You don't have to. That's my entire point. If you follow the procedure I listed using the posts I linked, you will get the same results I did.
Indeed, if you follow that procedure, it will show without ambiguity that the models are identical. Edit: FC actually does a binary file compare, so if as small a change as an extra enter key or space character is present, it will show the files to not be identical. You can verify this behavior too, without having to take my word for it.
If you were talking about some other file (say, the .asc file instead of the .txt), please say what it is and provide links to the relevant posts as I did.
Last edited:
I downloaded the files you posted, ran them then clicked Edit Net list for both files. I then did a line by line comparison of the two. There are differences in some of the parameters, and not small differences. I knew I couldn't post the files here after I had viewed them, and be believed.
http://www.diyaudio.com/forums/solid-state/151295-krill-next-generation.html#post1918262
http://www.diyaudio.com/forums/solid-state/151295-krill-next-generation-11.html#post2010053
The files in question are at these links. I didn't realize you couldn't find page one of this post or the later file you posted. I should have been more specific.
The best way is to build, measure and listen
But, 23-26mV across Re is really right for B/AB output stage, not any 5mV - simulation would be correct here.
Okay, so we're both talking about the same posts. That's good. You did say I changed the models, but note that every model referenced in both schematics is either a built-in LTspice model (1N4148), or a model that's in the attached KSA_MJLmodels.txt. You can verify that the two different KSA_MJLmodels.txt are identical by the procedure I mentioned earlier. Of course the netlists will be different, as I added an emitter ballast resistor to the drivers to demonstrate the difference between the behavior with the resistor as an open (actually 1e12 Ohms) and the 220 Ohm value you specified. Other than the addition of the driver emitter ballast resistor, the only other change was one originally made by you, which replaces the 0.22 Ohm output stage emitter resistors with 0.1 Ohm. That of course makes the nominal "off" currents of the output devices larger.
Rather than looking at the netlists, it's much easier to bring the .asc file into the schematic editor to compare. The netlist gets regenerated on every simulation anyway.
You'll find also that LTspice has an annoying habit of renumbering/renaming all nodes even when making rather trivial edits. I use it for schematic capture for PCB layout, and this makes it so back-annotation is impossible. If I need to make a change to my PCB netlist after the original import from LTspice, I need to do it in FreePCB using its netlist editor. Once that netlist is imported into FreePCB, there's no going back to LTspice to edit it, because of this property.
Last edited:
No, you are amazing. After so much time, I'm still asking myself WTF is at stake here?
Andy,
Sorry I came on a little strong earlier. I have run a compare 6 times now, and found a difference every time. I will continue to look into why this happens. I will run both files again in LTspice. If there is no difference in the results, then it doesn't matter to me if they are different.
Sorry I came on a little strong earlier. I have run a compare 6 times now, and found a difference every time. I will continue to look into why this happens. I will run both files again in LTspice. If there is no difference in the results, then it doesn't matter to me if they are different.
Not sure what's going on there. I re-ran and I'm showing them as identical. I'm using the freeware 7-zip software to extract the files. When extracting the file, the original time/date stamp is preserved. It shows 3/7/2009 6:30PM for both. I'm not sure if that ends up being adjusted for the user's local time or not. That file was originally created by ostripper (who created the first Krill LTspice sim AFAIK), and I edited it slightly back then because some syntax errors were showing up when running the simulation. If I remember right, the syntax errors were due to some comments in the file that didn't begin with the required * character.
Oh, and just to double-check, the original posts are #11 and #210.
Last edited:
I must apologize to Andy for my behavior and rant yesterday. I download both files again, ran a compare and it still showed a difference. I began to look for possible causes. I had noticed that the LTspice files had the same avatar as one of my cad programs. If I clicked on the file, this program would start up and try to open the file. I have now uninstalled that program. The program was BobCAD-CAM, a program I can now do without. It wrote CNC code from drawing files. I assume at this point that that program was making some change to a file it tried to open. Now that that program is gone, the files are fine.
I did check the sim results, and the difference between files did not change the results at all.
I had wondered why LTspice ran so slowly. It now runs several times faster. I don't know what BobCAD was doing, but LTspice didn't like it.
Now a question. Does LTspice have a math capability hidden away somewhere? It would be nice to just ask for minimums or maximums instead of using the courser to find them.
I did check the sim results, and the difference between files did not change the results at all.
I had wondered why LTspice ran so slowly. It now runs several times faster. I don't know what BobCAD was doing, but LTspice didn't like it.
Now a question. Does LTspice have a math capability hidden away somewhere? It would be nice to just ask for minimums or maximums instead of using the courser to find them.
Thanks for your post Steve.
I'll be wrapping up my participation in this thread with two posts. This one is the first. It will simply catalog the two simulation files we've been discussing and talk about what's the same and what's different about them. The second post will summarize my views on the Krill operation, which I'm sure will differ from yours. We'll simply have to "agree to disagree" on this, as the saying goes.
I reached this decision after seeing the time/date stamp on the KSA_MJLmodels.txt files below, and realizing they went back to March of this year.
First file, from post #11 = krill_again.zip.
Contents:
Krill_OPS.asc = LTspice simulation file
KSA_MJLmodels.txt = All models for the above (except 1N4148, which is built in).
Second file, from post #210 = Krill_ops.zip
Contents:
Krill_OPS.asc = LTspice simulation file
KSA_MJLmodels.txt = All models for the above (except 1N4148, which is built in).
