True, however uC net lists are .cir and not (yet?) allowed to be uploaded. And as I remarked earlier, the net list formats of uC and LT aren't compatible. Could a moderator add .cir to the list of allowable file extensions? I got plenty of examples to play with.
)
Good idea but much work. One file to exercise: zero global NFB hybrid amp with plenty of gain to add it. While in THD analysis mode, increase bias from 3.65 to 3.85 V for a better result.
Not without a lot of work. Check the link that Edmond posted earlier to his MicroCap models for an example of how to do it. Ian's models also use the LTspice VDMOS model temperature coefficients. This would require further work to emulate in MC. Ian's own temp.co work in LTSpice before the coefficients were added could provide an example.
Best wishes
David
The laterals are part of the uC mosfet library. 2SJ120, 2SK1058 etc. also are present as well as depletion-mode verticals D2530...3545.
This makes your results even less plausible, looks like a serious error in your interpretation or the simulation, or both. The IMD plot is for ~2 kHz, Return Ratio for 20 kHz harmonics is far less so distortion will be far worse, approximately in proportion to the loss of RR. So 2 PPM is physically unrealistic but consistent, PPB doesn't make sense. It's also massively better than anyone has ever approached. So you, fairly quickly, have far surpassed, say, Audio Precision with decades of research and a multi-million dollar R&D spend. Does this seem plausible to you?
If we look at your previous post #11 we find a very ordinary, primitive VAS and OPS, with slow transistors and it is supposedly sub PPM distortion.
Seriously?
Best wishes
David
The FFT is just an algorithm to compute the Fourier transform efficiently. In fact, everyone uses the FFT. The old-style transform just takes longer to first compute and then throw away unneeded intermediate results. Final results are the same.
But LTspice works for me; I see no need to switch to a product that probably will no longer maintained sooner or later.
If anything, I'll write a bridge from Altium Designer to ng-spice, so I don't need two copies of the same circuit.
Last edited:
Active component models are fully interchangeable. I liked both programs. In Microcap it is also convenient to draw in your own way, and even have interesting chips. For example, you can assign resistors not the resistance value, but the value of the previous resistor type R (R7).
Last edited:
Who cares it will no longer be maintained. To me version 10 is good enough, while later version are already a bit over-engineered with another bunch of new features I don't need. The latest version (of any program) is not always the best as well. What matters are the models which are supplied by third parties anyway.
Cheers, E.
I doubt the original SPICE ever used the DFT. I remember using it by hand a few times because the Fourier analysis (at the time) forced power of 2 points, which in itself would hint that it was using FFT's.
FFT
What do you mean by "The old-style transform"? The original algorithm, which involves N2 operation versus the Cooley-Tukey algorithm which reduces it to NlogN operations, or ...
do you mean a DFT is essentially a transform on complex numbers. So when applied on real numbers, half the number of operations and memory is wasted. But nobody will do it that way. Instead, one splits the array of real-valued data and stores it in an complex array of length N/2. (for details see: William H. Pess et al, Numerical Recipes in C, 2nd edition, pp. 510). This way it is just as efficient as, for example, the fast Hartley transform (FHT), which involves only real-valued data by design.
Cheers, E.
# a bit off-topic
What do you mean by "The old-style transform"? The original algorithm, which involves N2 operation versus the Cooley-Tukey algorithm which reduces it to NlogN operations, or ...
do you mean a DFT is essentially a transform on complex numbers. So when applied on real numbers, half the number of operations and memory is wasted. But nobody will do it that way. Instead, one splits the array of real-valued data and stores it in an complex array of length N/2. (for details see: William H. Pess et al, Numerical Recipes in C, 2nd edition, pp. 510). This way it is just as efficient as, for example, the fast Hartley transform (FHT), which involves only real-valued data by design.
Cheers, E.
Hi David,
Just busy with other things (DiAna and ....... tubes).
Cheers, E.
Keep DiAna going, I'm a fan!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.