free tool that can assemble the PDF's into one PDF and organize it like what you would find in a multi page manual?
Perhaps Jan, Scott or others could suggest something for the following.
My Schematic capture software generates multiple sheets. When I go to print the sheets to PDF using Bullzip, Bullzip generates individual PDF files. Is there a free tool that can assemble the PDF's into one PDF and organize it like what you would find in a multi page manual?
Indeed, that's also what I use for that.
Also allows you to string a bunch of JPGs into one PDF document.
jan
I was looking through the LTSpice changelog and I found that the HiSiM-HV model has been added. I haven't been able to find very accessible info, but according to this it seems that it supports subthreshold conduction:
SPICE Model Parameters for the HiSIM2 model - IC-CAP 2012 -Agilent EEsof Knowledge Center
EDIT: But according to the bottom of the page that parameter was actually removed later on. Maybe the newer parameters work better? Or did they just ruin the model?
SPICE Model Parameters for the HiSIM2 model - IC-CAP 2012 -Agilent EEsof Knowledge Center
EDIT: But according to the bottom of the page that parameter was actually removed later on. Maybe the newer parameters work better? Or did they just ruin the model?
Last edited:
I have recently purchased Bob's designing Audio Power Amplifiers to learn more about this hobby I'm interested. This may seem like a dumb question to many here, but....
On page 60 the second paragraph mentions using the darlington VAS circuit "The inclusion of Q12increases the input impedance of the new vas to about 40KOhm"
Could someone please show me mathematically how the figure of 40kohm input impedance was reached?
Regards
Simon
On page 60 the second paragraph mentions using the darlington VAS circuit "The inclusion of Q12increases the input impedance of the new vas to about 40KOhm"
Could someone please show me mathematically how the figure of 40kohm input impedance was reached?
Regards
Simon
Hi Simon,
Thanks for buying my book and I hope you enjoy it.
Q12 functions as an emitter follower stage. If an emitter follower transistor has a current gain of 100, the input impedance seen at its base will be roughly 100 times that seen by its emitter. This is because the signal voltages at the input and output of the EF are nearly the same, while the base signal current is about 1/100 that of the emitter signal current.
In Figure 3.7, the emitter of Q12 sees a load of 470 ohms in parallel with the load seen at the base of Q4. If Q4 has a current gain of 100, the load seen looking into its base would be on the order of 100 times the value of its emitter resistor of 22 ohms, or about 2200 ohms. So the load seen by the emitter of Q12 is roughly equal to the parallel combination of 470 ohms and 2200 ohms, which is about 390 ohms. That multiplied by 100 is almost 40k ohms.
Cheers,
Bob