The Krill_OPS.asc from post #210 differs from the Krill_OPS.asc from post #11 in two ways:
a) Output device emitter resistors were changed from 0.22 Ohms in the earlier file to 0.1 Ohms in the later file. This was to match Steve's change and has the effect of increasing the output stage "off" currents from the earlier simulation. Bias voltage adjustment is the same between the files.
b) A driver emitter resistor which is stepped from 220 Ohms to 1e12 Ohms (open circuit) was added to the later file. This is to illustrate the effect of this resistor on turnoff behavior.
As for the KSA_MJLmodels.txt files, they are absolutely bit-for-bit identical between the two simulations. This was verified by re-downloading the files and running the command-line file compare (FC) utility on the unzipped KSA_MJLmodels.txt from each one. If you are still getting indications of different KSA_MJLmodels.txt files here, something has gone wrong.
Of course, the two .asc simulation files are different as they must be, because of the changes described above in (a) and (b).
I'll be wrapping up my participation in this thread with two posts. This one is the first. It will simply catalog the two simulation files we've been discussing and talk about what's the same and what's different about them. The second post will summarize my views on the Krill operation, which I'm sure will differ from yours. We'll simply have to "agree to disagree" on this, as the saying goes.
I reached this decision after seeing the time/date stamp on the KSA_MJLmodels.txt files below, and realizing they went back to March of this year.
First file, from post #11 = krill_again.zip.
Contents:
Krill_OPS.asc = LTspice simulation file
KSA_MJLmodels.txt = All models for the above (except 1N4148, which is built in).
Second file, from post #210 = Krill_ops.zip
Contents:
Krill_OPS.asc = LTspice simulation file
KSA_MJLmodels.txt = All models for the above (except 1N4148, which is built in).
The Krill_OPS.asc from post #210 differs from the Krill_OPS.asc from post #11 in two ways:
a) Output device emitter resistors were changed from 0.22 Ohms in the earlier file to 0.1 Ohms in the later file. This was to match Steve's change and has the effect of increasing the output stage "off" currents from the earlier simulation. Bias voltage adjustment is the same between the files.
b) A driver emitter resistor which is stepped from 220 Ohms to 1e12 Ohms (open circuit) was added to the later file. This is to illustrate the effect of this resistor on turnoff behavior.
As for the KSA_MJLmodels.txt files, they are absolutely bit-for-bit identical between the two simulations. This was verified by re-downloading the files and running the command-line file compare (FC) utility on the unzipped KSA_MJLmodels.txt from each one. If you are still getting indications of different KSA_MJLmodels.txt files here, something has gone wrong.
Of course, the two .asc simulation files are different as they must be, because of the changes described above in (a) and (b).
Last edited:
As to my views on the Krill operation, this post summarizes them.
1) There is absolutely nothing wrong with this design from what I can see. The discussion so far has centered around how the circuit itself operates, which of course has led to much controversy. People wishing to build it should certainly do so, and as a bonus it seems that Steve has provided good support to the builders.
2) Except for negligibly small differences, the Krill biasing scheme behaves no differently than a standard power amp shunt regulator. In other words, it does not actually affect whether the output devices turn off or not. This was demonstrated by Glen in post #204. In that post, Glen compared two Krill variants. One had the Krill biasing scheme, and the other an ideal constant voltage source for the bias shunt regulator. In his results, the output stage currents between the two cases had a negligibly small difference. Ergo, the Krill biasing scheme has a negligible effect on turn-off behavior.
3) The soft turn-off behavior of the Krill is due entirely to the absence of a driver emitter bias resistor, as first mentioned by Glen in post #8. This was later shown by me in simulation in post #210, and then further expanded by Glen in post #229. Post #210 shows a graph with two traces. The blue trace, representing the driver emitter resistor as an open circuit, is standard Krill operation, in which the output device emitter current drops down to a couple of uA and exhibits the soft turn-off. The green trace, showing a 220 Ohm driver emitter resistor, shows the full turn-off per standard class AB operation. The different configurations simulated in post #204 and post #210 demonstrate that the soft turn-off behavior of the Krill is due entirely to the absence of a driver emitter resistor. That is, post #204 shows that the Krill biasing arrangement has negligible effect on turn-off, and post #210 shows that changing only the driver emitter resistor (between open-circuit and some other value) fully accounts for the Krill soft turn-off effect.
1) There is absolutely nothing wrong with this design from what I can see. The discussion so far has centered around how the circuit itself operates, which of course has led to much controversy. People wishing to build it should certainly do so, and as a bonus it seems that Steve has provided good support to the builders.
2) Except for negligibly small differences, the Krill biasing scheme behaves no differently than a standard power amp shunt regulator. In other words, it does not actually affect whether the output devices turn off or not. This was demonstrated by Glen in post #204. In that post, Glen compared two Krill variants. One had the Krill biasing scheme, and the other an ideal constant voltage source for the bias shunt regulator. In his results, the output stage currents between the two cases had a negligibly small difference. Ergo, the Krill biasing scheme has a negligible effect on turn-off behavior.
3) The soft turn-off behavior of the Krill is due entirely to the absence of a driver emitter bias resistor, as first mentioned by Glen in post #8. This was later shown by me in simulation in post #210, and then further expanded by Glen in post #229. Post #210 shows a graph with two traces. The blue trace, representing the driver emitter resistor as an open circuit, is standard Krill operation, in which the output device emitter current drops down to a couple of uA and exhibits the soft turn-off. The green trace, showing a 220 Ohm driver emitter resistor, shows the full turn-off per standard class AB operation. The different configurations simulated in post #204 and post #210 demonstrate that the soft turn-off behavior of the Krill is due entirely to the absence of a driver emitter resistor. That is, post #204 shows that the Krill biasing arrangement has negligible effect on turn-off, and post #210 shows that changing only the driver emitter resistor (between open-circuit and some other value) fully accounts for the Krill soft turn-off effect.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.