Krill - The Next Generation

[AKSA]

Syn08,

I think it WAS challenged if the outputs switched or not. That was the principle part of the challenge, in fact.

You say it does not matter. You may be right. Would you mind elucidating your reasons?

Any comparison with an EF triple would have required back to back testing. As I recall, no one had both amps and so the question was academic. Of itself, the question makes no factual statement, of course.

The distortion figure of 50ppm is certainly low, but wasn't it simply a simulation figure?

[syn08]

Originally Posted by AKSA

You say it does not matter. You may be right. Would you mind elucidating your reasons?

Any comparison with an EF triple would have required back to back testing. As I recall, no one had both amps and so the question was academic. Of itself, the question makes no factual statement, of course.

The distortion figure of 50ppm is certainly low, but wasn't it simply a simulation figure?

Krill - The Next Generation, I couldn't put it better.

For the rest, at this point, we are talking simulations only. That's why Steve posted the models, to align as much as possible the results from various simulators.

[Samuel Jayaraj]

Originally Posted by AKSA

Syn08,

Any comparison with an EF triple would have required back to back testing. As I recall, no one had both amps and so the question was academic. Of itself, the question makes no factual statement, of course.

I recall that OStripper tested both the Krill and EF triple at one point in time.

[AKSA]

Thanks, I will check all links.

Hugh

[andy_c]

Originally Posted by syn08

Krill - The Next Generation, I couldn't put it better.

For the rest, at this point, we are talking simulations only. That's why Steve posted the models, to align as much as possible the results from various simulators.

I challenged the "no device turn-off" claim earlier in this thread, and posted a simulation that anyone with the free LTspice could run. It showed currents in the nominally "off" device on the order of microamps - leakage currents and leftovers of stored charge. I do agree though, that whether or not they do turn off is pretty much irrelevant to the distortion performance, just as Scott said. Steve's posted simulation results showed a small current in the nominally "off" transistor also, but it was larger than the currents in the simulation I posted, maybe a few mA?

When I asked Steve to post the simulation files, he did so after a time. But when I asked him for specifics, he stated it was for a circuit that was different from the one he posted simulation graphs from, and that the original one for which he posted graphs was no longer available. The simulation file he posted was a netlist, which requires a time-consuming process to get back to a schematic from which an equivalent LTspice simulation could be created. So I asked him to post a schematic annotated with node names to help expedite this, hoping to then have him re-run and post the simulation results for this second circuit to try to correlate results. He then posted schematics and a netlist for a third circuit, which was a complete amplifier, not just the output stage whose behavior was in question. Having the extra front end circuitry of course makes the translation from netlist to LTspice simulation files much more time-consuming, in addition to their behavior being completely irrelevant to the issue at hand. Also, the output stage of this schematic was different from the first two simulations Steve posted. Further, the schematics of this third circuit were chopped up into pieces in which the irrelevant part (the input stage) was intact, but some of the relevant parts (parts of the output stage) missing. Portions of the output stage of this schematic was just garbled nonsense that one would have to be literally blind to not notice a problem.

At that point, I realized that Steve could, and likely would, continue this cat and mouse game indefinitely. So I decided to not waste any further time on it.

[Steve Dunlap]

Originally Posted by syn08

Not as far as I recall.

First, it was challanged that the outputs not switching off has any relevance for the crossover distortions. It does not.

Secondly, it was challanged that the Krill output stage will do anywhere significantly better (that is, to justify the extra complexity) compared to a EF triple (like the Locanthi), biased near the "Bernie Oliver" point.

Third, the posted performance (of 50ppm @ Ic=25mA, if memory serves) was challenged.

First, I don't recall stating that the outputs not switching off has any relevance for the crossover distortions. I did claim that the outputs were conducting through out the entire cycle of the wave form.

Secondly, I don't think using six parts in the bias circuit makes the output stage extra complex as opposed to 8 as in the D. Self amp. I did not claim the Krill did better than many other designs. I simply mentioned what I had measured on some of the amps I built.

Third, yes it was challenged. Since I never posted a schematic of that amp and never claimed it to be biased at 25ma, that challenge, as well as the conclusions drawn, is based on incomplete or incorrect information.

[Steve Dunlap]

Originally Posted by syn08

Krill - The Next Generation, I couldn't put it better.

For the rest, at this point, we are talking simulations only. That's why Steve posted the models, to align as much as possible the results from various simulators.

I,m not sure where that link was supposed to point.

Yes, that is why I posted my models and the schematic of the output stage.

[Steve Dunlap]

challenged the "no device turn-off" claim earlier in this thread, and posted a simulation that anyone with the free LTspice could run. It showed currents in the nominally "off" device on the order of microamps - leakage currents and leftovers of stored charge. I do agree though, that whether or not they do turn off is pretty much irrelevant to the distortion performance, just as Scott said. Steve's posted simulation results showed a small current in the nominally "off" transistor also, but it was larger than the currents in the simulation I posted, maybe a few mA?

As a follow up to my previous question, here is what my simulator says. This was run on the amp that I started this thread with and is the one I will be posting the schematic for later. The first 3 are at rated output of 50W into 8 ohms. This is the minimum current in each emitter resistor at the listed frequencies. You can see how this could lead me to believe the outputs stay on. These numbers lead me to believe GK might be correct at some sufficiently high frequency, but it should be well outside the audio band.

8.87827M amps Minimum at 20 KHz

9.98479M amps Minimum at 1K Hz

10.0355M amps Minimum at 20 Hz

9.99978M amps at DC with 28.3VDC at output with 8 ohm load

When I asked Steve to post the simulation files, he did so after a time.

Steve, take as long as you want. Seeing as how it took me over two weeks to respond to this, I've got no room to talk. I hope your basement thing gets taken care of soon, and that the maggies and other items are okay.

But when I asked him for specifics, he stated it was for a circuit that was different from the one he posted simulation graphs from, and that the original one for which he posted graphs was no longer available.

I threw everything out and started over. Fortunately, parts are cheap in simulation. I even changed the brand of some of the transistors, so these numbers won't be exactly the same as the previous amp.

I don't recall saying the original one was no longer available. "Threw every thing out" was simply a figure of speech.

The simulation file he posted was a netlist, which requires a time-consuming process to get back to a schematic from which an equivalent LTspice simulation could be created. So I asked him to post a schematic annotated with node names to help expedite this, hoping to then have him re-run and post the simulation results for this second circuit to try to correlate results. He then posted schematics and a netlist for a third circuit, which was a complete amplifier, not just the output stage whose behavior was in question. Having the extra front end circuitry of course makes the translation from netlist to LTspice simulation files much more time-consuming, in addition to their behavior being completely irrelevant to the issue at hand. Also, the output stage of this schematic was different from the first two simulations Steve posted. Further, the schematics of this third circuit were chopped up into pieces in which the irrelevant part (the input stage) was intact, but some of the relevant parts (parts of the output stage) missing. Portions of the output stage of this schematic was just garbled nonsense that one would have to be literally blind to not notice a problem.

At that point, I realized that Steve could, and likely would, continue this cat and mouse game indefinitely. So I decided to not waste any further time on it.

Guys, Steve has provided all the models, so that's a non-issue. They are in the netlist file attached to this post. Just have a look with a text editor. That's why I'm very optimistic about being able to duplicate his results.

I can reverse engineer the schematic from the netlist if need be.

Here is the output stage only. I simply deleted the VGS. Some of the node numbers were changed by the program when I did this. I have always used a higher voltage regulated supply on everything but the drivers and outputs, so I have left the 40V supplies in. If you choose to use only the 34V supplies, you will need to add the connections so the front of the OPS gets voltage. This will change the measurements I posted earlier, but not much. The DC on the input is to adjust the output offset. This was done before in the VGS. The 277 ohm resistor in series with the input represents the output impedance of the VGS. I put it there to be complete, it is not needed.
Attached Files
Export for Andy C2.pdf (53.0 KB, 66 views)

You have had the information for 28 days now. I didn't realize I was playing a cat and mouse game. Would you care to address this question?

What does surprise me is that changing the bias resistor from Rbias to 4200 ohm changed the results. Does anyone have an explanation for why this happens?

[tinitus]

Originally Posted by Steve Dunlap

I didn't realize I was playing a cat and mouse game.

I was indeed very surprised
Sounds like prejustice that wont change

[AKSA]

Jeez, it's on again......

More recriminations, more bile...... why not forget about it. Build the amp, listen to it, measure it, objectively report the results, refrain from absurd technical one upmanship.

In a world racked by poverty, war and disease it's not that important.

Still think it's a clever system, Steve, doffs me cap.